GPU Powered Macs and iPhones

[MacRumors]

http://www.macrumors.com/images/macrumorsthreadlogo.gif (http://www.macrumors.com)

Architosh (http://www.architosh.com/news/2008-07/0717_gpu_piranahs_opencl.html) points us to a Guardian.co.uk article (http://www.guardian.co.uk/technology/2008/jul/17/news.computing) from last week which details the upcoming trend of using GPUs (graphics processors) for day to day computing. As they point out, if you have a computer with either an ATI or nVidia graphics card, chances are you have more than 100 microprocessor cores waiting for use. While these cores have been optimized to deliver high performance graphics for games and video, there's an effort to harness these processors for general use. Those GPU cores are the piranhas of processing. Because there are so many of them, they can chomp through tens of gigabytes of data in a second. But it has to be the right kind of data - something that can be parcelled up and delivered in bite-sized chunks to each core. In many cases, almost as soon as they have started working, the GPU piranhas will be waiting for the next chunk of meat. Managing that is hard and often it is just easier for a developer to have all the software run on a regular CPU.Due to their specialized function, some tasks are better suited for GPU use. So far, research has focused on scientific tasks such as weather predictions, but there are efforts to standardize this programming.

Most industry support is focused around Apple's OpenCL specification which they announced (http://www.macrumors.com/2008/06/10/mac-os-x-snow-leopard-features-and-parallel-computing/) will be coming in the next major version of Mac OS X ("Snow Leopard"). Of course, not everyone is behind the initiative. As usual, Microsoft seems to have their own plans, and been involved in their own research on GPU computing.Michael Dimelow, director of marketing for media processing at ARM, said: "I don't think Microsoft will be sitting and watching. I would never underestimate Microsoft's ability to come up with alternative positions."Also relevant to Apple's recent mobile phone push is the fact that GPUs may provide handheld devices with extra computing power with less power consumption. According to the president of Khronos, GPUs can be 10 times more power-efficient than using a CPU. This can improve both video and audio performance on mobile devices.

Since the iPhone shares the underlying OS X codebase, these upcoming improvements in Mac OS X should trickle down to benefit the iPhone.

Article Link (http://www.macrumors.com/2008/07/22/gpu-powered-macs-and-iphones/)

[wrldwzrd89]

OpenCL looks simply amazing, if Apple and the rest of the industry can get the issues worked out. For example, I could see iTunes 8 for Snow Leopard utilizing this technology to drastically improve encoding speed. In some cases, improvements of 10x could be seen here alone.

Also, the GPU is a great place to look for hardware acceleration in video decoding. QuickTime X would be the perfect place for this, and such improvements would benefit iChat.

[andrewdale]

Sounds a lot like Adobe and CS4. Taking advantage of the GPU would be wonderful.

[ZeelessOne]

Kind of like NVIDIA's Tegra, GPU-powered mobile phones sounds like a great idea to me. The day-to-day computing on a phone is simple enough for a GPU, and then it has all that graphics power, all in one chip, saving power. Maybe this will trickle up to notebooks and desktops and we'll no longer have a two processor CPU-GPU combo, but one GPGPU or one CPU that's great at graphics!

[zelmo]

If the way the Folding app has been optimized for GPU's is any indication of the speed increases possible, this is pretty exciting stuff, particularly for the mobile market.

[rockinrocker]

If the way the Folding app has been optimized for GPU's is any indication of the speed increases possible, this is pretty exciting stuff, particularly for the mobile market.

could this mean the end of integrated graphics?

[Peace]

could this mean the end of integrated graphics?

It could also mean the beginning of newer,better integrated graphics.

[eidrunner247]

could this mean the end of integrated graphics?

I doubt it. Or at least not for a while. Integrated graphics are cheap and work decently. There may be a time when this new technology is filtered down into lower end computers, but of course, time will tell how soon that is.

[Beklim]

could this mean the end of integrated graphics?

I'm hoping it will be the beginning of replacable integrated graphics.

[ZeelessOne]

I'm hoping it will be the beginning of replacable integrated graphics.

That... doesn't make sense. One of the defining points of integrated graphics is that it's built in to the motherboard. If it's replaceable, it's just really wimpy dedicated graphics.

Computer are going down two routes. Either CPUs will become good at GPU functions (Intel Larrabee) or GPUs will become more general-purpose (GPGPU). GPGPUs have been more available in recent times, but Larrabee looks promising.

[ChrisA]

It could also mean the beginning of newer,better integrated graphics.

This is exactly what I was thinking. Once you have an OpenCL spec then the people who make GPUs can build them to run OpenCL well. I'm sure Intel is looking at this.

[sunfast]

I'm already excited about Snow Leopard. Could be quite a leap forward

[CarlHeanerd]

This technology is going to simply rock the industry. If Microsoft has their own plans to make a mockup, then this technology could become Mac exclusive, increasing the already great allure of Mac technology to all kinds of groups of people! At any rate, I want to see my Cinebench Score when this comes out!

[Saturn1217]

I just see this as another good reason to put dedicated graphics in the macbook. Or give us a 13 inch macbook pro. It's really all the same to me.

:apple: make it happen!!!

[B1gMac]

The notion that a piece of software could improve the performance specs of my current hardware makes me salivate.

Only :apple:

[wrldwzrd89]

I just see this as another good reason to put dedicated graphics in the macbook. Or give us a 13 inch macbook pro. It's really all the same to me.

:apple: make it happen!!!
As I see it, Apple could just as easily use Intel's upcoming GMA X4?00 (I don't remember what the ? is, but I know it's a number) in future MacBooks, which will probably work much better with OpenCL than the GMA X3100 does.

[herr_neumann]

I just see this as another good reason to put dedicated graphics in the macbook. Or give us a 13 inch macbook pro. It's really all the same to me.

:apple: make it happen!!!

I second the 13" MacBook Pro..... but we have been beating that issue since the demise of the 12" PowerBook

[Mousse]

I hope the folks who make Handbrake make it work with the GPU. It's one of the few programs I use that brings my mac to its knees and it "hotter than July.";)

If I can afford an 8-core Mac Pro, it's wouldn't be such a big deal, since those monster can run through a DVD faster than Mexican water through a first time tourist.:D:p

[Riemann Zeta]

Sounds great. But considering that all of NVIDIA's mobile GPUs are defective and will fail under their own massive thermal output, I hope Apple starts to go back to ATi, because once we start pushing those NVIDIA GPUs, they will start bursting into flames (figuratively). And I do have some reservations about this line: According to the president of Khronos, GPUs can be 10 times more power-efficient than using a CPU. This can improve both video and audio performance on mobile devices.
I think an entire order of magnitude of energy efficiency is wishful thinking. Perhaps highly integrated vector logic, like the itty-bitty PowerVR2 found in the iPhone/PDAs can be efficient, but most full-featured mobile GPUs--the ones that have hundreds of little vector units--use way more electricity than Intel's latest round of 45 nm Dual-Core CPUs and throw out about 80 deg of heat, necessitating active cooling. Once you need a couple of big badass fans just to keep an idling device from burning itself out, energy efficiency goes right out the window. That's why super high performance graphics arrays (the 256 parallel-VPU-core, 10th+ generation ones) will still be desktop-bound for quite a long time. Look at the latest generation of game-capable GPUs--they all need their own independent 300W power supplies, solid copper heat pipes and at least a couple of PCI slots for all the fans.

[iMacmatician]

OpenCL looks simply amazing, if Apple and the rest of the industry can get the issues worked out. For example, I could see iTunes 8 for Snow Leopard utilizing this technology to drastically improve encoding speed. In some cases, improvements of 10x could be seen here alone.You mean Media Me iTunes X? ;):D


A Core 2 Duo at 2.5 GHz is 20.00 GigaFLOPS using SSE.
2 quad-core Xeons at 3.2 GHz is 102.4 GigaFLOPS using SSE.
A GeForce 8600M GT is 91.20 GigaFLOPS.
A GeForce 8800 GT is 504.0 GigaFLOPS.
The Larrabee (Intel GPU using x86 cores) is expected to reach at least 960 GigaFLOPS.

Larrabee is 20x more efficient in terms of performance per watt than a Core 2 Duo, despite having half the single thread performance.

As I see it, Apple could just as easily use Intel's upcoming GMA X4?00 (I don't remember what the ? is, but I know it's a number) in future MacBooks, which will probably work much better with OpenCL than the GMA X3100 does.X4500

[rpp3po]

Cheers to all Macbook Pro owners, whose degenerate Nvidia GPU's are going to fail out-of-warranty.

[OS X Dude]

As I see it, Apple could just as easily use Intel's upcoming GMA X4?00 (I don't remember what the ? is, but I know it's a number) in future MacBooks, which will probably work much better with OpenCL than the GMA X3100 does.

The X4500 is the newer card from the Montevina platform. Along with up to 3.06 Penryn processors, 1066MHz FSB and WiMax support.

I hope WiMax finds its way in macs.

[wrldwzrd89]

You mean Media Me iTunes X? ;):D

A Core 2 Duo at 2.5 GHz is 20.00 GigaFLOPS using SSE.
2 quad-core Xeons at 3.2 GHz is 102.4 GigaFLOPS using SSE.
A GeForce 8600M GT is 91.20 GigaFLOPS.
A GeForce 8800 GT is 504.0 GigaFLOPS.
The Larrabee (Intel GPU using x86 cores) is expected to reach at least 960 GigaFLOPS.
Larrabee is 20x more efficient in terms of performance per watt than a Core 2 Duo, despite having half the single thread performance.
Wow, that's impressive. GPU performance is a resource we couldn't have found a better time to tap, in my humble opinion. :D
X4500
Thanks - just out of curiosity, is that info on Intel's site?

[Hattig]

It could also mean the beginning of newer,better integrated graphics.

It should do, but don't hold your breath for Intel to implement it.

nVidia and AMD both have integrated chipsets with more powerful and up-to-date GPUs on them. I don't know how suitable they are for OpenCL / CUDA at the moment.

Intel's Larrabee will come out eventually, but it might be 2010 or 2011 before its technology gets integrated into a chipset.

I'm hoping that this means that Apple will move to discrete graphics across the line again later this year in order to boost performance.

[SiliconAddict]

Umm not to rain on anyone parade but NVidia and Windows has been doing this for a while now, not sure about ATI though.

[wrldwzrd89]

Umm not to rain on anyone parade but NVidia and Windows has been doing this for a while now, not sure about ATI though.
nVidia came up with CUDA, an alternative to OpenCL. I was not aware that Windows was using it, though. ATI, on the other hand, was looking at OpenCL but had yet to implement full support for it.

[NovemberMike]

Wow, that's impressive. GPU performance is a resource we couldn't have found a better time to tap, in my humble opinion. :D


That doesn't necessarily mean too much. Assuming you can feed it the info fast enough it is going to be amazing, but any task that utilizes it will have to A) be highly parallelizable (not true for most tasks) and B) be limited by the processing power rather than the hard drive, RAM, user input speed etc.

This will be great for some tasks (handbrake, compiling code, video decoding) but don't expect it to make Safari run any faster.

Also, this stuff has been going on in Windows for at least a couple years. Folding@Home utilizes the GPU like this, and things like PS CS4 are supposed to be already utilizing the GPU on Windows (not sure how that is going on Macs).

[Eidorian]

While it isn't very amazing the Intel GMA X3100 does have hardware shaders to take advantage of their computation abilities.

You'll get a few more on the GMA X4500. The GPU based decoding is still a little up in the air right now.

[Hattig]

You mean Media Me iTunes X? ;):D


A Core 2 Duo at 2.5 GHz is 20.00 GigaFLOPS using SSE.
2 quad-core Xeons at 3.2 GHz is 102.4 GigaFLOPS using SSE.
A GeForce 8600M GT is 91.20 GigaFLOPS.
A GeForce 8800 GT is 504.0 GigaFLOPS.
The Larrabee (Intel GPU using x86 cores) is expected to reach at least 960 GigaFLOPS.

Larrabee is 20x more efficient in terms of performance per watt than a Core 2 Duo, despite having half the single thread performance.



AMD's current HD4700 gets 1.2 TFLOPS, today, at a quite affordable price. Quite a lot of power consumption too, but a lot less than Larrabee is rumoured to consume.

[themiracle]

I'm hoping that this means that Apple will move to discrete graphics across the line again later this year in order to boost performance.

Could be a part of the "products with features at price points that our competition can't match" deal.

[diamond.g]

As I see it, Apple could just as easily use Intel's upcoming GMA X4?00 (I don't remember what the ? is, but I know it's a number) in future MacBooks, which will probably work much better with OpenCL than the GMA X3100 does.

The X4500 would be far outclassed especially since it only had 10 unified shaders, compared to 32 for the 8600M GT. You may be correct on it being better than the X3100, although it shouldn't be by much, the X3100 has 8 unified shaders.

[Chilz0r]

Seeing all this nice stuff about GPU's is great and all but you have to remember that the GPU is suited to particular applications & tasks . It will be the combination of CPU's & GPU's working together, so it's not going to be the end all and be all of processing power. The deciding factor in this waiting game will be NVIDIA's track record and Apple's fantastic offering of GPU's we have all come to love and know, so we will see if this technology delivers or maybe we will see another MobileMe repeat?

[Hattig]

While it isn't very amazing the Intel GMA X3100 does have hardware shaders to take advantage of their computation abilities.

You'll get a few more on the GMA X4500. The GPU based decoding is still a little up in the air right now.

The big issue is that the shaders have additional functionality in modern graphics cards to allow them to work as generic calculation engines rather than just for the purposes of graphics. I think AMD made the leap between the HD2000 and the HD3000 series, although before that it was still good enough for Folding @ Home which had a client written directly for the AMD hardware using AMD's CTM (lower level than CUDA, but the same idea, probably harder to get to grips with).

[Eidorian]

The X4500 would be far outclassed especially since it only had 10 unified shaders, compared to 32 for the 8600M GT. You may be correct on it being better than the X3100, although it shouldn't be by much, the X3100 has 8 unified shaders.Keep in mind not all shaders are created equal. More shaders is still better though but I wouldn't compare different manufacturers on shader count alone.

[Joe The Dragon]

As I see it, Apple could just as easily use Intel's upcoming GMA X4?00 (I don't remember what the ? is, but I know it's a number) in future MacBooks, which will probably work much better with OpenCL than the GMA X3100 does.
but there use of system ram is not that good for this kind of setup.

[superman193]

what would we compare the GPU in the iPod touch / iPhone ?

Ati Rage vs iPhone/iPod gpu?

[wrldwzrd89]

but there use of system ram is not that good for this kind of setup.
True - integrated GPUs aren't that powerful to begin with. However, Intel's QuickPath Interconnect technology will reduce the penalty integrated graphics must pay to access system RAM - this will make a performance difference, but how big remains to be seen.

[Eidorian]

True - integrated GPUs aren't that powerful to begin with. However, Intel's QuickPath Interconnect technology will reduce the penalty integrated graphics must pay to access system RAM - this will make a performance difference, but how big remains to be seen.AMD's available IGP solutions are quite passable with their integrated memory controller usage. Then again some of them also have a small amount of dedicated DDR3 RAM as well.

[pgwalsh]

Would hate to leave out Toshiba's Spurs Engine multimedia powerhouse. Toshiba's put it into one of their laptops, so why not Apple who makes some serious multimedia applications. Didn't Apple mention some product transitions. It seems that you could have a board with integrated graphics, intel cpu and spurs engine to speed up multimedia.

[Michael CM1]

Someone with that article is using a misnomer. How is it that there are "hundreds of microprocessor cores" waiting to do my bidding yet my processor is a "dual-core" one? That makes absolutely no sense. Part of me thinks someone means "transistors" or something else.

[twoodcc]

well i hope apple adopts this soon, b/c microsoft isn't going to sit and do nothing about this

[Eidorian]

Someone with that article is using a misnomer. How is it that there are "hundreds of microprocessor cores" waiting to do my bidding yet my processor is a "dual-core" one? That makes absolutely no sense. Part of me thinks someone means "transistors" or something else.A unified GPU shader is a specialized processor. You can consider the many dozens to hundreds of them to be multiple cores.

[Chilz0r]

what would we compare the GPU in the iPod touch / iPhone ?

Ati Rage vs iPhone/iPod gpu?

Dude, like check out PowerVR chips, they r mad and I'm going to put one in my Mac Pro today!!!!

Umm not to rain on anyone parade but NVidia and Windows has been doing this for a while now, not sure about ATI though.

Yeah, that's definitely taken Vista places and don't get me started on NVIDIA, they can't even make SLi worth the extra dollars due to their "fantastic" drivers.

[babyj]

I don't think anyone is claiming the technology is brand new or the idea exclusive, but the big difference is Apple are pushing for an apparent open standard which will also be built in to Mac OS X and I assume it will also mean support is added to all the main compilers.

I'm guessing the other main part of Apple's plan is to have Mac OS X and all their own applications re-compiled to make us of it on release which will give their software and computers a big advantage over the competition if they don't keep up.

[diamond.g]

Keep in mind not all shaders are created equal. More shaders is still better though but I wouldn't compare different manufacturers on shader count alone.

It would be safe to say that Intel shaders wont be as fast as either ATI or Nvidia shaders. I can't seem to find a whole lot of info on the math capabilities (like you can of the others) on beyond3d. I guess we will have to wait till some more comprehensive benchmarks take place. Of course this would all be a moot point for Apple since they run different drivers anyways. Performance in Windows doesn't necessarily translate to performance in Mac.

[JonasLondon]

As long as it makes mobileMe more stable and Safari snappier, I'm happy.

[8CoreWhore]

Cheers to all Macbook Pro owners, whose degenerate Nvidia GPU's are going to fail out-of-warranty.
My C2D 2.16 MacBook Pro uses ATI Radeon X1600. Don't create misinformation:mad:.

[Salavat23]

If I can afford an 8-core Mac Pro, it's wouldn't be such a big deal, since those monster can run through a DVD faster than Mexican water through a first time tourist.:D:p

You mean faster than a Mexican hopping the border fence into the states. ;)

[ZeelessOne]

nVidia came up with CUDA, an alternative to OpenCL. I was not aware that Windows was using it, though. ATI, on the other hand, was looking at OpenCL but had yet to implement full support for it.

It's not so much used by Windows as Windows is the platform it's used on. I think it's been available for Linux just as long, but I distinctly remember it not being available for Mac OS, but looking at the CUDA downloads page, apparently it is.

There's also some stuff mulling around about CUDA being ported to Radeon cards, but I haven't been watching that closely.

Hopefully OpenGL 3.0 specs will be finalized in time for Snow Leopard. That will put all these DX10 cards in Macs to work! It'll put me to work too. I'll finally take the time to learn OpenGL.

[whooleytoo]

Just how much extra can they squeeze from the GPUs?

Doesn't just about everything (Quartz 2D, Core Image, Core Video) use the GPU already?

[Hattig]

Someone with that article is using a misnomer. How is it that there are "hundreds of microprocessor cores" waiting to do my bidding yet my processor is a "dual-core" one? That makes absolutely no sense. Part of me thinks someone means "transistors" or something else.

Modern graphics cards do have hundreds of "cores" but they're not like CPU cores which are very different. However for what the GPU cores are good at, they make the CPU cores look anaemic. Think of a single GPU core as being like a single G4 Altivec unit +/- some things.

[iMacmatician]

Intel's Larrabee will come out eventually, but it might be 2010 or 2011 before its technology gets integrated into a chipset.It's slated for a late 2009 release. Don't know about chipsets though.

AMD's current HD4700 gets 1.2 TFLOPS, today, at a quite affordable price. Quite a lot of power consumption too, but a lot less than Larrabee is rumoured to consume.I should have made this clear in my original post but...

The difference is that Larrabee's 960 gigaFLOPS is double-precision (used in CPUs), while the HD 4870's 1.2 teraFLOPS is single-precision. Regular GPUs take a big hit on double-precision. In the HD 4870's case, it's 1/5 the SP speed (for the GTX 280, it's 1/12). In other words, 240 double-precision gigaFLOPS.

Someone with that article is using a misnomer. How is it that there are "hundreds of microprocessor cores" waiting to do my bidding yet my processor is a "dual-core" one? That makes absolutely no sense. Part of me thinks someone means "transistors" or something else.That reminds me of this (http://forum.beyond3d.com/showpost.php?p=1141497&postcount=285).

And yes, I'm going to keep this crusade up until everyone ridicules anyone who refers to a single alu as a processor or a core. You either count a single bit xor as a "processor/core" you count something that actually is a processor as a processor.

[macduke]

Ok...say they release iPhone software 3.0 next summer after the release of Snow Leopard. How could the iPhone benefit? Does it have a video chip that could even take advantage of this technology? Or would there have to be a new version of the iPhone? Oh geeze...let the rumors begin!!

[Riemann Zeta]

Someone with that article is using a misnomer. How is it that there are "hundreds of microprocessor cores" waiting to do my bidding yet my processor is a "dual-core" one? That makes absolutely no sense. Part of me thinks someone means "transistors" or something else.
The "cores" in an Intel dual-core CPU are full featured general purpose CPUs in and of themselves--they have multiple out-of-order integral instruction units, floating point units and vector units, as well as much of "other stuff" for memory management, bus interfaces, etc... A GPU "micro-core" aka a "shader" is basically just a single vector unit; its purpose is to do linear algebra really quickly--it takes in an N-dimensional vector (which is composed the coefficients of a linear equation, e.g. Ax1 + Bx2 + Cx3 + ...) and does a linear transformation on that vector. This is great for calculating where things are and how things move in 3D, but not every type of OS code benefits from parallel vector processing. CPUs can easily perform other non-vectorized functions while GPUs can't, so it is far easier for programmers to write general purpose code for CPUs.

[commander.data]

True - integrated GPUs aren't that powerful to begin with. However, Intel's QuickPath Interconnect technology will reduce the penalty integrated graphics must pay to access system RAM - this will make a performance difference, but how big remains to be seen.
That doesn't make any sense. QuickPath increases the latency for IGP memory access. The reason is because when the memory controller is integrated in the chipset northbridge where the IGP is located, the IGP gets direct memory access. But then the memory controller is located off chip in the processor, then the IGP has to make a request from the northbridge to the CPU's IMC through QuickPath where it will then send information back to the northbridge and the IGP. No matter how efficient QuickPath is that is still an extra hop and will increase memory access latency for the IGP.

[emotion]

Just when NVidia's CUDA is shaping up to be the industry's mammoth paradigm. Interesting. Widespread adoption will depend on what Intel do for Larrabee then...

[iMacmatician]

Just when NVidia's CUDA is shaping up to be the industry's mammoth paradigm. Interesting. Widespread adoption will depend on what Intel do for Larrabee then...And Apple and Intel have a close relationship...

[emotion]

And Apple and Intel have a close relationship...

Precisely... It's unlikely the two haven't talked about this.

[diamond.g]

Just how much extra can they squeeze from the GPUs?

Doesn't just about everything (Quartz 2D, Core Image, Core Video) use the GPU already?If you think that stuff even remotely stresses anything other than the GMA line (GMA950, X3100) you are mistaken. Programable GPU's don't break any sweat when dealing with those things. You could probably show hundreds of thousands of Core Image/Video Windows and still not stress the GPU (you would probably run out of video memory first).

That doesn't make any sense. QuickPath increases the latency for IGP memory access. The reason is because when the memory controller is integrated in the chipset northbridge where the IGP is located, the IGP gets direct memory access. But then the memory controller is located off chip in the processor, then the IGP has to make a request from the northbridge to the CPU's IMC through QuickPath where it will then send information back to the northbridge and the IGP. No matter how efficient QuickPath is that is still an extra hop and will increase memory access latency for the IGP.
That is until they get the GPU embedded into the CPU (think Larabee).

[iMacmatician]

That is until they get the GPU embedded into the CPU (think Larabee).Larrabee IS a CPU. It apparently uses revised Pentium cores, a 512-bit SIMD unit, and features present on GPUs but not on CPUs.

[diamond.g]

Larrabee IS a CPU. It apparently uses revised Pentium cores, a 512-bit SIMD unit, and features present on GPUs but not on CPUs.
That doesn't change the fact that it will be integrated on/in to Nehalem cores.
It wont be a separate socket. There is supposed to be that option, but I douby Apple would go that route.

[iMacmatician]

That doesn't change the fact that it will be integrated on/in to Nehalem cores.No it won't (link please). Larrabee is discrete graphics.

[diamond.g]

No it won't (link please). Larrabee is discrete graphics.

Hmm, I could have sworn all fingers pointed to Intel basically using a stripped down (core wise) Larrabee for IGP duty. Shoot it would be the smart thing to do as it would basically be the reverse of what ATI/Nvidia does. It would get Larrabee into the hands of many and hopefully make the discreet part a bit more popular.


If I am wrong, sorry. I am still searching for anything, but an fast becoming sure of my wrongness.

EDIT: Found something on Electronista (http://www.electronista.com/articles/08/03/17/intel.on.nehalem.etc/)

[iMacmatician]

EDIT: Found something on Electronista (http://www.electronista.com/articles/08/03/17/intel.on.nehalem.etc/)That's motherboard integration.

I agree with the rest of your post though. Although Larrabee is 150~300 W, a version with just a few cores could easily be 10~20 W. Larrabee is supposed to scale to thousands of cores anyway.

[winterspan]

I'm hoping it will be the beginning of replacable integrated graphics.
"replaceable integrated graphics" is an oxymoron. The only reason we call them "integrated" is because they ARE integrated into the motherboard and therefore are unable to be replaced with newer units. This allows
Apple to save money by having Intel build cheapo graphics directly into the motherboard.

Also, the GPU is a great place to look for hardware acceleration in video decoding. QuickTime X would be the perfect place for this, and such improvements would benefit iChat.
Yep, hardware accelerated video decoding is already in place in all modern discerete graphics, although I'm unsure if Apple has been using this, or if that is indeed was Quicktime X is going to do for desktop OSX. I also believe Intel's integrated crap finally has integrated decoding as well with the new laptop platform.

Sounds great. But considering that all of NVIDIA's mobile GPUs are defective and will fail under their own massive thermal output, I hope Apple starts to go back to ATi...
That's a one-time manufacturing problem that is said to only affect a small amount of outstanding mobile GPUs. I've had 5+ over the years, and I haven't had an issue. I even recently replaced a Geforce Go 7900 with a mobile Quadro unit from Ebay, and it has been working great.
AFAIK, considering nVidia doesn't even have a fab, it could be the fault of the chinese/taiwanese fabrication facility.


Maybe this will trickle up to notebooks and desktops and we'll no longer have a two processor CPU-GPU combo, but one GPGPU or one CPU that's great at graphics!
Well, that is already starting to happen. Although not technically "merging" into one commmon unit, both Intel and AMD have projects on tap for 2009 that integrate CPU and CPU cores into the same package, and eventually into the same die even.
Although the first verions of these will only have the capabilities of integrated graphics, I'm sure eventually they will completely combine and we'll no longer see seperate high-end enthusiast graphics cards, as they'll be replaced by massively-parallel hybrid processors that take care of all processing duties on the sytem.



Also, this stuff has been going on in Windows for at least a couple years. Folding@Home utilizes the GPU like this, and things like PS CS4 are supposed to be already utilizing the GPU on Windows/
Umm not to rain on anyone parade but NVidia and Windows has been doing this for a while now, not sure about ATI though.

Both Nvidia and ATI have their own GPGPU software development kits, but Nvidia's CUDA has definitely been mentioned a lot more. Probably because they were the dominant force in GPUs recently as ATI didn't have competitive product offerings for a long time.
The SDKs have been available for Windows/Linx, but there haven't been ANY comemercial consumer applications using them, so I definitely wouldn't say "windows has been doing this for awhile". The Adobe Photoshop GPU presentation was an unofficial look into the Adobe's research labs, not something currently available for Windows, and folding@home is obviously a specialist scientific case. GPGPU is still largely something for research, although I have seen some examples of internal use at corporations.
Publicly at least, Apple is already way out in front of Microsoft having formed the OpenCL standards body to streamline GPGPU processing into a standard framework.


Keep in mind not all shaders are created equal. More shaders is still better though but I wouldn't compare different manufacturers on shader count alone.
definitely! Intel's IGPs have come a long way, but they are still crap compared to anything from AMD/nVidia.


True - integrated GPUs aren't that powerful to begin with. However, Intel's QuickPath Interconnect technology will reduce the penalty integrated graphics must pay to access system RAM - this will make a performance difference, but how big remains to be seen.
The IGP in current platforms has a direct connection to the memory throught the mem controller on the northbridge -- with nehalem, the IGP will have to go through the northbridge over quickpath to the CPU-integrated memory controller, which will then have to access the memory which is directly connected to the processor.
I can't imagine that this would improve the performance of the IGP through this mechanism, as the latency would appear to be cancel out any increase in memory bandwidth. I'm no expert however, so I'll have to do some more research.


Would hate to leave out Toshiba's Spurs Engine multimedia powerhouse. Toshiba's put it into one of their laptops, so why not Apple who makes some serious multimedia applications. Didn't Apple mention some product transitions. It seems that you could have a board with integrated graphics, intel cpu and spurs engine to speed up multimedia.
The "Spurs engine" is stupid. First of all, they should have used a full CELL BE chip. Secondly, it doesn't matter anyways because who one wants a proprietary, hard-to-program co-processor? Just like the "Physx" add-in card, this will fail because it won't have any extensive community support. For speeding up parllel computations, You want a vendor-neutral standards-based approach that will be compatible with a variety of hardware AKA OpenCL.


Just how much extra can they squeeze from the GPUs? Doesn't just about everything (Quartz 2D, Core Image, Core Video) use the GPU already?
Indeed, but you are referring to graphics-related processing. The term "GPGPU" refers to using the GPU to do general purpose calculations for tasks that are easily parallelized like video processing/encoding/decoding/iDCT, audio encoding, digital image processing, scientific simulations like fluid dynamics, protein folding, Oil/gas geology etc.



The difference is that Larrabee's 960 gigaFLOPS is double-precision (used in CPUs), while the HD 4870's 1.2 teraFLOPS is single-precision. Regular GPUs take a big hit on double-precision. In the HD 4870's case, it's 1/5 the SP speed (for the GTX 280, it's 1/12). In other words, 240 double-precision gigaFLOPS.
You are correct, but FP64/double precision is only really important for scientific and industrial simulation/computation since they need that level of accuracy. I don't think it will matter for most consumer use of GPGPU.


That doesn't change the fact that it will be integrated on/in to Nehalem cores. It wont be a separate socket. There is supposed to be that option, but I douby Apple would go that route.
There seems to be a common misunderstanding that Larabee will be introduced as a CPU-integrated graphics chip. Larabee is going to be an PCIe add-on board, like most GPUs. Intel's first products where they combine CPU and GPU cores into the same package or die are going to be using graphics technology from their current motherboard integrated graphics, NOT from Larabee ---- aLthough this will no doubt change down the road sometime.

[wrldwzrd89]

That doesn't make any sense. QuickPath increases the latency for IGP memory access. The reason is because when the memory controller is integrated in the chipset northbridge where the IGP is located, the IGP gets direct memory access. But then the memory controller is located off chip in the processor, then the IGP has to make a request from the northbridge to the CPU's IMC through QuickPath where it will then send information back to the northbridge and the IGP. No matter how efficient QuickPath is that is still an extra hop and will increase memory access latency for the IGP.
Hmm. I would have thought the exact opposite. Thanks for the information!

[AidenShaw]

That... doesn't make sense. One of the defining points of integrated graphics is that it's built in to the motherboard. If it's replaceable, it's just really wimpy dedicated graphics..

"Integrated" is built into the North Bridge (the memory controller of the chipset).

"Discrete" may be soldered to the motherboard, on a special form factor card like MXM, or on a PCIe card that's easily replaceable (except on an Apple).

[deputylove8]

Yup yup...agree

[chickenninja]

piranhas dont eat apples... DUH!

[diamond.g]

There seems to be a common misunderstanding that Larabee will be introduced as a CPU-integrated graphics chip. Larabee is going to be an PCIe add-on board, like most GPUs. Intel's first products where they combine CPU and GPU cores into the same package or die are going to be using graphics technology from their current motherboard integrated graphics, NOT from Larabee ---- aLthough this will no doubt change down the road sometime.

Ah, okay. I am really interested in seeing if Intel can actually hang with Nvidia/ATI in the discreet GPU market. The only twist I can think of is if Intel can get rasterization on the extinction list. Maybe they would have a chance then. As it stands Larrabee, while it sounds cool, just seems like it is going to get walked all over by ATI/Nvidia. Much like Intels first foray into the discreet graphics market.

[layte]

Publicly at least, Apple is already way out in front of Microsoft having formed the OpenCL standards body to streamline GPGPU processing into a standard framework.

DirectX 11 says hi.

http://www.gamasutra.com/php-bin/news_index.php?story=19522

Features include new shader technology that begins to allow developers to position GPUs as more general-purpose parallel processors, rather than being dedicated solely to graphics processing; better multi-threading capabilities; and hardware-based tesselation.

[iMacmatician]

You are correct, but FP64/double precision is only really important for scientific and industrial simulation/computation since they need that level of accuracy. I don't think it will matter for most consumer use of GPGPU.I think Larrabee's single-precision is 2x double-precision (although I'm not totally sure on this). That would give 1.92 teraFLOPS.

[blitzkrieg79]

So.... Basically Intel/AMD is again behind IBM... As I said few months ago, Cell processor will be a blueprint for future processor designs, main CPU surrounded by specialized cores... Nothing revolutionary considering IBM has been doing this for few years now... And with Cell 2 on horizon, IBM will be on the top of its game once again...