I disagree here. These chips are shared between servers and workstations. In both cases intel tends to be very conservative in favor of stability. Note how many things pushed them back already. They have high margins and limited competition in this segment.
Worldwide server shipments in 2011 were just over 8 million. That's around 2 million per quarter. That makes server and workstation volume worldwide just 3 million units per quarter. That's 12 million servers and workstations vs 340 million desktop/laptops - 3.5%. Intel would be much better off trying to get these fast chips into more mainstream form factors and lower price points.
A lot of the server chips aren't priced any higher than desktop ones so their margins can't be higher in all cases.
I forgot to add, it might fit in an imac in terms of tdp, yet they still won't use the same board. They didn't use Ivy Bridge E3s either. We'll see what actually happens, but I don't expect Apple to design extra logic boards just to accommodate a slightly wider range of cpu offerings in the imac. I could be wrong as they've turned less conservative on hardware choices lately.
They have to design the boards as it is though. The leaked model ID MP60 suggests they'll retain the headless form factor for now but Intel has no reason to show much interest in this segment:
"Furthermore, we've been told that the company is currently planning to change its media policy on the sampling side, as Intel would like to change the perception of the company from a component provider into a solutions provider. Naturally, this leaves the high-end desktop/workstation in a bit of a tight fit, as the roadmap will show."
Intel doesn't like allowing their rival companies to profit from their work:
That says Intel was required by the FTC to support PCIe until at least 2016 but they obviously tried to avoid it in the Atom machines at one point. Obviously motherboard manufacturers can put in their own PCIe support but imagine when Intel goes the solutions route instead of the component route and just ships the board with the CPU stuck to it.
This is where Thunderbolt comes in because that also blocks AMD for high-end peripherals.
That's the way all the companies will go too all the way up to servers and workstations. An SoC means you have to use their tech and soldered components means you can't just plug in a new upgrade and not buy another chip from Intel. Intel doesn't want you to keep a Mac Pro for 6 years and give money to NVidia to prolong its usefulness:
"It's a move that could make PC OEMs happy too. Soldering a component to a motherboard is cheaper than soldering a socket and then fitting that processor into the socket. The difference might only be pennies, but spread over millions of PCs, those pennies add up.
As far as the PC OEMs are concerned, killing off the PC upgrade market would be a good thing because it would push people to buy new PCs rather than upgrade their existing hardware. The PC industry is currently stagnant, partly because consumers and enterprise are making existing hardware last longer."
Both users will eventually have to realize that Apple needs to do something to assure desktops in the future.
I don't think it's a good idea to have the same chassis for both Mini and Pro. The Mini is designed as a low power desktop and server. The Mac Pro is designed as a high-end machine. When you try to pull them together, you obviously try to create the mid-range tower again but they won't do that.
While it is true that Intel is more conservative with the workstation chipsets that doesn't mean that they need to be stuck with USB 3 until the next major rev. USB 3 would be valuable even in the workstation market and could be implemented without much external change to the chipset. It might be out of character but I think market demand is strong enough that Intel could rev a chipset early.
This is exactly the problem the iMac currently is the only Mac that has the volume that could even remotely justify alternative motherboards. Conversely this is one of the reasons that I see that justifies a complete redesign of the Mac Pro, maybe even incorporating the Mac Pro and Mini concepts into the same chassis. They need to get the volume up on the Pro chassis. With sufficient sales of the general machine the lower volume "Pro" motherboard is less of an issues.
Now I'd be the first to admit that nobody would like this. The Mini users like their tiny foot point machine and the Pro users like the idea of their big format machines. Both users will eventually have to realize that Apple needs to do something to assure desktops in the future. The idea of a chassis with a low performance low power option, as seen in the Mini along with a high performance board option seems to be the only rational approach to dwindling sales.
They all ship with usb3. I said it's not part of the chipset. I didn't say shipping workstations don't include it. For now it's on the oems to include it. I don't expect multiple board designs in the 27" imac. I was extremely surprised they incorporated further custom display work, as it's potentially very expensive. I expected them to at least take a year of experience doing it with the retina macbook pros prior to moving up to doing so on a 27" display.
Worldwide server shipments in 2011 were just over 8 million. That's around 2 million per quarter. That makes server and workstation volume worldwide just 3 million units per quarter. That's 12 million servers and workstations vs 340 million desktop/laptops - 3.5%. Intel would be much better off trying to get these fast chips into more mainstream form factors and lower price points.
A lot of the server chips aren't priced any higher than desktop ones so their margins can't be higher in all cases.
They have to design the boards as it is though. The leaked model ID MP60 suggests they'll retain the headless form factor for now but Intel has no reason to show much interest in this segment:
I thought it was higher than that, but I didn't know where to find the statistic. Going back to what I said, I merely disagreed with the idea that they're trying to pull out of this market. When I mentioned different boards, it was an issue of Haswell compared to Ivy Bridge EP. I don't think they're going to offer options that cross between those two realms. They'll just figure that a number of their users will break down and eventually buy something. If it breaks too many points of compatibility, they'll wait longer, but they will still buy something unless they're already hitting bootcamp way too often.
"It's a move that could make PC OEMs happy too. Soldering a component to a motherboard is cheaper than soldering a socket and then fitting that processor into the socket. The difference might only be pennies, but spread over millions of PCs, those pennies add up.
This may not make them that happy if it pushes more support costs downstream and forces them to retain more stock to support the configurations offered as they can no longer swap out a cpu if an order changes. We shall see if it ends up being a true cost savings. I disagree that the upgrade market hits them. The upgrade crowd is more likely to just build their own machine. Most of the time to get a meaningful upgrade, they needed to change the board out anyway as cpu costs rarely plummet on a given generation.
I had a hard time reading that, either it is marketing BS reacting with my literal mind or my blood sugar is out of whack. In any event it kinda looks like USB 3 is at the end of its life. It is far to early to be turning out yet another USB standard.
Quote:
Originally Posted by Frank777
The question may not be whether they can put USB3 onboard, but which version of USB3 will be onboard.
It's confusing because instead of calling it "USB 3.1", they are calling it "USB 3.0 SuperSpeed". In any event, 10 Gbps USB will render Thunderbolt irrelevant in the consumer space. I'd rather see the new Mac Pros offer USB SS than TB, if it comes down to a question of either or.
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Most PC users who need to use a PCIe card will just buy a cheap tower. That leaves Mac users to support TB. It's going to wither on the vine.
It's confusing because instead of calling it "USB 3.1", they are calling it "USB 3.0 SuperSpeed". In any event, 10 Gbps USB will render Thunderbolt irrelevant in the consumer space. I'd rather see the new Mac Pros offer USB SS than TB, if it comes down to a question of either or.
There is no reason for it to be either or unless it comes to lazy engineering. With USB 3, they need to test third party chipsets either way. Without that what would they use in the next revision of the thunderbolt display?
Quote:
Originally Posted by Tallest Skil
Nope.
Thunderbolt does seem to have some adoption problems if intel views it as a mass market port. If they view it as something more like eSATA or even SAS, that's different. Increasing the speed or changing specs won't help that. I'm not confident that we'll see a lot of external PCIe devices outside of the specialty segments where these things are truly required. Note the Black Magic breakout boxes. I like things like that as they represent a complete certified solution. Any airflow and power requirements can be assessed by the engineers that design the card, and you get something that is actually certified to thunderbolt requirements. Unfortunately I don't think frequent changes will help manufacturer adoption rates.
Most PC users who need to use a PCIe card will just buy a cheap tower.
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
I don't see a reinvigoration of PCI. I see it as mostly things tacked onto the board. Much of what used to require cards is just attached to the logic board via IO hub at this point. The idea of breaking everything out into its own little box is just as out of focus as PCIe with mass market units. The edge PCI has is that it's often cheaper and more available, but development is very low on Macs at this point. Unless you need something like a capture card, you won't find much that is stable on OSX these days.
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Daisy-chaining, displays, the fact that Thunderbolt is 10Gb/s both ways simultaneously, the endless adaptability and backward compatibility of Thunderbolt to other port types compared to USB…
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Most PC users who need to use a PCIe card will just buy a cheap tower. That leaves Mac users to support TB. It's going to wither on the vine.
There is no reason why a solid state storage drive couldn't saturate USB3 or even TB for that matter. There is a popular misconception that these ports are some how invincible and won't ever be overwhelmed by some applications. The facts are rather the opposite each of these ports represents bandwidth limited serial ports over which data must flow, neither of them is actually a high capacity port when compared to a sixteen lane PCI Express 3 slot. TB has the advantage of bidirectional data transfer which can be a big win at times and doesn't suffer from protocol issues like USB does.
In summation each TB port gives one several advantages over USB 3. This leads to a preference for TB in high performance applications. Frankly this isn't much different than in the days of USB 1 or USB 2, these ports where always the low performance, low cost solutions to external connectivity.
PCI Express slots will be around for a very long time. There isn't a better solution on the horizon.
Now that doesn't mean Apple will support PCI Express in the future. However if they drop PCI Express completely it will be seen as a sign that Apples hardware is no longer professional use capable. The reality is there is nothing other than PCI Express for the less than mainstream computer expansion or I/O needs.
Quote:
Originally Posted by Marvin
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
The reality is there is nothing other than PCI Express for the less than mainstream computer expansion or I/O needs.
Thunderbolt is PCIe + displayport. It's not Thunderbolt vs PCIe but internal PCIe vs external. Even the group that controls the PCI spec proposed an external cable solution:
"The group is expected to deliver a standard system makers can implement in products before June 2013. The big issue here is proprietary versus industry standard, It's not clear third parties will have access to Thunderbolt on the same basis they get access to PCI Express."
"Apple is fine with the extra cost of the router chips, but we don’t need [the multiprotocol support] and a couple extra chips don't make business sense for us"
The only thing unique about Thunderbolt is that it uses one of the channels for display output. If they were both data, you'd get 20Gbps bi-directional or 40Gbps in 2014.
The problem with the PCI standard is there's so many competing companies. Apple and Intel are members of the group. Sometimes it's best to break away from the crowd and get the job done.
If they go the route of a PCI cable, they won't get display output and it would still need new drivers for plug and play. If Intel relinquishes control of the spec, quality control might suffer too.
If the PCI group come up with a slimmer port design and it works well, that's fine but it's going to have even worse adoption problems than TB.
But it doesn't really matter. As long as boxes like the Magma have decent support:
and there are enough TB peripherals, all that's missing is the GPU support but Apple's machines use decent GPUs anyway and it's possible to run external GPUs and other cards over Thunderbolt with the right drivers:
[VIDEO]
[VIDEO]
[VIDEO]
The 20Gbps upgrade should get rid of bottlenecks and it just needs some driver support from the vendors.
If the PCI group come up with a slimmer port design and it works well, that's fine but it's going to have even worse adoption problems than TB.
You don't know what they wish to implement. I'm not even sure of Intel's strategy for thunderbolt. Do they see it as something aimed at mass market adoption? Intel never suggested it would be cheap or any kind of replacement for usb. If the PCI group was trying to come up with something that would assign an open standard that could be applied across traditional offerings and mobile devices without a high cost per unit, they could really have something. You shouldn't be hung up on thunderbolt. Beyond that much of the time I read various numbers thrown out as to future specs, yet it remains to be seen what is offered in terms of PCI lanes from top to bottom. Right now you can get 80 lanes on an E5 workstation compared to 16 on LGA1155. I don't remember how many on notebooks. It's something like 12 or 16. Thunderbolt must pull from those for both display and data in addition to other ports.
Quote:
But it doesn't really matter. As long as boxes like the Magma have decent support:
and there are enough TB peripherals, all that's missing is the GPU support but Apple's machines use decent GPUs anyway and it's possible to run external GPUs and other cards over Thunderbolt with the right drivers:
GPU support is extremely unlikely, and I would recommend against suggesting that Magma solution to anyone. IPCI cards for the most part will not conform to thunderbolt's certification specs. Sonnet actually maintains a list of what has suitable drivers for their solution. That is the minimum level of support you would want. I don't see eGPUs becoming popular. You have the mini which is a budget machine, so the number of users who might go that route would be potentially limited. You have the imacs where it's unlikely to make sense over existing options unless some of the higher end desktop cards are certified. Workstation cards are more likely due to their margins, but they still aren't that likely.
Quote:
The 20Gbps upgrade should get rid of bottlenecks and it just needs some driver support from the vendors.
That isn't as big of a help as you think. You need a consistent standard if you want to see a lot of third parties start to develop for thunderbolt. If they're making spec changes two years in a row, this may just inhibit adoption. Manufacturers tend to hate fragmentation. With usb it's cheap to build and provides backwards compatibility with earlier versions of the standard.
As a media professional I need expansion and connectivity. No matter the amount of of native processing I'm going to need PCI slots (5-10) for external co-processing cards by various manufacturers. I'll be needing multiple Thunderbolt, Firewire, USB 2&3, and even SCSI (LTO archiving) and an obscene amount of RAM (64 Gig). The system needs to be a nexus for connectivity.
The system will be used for:
1) Audio (Both Music Recording and Film)
2) Video (Editing, FX and Colour correction for Indie Film and Video)
3) 3D (FX and 3D Printing)
4) Web and Print collateral
I'd be very happy with an updated design like the old IIx with it's multiple slots (NuBus in those days) An example from those day would be my ProTools system from '93
As a media professional I need expansion and connectivity. No matter the amount of of native processing I'm going to need PCI slots (5-10) for external co-processing cards by various manufacturers. I'll be needing multiple Thunderbolt, Firewire, USB 2&3, and even SCSI (LTO archiving) and an obscene amount of RAM (64 Gig). The system needs to be a nexus for connectivity.
Apple has never really offered that many PCI slots. You'd either end up with multiple machines, a different brand, or a different solution somewhere in there. I'm not questioning your requirements, just whether you'd be able to fulfill them that way with a Mac Pro.
The big fear I have in no PCI slots at all. The reality is you don't need a lot of slots to configure a machine for a specific usage. Apples problem is more of one of making those slots affordable.
Quote:
Originally Posted by hmm
Apple has never really offered that many PCI slots. You'd either end up with multiple machines, a different brand, or a different solution somewhere in there. I'm not questioning your requirements, just whether you'd be able to fulfill them that way with a Mac Pro.
The big fear I have in no PCI slots at all. The reality is you don't need a lot of slots to configure a machine for a specific usage. Apples problem is more of one of making those slots affordable.
It varies, but yeah it's possible to populate all available slots if you need specific editing hardware. One of the things I've noticed in the past couple years is that the only stable PCI cards left on OSX as of Mountain Lion are basically those with either higher margins or supported as part of a hardware + software solution. Some of those developers have put out thunderbolt and usb3 versions. Black Magic comes to mind. Sonnet has a list of cards with thunderbolt compliant drivers. It is possible that if you need more connections than that will support or multiple high bandwidth ports, you will soon have to look at Windows.
Comments
The volumes here are low too though:
http://www.idc.com/getdoc.jsp?containerId=prUS23347812#.UQfDmGjdzi0
Worldwide server shipments in 2011 were just over 8 million. That's around 2 million per quarter. That makes server and workstation volume worldwide just 3 million units per quarter. That's 12 million servers and workstations vs 340 million desktop/laptops - 3.5%. Intel would be much better off trying to get these fast chips into more mainstream form factors and lower price points.
A lot of the server chips aren't priced any higher than desktop ones so their margins can't be higher in all cases.
They have to design the boards as it is though. The leaked model ID MP60 suggests they'll retain the headless form factor for now but Intel has no reason to show much interest in this segment:
http://www.brightsideofnews.com/news/2012/9/11/intel-desktop-roadmap-i7-3970k-coming-in-q42c-i7-4900-in-q3-2013.aspx
"Furthermore, we've been told that the company is currently planning to change its media policy on the sampling side, as Intel would like to change the perception of the company from a component provider into a solutions provider. Naturally, this leaves the high-end desktop/workstation in a bit of a tight fit, as the roadmap will show."
Intel doesn't like allowing their rival companies to profit from their work:
http://www.electronista.com/articles/10/11/02/intel.oak.trail.required.to.carry.pcie.by.ftc/
That says Intel was required by the FTC to support PCIe until at least 2016 but they obviously tried to avoid it in the Atom machines at one point. Obviously motherboard manufacturers can put in their own PCIe support but imagine when Intel goes the solutions route instead of the component route and just ships the board with the CPU stuck to it.
This is where Thunderbolt comes in because that also blocks AMD for high-end peripherals.
That's the way all the companies will go too all the way up to servers and workstations. An SoC means you have to use their tech and soldered components means you can't just plug in a new upgrade and not buy another chip from Intel. Intel doesn't want you to keep a Mac Pro for 6 years and give money to NVidia to prolong its usefulness:
http://www.zdnet.com/intel-preparing-to-put-an-end-to-user-replaceable-cpus-7000008024/
"It's a move that could make PC OEMs happy too. Soldering a component to a motherboard is cheaper than soldering a socket and then fitting that processor into the socket. The difference might only be pennies, but spread over millions of PCs, those pennies add up.
As far as the PC OEMs are concerned, killing off the PC upgrade market would be a good thing because it would push people to buy new PCs rather than upgrade their existing hardware. The PC industry is currently stagnant, partly because consumers and enterprise are making existing hardware last longer."
I don't think it's a good idea to have the same chassis for both Mini and Pro. The Mini is designed as a low power desktop and server. The Mac Pro is designed as a high-end machine. When you try to pull them together, you obviously try to create the mid-range tower again but they won't do that.
Quote:
Originally Posted by wizard69
While it is true that Intel is more conservative with the workstation chipsets that doesn't mean that they need to be stuck with USB 3 until the next major rev. USB 3 would be valuable even in the workstation market and could be implemented without much external change to the chipset. It might be out of character but I think market demand is strong enough that Intel could rev a chipset early.
This is exactly the problem the iMac currently is the only Mac that has the volume that could even remotely justify alternative motherboards. Conversely this is one of the reasons that I see that justifies a complete redesign of the Mac Pro, maybe even incorporating the Mac Pro and Mini concepts into the same chassis. They need to get the volume up on the Pro chassis. With sufficient sales of the general machine the lower volume "Pro" motherboard is less of an issues.
Now I'd be the first to admit that nobody would like this. The Mini users like their tiny foot point machine and the Pro users like the idea of their big format machines. Both users will eventually have to realize that Apple needs to do something to assure desktops in the future. The idea of a chassis with a low performance low power option, as seen in the Mini along with a high performance board option seems to be the only rational approach to dwindling sales.
They all ship with usb3. I said it's not part of the chipset. I didn't say shipping workstations don't include it. For now it's on the oems to include it. I don't expect multiple board designs in the 27" imac. I was extremely surprised they incorporated further custom display work, as it's potentially very expensive. I expected them to at least take a year of experience doing it with the retina macbook pros prior to moving up to doing so on a 27" display.
Quote:
Originally Posted by Marvin
The volumes here are low too though:
http://www.idc.com/getdoc.jsp?containerId=prUS23347812#.UQfDmGjdzi0
Worldwide server shipments in 2011 were just over 8 million. That's around 2 million per quarter. That makes server and workstation volume worldwide just 3 million units per quarter. That's 12 million servers and workstations vs 340 million desktop/laptops - 3.5%. Intel would be much better off trying to get these fast chips into more mainstream form factors and lower price points.
A lot of the server chips aren't priced any higher than desktop ones so their margins can't be higher in all cases.
They have to design the boards as it is though. The leaked model ID MP60 suggests they'll retain the headless form factor for now but Intel has no reason to show much interest in this segment:
I thought it was higher than that, but I didn't know where to find the statistic. Going back to what I said, I merely disagreed with the idea that they're trying to pull out of this market. When I mentioned different boards, it was an issue of Haswell compared to Ivy Bridge EP. I don't think they're going to offer options that cross between those two realms. They'll just figure that a number of their users will break down and eventually buy something. If it breaks too many points of compatibility, they'll wait longer, but they will still buy something unless they're already hitting bootcamp way too often.
Quote:
http://www.zdnet.com/intel-preparing-to-put-an-end-to-user-replaceable-cpus-7000008024/
"It's a move that could make PC OEMs happy too. Soldering a component to a motherboard is cheaper than soldering a socket and then fitting that processor into the socket. The difference might only be pennies, but spread over millions of PCs, those pennies add up.
This may not make them that happy if it pushes more support costs downstream and forces them to retain more stock to support the configurations offered as they can no longer swap out a cpu if an order changes. We shall see if it ends up being a true cost savings. I disagree that the upgrade market hits them. The upgrade crowd is more likely to just build their own machine. Most of the time to get a meaningful upgrade, they needed to change the board out anyway as cpu costs rarely plummet on a given generation.
The question may not be whether they can put USB3 onboard, but which version of USB3 will be onboard.
I had a hard time reading that, either it is marketing BS reacting with my literal mind or my blood sugar is out of whack. In any event it kinda looks like USB 3 is at the end of its life. It is far to early to be turning out yet another USB standard.
Quote:
Originally Posted by Frank777
The question may not be whether they can put USB3 onboard, but which version of USB3 will be onboard.
It's confusing because instead of calling it "USB 3.1", they are calling it "USB 3.0 SuperSpeed". In any event, 10 Gbps USB will render Thunderbolt irrelevant in the consumer space. I'd rather see the new Mac Pros offer USB SS than TB, if it comes down to a question of either or.
Originally Posted by Junkyard Dawg
In any event, 10 Gbps USB will render Thunderbolt irrelevant in the consumer space.
Nope.
Quote:
Originally Posted by Tallest Skil
Nope.
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Most PC users who need to use a PCIe card will just buy a cheap tower. That leaves Mac users to support TB. It's going to wither on the vine.
Quote:
Originally Posted by Junkyard Dawg
It's confusing because instead of calling it "USB 3.1", they are calling it "USB 3.0 SuperSpeed". In any event, 10 Gbps USB will render Thunderbolt irrelevant in the consumer space. I'd rather see the new Mac Pros offer USB SS than TB, if it comes down to a question of either or.
There is no reason for it to be either or unless it comes to lazy engineering. With USB 3, they need to test third party chipsets either way. Without that what would they use in the next revision of the thunderbolt display?
Quote:
Originally Posted by Tallest Skil
Nope.
Thunderbolt does seem to have some adoption problems if intel views it as a mass market port. If they view it as something more like eSATA or even SAS, that's different. Increasing the speed or changing specs won't help that. I'm not confident that we'll see a lot of external PCIe devices outside of the specialty segments where these things are truly required. Note the Black Magic breakout boxes. I like things like that as they represent a complete certified solution. Any airflow and power requirements can be assessed by the engineers that design the card, and you get something that is actually certified to thunderbolt requirements. Unfortunately I don't think frequent changes will help manufacturer adoption rates.
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
Quote:
Originally Posted by Marvin
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
I don't see a reinvigoration of PCI. I see it as mostly things tacked onto the board. Much of what used to require cards is just attached to the logic board via IO hub at this point. The idea of breaking everything out into its own little box is just as out of focus as PCIe with mass market units. The edge PCI has is that it's often cheaper and more available, but development is very low on Macs at this point. Unless you need something like a capture card, you won't find much that is stable on OSX these days.
Originally Posted by Junkyard Dawg
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Daisy-chaining, displays, the fact that Thunderbolt is 10Gb/s both ways simultaneously, the endless adaptability and backward compatibility of Thunderbolt to other port types compared to USB…
Quote:
Originally Posted by Junkyard Dawg
Perhaps you can illuminate us as to why a consumer would choose TB over USB 3.0 SS? No external drive can saturate either bus, but TB costs significantly more and offers far fewer devices to choose from.
Most PC users who need to use a PCIe card will just buy a cheap tower. That leaves Mac users to support TB. It's going to wither on the vine.
There is no reason why a solid state storage drive couldn't saturate USB3 or even TB for that matter. There is a popular misconception that these ports are some how invincible and won't ever be overwhelmed by some applications. The facts are rather the opposite each of these ports represents bandwidth limited serial ports over which data must flow, neither of them is actually a high capacity port when compared to a sixteen lane PCI Express 3 slot. TB has the advantage of bidirectional data transfer which can be a big win at times and doesn't suffer from protocol issues like USB does.
In summation each TB port gives one several advantages over USB 3. This leads to a preference for TB in high performance applications. Frankly this isn't much different than in the days of USB 1 or USB 2, these ports where always the low performance, low cost solutions to external connectivity.
PCI Express slots will be around for a very long time. There isn't a better solution on the horizon.
Now that doesn't mean Apple will support PCI Express in the future. However if they drop PCI Express completely it will be seen as a sign that Apples hardware is no longer professional use capable. The reality is there is nothing other than PCI Express for the less than mainstream computer expansion or I/O needs.
Quote:
Originally Posted by Marvin
Most PC users are buying laptops so Thunderbolt is the only option for PCIe devices. It doesn't matter if it will be a popular standard, laptops are popular and Thunderbolt is the current successor to ExpressCard so it will be used where it works best and it will scale beyond 10Gbps.
By all means continue to caress your PCI slots while they and your second-hand 2009 tower are still relevant though.
Thunderbolt is PCIe + displayport. It's not Thunderbolt vs PCIe but internal PCIe vs external. Even the group that controls the PCI spec proposed an external cable solution:
http://www.eetimes.com/electronics-news/4217190/PCI-Express-takes-on-Thunderbolt
"The group is expected to deliver a standard system makers can implement in products before June 2013. The big issue here is proprietary versus industry standard, It's not clear third parties will have access to Thunderbolt on the same basis they get access to PCI Express."
"Apple is fine with the extra cost of the router chips, but we don’t need [the multiprotocol support] and a couple extra chips don't make business sense for us"
The only thing unique about Thunderbolt is that it uses one of the channels for display output. If they were both data, you'd get 20Gbps bi-directional or 40Gbps in 2014.
The problem with the PCI standard is there's so many competing companies. Apple and Intel are members of the group. Sometimes it's best to break away from the crowd and get the job done.
If they go the route of a PCI cable, they won't get display output and it would still need new drivers for plug and play. If Intel relinquishes control of the spec, quality control might suffer too.
If the PCI group come up with a slimmer port design and it works well, that's fine but it's going to have even worse adoption problems than TB.
But it doesn't really matter. As long as boxes like the Magma have decent support:
http://www.magma.com/thunderbolt-compatibility
and there are enough TB peripherals, all that's missing is the GPU support but Apple's machines use decent GPUs anyway and it's possible to run external GPUs and other cards over Thunderbolt with the right drivers:
[VIDEO]
[VIDEO]
[VIDEO]
The 20Gbps upgrade should get rid of bottlenecks and it just needs some driver support from the vendors.
Quote:
Originally Posted by Marvin
If the PCI group come up with a slimmer port design and it works well, that's fine but it's going to have even worse adoption problems than TB.
You don't know what they wish to implement. I'm not even sure of Intel's strategy for thunderbolt. Do they see it as something aimed at mass market adoption? Intel never suggested it would be cheap or any kind of replacement for usb. If the PCI group was trying to come up with something that would assign an open standard that could be applied across traditional offerings and mobile devices without a high cost per unit, they could really have something. You shouldn't be hung up on thunderbolt. Beyond that much of the time I read various numbers thrown out as to future specs, yet it remains to be seen what is offered in terms of PCI lanes from top to bottom. Right now you can get 80 lanes on an E5 workstation compared to 16 on LGA1155. I don't remember how many on notebooks. It's something like 12 or 16. Thunderbolt must pull from those for both display and data in addition to other ports.
Quote:
But it doesn't really matter. As long as boxes like the Magma have decent support:
http://www.magma.com/thunderbolt-compatibility
and there are enough TB peripherals, all that's missing is the GPU support but Apple's machines use decent GPUs anyway and it's possible to run external GPUs and other cards over Thunderbolt with the right drivers:
GPU support is extremely unlikely, and I would recommend against suggesting that Magma solution to anyone. IPCI cards for the most part will not conform to thunderbolt's certification specs. Sonnet actually maintains a list of what has suitable drivers for their solution. That is the minimum level of support you would want. I don't see eGPUs becoming popular. You have the mini which is a budget machine, so the number of users who might go that route would be potentially limited. You have the imacs where it's unlikely to make sense over existing options unless some of the higher end desktop cards are certified. Workstation cards are more likely due to their margins, but they still aren't that likely.
Quote:
The 20Gbps upgrade should get rid of bottlenecks and it just needs some driver support from the vendors.
That isn't as big of a help as you think. You need a consistent standard if you want to see a lot of third parties start to develop for thunderbolt. If they're making spec changes two years in a row, this may just inhibit adoption. Manufacturers tend to hate fragmentation. With usb it's cheap to build and provides backwards compatibility with earlier versions of the standard.
As a media professional I need expansion and connectivity. No matter the amount of of native processing I'm going to need PCI slots (5-10) for external co-processing cards by various manufacturers. I'll be needing multiple Thunderbolt, Firewire, USB 2&3, and even SCSI (LTO archiving) and an obscene amount of RAM (64 Gig). The system needs to be a nexus for connectivity.
The system will be used for:
1) Audio (Both Music Recording and Film)
2) Video (Editing, FX and Colour correction for Indie Film and Video)
3) 3D (FX and 3D Printing)
4) Web and Print collateral
I'd be very happy with an updated design like the old IIx with it's multiple slots (NuBus in those days) An example from those day would be my ProTools system from '93
IIx with:
1 x ProTools I/O card
2 x ProTools Farm card (DSP)
3 x Sample Cell card (Sampler on a card)
1 x Video output card.
Quote:
Originally Posted by scottglasel
As a media professional I need expansion and connectivity. No matter the amount of of native processing I'm going to need PCI slots (5-10) for external co-processing cards by various manufacturers. I'll be needing multiple Thunderbolt, Firewire, USB 2&3, and even SCSI (LTO archiving) and an obscene amount of RAM (64 Gig). The system needs to be a nexus for connectivity.
Apple has never really offered that many PCI slots. You'd either end up with multiple machines, a different brand, or a different solution somewhere in there. I'm not questioning your requirements, just whether you'd be able to fulfill them that way with a Mac Pro.
The big fear I have in no PCI slots at all. The reality is you don't need a lot of slots to configure a machine for a specific usage. Apples problem is more of one of making those slots affordable.
Quote:
Originally Posted by hmm
Apple has never really offered that many PCI slots. You'd either end up with multiple machines, a different brand, or a different solution somewhere in there. I'm not questioning your requirements, just whether you'd be able to fulfill them that way with a Mac Pro.
Quote:
Originally Posted by wizard69
The big fear I have in no PCI slots at all. The reality is you don't need a lot of slots to configure a machine for a specific usage. Apples problem is more of one of making those slots affordable.
It varies, but yeah it's possible to populate all available slots if you need specific editing hardware. One of the things I've noticed in the past couple years is that the only stable PCI cards left on OSX as of Mountain Lion are basically those with either higher margins or supported as part of a hardware + software solution. Some of those developers have put out thunderbolt and usb3 versions. Black Magic comes to mind. Sonnet has a list of cards with thunderbolt compliant drivers. It is possible that if you need more connections than that will support or multiple high bandwidth ports, you will soon have to look at Windows.
Quote:
Originally Posted by wizard69
The big fear I have in no PCI slots at all.
The advantage to USB3 moving to 10gbps is that between having faster USB3 and a Thunderbolt breakout box, 95% of users will be fine.
I really do think that at least one Pro option will lack any slots at all.
Apple has been waiting years to do something like this, and now they have the chance. They will take it.