what makes you feel thunderbolt is so great? I mean you've stated before that your opinion is if existing solutions are removed things will be forced forward. My opinion would be if this is really is a niche market, it's unlikely that many of these companies would allocate the necessary development costs to completely reworking something.
I really don't like redundancy. When I see a numbers row and a numpad on my keyboard, I don't like it. For years I've looked at the myriad of different ports and protocols and wished that one day there would be a more unified I/O setup that covered display data. Using magnets too, although that might not be feasible - the description is here:
It's all data so you really just need a single multi-protocol connection. Thunderbolt achieves that. That's what makes it better than all the other I/O standards besides the speed and latency.
As for PCI slots, I've never liked those because they dictate the size of the peripheral. If you need a FW800 port, you get a card that takes up an entire slot. With Thunderbolt, it's just a tiny plug.
PCI isn't plug and play either. You have to shut down your machine, open it up and install a peripheral. They used to do TV tuners this way. Once they realised how much of a pain it was to install, saw the dwindling desktop user base and got USB 2, they switched to USB.
It was a similar thing with A/V and Firewire. Video cameras could have shipped with PCI boards but it wouldn't have been a good solution. I can see Thunderbolt becoming the new standard on A/V equipment (low latency live capture over 100m optical for example) along with USB 3:
PCI slots have the speed advantage for now but it will take longer to evolve internal motherboard slots than an external port. A port change doesn't have to mean obsolete equipment but a change from PCIX to PCIe means an obsolete card. Then you have half-length slots, double-wide slots, different power limits and driver support. You still have driver support to contend with but on the Mac side, your potential audience is multiplied by 20 and you can test/develop the driver on a $600 computer.
We go through transitions over time. People didn't like IDE/PATA when it came on the scene because it wasn't as fast as SCSI but you didn't have to assign drive IDs and deal with conflicts. By the time SATA arrived in 2003, nobody had been talking about it for years.
The same has happened with optical drives being removed. The odd few jump up and shout 'hey, I was using that, you can't get rid of it' and looks around to realise he's in the small minority and can get an external drive. The entire Mac Pro audience is a tiny minority now and the people using the PCI slots for custom cards a fraction of them and people using cards other than I/O cards and cards that have TB equivalents a further fraction of them.
By the time you narrow it down to who this really affects, you will be down to an audience of less than 4 figures and there will be solutions for them in a couple of years so they can ride it out.
If Ivy doesn't make it into LGA2011 workstations prior to September/October of next year, which I think is likely, it seems unlikely that the next mac pro would use this. I think if they keep it going you're just going to see a late Sandy Bridge E rollout.
Tim Cook said Late 2013 for 'something great' so the wait will be more than a year from now. Sandy Bridge could be shipped any time so there would be no reason to say late 2013.
I was thinking about this some more and if they did use a custom co-processsor, they could probably do an iMac Pro without a GPU. They could actually use the co-processor as an IGP in the MP too for TB but its only remaining advantage over the iMac is being headless.
They could make an iMac with a Xeon chip (8-core/16-thread 77W IB for example), 4TB ports and have the co-processor run the graphics. It might not be as fast as a standard GPU but it would very fast. The current iteration can do real-time ray-tracing using 4 servers and hit 90FPS:
They don't say how many cards they use but it's suitable for graphics and fully programmable. They can update to the newer OpenGL versions without waiting on a GPU manufacturer offering support. They can support features that are exclusive to Quadro and FirePro cards. I'd prefer a Super-Cube as the headless form factor allows more usage scenarios but I think an iMac Pro would be fine too.
I really don't like redundancy. When I see a numbers row and a numpad on my keyboard, I don't like it. For years I've looked at the myriad of different ports and protocols and wished that one day there would be a more unified I/O setup that covered display data. Using magnets too, although that might not be feasible - the description is here:
This reminds me of how I wish all scripting APIs were python based, It's cleaner than having a dozen junk languages that came from the 1990s. Unfortunately they still have a place due to non-redundant functions. I'm not against a unified IO setup. I just do not see it happening if it costs many times what we pay for the current solution and has many hangups. It's not terribly efficient in data delivery meaning that you need much more raw bandwidth than you will use. Another option would be if all channels could be used for display bandwidth so as to be able to support current displayport protocols. I'm not saying it has to be as cheap or abundant as usb. It's just right now the implementation remains with the board manufacturers and intel has not provided a reference design for the higher end Xeon systems. At this time they aren't really pushing it as a unified solution. I've actually wondered if Apple would research wireless charging. At some point people may view consumer grade computer systems with wires as archaic. Unfortunately wireless standards for such things remain terribly inefficient. Wireless standards for usb and displayport were developed years ago. Neither has gained any real market adoption.
This is still one of the most detailed responses you've ever written to one of my posts . Thank you for that.
Quote:
PCI slots have the speed advantage for now but it will take longer to evolve internal motherboard slots than an external port. A port change doesn't have to mean obsolete equipment but a change from PCIX to PCIe means an obsolete card. Then you have half-length slots, double-wide slots, different power limits and driver support. You still have driver support to contend with but on the Mac side, your potential audience is multiplied by 20 and you can test/develop the driver on a $600 computer.
We go through transitions over time. People didn't like IDE/PATA when it came on the scene because it wasn't as fast as SCSI but you didn't have to assign drive IDs and deal with conflicts. By the time SATA arrived in 2003, nobody had been talking about it for years.
The same has happened with optical drives being removed. The odd few jump up and shout 'hey, I was using that, you can't get rid of it' and looks around to realise he's in the small minority and can get an external drive. The entire Mac Pro audience is a tiny minority now and the people using the PCI slots for custom cards a fraction of them and people using cards other than I/O cards and cards that have TB equivalents a further fraction of them.
I think getting anything close to what you'd like to see would require more backing from intel as well. If you recall last year many companies were complaining about the lack of an SDK for much of the year. I'm unsure how development was conducted on some of the early products, but it was noted. PCI-X is something Apple dropped really early even though the reference board designs all supported 1x PCI-X. Dell and some of the others maintained them for backward compliance. It was more of an Apple thing. I don't remember the initial SCSI implementations very well. Wasn't IDE /PATA implemented to cut costs including those around development? I see thunderbolt as something Intel would really need to push. Right now you can get a better ratio of cost : performance from mini sas on the high end (back to SCSI ) or cheaper products via usb3. A big thing would just be including it in more machines and ensuring backward compatibility with future plug generations like we have with usb. Something I truly abhor is e-waste, so I like things that can remain useful to someone as long as possible, even if it's not the original owner. Some people complain when their macbook pros are worth much less than they were last year due to the rMBP. I don't see how it matters. They're still very useful. If they're only a year old, someone else will gladly use them. People just need to get over the way they view macs as investments.
Quote:
By the time you narrow it down to who this really affects, you will be down to an audience of less than 4 figures and there will be solutions for them in a couple of years so they can ride it out.
Tim Cook said Late 2013 for 'something great' so the wait will be more than a year from now. Sandy Bridge could be shipped any time so there would be no reason to say late 2013.
I was thinking about this some more and if they did use a custom co-processsor, they could probably do an iMac Pro without a GPU. They could actually use the co-processor as an IGP in the MP too for TB but its only remaining advantage over the iMac is being headless.
They could make an iMac with a Xeon chip (8-core/16-thread 77W IB for example), 4TB ports and have the co-processor run the graphics. It might not be as fast as a standard GPU but it would very fast. The current iteration can do real-time ray-tracing using 4 servers and hit 90FPS:
They don't say how many cards they use but it's suitable for graphics and fully programmable. They can update to the newer OpenGL versions without waiting on a GPU manufacturer offering support. They can support features that are exclusive to Quadro and FirePro cards. I'd prefer a Super-Cube as the headless form factor allows more usage scenarios but I think an iMac Pro would be fine too.
A couple years isn't that bad to ride out, but you're drifting away from facts again with the concept of a 77W 8 core IB chip. None exist. The 8-10 core chips are supposedly quite hot, just like Sandy. 77W is the cap on the desktop chips, but the the mini-server chips go a bit higher. Ivy Bridge EP with 10 cores at the high end will be much hotter. There are estimates of tdp as high as 150+ W. The downside i've mentioned before about building such a thing into an imac is that you end up with a totally different design anyway. The boards are totally different, and none of the reference board designs are likely to be appropriate for such a machine. This means they would have quite a lot of work just to produce such an item. I'm not sure what the issue with OpenGL versions is here. Cards on the market do offer support for newer OpenGL versions than you have under Mountain Lion.
I have no idea how costly those co-processor cards will be. Given that they seem to be intel's response to tesla computing, they wouldn't have to be that cheap to find a market. That video made me laugh. The comments are even funnier. Note that it's not the notebook doing the rendering work there, but I really do like that they raytraced wolfenstein. That is just so silly. It required a server with 4 knights ferry cards to feed the laptop. They're not housed within it. I just really like the concept of raytraced wolfenstein.
It's not terribly efficient in data delivery meaning that you need much more raw bandwidth than you will use.
I'm thinking eventually there won't be an issue of bandwidth allocation. Think of a single optical channel that operates at 1Tbit/s. It gets split into 5 ports and these ports would be tiny dots around the size of an earphone jack. There can be a magnetic ring round them that can be used to hold it in place and provide power like magsafe. The plug ends would be circular with no wrong way to connect them. Bandwidth is shared across all ports and 1Tbit/s for all of them is a lot. The circular design would make the magnetic connection stronger to avoid accidental disconnects and the wires could be pretty thin, especially ones that didn't require power. This can even replace headphone jacks so you get optical audio quality and no broken jacks. It would have a simple in-line converter for 3.5mm.
Right now you can get a better ratio of cost : performance from mini sas on the high end (back to SCSI) or cheaper products via usb3.
I don't think a unified solution is possible with electrical connections but optical is just a light signal so a multi-protocol optical connection would automatically be universal because it just needs a basic connector. For now, the closest we will get to unification is Thunderbolt + USB 3, which I think is fine.
A couple years isn't that bad to ride out, but you're drifting away from facts again with the concept of a 77W 8 core IB chip. None exist. The 8-10 core chips are supposedly quite hot, just like Sandy. 77W is the cap on the desktop chips, but the the mini-server chips go a bit higher.
They don't have to use the higher powered chips though. Apple already makes the decision they don't want to offer the faster CPUs. They could even stick with a 6-core. The co-processor would make up the performance lost from having more full Xeon cores.
The downside i've mentioned before about building such a thing into an imac is that you end up with a totally different design anyway. The boards are totally different, and none of the reference board designs are likely to be appropriate for such a machine.
HP managed to put a Xeon in their AIO Z1 so it's possible. It might be a pretty expensive machine though.
I have no idea how costly those co-processor cards will be. Given that they seem to be intel's response to tesla computing, they wouldn't have to be that cheap to find a market.
The PCI boards might be expensive but they are being sold to a very small group of people and they are standalone products with their own memory and packaging and run Linux OS. Apple would get the chips from Intel and put them on the motherboard and have them use system memory as well as run OS X processes directly. Although the MP audience is small, it'll still be 100k units per quarter, which is probably among the higher orders for these kind of chips.
If you think of a single supercomputer installation, they'd order a few thousand PCI cards. The amount of those kind of installations worldwide is very small. NVidia Tesla launched 2007. This article here says only 150,000 units total have been shipped:
Even the world's 2nd fastest supercomputer only has 7,168 Tesla GPUs. Compare total Tesla units over 3 years to Apple shipping 150,000 machines with a co-processor in 3 months and they can easily get the prices down. In an iMac, it's even better because that's Apple's best selling desktop.
I'm thinking eventually there won't be an issue of bandwidth allocation. Think of a single optical channel that operates at 1Tbit/s. It gets split into 5 ports and these ports would be tiny dots around the size of an earphone jack. There can be a magnetic ring round them that can be used to hold it in place and provide power like magsafe. The plug ends would be circular with no wrong way to connect them. Bandwidth is shared across all ports and 1Tbit/s for all of them is a lot. The circular design would make the magnetic connection stronger to avoid accidental disconnects and the wires could be pretty thin, especially ones that didn't require power. This can even replace headphone jacks so you get optical audio quality and no broken jacks. It would have a simple in-line converter for 3.5mm.
I don't think a unified solution is possible with electrical connections but optical is just a light signal so a multi-protocol optical connection would automatically be universal because it just needs a basic connector. For now, the closest we will get to unification is Thunderbolt + USB 3, which I think is fine.
I don't think thunderbolt will displace the use of fibre channel HBAs and mini-SAS at the moment. There are a lot of existing things using mini-SAS with greater bandwidth, and it's a very small connector. I agree a universal one would be great. I just don't see it as something that will be out soon, especially as the current one has to provide more bandwidth to make up for inefficiencies in delivery. It'll be interesting to see what it's like when they make a true optical version, but costs might be an issue for now. The cost of cabling is also important if you want to reach a really broad market. Intel seemed concerned about this as they did mention it would come down to how much bandwidth people were willing to pay for. I still see it as thunderbolt kool-aid for now. Intel likes to propagate far out numbers, and I don't really trust them.
Quote:
They don't have to use the higher powered chips though. Apple already makes the decision they don't want to offer the faster CPUs. They could even stick with a 6-core. The co-processor would make up the performance lost from having more full Xeon cores.
HP managed to put a Xeon in their AIO Z1 so it's possible. It might be a pretty expensive machine though.
I think you may not have looked at HP's hardware specifications. HP has a big workstation market. This was somewhat aimed at the lighter end of it. They're defined by features and serviceability in an AIO rather than raw power. The most powerful cpu is an E3-1280. It's one of the lighter Sandy Bridge cpus rather than Sandy Bridge E. Sandy Bridge E may have been too hot. Note that when you opt for one of the Xeons, integrated graphics stops being a configuration option. I think their entry configuration is pretty bleh for what it costs. It appears to be beautifully engineered, yet starting with an i3 and Intel HD2000 graphics is really sub par for its price. Configured to the minimum specs I'd personally want pushes it to around $4k. Some of their upgrade steps are oddly structured. A Quadro 1000m is around $400 for a very cheap card. The 3000m is only a couple hundred more. The 4000m is about $600 on top of that, and that's where you're really into the performance of a lower workstation card. It still wouldn't perform at the level of a desktop Quadro which retails for around $700-800. Anything below the 3000m is pretty bad for the market HP seems to want (I'm assuming CAD design market). That Xeon isn't like the ones in the Mac Pro. It's based on desktop parts. You get 20 PCI lanes total. The Ivy version also tops out at 4 cores. Haswell is unlikely to up the cores. It's an architectural change rather than a die shrink, and processors of that class tend to chase ghz more than cores. In the case of high end Sandy E, the 8 core units are still clocked lower. It's just that the change in architecture allowed them to be clocked at a rate that wouldn't appear laughably low. Anyway I was just trying to offer some perspective on that. I don't think a 6 core will be possible next year in that socket type for the Z1 or the imac.
Quote:
The PCI boards might be expensive but they are being sold to a very small group of people and they are standalone products with their own memory and packaging and run Linux OS. Apple would get the chips from Intel and put them on the motherboard and have them use system memory as well as run OS X processes directly. Although the MP audience is small, it'll still be 100k units per quarter, which is probably among the higher orders for these kind of chips.
If you think of a single supercomputer installation, they'd order a few thousand PCI cards. The amount of those kind of installations worldwide is very small. NVidia Tesla launched 2007. This article here says only 150,000 units total have been shipped:
Even the world's 2nd fastest supercomputer only has 7,168 Tesla GPUs. Compare total Tesla units over 3 years to Apple shipping 150,000 machines with a co-processor in 3 months and they can easily get the prices down. In an iMac, it's even better because that's Apple's best selling desktop.
NVidia may not ship that many of them, but the margins are most likely quite high. It was interesting R&D on their part. They leveraged in on cost and power consumption. It's not to say that tesla cards are cheap. It's just that for NVidia's market, the comparison was what that kind of power would cost if powered by x86 cores from intel. That was a cool article overall. I've always liked NVidia's way of thinking. I like AMD too. Both are smaller companies relative to their competition, and they come out with some very cool ideas. While I could see such a thing bringing prices down assuming enough capacity exists to turn out this many, I don't think Apple would be the one to make such a bold move in the workstation market. I'm still wondering where you saw a late 2013 quote. The only ones I can turn up are quite ambiguous and merely point to 2013 without even truly confirming a mac pro.
I don't think thunderbolt will displace the use of fibre channel HBAs and mini-SAS at the moment. There are a lot of existing things using mini-SAS with greater bandwidth, and it's a very small connector.
These are storage-only interconnects, Thunderbolt is PCI so has much wider uses. Concerning bandwidth, you still have to have storage that would bottleneck TB. OWC has a product called Jupiter:
For fast high capacity storage, you'd need multiple 15k drives in RAID 0. You wouldn't want more than 4 drives in RAID 0 and even then you'd put them in RAID 10 so you need 8 drives and you'll still only get 800MB/s, which TB can handle. With SSD, you can exceed those speeds but the types of files that need over 1000MB/s use up a lot of space (90 minutes at 300MB/s = 1.6TB) so it makes the solution expensive. $5,000 for the box + 8 x $500 for 2TB of SSD = $9,000. Not many people will be buying a $3,000 MP along with $9,000 of storage.
Even OWC make a Thunderbolt adaptor to allow connection to the SAS and the CEO says:
“With the advanced processors modern MacBook Airs, MacBook Pros, iMacs, and even the Mac mini offer, these machines are very capable of running advanced applications and processes that rival the Mac Pro,” said Larry O’Connor, Founder and CEO, Other World Computing. “Now with the Mercury Helios, users of these non-PCIe slot equipped machines can tap into the power of PCIe cards and experience capabilities previously unattainable.”
assuming enough capacity exists to turn out this many, I don't think Apple would be the one to make such a bold move in the workstation market.
This is the same Apple that brought us the personal computer, the iPod, iPhone, iPad, the first usable unix desktop, OpenCL. Yeah, they aren't the type to make bold moves. Yield could be an issue if they aimed for 50-core chips but they don't need to do that. A 24-36 core co-processor would work ok.
I'm still wondering where you saw a late 2013 quote. The only ones I can turn up are quite ambiguous and merely point to 2013 without even truly confirming a mac pro.
Tim Cook: "Our pro customers are really important to us... don't worry as we're working on something really great for later next year."
If he meant early 2013, he'd have just said "next year". If they intended on using Sandy Bridge, they would have used it.
These are storage-only interconnects, Thunderbolt is PCI so has much wider uses. Concerning bandwidth, you still have to have storage that would bottleneck TB. OWC has a product called Jupiter:
For fast high capacity storage, you'd need multiple 15k drives in RAID 0. You wouldn't want more than 4 drives in RAID 0 and even then you'd put them in RAID 10 so you need 8 drives and you'll still only get 800MB/s, which TB can handle. With SSD, you can exceed those speeds but the types of files that need over 1000MB/s use up a lot of space (90 minutes at 300MB/s = 1.6TB) so it makes the solution expensive. $5,000 for the box + 8 x $500 for 2TB of SSD = $9,000. Not many people will be buying a $3,000 MP along with $9,000 of storage.
Even OWC make a Thunderbolt adaptor to allow connection to the SAS and the CEO says:
“With the advanced processors modern MacBook Airs, MacBook Pros, iMacs, and even the Mac mini offer, these machines are very capable of running advanced applications and processes that rival the Mac Pro,” said Larry O’Connor, Founder and CEO, Other World Computing. “Now with the Mercury Helios, users of these non-PCIe slot equipped machines can tap into the power of PCIe cards and experience capabilities previously unattainable.”
That OWC device looks like it was built well enough. In terms of drives, populating via 15k SAS drives wasn't what I meant at all. I was thinking of something more like an Atto Roc with parity stripes. You can put together a fairly robust solution this way using something like RE4s and a dumb box with mini sas out. I've been enjoying this conversation and I don't really with to kill it, but no one with half a brain would spend $9000 for a RAID populated with 2TB worth of SSD when existing SATA enterprise drives would be a superior choice in terms of reliability and cost per GB. Someone in IT suggesting that would likely find themselves unemployed. Video editing is a very common use case since we're talking about Macs and high bandwidth requirements. For those guys there is really no reason to ever even consider SSDs for large Raid volumes when you can do much better with typical HDDs. It's just when you're going into the 8TB+ realm, you should be looking at hardware controllers rather than a software raid. All of this needs to be backed up anyway. In terms of raw speed, you can get around 135MB/s sequential write from modern HDDs. You get less in parity stripe mode, but the potential for downtime is lessened. In either case the RAID requires backup. It's just that this is the difference between building a massive raid for $4k~ and building a tiny one with no inherent fault tolerance for $9k. I don't see the potential use case for yours. I don't even think it will happen with NAND. Note I'm not saying that solid state isn't the way forward. I'm just not confident in the future of NAND technology.
Quote:
That's the point though. Apple would ship higher volume so they don't need to make the margins so high.
This is the same Apple that brought us the personal computer, the iPod, iPhone, iPad, the first usable unix desktop, OpenCL. Yeah, they aren't the type to make bold moves. Yield could be an issue if they aimed for 50-core chips but they don't need to do that. A 24-36 core co-processor would work ok.
They weren't the first to bring out the personal computer. They brought out a personal computer. A few others had similar solutions that were basically concurrent, but that doesn't really matter. The point I wished to make was that they aren't that aggressive with workstation technology. Apple's behavior is very conservative unless it involves something that they think they can package. They were conservative on usb3. With thunerbolt it gave them a nice docking station for their notebook line. I couldn't see them backing such a thing unless they were trying to increase market adoption to eventually bring it into their portables. In this kind of situation I could see them trying to groom such a package for their volume lines, but I don't know enough about the power consumption or size of that technology to know anything about its feasibility. It sounds like we're talking about much smaller/simpler cores compared to x86. I'll do more reading on it later.
Quote:
Tim Cook: "Our pro customers are really important to us... don't worry as we're working on something really great for later next year."
If he meant early 2013, he'd have just said "next year". If they intended on using Sandy Bridge, they would have used it.
Whenever an executive makes a comment like this, people read into each word. I think it was intended more as reassurance for PR reasons than anything. In terms of Sandy vs ivy, same chipsets, but Ivy gets a boost in maximum core count. There's just nothing interesting specifically tied to Ivy E, and when it does come out it's likely to be like nehalem where Sandy Bridge E components still fill in the lower end of intel's lineup.
I've been enjoying this conversation and I don't really with to kill it, but no one with half a brain would spend $9000 for a RAID populated with 2TB worth of SSD when existing SATA enterprise drives would be a superior choice in terms of reliability and cost per GB.
In terms of raw speed, you can get around 135MB/s sequential write from modern HDDs.
The SSD is the only way to max the bandwidth. Someone here said they spent $30k on a RAM drive so people have their reasons.
As for standard drives, you still need them in RAID 0 and you won't max out Thunderbolt's bandwidth. Plus are you talking about RAIDing the internal drives or external? I hope you ain't making a rat's nest at the back or having external boxes lying around everywhere with separate power supplies.
Whenever an executive makes a comment like this, people read into each word.
He said 'later', so I assumed it meant later. Wouldn't you have to read into it to assume it meant earlier? I'd give up hope of hearing another word about the MP from Apple for another year at least.
For the next year, we're going to get as many threads talking about what's going to happen but no matter what they do, people won't be happy because they either:
- don't redesign it and stick with Ivy Bridge in which case it's a minor performance bump after 3 years and no Thunderbolt support
- do redesign it in which case, they will gut the thing and put Thunderbolt on, possibly include an MIC and not have PCI slots (unless they can reconcile the display issue)
- ditch it in favour of a souped up iMac of some kind in which case, we'll get the usual raft of comments on Apple dumbing things down
- discontinue it in which case Apple will be described as abandoning Pros
It doesn't matter what they do, it will be late and some group of people won't like it.
The SSD is the only way to max the bandwidth. Someone here said they spent $30k on a RAM drive so people have their reasons.
As for standard drives, you still need them in RAID 0 and you won't max out Thunderbolt's bandwidth. Plus are you talking about RAIDing the internal drives or external? I hope you ain't making a rat's nest at the back or having external boxes lying around everywhere with separate power supplies.
$30k on a ram drive is pretty specialized. Conversely there are many people who drop a couple grand on storage.
Do not make me hunt down DAS units with mini SAS to prove they exist :P. You really can't pull that kind of bandwidth internally. Apple has one RAID card, and it's a poorly supported piece of trash. It's one of those things where if they couldn't do it right, they should have left it to third party solutions. Atto or Areca would have been better. I made no mention of a rat's nest anyway. DAS unit with 1 mini SAS out the back ----> Roc on the back of the mac pro. You make it sound like high bandwidth only came around with SSDs, yet SSDs typically aren't suitable for RAIDs at all, especially anything with striped parity. I'm not sure how they'd hold up with dedicated parity drives, but who really uses those variants anyway? I said if you need high bandwidth and a lot of storage you'd go with something of that sort. SSDs are nonsensical for such things due to limited write cycles and other problems + much higher cost per gigabyte. I really really don't understand your logic here. What would be the case for using an array of SSDs compared to HDDs here? You'd have two boxes either way as the SSDs would quickly max out the internal SATA bus. You can do an internal card, and there are good reasons to populate the internal bays. I was saying if you need a high bandwidth solution, it quickly becomes impractical. You wouldn't really stuff thousands of dollars worth of SSDs internally anyway (at least I wouldn't). The time it becomes a mess is when you have many connections out the front and back with a poor routing method.
i feel like we're almost arguing semantics at this point. My initial point was that as a consumer grade solution usb3 is cheaper. If you're in need of the maximum possible bandwidth and less cost sensitive, mini-SAS has existed longer, so there may not be a good reason to migrate there. Just for reference...
Realistically you can set something like this up cheaper if you're careful on what you buy. The express34 ports don't have as much bandwidth. They're just a notebook solution. I linked to a populated DAS. I'd probably buy a bare one myself and populate it with WD enterprise firmware drives, verifying identical firmware across them prior to setup. If you need large amounts of fast storage, this is a feasible solution. Actually I'm not sure of the bay count on the one I linked, but there are many sizes, and they come in horizontal rackmount optional form factors too. I was saying for storage if you need large quantities of performance storage, this remains a better solution than dropping that much on SSDs. If you're talking about consumer grade solutions, usb3 is cheaper for nearly the same performance you'll get from current thunderbolt options. Intel has yet to really push adoption. Wasn't thunderbolt on its data channels tested at a max of 650-800MB/s? Beyond that are there any DAS thunderbolt units that are big enough to push that without SSDs? I wouldn't suggest Raid 0 anyway for anything time crucial. I'd say go with Raid 10 for some measure of fault tolerance. At that point you are often better off allocating more drives to a bigger 5. I'm not sure what rebuild times would be like on one of those, but retrieving a Raid 0 from a NAS or tape backup would take you a very long time.
Back to what I was saying before, you have customers who have potential customers who are looking at upgrades and new ones that are putting together such solutions. Splitting that further you have different prioritizations of cost and performance requirements. I keep trying to make the point that out of the potential matrix of solutions you could draw from this, thunderbolt taken as a connection + current selection of devices falls in the middle here. Beyond that with SSDs it wasn't just a cost issue. They really aren't designed to hold up to such use (yet).
Quote:
He said 'later', so I assumed it meant later. Wouldn't you have to read into it to assume it meant earlier? I'd give up hope of hearing another word about the MP from Apple for another year at least.
For the next year, we're going to get as many threads talking about what's going to happen but no matter what they do, people won't be happy because they either:
- don't redesign it and stick with Ivy Bridge in which case it's a minor performance bump after 3 years and no Thunderbolt support
- do redesign it in which case, they will gut the thing and put Thunderbolt on, possibly include an MIC and not have PCI slots (unless they can reconcile the display issue)
- ditch it in favour of a souped up iMac of some kind in which case, we'll get the usual raft of comments on Apple dumbing things down
- discontinue it in which case Apple will be described as abandoning Pros
It doesn't matter what they do, it will be late and some group of people won't like it.
I figured later as something other than soon. Later next year doesn't necessarily mean "late next year". Next year is later than this year. There is little context here and it makes little sense to dissect this as an oral contract.
You make it sound like high bandwidth only came around with SSDs. I really really don't understand your logic here.
Wasn't thunderbolt on its data channels tested at a max of 650-800MB/s? Beyond that are there any DAS thunderbolt units that are big enough to push that without SSDs?
What would be the case for using an array of SSDs compared to HDDs here?
You answer your own question here. You can't easily push the limits of Thunderbolt without SSD, which you've already said is not practical. That would indicate Thunderbolt is fine for this setup.
SAS can outperform TB with quad RAID 0 SSDs as shown here:
They also hook up two TB drives to the rMBP and manage 1.38GB/s. They said if they had two of the SSD drives they'd expect closer to 1.8GB/s but they'd need to test it to verify.
They are test scenarios that aren't likely to be used in a practical scenario but it shows that TB is up to the task and it works on all Macs, not just a model people aren't buying in large numbers.
If your work is storage limited then you can buy a $600 Mini, an $1100 Pegasus with 4x SSDs and you will be able to cope with workflows that a stock $3000 Mac Pro couldn't cope with.
Later next year doesn't necessarily mean "late next year". Next year is later than this year. There is little context here and it makes little sense to dissect this as an oral contract.
That doesn't explain why they'd wait until next year before using Sandy Bridge when they can use it now. The only reason for the delay has to be that they plan to use something that's not out yet i.e Ivy Bridge. It would be a stretch to say they are waiting on Knights Ferry as there's no indication this is the route they will take yet but I think SB is off the table or it would be shipping.
That doesn't explain why they'd wait until next year before using Sandy Bridge when they can use it now. The only reason for the delay has to be that they plan to use something that's not out yet i.e Ivy Bridge. It would be a stretch to say they are waiting on Knights Ferry as there's no indication this is the route they will take yet but I think SB is off the table or it would be shipping.
The Pegasus box that Apple ships isn't aimed at the market I suggested, and RAID 0 requires that you accept the potential for a day of downtime if a disk fails. Beyond that if you're migrating data on and off constantly due to small disks on your fast storage, you run into the issue that SSDs have limited write cycles. You're likely to kill that large storage investment quite fast.
In terms of Sandy vs Ivy, my only guess would be they wanted to kill it and didn't assign anyone to create an updated reference design. This is not a project that requires many people. Even board designs don't usually involve a large team. If they had a couple engineers on it, you'd have something. As it is, this was likely something that was abandoned due to uncertainty. Perhaps the rumors were true regardless of whether the articles themselves involved bloggers looking for page hits. There isn't any real reason to skip a generation when they both plug into the same chipset. Either way you'll have the same board design. I do not buy into your theories of early access, especially on a line where they are likely to be relatively conservative. We could see a 16 core mac pro. It's likely that chips for a 20 core would exceed Apple's chip budgeting at their desired margins. I think you place too much confidence in Apple's semantics. That may be the reason we disagree on this one. I think they just never allocated anyone to such a project.
You answer your own question here. You can't easily push the limits of Thunderbolt without SSD, which you've already said is not practical. That would indicate Thunderbolt is fine for this setup.
SAS can outperform TB with quad RAID 0 SSDs as shown here: http://www.barefeats.com/tbolt01.html
but performance varies. Anand did a test with multiple SSD and managed closer to 1GB/s: http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/11
They also hook up two TB drives to the rMBP and manage 1.38GB/s. They said if they had two of the SSD drives they'd expect closer to 1.8GB/s but they'd need to test it to verify.
They are test scenarios that aren't likely to be used in a practical scenario but it shows that TB is up to the task and it works on all Macs, not just a model people aren't buying in large numbers.
There is no doubt TB is speedy and flexible. The problem on most Apple hardware is that TB is your only solution to high speed I/O.
If your work is storage limited then you can buy a $600 Mini, an $1100 Pegasus with 4x SSDs and you will be able to cope with workflows that a stock $3000 Mac Pro couldn't cope with.
Well no you will still have bottle necks on the Mini the Pro doesn't have.
That doesn't explain why they'd wait until next year before using Sandy Bridge when they can use it now. The only reason for the delay has to be that they plan to use something that's not out yet i.e Ivy Bridge.
I don't see that, Ivy Bridge isn't going to be all that much more powerful than Sandy Bridge and frankly we could wait forever to see an Ivy Bridge based XEON. If the Pro delay is related to anything processor wise it won't be for Ivy Bridge.
It would be a stretch to say they are waiting on Knights Ferry as there's no indication this is the route they will take yet but I think SB is off the table or it would be shipping.
Well I agree SB is off the table in the sense that it is the only processor in the machine. However Intel intends for MiC to be a family of devices as part of their foray into high performance computing. Even if Apple forgoes the 62 core coprocessor there are other chips that will be coming on line targeting high performance computing. For example Intel expects to integrate Infiniband into a multicore processor chip.
The latest rumor is that Intel will launch Xeon Phi and it high performance computing initiative in the November/December time frame. If so this would be reason enough to speculate that the Mac Pro, at least, is waiting for this hardware to launch. I just don't see Apple waiting until October of 2013 and hope that Intel has an Ivy Bridge based Xeon ready at that time.
Lastly Intel is basically in love with Apple. Without Apple there would be nobody innovating with their hardware. So I have no doubt at all that Apple and Intel have a Skunk Works of sorts working on new technology for the Mac Pro. In this regard I wouldn't be surprised at all that Intel and Apple have some sort of exclusive deal going for a MiC chip of some sort, much like the way Thunderbolt was handled.
RAID 0 requires that you accept the potential for a day of downtime if a disk fails. SSDs have limited write cycles.
No RAID 0 and no SSD means Thunderbolt is fast enough. You were saying TB is not fast enough to replace Mini-SAS but it is unless you use SSD in RAID 0, which as you say, wouldn't be the case.
Early access to what? The only thing I suggested they could get early access to was a new Thunderbolt controller. They got early access to the first one among other things from Intel:
The problem on most Apple hardware is that TB is your only solution to high speed I/O.
I don't see that being a problem. I'd say it would be a worse situation if they made Mini-SAS, Fibre Channel or Infini-band exclusively on a machine that makes up 5% or less of their lineup while not offering a significant advantage over Thunderbolt (taking into consideration its roadmap).
A multi-protocol optical interconnect is the way forward so everything else that's not that will die out and shouldn't be supported.
Well no you will still have bottle necks on the Mini the Pro doesn't have.
Not if the work is storage limited. If you are editing high-bitrate intra-frame compressed video, the Mini doesn't have a bottleneck if you have fast storage. When it comes to encoding it will be slower but you can still do it. Without fast storage, the Mac Pro wouldn't be able to do the edit at all. In that scenario, the stock Mac Pro couldn't do the job that a Mac Mini could do. Obviously adding storage is trivial but the point is that the MP itself is less important than the connection and storage.
The latest rumor is that Intel will launch Xeon Phi and it high performance computing initiative in the November/December time frame. If so this would be reason enough to speculate that the Mac Pro, at least, is waiting for this hardware to launch. I just don't see Apple waiting until October of 2013 and hope that Intel has an Ivy Bridge based Xeon ready at that time.
I think it will be in the 2nd half of the year at the earliest that we find out what's going on. Even though people don't see the last update as an update, it was. There's no way they would obsolete that refresh within 6-8 months with a radical machine.
I don't think they would use the MIC along with a DP system so I think it makes sense to wait for Ivy Bridge. Q3 runs July - Sept so not as far out as October 2013 but still around a year from now.
We know the cycle:
iMac/Mini - Aug/Sept (might be announced at iPhone launch though due to major iMac redesign)
iPhone - Sept/Oct (I'd say Oct 5th)
iPad - March
MBP/MBA (Haswell) - May/June
WWDC June - announce redesigned MP, possibly exclusive Ivy Bridge launch, Haswell iMac/Mini similar timeframe
No RAID 0 and no SSD means Thunderbolt is fast enough. You were saying TB is not fast enough to replace Mini-SAS but it is unless you use SSD in RAID 0, which as you say, wouldn't be the case.
While I enjoy our conversations, sometimes I'm unsure of how to explain certain things in a way that keeps the discussion on track. I think you're referring to something like the Promise Pegasus populated with ssds right? I included a wider product range in my comparison where this wouldn't necessarily be the case. My point was if you aren't going to reach such speeds, usb3 is the cheaper solution. You can set it up for less in most cases. With the Pegasus raid I don't trust them. They may go cheap on the installed drives. I'd rather buy my own, although Promise does make some solid professional and enterprise grade hardware. If you're putting together a direct attached storage solution, there are tons of inexpensive options going the usb route. If you're going to spend the cash on a fast/robust solution, you can do better than what is offered currently over thunderbolt. To make it a viable option, it needs more options that can compete in enough points wiht usb3 on the low end or fibre/sas on the higher end. I called the $9k worth of ssds nonsense because you can do much better with a large das array populated by hdds. Noise is a factor with any of them. The Promise DAS isn't silent, and neither are the others. If you buy one that doesn't use trash fans to save a few dollars, it's not too bad. Some of the lower end solutions use either sleeve fans or ones with cheap bearings to save a couple dollars per fan. They'll get noisy as they wear. Hopefully costs and adoption rates will trend in a favorable direction so that we have solid thunderbolt options. Right now I see it as new and not a perfect fit against existing solutions. I see it more something that should be a volume solution rather than a niche one. It currently aligns better with computers that have integrated graphics within the cpu package. It offers lower performance relative to its cost against current solutions seen in desktops. To me the place to really push adoption is on the mobile end. It just has to better compete in price and available peripherals with usb3. Pushing adoption from the workstation end makes no sense to me. It's something that would be better off leveraged into that space rather than forced given the inherent misalignment and pre-existing solutions.
Apple used them for some of their superdrives and Lacie used them too. There are other quality options for bundled drives but it's best at this stage to leave them outside as optional extras.
It made sense that the old-style MBPs kept them because they'd have to redesign them and retool the manufacturing line just for models that they'd discontinue. When the prices drop, the old-style models will just disappear.
With the iMac, they could do a similar thing - leave the lineup the same except for the highest model and make it Retina with SSD and no optical but I don't think that would be a good move.
Apple used them for some of their superdrives and Lacie used them too. There are other quality options for bundled drives but it's best at this stage to leave them outside as optional extras.
It made sense that the old-style MBPs kept them because they'd have to redesign them and retool the manufacturing line just for models that they'd discontinue. When the prices drop, the old-style models will just disappear.
With the iMac, they could do a similar thing - leave the lineup the same except for the highest model and make it Retina with SSD and no optical but I don't think that would be a good move.
These things are not uncommon. Around a decade ago lcd display quality still sucked unless you wanted to pay $3000 for a 20" display, yet crts had been consistently bid down to where they were made with very cheap parts due to thinning margins. I am not surprised they're moving on.
Comments
I really don't like redundancy. When I see a numbers row and a numpad on my keyboard, I don't like it. For years I've looked at the myriad of different ports and protocols and wished that one day there would be a more unified I/O setup that covered display data. Using magnets too, although that might not be feasible - the description is here:
http://forums.appleinsider.com/t/86068/wireless-power-charging-technology-may-unplug-apples-iphone#post_1238133
It's all data so you really just need a single multi-protocol connection. Thunderbolt achieves that. That's what makes it better than all the other I/O standards besides the speed and latency.
As for PCI slots, I've never liked those because they dictate the size of the peripheral. If you need a FW800 port, you get a card that takes up an entire slot. With Thunderbolt, it's just a tiny plug.
PCI isn't plug and play either. You have to shut down your machine, open it up and install a peripheral. They used to do TV tuners this way. Once they realised how much of a pain it was to install, saw the dwindling desktop user base and got USB 2, they switched to USB.
It was a similar thing with A/V and Firewire. Video cameras could have shipped with PCI boards but it wouldn't have been a good solution. I can see Thunderbolt becoming the new standard on A/V equipment (low latency live capture over 100m optical for example) along with USB 3:
http://www.engadget.com/2012/04/16/blackmagic-cinema-camera-packs-feature-film-2-5k-quality-touc/
PCI slots have the speed advantage for now but it will take longer to evolve internal motherboard slots than an external port. A port change doesn't have to mean obsolete equipment but a change from PCIX to PCIe means an obsolete card. Then you have half-length slots, double-wide slots, different power limits and driver support. You still have driver support to contend with but on the Mac side, your potential audience is multiplied by 20 and you can test/develop the driver on a $600 computer.
We go through transitions over time. People didn't like IDE/PATA when it came on the scene because it wasn't as fast as SCSI but you didn't have to assign drive IDs and deal with conflicts. By the time SATA arrived in 2003, nobody had been talking about it for years.
The same has happened with optical drives being removed. The odd few jump up and shout 'hey, I was using that, you can't get rid of it' and looks around to realise he's in the small minority and can get an external drive. The entire Mac Pro audience is a tiny minority now and the people using the PCI slots for custom cards a fraction of them and people using cards other than I/O cards and cards that have TB equivalents a further fraction of them.
By the time you narrow it down to who this really affects, you will be down to an audience of less than 4 figures and there will be solutions for them in a couple of years so they can ride it out.
Tim Cook said Late 2013 for 'something great' so the wait will be more than a year from now. Sandy Bridge could be shipped any time so there would be no reason to say late 2013.
I was thinking about this some more and if they did use a custom co-processsor, they could probably do an iMac Pro without a GPU. They could actually use the co-processor as an IGP in the MP too for TB but its only remaining advantage over the iMac is being headless.
They could make an iMac with a Xeon chip (8-core/16-thread 77W IB for example), 4TB ports and have the co-processor run the graphics. It might not be as fast as a standard GPU but it would very fast. The current iteration can do real-time ray-tracing using 4 servers and hit 90FPS:
They don't say how many cards they use but it's suitable for graphics and fully programmable. They can update to the newer OpenGL versions without waiting on a GPU manufacturer offering support. They can support features that are exclusive to Quadro and FirePro cards. I'd prefer a Super-Cube as the headless form factor allows more usage scenarios but I think an iMac Pro would be fine too.
Quote:
Originally Posted by Marvin
I really don't like redundancy. When I see a numbers row and a numpad on my keyboard, I don't like it. For years I've looked at the myriad of different ports and protocols and wished that one day there would be a more unified I/O setup that covered display data. Using magnets too, although that might not be feasible - the description is here:
http://forums.appleinsider.com/t/86068/wireless-power-charging-technology-may-unplug-apples-iphone#post_1238133
This reminds me of how I wish all scripting APIs were python based, It's cleaner than having a dozen junk languages that came from the 1990s. Unfortunately they still have a place due to non-redundant functions. I'm not against a unified IO setup. I just do not see it happening if it costs many times what we pay for the current solution and has many hangups. It's not terribly efficient in data delivery meaning that you need much more raw bandwidth than you will use. Another option would be if all channels could be used for display bandwidth so as to be able to support current displayport protocols. I'm not saying it has to be as cheap or abundant as usb. It's just right now the implementation remains with the board manufacturers and intel has not provided a reference design for the higher end Xeon systems. At this time they aren't really pushing it as a unified solution. I've actually wondered if Apple would research wireless charging. At some point people may view consumer grade computer systems with wires as archaic. Unfortunately wireless standards for such things remain terribly inefficient. Wireless standards for usb and displayport were developed years ago. Neither has gained any real market adoption.
This is still one of the most detailed responses you've ever written to one of my posts
Quote:
PCI slots have the speed advantage for now but it will take longer to evolve internal motherboard slots than an external port. A port change doesn't have to mean obsolete equipment but a change from PCIX to PCIe means an obsolete card. Then you have half-length slots, double-wide slots, different power limits and driver support. You still have driver support to contend with but on the Mac side, your potential audience is multiplied by 20 and you can test/develop the driver on a $600 computer.
We go through transitions over time. People didn't like IDE/PATA when it came on the scene because it wasn't as fast as SCSI but you didn't have to assign drive IDs and deal with conflicts. By the time SATA arrived in 2003, nobody had been talking about it for years.
The same has happened with optical drives being removed. The odd few jump up and shout 'hey, I was using that, you can't get rid of it' and looks around to realise he's in the small minority and can get an external drive. The entire Mac Pro audience is a tiny minority now and the people using the PCI slots for custom cards a fraction of them and people using cards other than I/O cards and cards that have TB equivalents a further fraction of them.
I think getting anything close to what you'd like to see would require more backing from intel as well. If you recall last year many companies were complaining about the lack of an SDK for much of the year. I'm unsure how development was conducted on some of the early products, but it was noted. PCI-X is something Apple dropped really early even though the reference board designs all supported 1x PCI-X. Dell and some of the others maintained them for backward compliance. It was more of an Apple thing. I don't remember the initial SCSI implementations very well. Wasn't IDE /PATA implemented to cut costs including those around development? I see thunderbolt as something Intel would really need to push. Right now you can get a better ratio of cost : performance from mini sas on the high end (back to SCSI
Quote:
By the time you narrow it down to who this really affects, you will be down to an audience of less than 4 figures and there will be solutions for them in a couple of years so they can ride it out.
Tim Cook said Late 2013 for 'something great' so the wait will be more than a year from now. Sandy Bridge could be shipped any time so there would be no reason to say late 2013.
I was thinking about this some more and if they did use a custom co-processsor, they could probably do an iMac Pro without a GPU. They could actually use the co-processor as an IGP in the MP too for TB but its only remaining advantage over the iMac is being headless.
They could make an iMac with a Xeon chip (8-core/16-thread 77W IB for example), 4TB ports and have the co-processor run the graphics. It might not be as fast as a standard GPU but it would very fast. The current iteration can do real-time ray-tracing using 4 servers and hit 90FPS:
They don't say how many cards they use but it's suitable for graphics and fully programmable. They can update to the newer OpenGL versions without waiting on a GPU manufacturer offering support. They can support features that are exclusive to Quadro and FirePro cards. I'd prefer a Super-Cube as the headless form factor allows more usage scenarios but I think an iMac Pro would be fine too.
A couple years isn't that bad to ride out, but you're drifting away from facts again with the concept of a 77W 8 core IB chip. None exist. The 8-10 core chips are supposedly quite hot, just like Sandy. 77W is the cap on the desktop chips, but the the mini-server chips go a bit higher. Ivy Bridge EP with 10 cores at the high end will be much hotter. There are estimates of tdp as high as 150+ W. The downside i've mentioned before about building such a thing into an imac is that you end up with a totally different design anyway. The boards are totally different, and none of the reference board designs are likely to be appropriate for such a machine. This means they would have quite a lot of work just to produce such an item. I'm not sure what the issue with OpenGL versions is here. Cards on the market do offer support for newer OpenGL versions than you have under Mountain Lion.
I have no idea how costly those co-processor cards will be. Given that they seem to be intel's response to tesla computing, they wouldn't have to be that cheap to find a market. That video made me laugh. The comments are even funnier. Note that it's not the notebook doing the rendering work there, but I really do like that they raytraced wolfenstein. That is just so silly. It required a server with 4 knights ferry cards to feed the laptop. They're not housed within it. I just really like the concept of raytraced wolfenstein.
I'm thinking eventually there won't be an issue of bandwidth allocation. Think of a single optical channel that operates at 1Tbit/s. It gets split into 5 ports and these ports would be tiny dots around the size of an earphone jack. There can be a magnetic ring round them that can be used to hold it in place and provide power like magsafe. The plug ends would be circular with no wrong way to connect them. Bandwidth is shared across all ports and 1Tbit/s for all of them is a lot. The circular design would make the magnetic connection stronger to avoid accidental disconnects and the wires could be pretty thin, especially ones that didn't require power. This can even replace headphone jacks so you get optical audio quality and no broken jacks. It would have a simple in-line converter for 3.5mm.
I don't think a unified solution is possible with electrical connections but optical is just a light signal so a multi-protocol optical connection would automatically be universal because it just needs a basic connector. For now, the closest we will get to unification is Thunderbolt + USB 3, which I think is fine.
They don't have to use the higher powered chips though. Apple already makes the decision they don't want to offer the faster CPUs. They could even stick with a 6-core. The co-processor would make up the performance lost from having more full Xeon cores.
HP managed to put a Xeon in their AIO Z1 so it's possible. It might be a pretty expensive machine though.
The PCI boards might be expensive but they are being sold to a very small group of people and they are standalone products with their own memory and packaging and run Linux OS. Apple would get the chips from Intel and put them on the motherboard and have them use system memory as well as run OS X processes directly. Although the MP audience is small, it'll still be 100k units per quarter, which is probably among the higher orders for these kind of chips.
If you think of a single supercomputer installation, they'd order a few thousand PCI cards. The amount of those kind of installations worldwide is very small. NVidia Tesla launched 2007. This article here says only 150,000 units total have been shipped:
http://vr-zone.com/articles/how-the-geforce-gtx-690-is-a-prime-example-of-nvidia-reshaped-/15786.html#ixzz1uOIhM4xQ
Even the world's 2nd fastest supercomputer only has 7,168 Tesla GPUs. Compare total Tesla units over 3 years to Apple shipping 150,000 machines with a co-processor in 3 months and they can easily get the prices down. In an iMac, it's even better because that's Apple's best selling desktop.
Quote:
Originally Posted by Marvin
I'm thinking eventually there won't be an issue of bandwidth allocation. Think of a single optical channel that operates at 1Tbit/s. It gets split into 5 ports and these ports would be tiny dots around the size of an earphone jack. There can be a magnetic ring round them that can be used to hold it in place and provide power like magsafe. The plug ends would be circular with no wrong way to connect them. Bandwidth is shared across all ports and 1Tbit/s for all of them is a lot. The circular design would make the magnetic connection stronger to avoid accidental disconnects and the wires could be pretty thin, especially ones that didn't require power. This can even replace headphone jacks so you get optical audio quality and no broken jacks. It would have a simple in-line converter for 3.5mm.
I don't think a unified solution is possible with electrical connections but optical is just a light signal so a multi-protocol optical connection would automatically be universal because it just needs a basic connector. For now, the closest we will get to unification is Thunderbolt + USB 3, which I think is fine.
I don't think thunderbolt will displace the use of fibre channel HBAs and mini-SAS at the moment. There are a lot of existing things using mini-SAS with greater bandwidth, and it's a very small connector. I agree a universal one would be great. I just don't see it as something that will be out soon, especially as the current one has to provide more bandwidth to make up for inefficiencies in delivery. It'll be interesting to see what it's like when they make a true optical version, but costs might be an issue for now. The cost of cabling is also important if you want to reach a really broad market. Intel seemed concerned about this as they did mention it would come down to how much bandwidth people were willing to pay for. I still see it as thunderbolt kool-aid for now. Intel likes to propagate far out numbers, and I don't really trust them.
Quote:
They don't have to use the higher powered chips though. Apple already makes the decision they don't want to offer the faster CPUs. They could even stick with a 6-core. The co-processor would make up the performance lost from having more full Xeon cores.
HP managed to put a Xeon in their AIO Z1 so it's possible. It might be a pretty expensive machine though.
I think you may not have looked at HP's hardware specifications. HP has a big workstation market. This was somewhat aimed at the lighter end of it. They're defined by features and serviceability in an AIO rather than raw power. The most powerful cpu is an E3-1280. It's one of the lighter Sandy Bridge cpus rather than Sandy Bridge E. Sandy Bridge E may have been too hot. Note that when you opt for one of the Xeons, integrated graphics stops being a configuration option. I think their entry configuration is pretty bleh for what it costs. It appears to be beautifully engineered, yet starting with an i3 and Intel HD2000 graphics is really sub par for its price. Configured to the minimum specs I'd personally want pushes it to around $4k. Some of their upgrade steps are oddly structured. A Quadro 1000m is around $400 for a very cheap card. The 3000m is only a couple hundred more. The 4000m is about $600 on top of that, and that's where you're really into the performance of a lower workstation card. It still wouldn't perform at the level of a desktop Quadro which retails for around $700-800. Anything below the 3000m is pretty bad for the market HP seems to want (I'm assuming CAD design market). That Xeon isn't like the ones in the Mac Pro. It's based on desktop parts. You get 20 PCI lanes total. The Ivy version also tops out at 4 cores. Haswell is unlikely to up the cores. It's an architectural change rather than a die shrink, and processors of that class tend to chase ghz more than cores. In the case of high end Sandy E, the 8 core units are still clocked lower. It's just that the change in architecture allowed them to be clocked at a rate that wouldn't appear laughably low. Anyway I was just trying to offer some perspective on that. I don't think a 6 core will be possible next year in that socket type for the Z1 or the imac.
Quote:
The PCI boards might be expensive but they are being sold to a very small group of people and they are standalone products with their own memory and packaging and run Linux OS. Apple would get the chips from Intel and put them on the motherboard and have them use system memory as well as run OS X processes directly. Although the MP audience is small, it'll still be 100k units per quarter, which is probably among the higher orders for these kind of chips.
If you think of a single supercomputer installation, they'd order a few thousand PCI cards. The amount of those kind of installations worldwide is very small. NVidia Tesla launched 2007. This article here says only 150,000 units total have been shipped:
http://vr-zone.com/articles/how-the-geforce-gtx-690-is-a-prime-example-of-nvidia-reshaped-/15786.html#ixzz1uOIhM4xQ
Even the world's 2nd fastest supercomputer only has 7,168 Tesla GPUs. Compare total Tesla units over 3 years to Apple shipping 150,000 machines with a co-processor in 3 months and they can easily get the prices down. In an iMac, it's even better because that's Apple's best selling desktop.
NVidia may not ship that many of them, but the margins are most likely quite high. It was interesting R&D on their part. They leveraged in on cost and power consumption. It's not to say that tesla cards are cheap. It's just that for NVidia's market, the comparison was what that kind of power would cost if powered by x86 cores from intel. That was a cool article overall. I've always liked NVidia's way of thinking. I like AMD too. Both are smaller companies relative to their competition, and they come out with some very cool ideas. While I could see such a thing bringing prices down assuming enough capacity exists to turn out this many, I don't think Apple would be the one to make such a bold move in the workstation market. I'm still wondering where you saw a late 2013 quote. The only ones I can turn up are quite ambiguous and merely point to 2013 without even truly confirming a mac pro.
These are storage-only interconnects, Thunderbolt is PCI so has much wider uses. Concerning bandwidth, you still have to have storage that would bottleneck TB. OWC has a product called Jupiter:
For fast high capacity storage, you'd need multiple 15k drives in RAID 0. You wouldn't want more than 4 drives in RAID 0 and even then you'd put them in RAID 10 so you need 8 drives and you'll still only get 800MB/s, which TB can handle. With SSD, you can exceed those speeds but the types of files that need over 1000MB/s use up a lot of space (90 minutes at 300MB/s = 1.6TB) so it makes the solution expensive. $5,000 for the box + 8 x $500 for 2TB of SSD = $9,000. Not many people will be buying a $3,000 MP along with $9,000 of storage.
Even OWC make a Thunderbolt adaptor to allow connection to the SAS and the CEO says:
“With the advanced processors modern MacBook Airs, MacBook Pros, iMacs, and even the Mac mini offer, these machines are very capable of running advanced applications and processes that rival the Mac Pro,” said Larry O’Connor, Founder and CEO, Other World Computing. “Now with the Mercury Helios, users of these non-PCIe slot equipped machines can tap into the power of PCIe cards and experience capabilities previously unattainable.”
That's the point though. Apple would ship higher volume so they don't need to make the margins so high.
This is the same Apple that brought us the personal computer, the iPod, iPhone, iPad, the first usable unix desktop, OpenCL. Yeah, they aren't the type to make bold moves. Yield could be an issue if they aimed for 50-core chips but they don't need to do that. A 24-36 core co-processor would work ok.
Tim Cook: "Our pro customers are really important to us... don't worry as we're working on something really great for later next year."
If he meant early 2013, he'd have just said "next year". If they intended on using Sandy Bridge, they would have used it.
Quote:
Originally Posted by Marvin
These are storage-only interconnects, Thunderbolt is PCI so has much wider uses. Concerning bandwidth, you still have to have storage that would bottleneck TB. OWC has a product called Jupiter:
For fast high capacity storage, you'd need multiple 15k drives in RAID 0. You wouldn't want more than 4 drives in RAID 0 and even then you'd put them in RAID 10 so you need 8 drives and you'll still only get 800MB/s, which TB can handle. With SSD, you can exceed those speeds but the types of files that need over 1000MB/s use up a lot of space (90 minutes at 300MB/s = 1.6TB) so it makes the solution expensive. $5,000 for the box + 8 x $500 for 2TB of SSD = $9,000. Not many people will be buying a $3,000 MP along with $9,000 of storage.
Even OWC make a Thunderbolt adaptor to allow connection to the SAS and the CEO says:
“With the advanced processors modern MacBook Airs, MacBook Pros, iMacs, and even the Mac mini offer, these machines are very capable of running advanced applications and processes that rival the Mac Pro,” said Larry O’Connor, Founder and CEO, Other World Computing. “Now with the Mercury Helios, users of these non-PCIe slot equipped machines can tap into the power of PCIe cards and experience capabilities previously unattainable.”
That OWC device looks like it was built well enough. In terms of drives, populating via 15k SAS drives wasn't what I meant at all. I was thinking of something more like an Atto Roc with parity stripes. You can put together a fairly robust solution this way using something like RE4s and a dumb box with mini sas out. I've been enjoying this conversation and I don't really with to kill it, but no one with half a brain would spend $9000 for a RAID populated with 2TB worth of SSD when existing SATA enterprise drives would be a superior choice in terms of reliability and cost per GB. Someone in IT suggesting that would likely find themselves unemployed. Video editing is a very common use case since we're talking about Macs and high bandwidth requirements. For those guys there is really no reason to ever even consider SSDs for large Raid volumes when you can do much better with typical HDDs. It's just when you're going into the 8TB+ realm, you should be looking at hardware controllers rather than a software raid. All of this needs to be backed up anyway. In terms of raw speed, you can get around 135MB/s sequential write from modern HDDs. You get less in parity stripe mode, but the potential for downtime is lessened. In either case the RAID requires backup. It's just that this is the difference between building a massive raid for $4k~ and building a tiny one with no inherent fault tolerance for $9k. I don't see the potential use case for yours. I don't even think it will happen with NAND. Note I'm not saying that solid state isn't the way forward. I'm just not confident in the future of NAND technology.
Quote:
That's the point though. Apple would ship higher volume so they don't need to make the margins so high.
This is the same Apple that brought us the personal computer, the iPod, iPhone, iPad, the first usable unix desktop, OpenCL. Yeah, they aren't the type to make bold moves. Yield could be an issue if they aimed for 50-core chips but they don't need to do that. A 24-36 core co-processor would work ok.
They weren't the first to bring out the personal computer. They brought out a personal computer. A few others had similar solutions that were basically concurrent, but that doesn't really matter. The point I wished to make was that they aren't that aggressive with workstation technology. Apple's behavior is very conservative unless it involves something that they think they can package. They were conservative on usb3. With thunerbolt it gave them a nice docking station for their notebook line. I couldn't see them backing such a thing unless they were trying to increase market adoption to eventually bring it into their portables. In this kind of situation I could see them trying to groom such a package for their volume lines, but I don't know enough about the power consumption or size of that technology to know anything about its feasibility. It sounds like we're talking about much smaller/simpler cores compared to x86. I'll do more reading on it later.
Quote:
Tim Cook: "Our pro customers are really important to us... don't worry as we're working on something really great for later next year."
If he meant early 2013, he'd have just said "next year". If they intended on using Sandy Bridge, they would have used it.
Whenever an executive makes a comment like this, people read into each word. I think it was intended more as reassurance for PR reasons than anything. In terms of Sandy vs ivy, same chipsets, but Ivy gets a boost in maximum core count. There's just nothing interesting specifically tied to Ivy E, and when it does come out it's likely to be like nehalem where Sandy Bridge E components still fill in the lower end of intel's lineup.
The SSD is the only way to max the bandwidth. Someone here said they spent $30k on a RAM drive so people have their reasons.
As for standard drives, you still need them in RAID 0 and you won't max out Thunderbolt's bandwidth. Plus are you talking about RAIDing the internal drives or external? I hope you ain't making a rat's nest at the back or having external boxes lying around everywhere with separate power supplies.
Exactly.
Yeah the Intel MIC isn't using full Xeon cores. It's like a GPU but the cores run general purpose x86 code.
He said 'later', so I assumed it meant later. Wouldn't you have to read into it to assume it meant earlier? I'd give up hope of hearing another word about the MP from Apple for another year at least.
For the next year, we're going to get as many threads talking about what's going to happen but no matter what they do, people won't be happy because they either:
- don't redesign it and stick with Ivy Bridge in which case it's a minor performance bump after 3 years and no Thunderbolt support
- do redesign it in which case, they will gut the thing and put Thunderbolt on, possibly include an MIC and not have PCI slots (unless they can reconcile the display issue)
- ditch it in favour of a souped up iMac of some kind in which case, we'll get the usual raft of comments on Apple dumbing things down
- discontinue it in which case Apple will be described as abandoning Pros
It doesn't matter what they do, it will be late and some group of people won't like it.
Quote:
Originally Posted by Marvin
The SSD is the only way to max the bandwidth. Someone here said they spent $30k on a RAM drive so people have their reasons.
As for standard drives, you still need them in RAID 0 and you won't max out Thunderbolt's bandwidth. Plus are you talking about RAIDing the internal drives or external? I hope you ain't making a rat's nest at the back or having external boxes lying around everywhere with separate power supplies.
$30k on a ram drive is pretty specialized. Conversely there are many people who drop a couple grand on storage.
Do not make me hunt down DAS units with mini SAS to prove they exist :P. You really can't pull that kind of bandwidth internally. Apple has one RAID card, and it's a poorly supported piece of trash. It's one of those things where if they couldn't do it right, they should have left it to third party solutions. Atto or Areca would have been better. I made no mention of a rat's nest anyway. DAS unit with 1 mini SAS out the back ----> Roc on the back of the mac pro. You make it sound like high bandwidth only came around with SSDs, yet SSDs typically aren't suitable for RAIDs at all, especially anything with striped parity. I'm not sure how they'd hold up with dedicated parity drives, but who really uses those variants anyway? I said if you need high bandwidth and a lot of storage you'd go with something of that sort. SSDs are nonsensical for such things due to limited write cycles and other problems + much higher cost per gigabyte. I really really don't understand your logic here. What would be the case for using an array of SSDs compared to HDDs here? You'd have two boxes either way as the SSDs would quickly max out the internal SATA bus. You can do an internal card, and there are good reasons to populate the internal bays. I was saying if you need a high bandwidth solution, it quickly becomes impractical. You wouldn't really stuff thousands of dollars worth of SSDs internally anyway (at least I wouldn't). The time it becomes a mess is when you have many connections out the front and back with a poor routing method.
i feel like we're almost arguing semantics at this point. My initial point was that as a consumer grade solution usb3 is cheaper. If you're in need of the maximum possible bandwidth and less cost sensitive, mini-SAS has existed longer, so there may not be a good reason to migrate there. Just for reference...
DAS
Mini SAS express 34 card
SAS HBA
Realistically you can set something like this up cheaper if you're careful on what you buy. The express34 ports don't have as much bandwidth. They're just a notebook solution. I linked to a populated DAS. I'd probably buy a bare one myself and populate it with WD enterprise firmware drives, verifying identical firmware across them prior to setup. If you need large amounts of fast storage, this is a feasible solution. Actually I'm not sure of the bay count on the one I linked, but there are many sizes, and they come in horizontal rackmount optional form factors too. I was saying for storage if you need large quantities of performance storage, this remains a better solution than dropping that much on SSDs. If you're talking about consumer grade solutions, usb3 is cheaper for nearly the same performance you'll get from current thunderbolt options. Intel has yet to really push adoption. Wasn't thunderbolt on its data channels tested at a max of 650-800MB/s? Beyond that are there any DAS thunderbolt units that are big enough to push that without SSDs? I wouldn't suggest Raid 0 anyway for anything time crucial. I'd say go with Raid 10 for some measure of fault tolerance. At that point you are often better off allocating more drives to a bigger 5. I'm not sure what rebuild times would be like on one of those, but retrieving a Raid 0 from a NAS or tape backup would take you a very long time.
Back to what I was saying before, you have customers who have potential customers who are looking at upgrades and new ones that are putting together such solutions. Splitting that further you have different prioritizations of cost and performance requirements. I keep trying to make the point that out of the potential matrix of solutions you could draw from this, thunderbolt taken as a connection + current selection of devices falls in the middle here. Beyond that with SSDs it wasn't just a cost issue. They really aren't designed to hold up to such use (yet).
Quote:
He said 'later', so I assumed it meant later. Wouldn't you have to read into it to assume it meant earlier? I'd give up hope of hearing another word about the MP from Apple for another year at least.
For the next year, we're going to get as many threads talking about what's going to happen but no matter what they do, people won't be happy because they either:
- don't redesign it and stick with Ivy Bridge in which case it's a minor performance bump after 3 years and no Thunderbolt support
- do redesign it in which case, they will gut the thing and put Thunderbolt on, possibly include an MIC and not have PCI slots (unless they can reconcile the display issue)
- ditch it in favour of a souped up iMac of some kind in which case, we'll get the usual raft of comments on Apple dumbing things down
- discontinue it in which case Apple will be described as abandoning Pros
It doesn't matter what they do, it will be late and some group of people won't like it.
I figured later as something other than soon. Later next year doesn't necessarily mean "late next year". Next year is later than this year. There is little context here and it makes little sense to dissect this as an oral contract.
You answer your own question here. You can't easily push the limits of Thunderbolt without SSD, which you've already said is not practical. That would indicate Thunderbolt is fine for this setup.
SAS can outperform TB with quad RAID 0 SSDs as shown here:
http://www.barefeats.com/tbolt01.html
but performance varies. Anand did a test with multiple SSD and managed closer to 1GB/s:
http://www.anandtech.com/show/6023/the-nextgen-macbook-pro-with-retina-display-review/11
They also hook up two TB drives to the rMBP and manage 1.38GB/s. They said if they had two of the SSD drives they'd expect closer to 1.8GB/s but they'd need to test it to verify.
They are test scenarios that aren't likely to be used in a practical scenario but it shows that TB is up to the task and it works on all Macs, not just a model people aren't buying in large numbers.
If your work is storage limited then you can buy a $600 Mini, an $1100 Pegasus with 4x SSDs and you will be able to cope with workflows that a stock $3000 Mac Pro couldn't cope with.
That doesn't explain why they'd wait until next year before using Sandy Bridge when they can use it now. The only reason for the delay has to be that they plan to use something that's not out yet i.e Ivy Bridge. It would be a stretch to say they are waiting on Knights Ferry as there's no indication this is the route they will take yet but I think SB is off the table or it would be shipping.
Quote:
Originally Posted by Marvin
That doesn't explain why they'd wait until next year before using Sandy Bridge when they can use it now. The only reason for the delay has to be that they plan to use something that's not out yet i.e Ivy Bridge. It would be a stretch to say they are waiting on Knights Ferry as there's no indication this is the route they will take yet but I think SB is off the table or it would be shipping.
The Pegasus box that Apple ships isn't aimed at the market I suggested, and RAID 0 requires that you accept the potential for a day of downtime if a disk fails. Beyond that if you're migrating data on and off constantly due to small disks on your fast storage, you run into the issue that SSDs have limited write cycles. You're likely to kill that large storage investment quite fast.
In terms of Sandy vs Ivy, my only guess would be they wanted to kill it and didn't assign anyone to create an updated reference design. This is not a project that requires many people. Even board designs don't usually involve a large team. If they had a couple engineers on it, you'd have something. As it is, this was likely something that was abandoned due to uncertainty. Perhaps the rumors were true regardless of whether the articles themselves involved bloggers looking for page hits. There isn't any real reason to skip a generation when they both plug into the same chipset. Either way you'll have the same board design. I do not buy into your theories of early access, especially on a line where they are likely to be relatively conservative. We could see a 16 core mac pro. It's likely that chips for a 20 core would exceed Apple's chip budgeting at their desired margins. I think you place too much confidence in Apple's semantics. That may be the reason we disagree on this one. I think they just never allocated anyone to such a project.
Well I agree SB is off the table in the sense that it is the only processor in the machine. However Intel intends for MiC to be a family of devices as part of their foray into high performance computing. Even if Apple forgoes the 62 core coprocessor there are other chips that will be coming on line targeting high performance computing. For example Intel expects to integrate Infiniband into a multicore processor chip.
The latest rumor is that Intel will launch Xeon Phi and it high performance computing initiative in the November/December time frame. If so this would be reason enough to speculate that the Mac Pro, at least, is waiting for this hardware to launch. I just don't see Apple waiting until October of 2013 and hope that Intel has an Ivy Bridge based Xeon ready at that time.
Lastly Intel is basically in love with Apple. Without Apple there would be nobody innovating with their hardware. So I have no doubt at all that Apple and Intel have a Skunk Works of sorts working on new technology for the Mac Pro. In this regard I wouldn't be surprised at all that Intel and Apple have some sort of exclusive deal going for a MiC chip of some sort, much like the way Thunderbolt was handled.
No RAID 0 and no SSD means Thunderbolt is fast enough. You were saying TB is not fast enough to replace Mini-SAS but it is unless you use SSD in RAID 0, which as you say, wouldn't be the case.
Early access to what? The only thing I suggested they could get early access to was a new Thunderbolt controller. They got early access to the first one among other things from Intel:
http://www.appleinsider.com/articles/11/02/24/intel_details_thunderbolt_as_exclusive_to_apple_until_2012.html
I don't see that being a problem. I'd say it would be a worse situation if they made Mini-SAS, Fibre Channel or Infini-band exclusively on a machine that makes up 5% or less of their lineup while not offering a significant advantage over Thunderbolt (taking into consideration its roadmap).
A multi-protocol optical interconnect is the way forward so everything else that's not that will die out and shouldn't be supported.
Not if the work is storage limited. If you are editing high-bitrate intra-frame compressed video, the Mini doesn't have a bottleneck if you have fast storage. When it comes to encoding it will be slower but you can still do it. Without fast storage, the Mac Pro wouldn't be able to do the edit at all. In that scenario, the stock Mac Pro couldn't do the job that a Mac Mini could do. Obviously adding storage is trivial but the point is that the MP itself is less important than the connection and storage.
I think it will be in the 2nd half of the year at the earliest that we find out what's going on. Even though people don't see the last update as an update, it was. There's no way they would obsolete that refresh within 6-8 months with a radical machine.
I don't think they would use the MIC along with a DP system so I think it makes sense to wait for Ivy Bridge. Q3 runs July - Sept so not as far out as October 2013 but still around a year from now.
We know the cycle:
iMac/Mini - Aug/Sept (might be announced at iPhone launch though due to major iMac redesign)
iPhone - Sept/Oct (I'd say Oct 5th)
iPad - March
MBP/MBA (Haswell) - May/June
WWDC June - announce redesigned MP, possibly exclusive Ivy Bridge launch, Haswell iMac/Mini similar timeframe
Quote:
Originally Posted by Marvin
No RAID 0 and no SSD means Thunderbolt is fast enough. You were saying TB is not fast enough to replace Mini-SAS but it is unless you use SSD in RAID 0, which as you say, wouldn't be the case.
While I enjoy our conversations, sometimes I'm unsure of how to explain certain things in a way that keeps the discussion on track. I think you're referring to something like the Promise Pegasus populated with ssds right? I included a wider product range in my comparison where this wouldn't necessarily be the case. My point was if you aren't going to reach such speeds, usb3 is the cheaper solution. You can set it up for less in most cases. With the Pegasus raid I don't trust them. They may go cheap on the installed drives. I'd rather buy my own, although Promise does make some solid professional and enterprise grade hardware. If you're putting together a direct attached storage solution, there are tons of inexpensive options going the usb route. If you're going to spend the cash on a fast/robust solution, you can do better than what is offered currently over thunderbolt. To make it a viable option, it needs more options that can compete in enough points wiht usb3 on the low end or fibre/sas on the higher end. I called the $9k worth of ssds nonsense because you can do much better with a large das array populated by hdds. Noise is a factor with any of them. The Promise DAS isn't silent, and neither are the others. If you buy one that doesn't use trash fans to save a few dollars, it's not too bad. Some of the lower end solutions use either sleeve fans or ones with cheap bearings to save a couple dollars per fan. They'll get noisy as they wear. Hopefully costs and adoption rates will trend in a favorable direction so that we have solid thunderbolt options. Right now I see it as new and not a perfect fit against existing solutions. I see it more something that should be a volume solution rather than a niche one. It currently aligns better with computers that have integrated graphics within the cpu package. It offers lower performance relative to its cost against current solutions seen in desktops. To me the place to really push adoption is on the mobile end. It just has to better compete in price and available peripherals with usb3. Pushing adoption from the workstation end makes no sense to me. It's something that would be better off leveraged into that space rather than forced given the inherent misalignment and pre-existing solutions.
Quote:
Originally Posted by Marvin
We know the cycle:
iMac/Mini - Aug/Sept (might be announced at iPhone launch though due to major iMac redesign)
iPhone - Sept/Oct (I'd say Oct 5th)
iPad - March
MBP/MBA (Haswell) - May/June
WWDC June - announce redesigned MP, possibly exclusive Ivy Bridge launch, Haswell iMac/Mini similar timeframe
Here's to you're being right about the new iMacs coming soon. Would love a more powerful mini that could be linked to provide more power.
That said, the store is down in the US and Japan (haven't checked elsewhere). Likely just maintenance.
We'll see.
Apple's desktop line needs a re-boot.
All 3 dekstops are out of date and overpriced and underspecced.
It's an all out joke.
The Macbook Pro retina makes the rest of the Mac line (bar the Air) look antiquated.
A slender iMac with ivy in it? whoop de doo? What took it so long.
If it's got retina in it? We may do business, Apple.
After the retina Pro? I'm not a laptop fan...but after playing with that? No going back.
Sorry, overpriced and under value from the Mini to the Pro workstation (for want of a better name. 'Joke' comes to mind.)
Lemon Bon Bon.
http://www.engadget.com/2012/08/27/sony-optiarc-closure/
Apple used them for some of their superdrives and Lacie used them too. There are other quality options for bundled drives but it's best at this stage to leave them outside as optional extras.
It made sense that the old-style MBPs kept them because they'd have to redesign them and retool the manufacturing line just for models that they'd discontinue. When the prices drop, the old-style models will just disappear.
With the iMac, they could do a similar thing - leave the lineup the same except for the highest model and make it Retina with SSD and no optical but I don't think that would be a good move.
Quote:
Originally Posted by Marvin
On the subject of optical drives, Sony is ditching the PC business:
http://www.engadget.com/2012/08/27/sony-optiarc-closure/
Apple used them for some of their superdrives and Lacie used them too. There are other quality options for bundled drives but it's best at this stage to leave them outside as optional extras.
It made sense that the old-style MBPs kept them because they'd have to redesign them and retool the manufacturing line just for models that they'd discontinue. When the prices drop, the old-style models will just disappear.
With the iMac, they could do a similar thing - leave the lineup the same except for the highest model and make it Retina with SSD and no optical but I don't think that would be a good move.
These things are not uncommon. Around a decade ago lcd display quality still sucked unless you wanted to pay $3000 for a 20" display, yet crts had been consistently bid down to where they were made with very cheap parts due to thinning margins. I am not surprised they're moving on.