Do you think that will happen sooner rather than later or not for at least say five years?
The two things to look at is the next DRAM standard which in one variant the use of the highest spec RAM would require that the RAM be soldered onto the motherboard. A little farther off is Intel/Microns R&D efforts into 3D RAM.
I have no idea when either of these will become common place, but to deliver the highest performance sockets will have to go. Note that the Mini has never been Apples high performance machine so it could go for years using the standard approach. On the flips side Intels 3D memory would be a huge advantage in a Mini class machine as it is a denser and cooler memory.
Note that it very well may be possible for a company like Apple to offer one bank of high speed RAM, soldered on and another in conventional sockets running slower devices. This would preserve expandability at the expense of uniform performance.
As long there always is an entry level machine offered my Apple be it the Mini or otherwise or a higher performing Mini type machine, I'll be happy.
I suspect we will see rapid increases in capability of machines like the Mini. Haswell will be a big part ofthat along with other technologies. Fusion drive is just one example of ways that the performance of the Mini can be boosted at relatively low cost. Pack more RAM and GPU performance into the box each year and eventually the machine will loose its reputation as a lackluster device.
I suspect we will see rapid increases in capability of machines like the Mini. Haswell will be a big part ofthat along with other technologies. Fusion drive is just one example of ways that the performance of the Mini can be boosted at relatively low cost. Pack more RAM and GPU performance into the box each year and eventually the machine will loose its reputation as a lackluster device.
Right now the fusion drive adds too much to the cost of the machine. Oddly Apple has maintained fairly static pricing on ram and hard drive upgrades for quite some time. I do expect the mini to leverage a wider range of tasks simply because hardware has outpaced increases in software demands in a number of areas.
Right now the fusion drive adds too much to the cost of the machine. Oddly Apple has maintained fairly static pricing on ram and hard drive upgrades for quite some time. I do expect the mini to leverage a wider range of tasks simply because hardware has outpaced increases in software demands in a number of areas.
They should have added the Fusion drive to the base model and cut the cost to $200 on the quad-core model. Leave the plain solid state at $300.
They should have added the Fusion drive to the base model and cut the cost to $200 on the quad-core model. Leave the plain solid state at $300.
I could see that. Apple is very hungry for high margins, yet the configuration prices are just out of alignment with the mini line. They tack on a large percentage of the base cost.
They should have added the Fusion drive to the base model and cut the cost to $200 on the quad-core model. Leave the plain solid state at $300.
I could see that. Apple is very hungry for high margins, yet the configuration prices are just out of alignment with the mini line. They tack on a large percentage of the base cost.
Then Apple wonders why Mini sales are declining! The cost of the mobile processors in the Mini is significant thus I think they try to recover margins by making upgrades excessively expensive. The problem is it is easier for most people to buy base hardware and upgrade the DIY way. It is already possible to DIY a fusion drive solution so the appeal of the package deal isn't going to be all that great.
Fusion drive is a big marketing thrust and for many wont always e the best solution. I still think many users would be better off simply buying a larger SSD to install apps and the OS onto and put bulk storage on a magnetic disk. In other words Fusion Drive is good tech for a number of uses but maybe isn't he best solution for everybody. Especially with the cost of SSD capacity dropping quickly.
Then Apple wonders why Mini sales are declining! The cost of the mobile processors in the Mini is significant thus I think they try to recover margins by making upgrades excessively expensive. The problem is it is easier for most people to buy base hardware and upgrade the DIY way. It is already possible to DIY a fusion drive solution so the appeal of the package deal isn't going to be all that great.
Fusion drive is a big marketing thrust and for many wont always e the best solution. I still think many users would be better off simply buying a larger SSD to install apps and the OS onto and put bulk storage on a magnetic disk. In other words Fusion Drive is good tech for a number of uses but maybe isn't he best solution for everybody. Especially with the cost of SSD capacity dropping quickly.
They used a fairly expensive cpu in the mid range models. The cpu there costs more than the one used in the entry mac pro. Achieving equivalent performance in a desktop component is cheaper. You might lose hyperthreading, but while it can be useful in real world situations, its contribution to benchmarking tests is somewhat inflated. I think the mini quickly drifts out of focus when you configure it with a fusion drive and add on the total cost of peripheral devices. I tend to interpret things in their entirety, and some of these upgrades quickly alter the potential target market.
What I meant to say that DRAM in sockets will soon come to an end. At least in sockets as we know them today. The reason being that engineers can better control the electrical characteristics of the circuitry and thus communicate much faster. Plus as speeds increase wire distance becomes very significant.
So expect in the future to see high end machines that only support RAM soldered onto the motherboard.
Sorry for twisting up the logic in the previous post. Sometimes it is better to hit the sack then to try to do anything constructive online.
Distance certainly does matter...putting RAM closer to the CPU, for instance helps as does a substantial on chip cache...but I wonder about soldering RAM in high end machines. I certainly cold foresee a new physical interface though.
If a high end machine has a RAM module go belly up, it simply would not be an acceptable to say "get a new motherboard". RAM does go bad for any number of reasons. Having spent time watching customers buying various products at an Apple Store, I can understand Apple's movement to soldered RAM on some of the lower end consumer products. Truth be told, those customers are certainly the turn it on, use it type. They probably don't know what OS is installed on it or how much memory it was that the sales rep recommended they get for their intended use. There's nothing wrong with that, but the concept makes me nervous when applied to higher end solutions. What's next? A Mac Pro in a sealed box that can't be opened? Yikes!
They used a fairly expensive cpu in the mid range models.
Exactly! Thus they try to make up margins by making expensive upgrades almost mandatory. The Mini has always been stripped down with respect to RAM and hard disk space. Many upgrade the DIY route but I suspect more than enough buy Apples expensive solutions to keep margins healthy.
The cpu there costs more than the one used in the entry mac pro. Achieving equivalent performance in a desktop component is cheaper.
Or they could go AMD. Yeah I know CPU performance from AMD isn't anywhere nears as good as Intels processors but we aren't getting good CPU performance in the Mini as it is.
I don't knock Apple for making a low power platforms just get frustrated with the extra cost and the limited performance relative to a machine with a nice desktop processor. A lot of this is marketing mumble jumble from Intel too as there isn't a huge difference in processor design anymore when it comes to the laptop / desktop divide.
You might lose hyperthreading, but while it can be useful in real world situations, its contribution to benchmarking tests is somewhat inflated. I think the mini quickly drifts out of focus when you configure it with a fusion drive and add on the total cost of peripheral devices. I tend to interpret things in their entirety, and some of these upgrades quickly alter the potential target market.
I really have trouble grasping what Apples target market is now for the $799 model. The entry level machine is just that but the next step up is a bit of a different ball game. On the face of it the $799 Mini just seems to be awfully expensive for what you are getting. As it is we sit hear talking about the Mini and frankly I'm thinking I might upgrade my iPad before even considering a Mini. It is funny but my interest in keeping up with the desktop/laptop technology is really plunging, no it is more about buying for a task at hand and that is it. IPad is significantly more useful to me right now.
However one of the reasons I prefer to see an XMac with a couple of slots is that it allows users to adapt the machine to their specific needs. Thus is FireWire is really needed they can just plug in a card for that. I'm still not a fan of dongles and would much prefer in the box slots. Even here though I'm flexible, put four or six TB ports in an XMac and I might change my mind. The big problem though is that some hardware will never be adopted for TB.
This is fairly consistent with what I'm reading but there is one big qualifier here, Intel apparently has three different GPU implementations planned one with significant performance. We really don't know what type of GPU performance those lower power units will have.
For a midrange machine I don't see the need for a discrete GPU going away anytime soon either. On the Mini Haswell might do the trick if Apple would actually implement at least one machine with a performance chip.
The other thing here is TB and integration with that port. My understanding is that the GPU has to be on the motherboard unless Apple implements some sort of extended PCI Express port. Because of this and the desire to control cost I beleive Apple would have little choice but to glue the GPU to the motherboard.
That sounds good on the surface but Linux has significant issues with respect to the desktop environment. ADOBE would basically have to have their own distro. Even then GNOME, KDE and a bunch of other desktop environments leave a lot to be desired. Even Linus has been vocal about this lately.
As a side note one of the reasons I purchased a Mac Book Pro in 2008 was to have a far more stable desktop environment. For the mot part this has worked out really wel for me. Add in Apples integration with iPhone or iOS and you have a very pleasing platform. I wouldn't go back to using Linux as a primary desktop machine anytime soon but it is fantastic for servers and things that don't run on the Mac at all (LinuxCNC).
My biggest fear or concern is the freedom Mac OS provides me. I can get by with a locked down platform on my iOS devices but not on my Mac. The ability to dip into the supplied utilities, Python and other tools is invaluable to me. On the other hand an ARM based AIR, with twelve threads of performance (cores if your will) burning maybe 17 watts of power total, is very appealing to me. On top of that an AIR selling for much less due to not having to pay Intels margins is even nicer.
The reason I still prefer discreet graphics processors in a slot is it gives you the option to change it out in a year or two if/when something strikingly better comes along. That said, I would not grumble too loudly if Apple were to solder a good GPU on the board, one that includes more VRAM than is currently their practice. The units are available, indeed, they are commonplace in the PC community.
I don't think Adobe would want to, or has the capability of creating their own distro, but, yes, there would have to be a reference standard distro of Linux which Adobe supports and works with the developers to keep things moving along smoothly. This could actually be a good thing for Linux as there are simply too many distros which are splitting the development efforts. The result is that Linux, as a whole, remains a marginalized desktop OS. Canonical have made slow progress, but have not, as yet, fulfilled their stated objective of bringing the OS X experience to Linux. Adobe would become a king maker when it chose the distro to support, assuming they made a sound choice.
About slots, yes, slots are freedom. They enable the user to truly customize the machine to their needs/desires. Slots are good! :-)
Cheers
P.S. Even with all this, I am taking a hard look at the newly released Mini. It's a shame that Apple doesn't offer a BTO model with the GPU out of the Retina MBP. That would do quite nicely for the time being. The question is just how long Apple might take to release a Haswell Mini next year and whether they would offer such a GPU as a BTO then.
GPU's in expansion slots are a good thin but I'm not one to upgrade a GPU after buying a machine. So for my needs a soldered in GPU is not a big deal. However I see Apples problem in how they would support the TB connector. Right now TB would require an "on the motherboard GPU" to support signal routing. The other thing an on the motherboard GPU does is that it locks in power dissipation which allows tighter control of the power supply size.
The really big problem with Linux is the GUI which is split into so many competing factions that no real progress has been made. Further not a one of the GUIs has the elegance of either MS or Apple solution. Ideally what somebody like Adobe would do is to is to start over completely with a entirely new approach and shepherd it to become the Linux standard.
As to slots they certainly are good as long as you don't have to pay Mac Pro prices for them. Frankly technology has progressed to the point where it might be time to debut a new card standard, or simply adopt the half height card format with a new bracket.
As to a new computer, I'm sitting here in front of my 2008 MBP becoming increasingly frustrated with the keyboard. At this point I simply don't know which way I would go if a replacement is needed. Haswell would certainly make for a much better Mini.
Quote:
Originally Posted by RBR
The reason I still prefer discreet graphics processors in a slot is it gives you the option to change it out in a year or two if/when something strikingly better comes along. That said, I would not grumble too loudly if Apple were to solder a good GPU on the board, one that includes more VRAM than is currently their practice. The units are available, indeed, they are commonplace in the PC community.
I don't think Adobe would want to, or has the capability of creating their own distro, but, yes, there would have to be a reference standard distro of Linux which Adobe supports and works with the developers to keep things moving along smoothly. This could actually be a good thing for Linux as there are simply too many distros which are splitting the development efforts. The result is that Linux, as a whole, remains a marginalized desktop OS. Canonical have made slow progress, but have not, as yet, fulfilled their stated objective of bringing the OS X experience to Linux. Adobe would become a king maker when it chose the distro to support, assuming they made a sound choice.
About slots, yes, slots are freedom. They enable the user to truly customize the machine to their needs/desires. Slots are good! :-)
Cheers
P.S. Even with all this, I am taking a hard look at the newly released Mini. It's a shame that Apple doesn't offer a BTO model with the GPU out of the Retina MBP. That would do quite nicely for the time being. The question is just how long Apple might take to release a Haswell Mini next year and whether they would offer such a GPU as a BTO then.
GPU's in expansion slots are a good thin but I'm not one to upgrade a GPU after buying a machine. So for my needs a soldered in GPU is not a big deal. However I see Apples problem in how they would support the TB connector. Right now TB would require an "on the motherboard GPU" to support signal routing. The other thing an on the motherboard GPU does is that it locks in power dissipation which allows tighter control of the power supply size.
The really big problem with Linux is the GUI which is split into so many competing factions that no real progress has been made. Further not a one of the GUIs has the elegance of either MS or Apple solution. Ideally what somebody like Adobe would do is to is to start over completely with a entirely new approach and shepherd it to become the Linux standard.
As to slots they certainly are good as long as you don't have to pay Mac Pro prices for them. Frankly technology has progressed to the point where it might be time to debut a new card standard, or simply adopt the half height card format with a new bracket.
As to a new computer, I'm sitting here in front of my 2008 MBP becoming increasingly frustrated with the keyboard. At this point I simply don't know which way I would go if a replacement is needed. Haswell would certainly make for a much better Mini.
My understanding is that Thunderbolt must have "direct access" to the graphics processor rather than it needing to be on the motherboard. For example, a PCIe expansion card can not add TB to a system that does not have it. The TB chip needs to be on the MB and configured to directly address the graphics processor whether it is on the MB or in a PCIe slot.There are examples of PC MBs on the market that have TB and a PCIe slot for the graphics card.
Still, I will take an on the MB graphics processor if that is the way Apple chooses to go with the next Mini. I am going to "pull the trigger" on a current Mini in the next few days as I just can't wait. I do hope Apple gets the message though.
Cheers
[Edit] P.S. Have you tried looking for a replacement keyboard for your 2008 MBP? I just picked up Late 2008 (Unibody) MBP and it seems to be a decent enough machine to suffice for a while longer. I am going to put a SSD in it to help it out and it is supposed to run Mountain Lion and all the current software.
My understanding is that Thunderbolt must have "direct access" to the graphics processor rather than it needing to be on the motherboard.
The iMac GPU is in an MXM slot so that's an option e.g putting a GPU where the second hard drive would be but the more processing going on in there, the more heat there is. Dedicated GPUs apparently copy the framebuffer into the integrated graphics and pass it out through Thunderbolt, likely via the drivers.
Apple can gain a bit of performance by putting another dedicated GPU in but will the cost, complexity and reliability be worth doing from Apple's point of view? Most customers just settle for an iMac and Apple makes more profit from them. I'd rather be able to take full advantage of the Thunderbolt port than have a dedicated GPU inside. I like the size of this Thunderbolt PCI box:
They have a bigger full-length one for the Red Rocket ($399 and $699 respectively). I'd like to be able to get a box and 100W GPU for under $399. While it doesn't sound like a lot of power vs high-end desktop GPUs that can go to over double that, the Mini GPU allocation is something like 10W so 100W offers a decent power boost. Being able to have that outside, plug-and-play and upgradeable would be very useful. It doesn't work properly yet though and that solution might never come about.
It wouldn't compromise the design of the Mini for people who don't need it and it would help 3rd parties like NVidia/AMD as it's not just for Mac users. It's for the 50% of all people who get suckered with Intel's poor GPUs. That's 200 million people a year or more. Only a fraction will have Thunderbolt and a fraction will want the upgrade at that price but if 5 million (there are tens of millions of gamers) go for it and the margins are 25% on a $400 product, there's half a billion dollars profit a year sitting waiting for somebody. NVidia only makes $100m profit a quarter.
The iMac GPU is in an MXM slot so that's an option e.g putting a GPU where the second hard drive would be but the more processing going on in there, the more heat there is. Dedicated GPUs apparently copy the framebuffer into the integrated graphics and pass it out through Thunderbolt, likely via the drivers.
Apple can gain a bit of performance by putting another dedicated GPU in but will the cost, complexity and reliability be worth doing from Apple's point of view? Most customers just settle for an iMac and Apple makes more profit from them. I'd rather be able to take full advantage of the Thunderbolt port than have a dedicated GPU inside. I like the size of this Thunderbolt PCI box:
They have a bigger full-length one for the Red Rocket ($399 and $699 respectively). I'd like to be able to get a box and 100W GPU for under $399. While it doesn't sound like a lot of power vs high-end desktop GPUs that can go to over double that, the Mini GPU allocation is something like 10W so 100W offers a decent power boost. Being able to have that outside, plug-and-play and upgradeable would be very useful. It doesn't work properly yet though and that solution might never come about.
It wouldn't compromise the design of the Mini for people who don't need it and it would help 3rd parties like NVidia/AMD as it's not just for Mac users. It's for the 50% of all people who get suckered with Intel's poor GPUs. That's 200 million people a year or more. Only a fraction will have Thunderbolt and a fraction will want the upgrade at that price but if 5 million (there are tens of millions of gamers) go for it and the margins are 25% on a $400 product, there's half a billion dollars profit a year sitting waiting for somebody. NVidia only makes $100m profit a quarter.
The expansion chassis are appealing in several regards. They allow changing PCIe cards in closed platforms such as the iMac or Mini, but...the problem is that Thunderbolt is not fast enough to use a moderately fast graphics card. The bus would be rapidly saturated as it is essentially a 1X PCIe lane whereas most graphics cards on motherboards are on 8X if not 16X PCIe slots (ignoring, for the moment, the Crossfire and other multiple card setups). OWC, among others, offer expansion chassis, but there do not appear to be any compatible graphics cards at the moment though the larger Red Rocket one you mention might offer greater choice as it is a full length card compatible chassis. There is one interesting possibility that is available NOW though. You can install a bootable PCIe SSD (not a SATA drive) in the Thunderbolt expansion chassis. Lloyd Chambers & OWC have tested this configuration and it is appreciably faster than a SATA 3 SSD installed on the Mini's internal SATA port. The fly in the ointment though is that this apprach is hundreds of dollars more expensive to implement than an actual PCIe slot and is bandwidth limited.
Unfortunately, Thunderbolt has not received much love either in the PC community or the developer community at this point. The problem is cost. It simply costs a lot more than USB 3 and, so far, most have opted to use USB 3 as it is "good enough" for much external storage or else go the NAS storage route. To to it off, Thunderbolt II (or whatever Intel may call it) is just around the corner. It is supposed to be announce late this year or very early next year and is supposed to be much faster than Thunderbolt (I). A number of vendors are only now about to bring products to market for version I. I am concerned that Thunderbolt is going to repeat the cycle of Firewire, only in a shorter time span, and become a dead end, expensive, marginalized product. If Thunderbolt is to succeed, they simply must get the price of it down to a level where it will be the connection of choice.
The simple truth is that Thunderbolt as we know it is not fast enough for most graphics processors.
I wouldn't go back to using Linux as a primary desktop machine anytime soon but it is fantastic for servers and things that don't run on the Mac at all (LinuxCNC).
I am currently constructing a CNC box with a breakout board, a cluster of Gecko drives, power supplies and an embedded computer.
The last of these is an Intel Atom D525MW board with an attached SSD with Ubuntu 10.04/LinuxCNC 2.5 on it. It never occurred to me to use a Mac Mini (even though it can run Linux) on account of the decade-old decision by Apple to drop the parallel port (other PC board makers seem to more and more be dropping it now, too).
I also noted yor comment about the Mini being used as part of embedded machine control, however, and that got me thinking a bit - presumably that's not for stuff that uses the old parallel port (which LinuuxCNC supports).
A quick check of the Mini's specs didn't enlighten me wrt the presence or otherwise of a PCI slot into which one might be able to plug a parallel port card and thus be able to use the Mini for LinuxCNC control.
Does the Mini in fact have such a slot buried in its not-very-deep depths? Or is there another way around this that I don't know about (e.g., via TB)?
the problem is that Thunderbolt is not fast enough to use a moderately fast graphics card.
The GPU just operates a bit slower than its peak but there is less point going for a high-end card for gaming performance - for compute, you copy the data over anyway so the connection won't cause an issue. Here is a TB case running a Radeon 6770 GPU:
[VIDEO]
That was 720p, high quality. Average FPS was 30. Here is the same card running natively:
[VIDEO]
The average comes out at 35FPS and that had all settings the same except anisotropic filtering disabled instead of 4x over TB. Worst case it's 85%. This has been tested on the highest end cards you can buy by blocking lanes and the same thing was found. There is a performance hit but the card is perfectly usable. A higher-end card was tested here:
They said the box would be $199 and could have an option with a Radeon 7750 for $199 + card retail cost minus $5-10 so possibly $299. There's a whole forum section here about this sort of thing:
I am concerned that Thunderbolt is going to repeat the cycle of Firewire, only in a shorter time span, and become a dead end, expensive, marginalized product.
Thunderbolt has the advantage in that it is a form of PCI, which means it can do very advanced things that USB 3 can never do. Firewire was similar in that it became the industry standard for video/audio capture, especially real-time due to the low latency.
Thunderbolt will be held back a bit due to it being an Intel technology and AMD will push USB 3 as an alternative. USB 3 isn't a competitor to TB really in much the same way USB isn't a competitor to ExpressCard. They have different uses. It certainly won't be a failed standard simply because it has a smaller market appeal.
I will have to follow the MSI link, but, from the reference to using a 150 watt card, MSI must have provided a power supply for their expansion chassis. All the rest of the TB expansion chassis I have seen were restricted to the power supplied by the TB port which meant that there really were not any graphics cards that were useful at that low a power level and then there is the matter of most of the rest being half length chassis which presents problems of its own.
Thank you for digging this up.
This is the first TB expansion chassis I have seen that appears to provide a reasonably practical solution to the new Mini's limited graphics capability. Though not "cheap", the price looks to be much more reasonable than the alternative chassis which are of a lesser capability in any event.
I can see adding this to the Mini I am in the process of getting. :yahoo
Lloyd Chambers' recent tests of the new Mini, in conjunction with OWC, who supplied the RAM and SSD used in the tests, showed the capability of the new Mini, but also its Achilles' heel, graphics. In any of the tests which utilized the graphics processor the Mini came out a poor second to the Retina MBP and, of course, the Mac Pro.It surely would be interesting to see those same tests performed with the MSI chassis and graphics card. Chambers now thinks the next Mac Pro release will be late 2013, which makes going ahead with the Mini in the mean time seem even more realistic. You'll recall that Tim Cook only said there would be a Mac Pro "next year". http://macperformanceguide.com/
They said the box would be $199 and could have an option with a Radeon 7750 for $199 + card retail cost minus $5-10 so possibly $299. There's a whole forum section here about this sort of thing:
Here is a video of someone who has hooked up a Quadro 4000 to a Mini Server:
I watched the video. I'm amazed that worked. If I specifically wanted CUDA processing, I'd try for something Tesla based at this point. At comparable price points, you get more ram with the teslas. This would be necessary if you wanted to deal with anything involving significant displacement and seamless textures. That's obviously not necessary for everyone. It's interesting that this is the first seemingly credible solution that I've seen, as the theoretical cost isn't completely out of focus with the cost of the computer hardware.
Quote:
Originally Posted by RBR
This is the first TB expansion chassis I have seen that appears to provide a reasonably practical solution to the new Mini's limited graphics capability. Though not "cheap", the price looks to be much more reasonable than the alternative chassis which are of a lesser capability in any event.
I just see the mini as being in poor alignment with such requirements, but this is the cheapest I've seen for such a solution. It's still interesting either way. I didn't expect anything remotely cost effective given that we're still firmly in the realm of niche products.
Comments
The two things to look at is the next DRAM standard which in one variant the use of the highest spec RAM would require that the RAM be soldered onto the motherboard. A little farther off is Intel/Microns R&D efforts into 3D RAM.
I have no idea when either of these will become common place, but to deliver the highest performance sockets will have to go. Note that the Mini has never been Apples high performance machine so it could go for years using the standard approach. On the flips side Intels 3D memory would be a huge advantage in a Mini class machine as it is a denser and cooler memory.
Note that it very well may be possible for a company like Apple to offer one bank of high speed RAM, soldered on and another in conventional sockets running slower devices. This would preserve expandability at the expense of uniform performance.
I suspect we will see rapid increases in capability of machines like the Mini. Haswell will be a big part ofthat along with other technologies. Fusion drive is just one example of ways that the performance of the Mini can be boosted at relatively low cost. Pack more RAM and GPU performance into the box each year and eventually the machine will loose its reputation as a lackluster device.
Quote:
Originally Posted by Winter
As long there always is an entry level machine offered my Apple be it the Mini or otherwise or a higher performing Mini type machine, I'll be happy.
Which begs the question of why Apple is rebooting the Mac Pro next year.
It seems easy to build a larger (taller) Mini with access to RAM and Hard Drive, and a discrete graphics slot.
With Thunderbolt expansion poised for takeoff, a machine like that would satisfy most Pro needs.
I know that's probably too close to the mythical midrange Mac, but it's not midrange if there's no Pro, right?
Quote:
Originally Posted by wizard69
I suspect we will see rapid increases in capability of machines like the Mini. Haswell will be a big part ofthat along with other technologies. Fusion drive is just one example of ways that the performance of the Mini can be boosted at relatively low cost. Pack more RAM and GPU performance into the box each year and eventually the machine will loose its reputation as a lackluster device.
Right now the fusion drive adds too much to the cost of the machine. Oddly Apple has maintained fairly static pricing on ram and hard drive upgrades for quite some time. I do expect the mini to leverage a wider range of tasks simply because hardware has outpaced increases in software demands in a number of areas.
They should have added the Fusion drive to the base model and cut the cost to $200 on the quad-core model. Leave the plain solid state at $300.
Quote:
Originally Posted by Winter
They should have added the Fusion drive to the base model and cut the cost to $200 on the quad-core model. Leave the plain solid state at $300.
I could see that. Apple is very hungry for high margins, yet the configuration prices are just out of alignment with the mini line. They tack on a large percentage of the base cost.
Then Apple wonders why Mini sales are declining! The cost of the mobile processors in the Mini is significant thus I think they try to recover margins by making upgrades excessively expensive. The problem is it is easier for most people to buy base hardware and upgrade the DIY way. It is already possible to DIY a fusion drive solution so the appeal of the package deal isn't going to be all that great.
Fusion drive is a big marketing thrust and for many wont always e the best solution. I still think many users would be better off simply buying a larger SSD to install apps and the OS onto and put bulk storage on a magnetic disk. In other words Fusion Drive is good tech for a number of uses but maybe isn't he best solution for everybody. Especially with the cost of SSD capacity dropping quickly.
Quote:
Originally Posted by wizard69
Then Apple wonders why Mini sales are declining! The cost of the mobile processors in the Mini is significant thus I think they try to recover margins by making upgrades excessively expensive. The problem is it is easier for most people to buy base hardware and upgrade the DIY way. It is already possible to DIY a fusion drive solution so the appeal of the package deal isn't going to be all that great.
Fusion drive is a big marketing thrust and for many wont always e the best solution. I still think many users would be better off simply buying a larger SSD to install apps and the OS onto and put bulk storage on a magnetic disk. In other words Fusion Drive is good tech for a number of uses but maybe isn't he best solution for everybody. Especially with the cost of SSD capacity dropping quickly.
They used a fairly expensive cpu in the mid range models. The cpu there costs more than the one used in the entry mac pro. Achieving equivalent performance in a desktop component is cheaper. You might lose hyperthreading, but while it can be useful in real world situations, its contribution to benchmarking tests is somewhat inflated. I think the mini quickly drifts out of focus when you configure it with a fusion drive and add on the total cost of peripheral devices. I tend to interpret things in their entirety, and some of these upgrades quickly alter the potential target market.
Quote:
Originally Posted by wizard69
YIKES!
Pardon my conflagration above.
What I meant to say that DRAM in sockets will soon come to an end. At least in sockets as we know them today. The reason being that engineers can better control the electrical characteristics of the circuitry and thus communicate much faster. Plus as speeds increase wire distance becomes very significant.
So expect in the future to see high end machines that only support RAM soldered onto the motherboard.
Sorry for twisting up the logic in the previous post. Sometimes it is better to hit the sack then to try to do anything constructive online.
Distance certainly does matter...putting RAM closer to the CPU, for instance helps as does a substantial on chip cache...but I wonder about soldering RAM in high end machines. I certainly cold foresee a new physical interface though.
If a high end machine has a RAM module go belly up, it simply would not be an acceptable to say "get a new motherboard". RAM does go bad for any number of reasons. Having spent time watching customers buying various products at an Apple Store, I can understand Apple's movement to soldered RAM on some of the lower end consumer products. Truth be told, those customers are certainly the turn it on, use it type. They probably don't know what OS is installed on it or how much memory it was that the sales rep recommended they get for their intended use. There's nothing wrong with that, but the concept makes me nervous when applied to higher end solutions. What's next? A Mac Pro in a sealed box that can't be opened? Yikes!
I don't knock Apple for making a low power platforms just get frustrated with the extra cost and the limited performance relative to a machine with a nice desktop processor. A lot of this is marketing mumble jumble from Intel too as there isn't a huge difference in processor design anymore when it comes to the laptop / desktop divide.
I really have trouble grasping what Apples target market is now for the $799 model. The entry level machine is just that but the next step up is a bit of a different ball game. On the face of it the $799 Mini just seems to be awfully expensive for what you are getting. As it is we sit hear talking about the Mini and frankly I'm thinking I might upgrade my iPad before even considering a Mini. It is funny but my interest in keeping up with the desktop/laptop technology is really plunging, no it is more about buying for a task at hand and that is it. IPad is significantly more useful to me right now.
Quote:
Originally Posted by wizard69
Yep, evolutionary dead end.
However one of the reasons I prefer to see an XMac with a couple of slots is that it allows users to adapt the machine to their specific needs. Thus is FireWire is really needed they can just plug in a card for that. I'm still not a fan of dongles and would much prefer in the box slots. Even here though I'm flexible, put four or six TB ports in an XMac and I might change my mind. The big problem though is that some hardware will never be adopted for TB.
This is fairly consistent with what I'm reading but there is one big qualifier here, Intel apparently has three different GPU implementations planned one with significant performance. We really don't know what type of GPU performance those lower power units will have.
For a midrange machine I don't see the need for a discrete GPU going away anytime soon either. On the Mini Haswell might do the trick if Apple would actually implement at least one machine with a performance chip.
The other thing here is TB and integration with that port. My understanding is that the GPU has to be on the motherboard unless Apple implements some sort of extended PCI Express port. Because of this and the desire to control cost I beleive Apple would have little choice but to glue the GPU to the motherboard.
That sounds good on the surface but Linux has significant issues with respect to the desktop environment. ADOBE would basically have to have their own distro. Even then GNOME, KDE and a bunch of other desktop environments leave a lot to be desired. Even Linus has been vocal about this lately.
As a side note one of the reasons I purchased a Mac Book Pro in 2008 was to have a far more stable desktop environment. For the mot part this has worked out really wel for me. Add in Apples integration with iPhone or iOS and you have a very pleasing platform. I wouldn't go back to using Linux as a primary desktop machine anytime soon but it is fantastic for servers and things that don't run on the Mac at all (LinuxCNC).
My biggest fear or concern is the freedom Mac OS provides me. I can get by with a locked down platform on my iOS devices but not on my Mac. The ability to dip into the supplied utilities, Python and other tools is invaluable to me. On the other hand an ARM based AIR, with twelve threads of performance (cores if your will) burning maybe 17 watts of power total, is very appealing to me. On top of that an AIR selling for much less due to not having to pay Intels margins is even nicer.
The reason I still prefer discreet graphics processors in a slot is it gives you the option to change it out in a year or two if/when something strikingly better comes along. That said, I would not grumble too loudly if Apple were to solder a good GPU on the board, one that includes more VRAM than is currently their practice. The units are available, indeed, they are commonplace in the PC community.
I don't think Adobe would want to, or has the capability of creating their own distro, but, yes, there would have to be a reference standard distro of Linux which Adobe supports and works with the developers to keep things moving along smoothly. This could actually be a good thing for Linux as there are simply too many distros which are splitting the development efforts. The result is that Linux, as a whole, remains a marginalized desktop OS. Canonical have made slow progress, but have not, as yet, fulfilled their stated objective of bringing the OS X experience to Linux. Adobe would become a king maker when it chose the distro to support, assuming they made a sound choice.
About slots, yes, slots are freedom. They enable the user to truly customize the machine to their needs/desires. Slots are good! :-)
Cheers
P.S. Even with all this, I am taking a hard look at the newly released Mini. It's a shame that Apple doesn't offer a BTO model with the GPU out of the Retina MBP. That would do quite nicely for the time being. The question is just how long Apple might take to release a Haswell Mini next year and whether they would offer such a GPU as a BTO then.
GPU's in expansion slots are a good thin but I'm not one to upgrade a GPU after buying a machine. So for my needs a soldered in GPU is not a big deal. However I see Apples problem in how they would support the TB connector. Right now TB would require an "on the motherboard GPU" to support signal routing. The other thing an on the motherboard GPU does is that it locks in power dissipation which allows tighter control of the power supply size.
The really big problem with Linux is the GUI which is split into so many competing factions that no real progress has been made. Further not a one of the GUIs has the elegance of either MS or Apple solution. Ideally what somebody like Adobe would do is to is to start over completely with a entirely new approach and shepherd it to become the Linux standard.
As to slots they certainly are good as long as you don't have to pay Mac Pro prices for them. Frankly technology has progressed to the point where it might be time to debut a new card standard, or simply adopt the half height card format with a new bracket.
As to a new computer, I'm sitting here in front of my 2008 MBP becoming increasingly frustrated with the keyboard. At this point I simply don't know which way I would go if a replacement is needed. Haswell would certainly make for a much better Mini.
Quote:
Originally Posted by RBR
The reason I still prefer discreet graphics processors in a slot is it gives you the option to change it out in a year or two if/when something strikingly better comes along. That said, I would not grumble too loudly if Apple were to solder a good GPU on the board, one that includes more VRAM than is currently their practice. The units are available, indeed, they are commonplace in the PC community.
I don't think Adobe would want to, or has the capability of creating their own distro, but, yes, there would have to be a reference standard distro of Linux which Adobe supports and works with the developers to keep things moving along smoothly. This could actually be a good thing for Linux as there are simply too many distros which are splitting the development efforts. The result is that Linux, as a whole, remains a marginalized desktop OS. Canonical have made slow progress, but have not, as yet, fulfilled their stated objective of bringing the OS X experience to Linux. Adobe would become a king maker when it chose the distro to support, assuming they made a sound choice.
About slots, yes, slots are freedom. They enable the user to truly customize the machine to their needs/desires. Slots are good! :-)
Cheers
P.S. Even with all this, I am taking a hard look at the newly released Mini. It's a shame that Apple doesn't offer a BTO model with the GPU out of the Retina MBP. That would do quite nicely for the time being. The question is just how long Apple might take to release a Haswell Mini next year and whether they would offer such a GPU as a BTO then.
Quote:
Originally Posted by wizard69
GPU's in expansion slots are a good thin but I'm not one to upgrade a GPU after buying a machine. So for my needs a soldered in GPU is not a big deal. However I see Apples problem in how they would support the TB connector. Right now TB would require an "on the motherboard GPU" to support signal routing. The other thing an on the motherboard GPU does is that it locks in power dissipation which allows tighter control of the power supply size.
The really big problem with Linux is the GUI which is split into so many competing factions that no real progress has been made. Further not a one of the GUIs has the elegance of either MS or Apple solution. Ideally what somebody like Adobe would do is to is to start over completely with a entirely new approach and shepherd it to become the Linux standard.
As to slots they certainly are good as long as you don't have to pay Mac Pro prices for them. Frankly technology has progressed to the point where it might be time to debut a new card standard, or simply adopt the half height card format with a new bracket.
As to a new computer, I'm sitting here in front of my 2008 MBP becoming increasingly frustrated with the keyboard. At this point I simply don't know which way I would go if a replacement is needed. Haswell would certainly make for a much better Mini.
My understanding is that Thunderbolt must have "direct access" to the graphics processor rather than it needing to be on the motherboard. For example, a PCIe expansion card can not add TB to a system that does not have it. The TB chip needs to be on the MB and configured to directly address the graphics processor whether it is on the MB or in a PCIe slot.There are examples of PC MBs on the market that have TB and a PCIe slot for the graphics card.
Still, I will take an on the MB graphics processor if that is the way Apple chooses to go with the next Mini. I am going to "pull the trigger" on a current Mini in the next few days as I just can't wait. I do hope Apple gets the message though.
Cheers
[Edit] P.S. Have you tried looking for a replacement keyboard for your 2008 MBP? I just picked up Late 2008 (Unibody) MBP and it seems to be a decent enough machine to suffice for a while longer. I am going to put a SSD in it to help it out and it is supposed to run Mountain Lion and all the current software.
The iMac GPU is in an MXM slot so that's an option e.g putting a GPU where the second hard drive would be but the more processing going on in there, the more heat there is. Dedicated GPUs apparently copy the framebuffer into the integrated graphics and pass it out through Thunderbolt, likely via the drivers.
Apple can gain a bit of performance by putting another dedicated GPU in but will the cost, complexity and reliability be worth doing from Apple's point of view? Most customers just settle for an iMac and Apple makes more profit from them. I'd rather be able to take full advantage of the Thunderbolt port than have a dedicated GPU inside. I like the size of this Thunderbolt PCI box:
http://www.engadget.com/2012/11/13/mlogic-mlink-thunderbolt-chassis/
They have a bigger full-length one for the Red Rocket ($399 and $699 respectively). I'd like to be able to get a box and 100W GPU for under $399. While it doesn't sound like a lot of power vs high-end desktop GPUs that can go to over double that, the Mini GPU allocation is something like 10W so 100W offers a decent power boost. Being able to have that outside, plug-and-play and upgradeable would be very useful. It doesn't work properly yet though and that solution might never come about.
It wouldn't compromise the design of the Mini for people who don't need it and it would help 3rd parties like NVidia/AMD as it's not just for Mac users. It's for the 50% of all people who get suckered with Intel's poor GPUs. That's 200 million people a year or more. Only a fraction will have Thunderbolt and a fraction will want the upgrade at that price but if 5 million (there are tens of millions of gamers) go for it and the margins are 25% on a $400 product, there's half a billion dollars profit a year sitting waiting for somebody. NVidia only makes $100m profit a quarter.
Quote:
Originally Posted by Marvin
The iMac GPU is in an MXM slot so that's an option e.g putting a GPU where the second hard drive would be but the more processing going on in there, the more heat there is. Dedicated GPUs apparently copy the framebuffer into the integrated graphics and pass it out through Thunderbolt, likely via the drivers.
Apple can gain a bit of performance by putting another dedicated GPU in but will the cost, complexity and reliability be worth doing from Apple's point of view? Most customers just settle for an iMac and Apple makes more profit from them. I'd rather be able to take full advantage of the Thunderbolt port than have a dedicated GPU inside. I like the size of this Thunderbolt PCI box:
http://www.engadget.com/2012/11/13/mlogic-mlink-thunderbolt-chassis/
They have a bigger full-length one for the Red Rocket ($399 and $699 respectively). I'd like to be able to get a box and 100W GPU for under $399. While it doesn't sound like a lot of power vs high-end desktop GPUs that can go to over double that, the Mini GPU allocation is something like 10W so 100W offers a decent power boost. Being able to have that outside, plug-and-play and upgradeable would be very useful. It doesn't work properly yet though and that solution might never come about.
It wouldn't compromise the design of the Mini for people who don't need it and it would help 3rd parties like NVidia/AMD as it's not just for Mac users. It's for the 50% of all people who get suckered with Intel's poor GPUs. That's 200 million people a year or more. Only a fraction will have Thunderbolt and a fraction will want the upgrade at that price but if 5 million (there are tens of millions of gamers) go for it and the margins are 25% on a $400 product, there's half a billion dollars profit a year sitting waiting for somebody. NVidia only makes $100m profit a quarter.
The expansion chassis are appealing in several regards. They allow changing PCIe cards in closed platforms such as the iMac or Mini, but...the problem is that Thunderbolt is not fast enough to use a moderately fast graphics card. The bus would be rapidly saturated as it is essentially a 1X PCIe lane whereas most graphics cards on motherboards are on 8X if not 16X PCIe slots (ignoring, for the moment, the Crossfire and other multiple card setups). OWC, among others, offer expansion chassis, but there do not appear to be any compatible graphics cards at the moment though the larger Red Rocket one you mention might offer greater choice as it is a full length card compatible chassis. There is one interesting possibility that is available NOW though. You can install a bootable PCIe SSD (not a SATA drive) in the Thunderbolt expansion chassis. Lloyd Chambers & OWC have tested this configuration and it is appreciably faster than a SATA 3 SSD installed on the Mini's internal SATA port. The fly in the ointment though is that this apprach is hundreds of dollars more expensive to implement than an actual PCIe slot and is bandwidth limited.
Unfortunately, Thunderbolt has not received much love either in the PC community or the developer community at this point. The problem is cost. It simply costs a lot more than USB 3 and, so far, most have opted to use USB 3 as it is "good enough" for much external storage or else go the NAS storage route. To to it off, Thunderbolt II (or whatever Intel may call it) is just around the corner. It is supposed to be announce late this year or very early next year and is supposed to be much faster than Thunderbolt (I). A number of vendors are only now about to bring products to market for version I. I am concerned that Thunderbolt is going to repeat the cycle of Firewire, only in a shorter time span, and become a dead end, expensive, marginalized product. If Thunderbolt is to succeed, they simply must get the price of it down to a level where it will be the connection of choice.
The simple truth is that Thunderbolt as we know it is not fast enough for most graphics processors.
I am currently constructing a CNC box with a breakout board, a cluster of Gecko drives, power supplies and an embedded computer.
The last of these is an Intel Atom D525MW board with an attached SSD with Ubuntu 10.04/LinuxCNC 2.5 on it. It never occurred to me to use a Mac Mini (even though it can run Linux) on account of the decade-old decision by Apple to drop the parallel port (other PC board makers seem to more and more be dropping it now, too).
I also noted yor comment about the Mini being used as part of embedded machine control, however, and that got me thinking a bit - presumably that's not for stuff that uses the old parallel port (which LinuuxCNC supports).
A quick check of the Mini's specs didn't enlighten me wrt the presence or otherwise of a PCI slot into which one might be able to plug a parallel port card and thus be able to use the Mini for LinuxCNC control.
Does the Mini in fact have such a slot buried in its not-very-deep depths? Or is there another way around this that I don't know about (e.g., via TB)?
Cheers,
Alex.
The GPU just operates a bit slower than its peak but there is less point going for a high-end card for gaming performance - for compute, you copy the data over anyway so the connection won't cause an issue. Here is a TB case running a Radeon 6770 GPU:
[VIDEO]
That was 720p, high quality. Average FPS was 30. Here is the same card running natively:
[VIDEO]
The average comes out at 35FPS and that had all settings the same except anisotropic filtering disabled instead of 4x over TB. Worst case it's 85%. This has been tested on the highest end cards you can buy by blocking lanes and the same thing was found. There is a performance hit but the card is perfectly usable. A higher-end card was tested here:
http://www.tomshardware.com/reviews/pci-express-graphics-thunderbolt,3263-6.html
http://www.tomshardware.com/reviews/pci-express-graphics-thunderbolt,3263-7.html
Someone seems to have contacted an MSI rep about the GUS box:
http://forum-en.msi.com/index.php?topic=157764.0
They said the box would be $199 and could have an option with a Radeon 7750 for $199 + card retail cost minus $5-10 so possibly $299. There's a whole forum section here about this sort of thing:
http://forum.notebookreview.com/e-gpu-external-graphics-discussion/
Here is a video of someone who has hooked up a Quadro 4000 to a Mini Server:
[VIDEO]
Not many details but they say it's a solution coming soon. It'll take a bit of time and a few revisions to get something that works well.
Thunderbolt has the advantage in that it is a form of PCI, which means it can do very advanced things that USB 3 can never do. Firewire was similar in that it became the industry standard for video/audio capture, especially real-time due to the low latency.
Thunderbolt will be held back a bit due to it being an Intel technology and AMD will push USB 3 as an alternative. USB 3 isn't a competitor to TB really in much the same way USB isn't a competitor to ExpressCard. They have different uses. It certainly won't be a failed standard simply because it has a smaller market appeal.
Marvin,
A couple of comments:
!!!!!!!!!!!!!!!!!!!!
You silver tongued devil!
I will have to follow the MSI link, but, from the reference to using a 150 watt card, MSI must have provided a power supply for their expansion chassis. All the rest of the TB expansion chassis I have seen were restricted to the power supplied by the TB port which meant that there really were not any graphics cards that were useful at that low a power level and then there is the matter of most of the rest being half length chassis which presents problems of its own.
Thank you for digging this up.
This is the first TB expansion chassis I have seen that appears to provide a reasonably practical solution to the new Mini's limited graphics capability. Though not "cheap", the price looks to be much more reasonable than the alternative chassis which are of a lesser capability in any event.
I can see adding this to the Mini I am in the process of getting. :yahoo
Lloyd Chambers' recent tests of the new Mini, in conjunction with OWC, who supplied the RAM and SSD used in the tests, showed the capability of the new Mini, but also its Achilles' heel, graphics. In any of the tests which utilized the graphics processor the Mini came out a poor second to the Retina MBP and, of course, the Mac Pro.It surely would be interesting to see those same tests performed with the MSI chassis and graphics card. Chambers now thinks the next Mac Pro release will be late 2013, which makes going ahead with the Mini in the mean time seem even more realistic. You'll recall that Tim Cook only said there would be a Mac Pro "next year". http://macperformanceguide.com/
Cheers
Quote:
Originally Posted by Marvin
They said the box would be $199 and could have an option with a Radeon 7750 for $199 + card retail cost minus $5-10 so possibly $299. There's a whole forum section here about this sort of thing:
http://forum.notebookreview.com/e-gpu-external-graphics-discussion/
Here is a video of someone who has hooked up a Quadro 4000 to a Mini Server:
I watched the video. I'm amazed that worked. If I specifically wanted CUDA processing, I'd try for something Tesla based at this point. At comparable price points, you get more ram with the teslas. This would be necessary if you wanted to deal with anything involving significant displacement and seamless textures. That's obviously not necessary for everyone. It's interesting that this is the first seemingly credible solution that I've seen, as the theoretical cost isn't completely out of focus with the cost of the computer hardware.
Quote:
Originally Posted by RBR
This is the first TB expansion chassis I have seen that appears to provide a reasonably practical solution to the new Mini's limited graphics capability. Though not "cheap", the price looks to be much more reasonable than the alternative chassis which are of a lesser capability in any event.
I just see the mini as being in poor alignment with such requirements, but this is the cheapest I've seen for such a solution. It's still interesting either way. I didn't expect anything remotely cost effective given that we're still firmly in the realm of niche products.