AMD CPU speculation... and expert conjecture

[amdfangirl]

I wish AMD would just go ahead and kill AM3+.

Make an FX 4 module CPU for FM2+.

 

[jdwii]

Spoiler

afaik, cpu and gpu integration on same die is to minimize the time taken to communicate between cpu and gpu and bandwidth thus removing communication bottleneck
which means more performance in gpu assisted tasks
thus for processing tasks which needs tons of data exchange between cpu and gpu will get huge benefit from this tech
for example hardware video conversion

but

gaming does not require such communications speeds (only productive works needs higher cpu-gpu communication), gaming needs higher gpu-vram speed which is not possible using system ram, also not to forget power requirement of gpus
this is why we say that pcie 2.0 x16 is enough for gaming, which means cpu-gpu communication is not a bottleneck in gaming

thus, by this i mean to say that imo we won't see death of dedicated gaming GPUs as far as gaming is alive, but workstation cards may suffer (not sure about higher end workstation cards) and we may see cpu/apu designed for workstations (if we can get enough ram bandwidth)

1) The "about 10x faster" claim only considered CPU-->GPU communication. Thus applies to gaming. For compute one has to consider also CPU-->GPU and other nuisances, increasing the performance gap.

2) Only because ancient programs used less than 640Kb doesn't mean that 640Kb are enough for everybody. Current games are developed with current hardware limits in mind. Future games doesn't need to be limited by constraints of today hardware.

3) The slow PCie3 is already bottlenecking current games. Developers alleviate this bottleneck by using tricks such as low resolution textures, texture quilting (building large textures by repetition of small textures), texture compression... E.g. unreal engine uses texture compressed at 1:4 or 1:8 ratio (depends of DXTn format used) before sending them by the PCIe slot. The compressed textures are then decompressed by the GPU.

4) Another important point is that gaming GPUs are not designed in a vacuum, but use the same basic architecture used for compute GPUs. R&D costs are then distributed between all the cards that use the architecture. If you develop an architecture for use exclusive on a gaming card, then all the cost have to be transfered to those cards, which implies ultra-expensive cards that no gamer would purchase.

This is essentially the same reason why AMD has not released Steamroller FX CPU for gamers. Previous FX CPUs shared the R&D cost with the Opterons used in servers and HPC. But Steamroller is not competitive enough for server/HPC CPUs; thus, once Steamroller architecture was dropped from Opteron CPUs, it became evident that AMD couldn't release a FX brand because a mere question of costs.

I recall that I explained this when I claimed the past year, that no FX Steamroller was coming to the deskopt. Several posters ignored my point and claimed "wait for the roadmap". The 2014/2015 desktops roadmap gave me the reason. I did the math. AMD also did.

5) Only because current PCs use slow DDR3 for "system ram" doesn't imply that "system ram" has to be slower than "vram". The PS4 uses GDDR5 as "system ram". The next year Intel releases a 'CPU' with 8--16 GB of MCDRAM with a sustained bandwidth of 500GB/s. The Nvidia Titan VRAM peaks at 288GB/s.

The Nvidia Research Team that did make the above claim is designing an ultra-high performance APU with a "system ram" of 1.6TB/s of bandwidth. I.e. more than 5x the bandwidth of the GDDR5 on the Nvidia Titan designed by the same Team.

6) Another point is that games evolve towards GPU computing. Ancient games used the CPU for everything, and the GPU only for basic display. But current GPUs evolved towards offloading the CPU for 3D graphics computation

The next step is towards offloading the physics and AI computations also to the GPU. In fact, the APU used in the PS4 has been specifically designed to compute the physics and the AI on the GPU. Watch Sony talks. Existent games and demos are already computing the physics on the PS4 GPU.

==============================================

Once again:

The past is CPU + dGPU

The future is high-performance APU

The transition for AMD is clearly APU + dGPU

Why do you believe that AMD is enabling APU-dGPU Crossfire? Why do you believe that MANTLE has asymmetric multi-GPU support? Why do you believe that AMD is giving talks about using the dGPU for rendering whereas offloading the post-processing to the APU?

http://gearnuke.com/in-depth-look-at-amd-mantle/

Yeah that makes way more since i almost find it impossible to get more power from 1 die vs 2. Seems a little weird to state such a thing

Except that maths says otherwise. What happens is that Nvidia research Team did the math before doing the claim. I also did.

Well saying it doesn't make it so. unless you show me how gaming is faster on 1 die vs 2 it doesn't make sense.

 

[colinp]

I'm sure the time will come when it will be possible to do fully photo realistic rendering at 60fps in 3D, at 4K or higher resolution on a single die for less than 100W. At that point, the case for a dGPU has gone, but how long will it be before that becomes a reality?

 

[wh3resmycar]

I'm sure the time will come when it will be possible to do fully photo realistic rendering at 60fps in 3D, at 4K or higher resolution on a single die for less than 100W. At that point, the case for a dGPU has gone, but how long will it be before that becomes a reality?

by the time it happens the world would've moved on to 8k or better where the dgpu will come into play again. the way these tech companies play the numbers usually go higher each generation.

 

[palladin9479]

I'm sure the time will come when it will be possible to do fully photo realistic rendering at 60fps in 3D, at 4K or higher resolution on a single die for less than 100W. At that point, the case for a dGPU has gone, but how long will it be before that becomes a reality?

When that's possible we would of already moved into fully immerse VR where imagery is no longer rendered onto a single 2D flat surface. They have already been experimenting with this for awhile now. Remember anything you can do with integrated you can do 4~5x more with external. The only question then becomes, do you have an application that could use that kind of power?

 

[colinp]

When that's possible we would of already moved into fully immerse VR where imagery is no longer rendered onto a single 2D flat surface. They have already been experimenting with this for awhile now. Remember anything you can do with integrated you can do 4~5x more with external. The only question then becomes, do you have an application that could use that kind of power?

By the way, I agree with the above two points.

When it comes to displays, we're already reaching the limits of the human eye, so resolutions won't increase forever, but of course there will be next gen technologies like 3D displays etc. Eventually display tech will top out, but it will take a long time.

Alongside this you'll have a greater push for realism. So, things like real world materials like metal, concrete, glass, vegetation, hair, dust, mud, water, weather etc. that all behave exactly as they do in the real world when you interact with them in some way.

And then ever more realistic AI.... and so on and so on.

And to do all this in a single 100w socket is going to be waaay off in the future.

 

[de5_Roy]

this.. it's sad that some folks here are still justifying Piledrivers to people. and i pity the folks who actually fall into this trap. the fake security of having 6-8 cores "since consoles are using an 8 core AMD chip" and "games are starting to use MOAR COARS", then you'll be stuck with a platform with no upgrade path.

i question using "futureproof" with any cpu(s), not just pd.
"upgradability" or "upgrade path" was "real" (i.e. worth doing) before, but it's going the same way as "futureproofing" now. both intel and amd have changed socket after two gens. unlike intel, amd maintained socket compatibility for socket fm2 cpus/apus for socket fm2+, but users won't get anything significant from using a richland apu on a socket fm2+ motherboard. it was similar with sb core i5/i7 owners with lga1155 motherboards, ivb didn't offer significant gains. same thing will happen with haswell's convoluted upgrade path.

i don't know how consoles' cpus operate... from what i've read, there's no way the console games could use (not load) all 8 cores. windows/linux execute software threads differently.

however, 2 core cpus are more or less done, in terms of dt gaming. their only relevance is in small size pcs where heat and power issues are significant and because intel won't sell cheap 4 cores (but baytrail cra.p.u.s). 4 cores cpus will hit a limit too. within next year or two.

 

[palladin9479]

however, 2 core cpus are more or less done, in terms of dt gaming. their only relevance is in small size pcs where heat and power issues are significant and because intel won't sell cheap 4 cores (but baytrail cra.p.u.s). 4 cores cpus will hit a limit too. within next year or two.

Dual cores ended their practical use over a year ago. Four cores are standard now and will probably stay as such for another 3~5 years depending on how software evolved. The reason why you want at least four is that if heavy programs are using 2~3 you want at least one spare to execute background and interface tasks. This is where benchmark programs REALLY mislead everyone, they produce a picture of a platform with everything turned off and some program running without any random human interaction. This is all done so as to sterilize the environment but in real world use you won't have background services disabled and often the user will have other things going on. They will rarely run two heavy programs at once (unless their streaming) so you typically want 1~2 more cores then the maximum your heaviest program will use. This is why I went with an fx8350, some of the stuff I do will actually use six cores frequently and I wanted a few spare for additional tasks. Most users will be happy with a 4 or 6 core CPU depending.

 

[juanrga]

And yet you can still get pretty good scaling up to quad SLI. When you can fit 4 Hawaii GPUs and an i7 on a single die I'll worry about PCIe speeds. And you can always combat that bandwidth limit by increasing the local memory size.

4x Hawaii is 24.8 billion transistors
Haswell GT2 is 1.4B

26.2 Billion transistors. Yeah that's going to be a long ass while to fit on a single die. Maybe at 7nm, a full 3 nodes ahead of where we are now..

This incorrect argument was corrected before. Again, the APU die diagram shows that about 90% of the die space is GPU. This implies a discrete GPU would have only ~11% more transistors than the APU, but the small theoretical advantage in performance is completely outperformed by the interconnect bottleneck, doing that the APU was much faster than the discrete graphics card with 11% more transistors. In fact the APU is about 10x faster.

I also mentioned what node are considering for the APU. I repeat: 10nm.

Regarding your comparison with Hawai. Let us do some math:

R9-290X: 704 GFLOPS (DP)

Thus 4x Hawai, assuming perfect scaling are 2816 GFLOPS

They rate their ultra-high-performance APU at 20.48 TFLOPS (DP). This is 20480 GFLOPS. This is 7.27x the raw gflops of 4x Hawai discrete cards.

But this is assuming perfect scaling, not interconnect bottleneck... In reality the above number has to be multiplied by a factor of ten at least.

Conclusion: their ultra-high-performance APU is about 73 times faster than four R9-290X working in parallel.

 

[bigal80ak]

wow feel like I am jumping in to a fight about question or two. First in the drop box why is one of the options how to get 99% isopropyl alcohol really? Second I have been looking up info as brother wants new pc and I give him my old and upgrade mine. When will AMD release new CPU and I mean CPU not APU? Cant find out much or are they waiting till it gets closer to ddr4 release? That would be a nice jump but just some good info on AMD new CPU as in release, power, consumption price new tech would be great.

Thanks. and 3-5 years for 4 core dominance I don't know a lot of pc at business's I see are still running dual at 3 I think it will be more like 5-8. Cant wait till I am 50 and I have a 10ghz 18 core haha.

 

[juanrga]

I want remark that the factor of 10x that I am always applying to performance (APU ~ 10x dGPU) is only valid for basic gaming and some light compute tasks. For heavy compute tasks the factor is much greater.

In some cases the APU is up to 120x faster than a GPU with the same number of cores. This is the reason why a 80 core APU is up to 6 times faster than a discrete GPU with 1600 cores.

On the Efficacy of a Fused CPU+GPU Processor (or APU) for Parallel Computing

120 = (1600/80) x 6

But this 120x performance factor is only for some heavy compute tasks. I will continue using the conservative 10x factor in APU vs GPU comparisons.

 

[juanrga]

wow feel like I am jumping in to a fight about question or two. First in the drop box why is one of the options how to get 99% isopropyl alcohol really? Second I have been looking up info as brother wants new pc and I give him my old and upgrade mine. When will AMD release new CPU and I mean CPU not APU? Cant find out much or are they waiting till it gets closer to ddr4 release? That would be a nice jump but just some good info on AMD new CPU as in release, power, consumption price new tech would be great.

Thanks. and 3-5 years for 4 core dominance I don't know a lot of pc at business's I see are still running dual at 3 I think it will be more like 5-8. Cant wait till I am 50 and I have a 10ghz 18 core haha.

The probability that AMD releases a new CPU (I suppose you mean FX-style) is something as 0.01% or so. FX-line is extended towards 2015 and doesn't receive Steamroller or excavator updates.

Post-excavator uarch will be 99.9% an APU, probably with up to 8 CPU cores inside.

 

[de5_Roy]

When will AMD release new CPU and I mean CPU not APU?

other than the die-harvested ones with igpu fused off... can't say... amd is mum about future cpu releases.

Cant find out much or are they waiting till it gets closer to ddr4 release?

carrizo apu/soc will have ddr4 support.

That would be a nice jump but just some good info on AMD new CPU as in release, power, consumption price new tech would be great.

amd has nothing new to show off this year (that i know of). they are limited by resources, market condition, foundry(s) etc.

 

[bigal80ak]

Thanks guys for the info its a shame but thanks.

 

[8350rocks]

Thanks guys for the info its a shame but thanks.

All I can say is, I have word it is coming...just no definite time table.

 

[colinp]

Yaay.

A new chip is coming.

Can't tell you if it'll be 2015, 2016 or beyond.

Can't tell you if it'll be AM3+, FM2+ or something else.

Can't tell you if it'll have L3 and / or integrated graphics.

Can't tell you if it's Excavator, Big Cat or something else.

Can't tell you if it'll be quad, octo or deca core or module.

Can't tell you if it'll use DDR3, DDR4 or something else.

Can't tell you the TDP either.

But hold onto your hats, don't let your socks get blown off, something's coming for sure.

 

[de5_Roy]

^^ bob dylan?

 

[colinp]

^^ bob dylan?

It's from one of his lesser known tracks, called, "AMD Roadmap Blues".

In the music video, Dylan is in some run down alleyway in a business park in Austin, flipping through a series of about 50 boards, all saying the same thing:

All I can say is, I have word it is coming...just no definite time table

 

[juanrga]

Yaay.

A new chip is coming.

Can't tell you if it'll be 2015, 2016 or beyond.

Can't tell you if it'll be AM3+, FM2+ or something else.

Can't tell you if it'll have L3 and / or integrated graphics.

Can't tell you if it's Excavator, Big Cat or something else.

Can't tell you if it'll be quad, octo or deca core or module.

Can't tell you if it'll use DDR3, DDR4 or something else.

Can't tell you the TDP either.

But hold onto your hats, don't let your socks get blown off, something's coming for sure.

Can't tell you if it'll use FD-SOI, Finfet on SOI or something else.

 

[palladin9479]

I'm kinda disappointed AMD withheld the A8-7600 release for so long. It should be out sometime this summer or August but seeing as they had chips in reviewers hands a few months ago, this is definitely on purpose. The A8-7600 was insanely good for it's price point, coupled with DDR-2133 memory it give monster performance for a really cheap cost / energy usage.

 

[juanrga]

I recall some people on forums claiming that AMD was going to release a 8-core Steamroller FX, because a "friend at AMD said so". I recall some people recommending to purchase AM3+ platform because it could be upgraded to FX-Steamroller. But this did never happen. This is from a pair of years ago:

OBR’s the news breaker who acquired some internal documents of AMD which show that the company would discontinue its performance processors for desktop market after launch of Vishera. From them, AMD would focus more towards the development of its APU’s (Accelerated Processing Units) which are doing quite well against competitor Intel.

AMD’s performance offering such as Bulldozer didn’t do well against Intel’s Sandy Bridge and same is being said about the upcoming Piledriver architecture which doesn’t has sufficient IPC improvements to tackle Haswell or even the recently launched Ivy Bridge processors. The last of performance processors we would see from AMD would be the FX 8200/8300 series after which the series would be discontinued.

We still can’t confirm how legitimate the source is since there has been no word on this by AMD themselves but for now it seems as if Steamroller and Excavator are just names to be remembered in the performance market, their core architecture would still be used in APU development. On AMD’s APU’s forefront, we would be looking at Trinity and later on Kaveri which would be the 3rd generation APU based on 28nm steamroller architecture featuring new HSA improvements and Sea Islands IGP chipsets

Guess what, everything said in the internal documents was right. Steamroller and Excavator are used for Kaveri and Carrizo APUs. In rigor the FX-8350 was not the last of the series, because AMD in last minute decided to launch the FX-9000 series, but we can consider those as overclocked FX-8350.

What we know today? We know that AMD is pushing APUs and HSA for the future

http://www.arnnet.com.au/article/542937/hsa_maximises_cpu_gpu_combination_amd_/

We also know which is the internal codename of the Carrizo APU succesor (Basilisk), what socket will use...

We also know that AMD is working in stacked RAM

Spoiler

 

[jimmysmitty]

AMD tweet: "Something SMALL is coming and it’s BIG news! Can you guess what it is? Click for a hint!"

Mobile Kaveri I guess

Considering how parts of the CPU are not "bright" I would say it is some UMD CPU. Maybe it has the ability to turn off certain parts like Intels Atom does when it is not used and that is why certain parts are dim?

Game consoles spur AMD's x86 processor market share
http://www.pcworld.com/article/2148620/game-consoles-spur-amds-x86-processor-market-share.html

The latest Project CARS trailer looks spectacular
http://techreport.com/news/26374/the-latest-project-cars-trailer-looks-spectacular
windows version will support mantle api.

catalyst 14.4 out of release candidate
http://www.techpowerup.com/200230/amd-catalyst-14-4-whql-released.html

Wait..... how do console sales factor into PC sales? I know that they use PC parts but you cannot use it like a PC and therefore it is a specialized PC or better yet a console.

This is like taking Android into account as OS market share yet Android is a watered down meh OS. My S4 just got Kit Kat and overall it feels just like 4.3 Jelly bean. Maybe a few differences in looks but still the same thing.

 

[anxiousinfusion]

What if you want your i5 and you want to upgrade your GTX 650? But you can't because it's all on one dye. Now you have to buy a whole new APU and possibly upgrade your motherboard. It makes no sense.

As a user, I prefer the idea of upgrading both my CPU and GPU with a single component replacement. Two for one, how cool is that? I think you're just seeing the glass as half-empty.

The evolution of computing has always been towards integration. Do you know that some years ago the floating point unit was an external coprocessor attached to a special socket?

I second this. Was it here that I did a write up on the trend of integration? We've seen it happen to everything, the controller boards for hard drives... dedicated to on-board audio. Mainstream integration of the GPU is only the next logical step before we see the next component merger which I personally believe will be between memory and storage.

I wish AMD would just go ahead and kill AM3+.

Make an FX 4 module CPU for FM2+.

They already have (kind of)! http://vr-zone.com/articles/amd-fx-670k-appears-wild/74597.html

 

[palladin9479]

As a user, I prefer the idea of upgrading both my CPU and GPU with a single component replacement. Two for one, how cool is that? I think you're just seeing the glass as half-empty.

I second this. Was it here that I did a write up on the trend of integration? We've seen it happen to everything, the controller boards for hard drives... dedicated to on-board audio. Mainstream integration of the GPU is only the next logical step before we see the next component merger which I personally believe will be between memory and storage.

And physics dictates that you'll get less then half the ability of both. Die sized can only get so big and there is a limit to the amount of power you can put into them. Two chips will always have a higher size and power ceiling then a single chip. Above I gave a break down of the performance characteristics of the most powerful APU vs two discrete solutions, especially the external GPU. Also juan was horribly incorrect in his statement, the Xeon Phi is nothing but a giant array of external SIMD FPUs. The same thing with the NVidia and AMD dGPU's, they are just giant external vector coprocessors. The internal FPU's on both the Intel and AMD CPU's are perfectly fine at running the code required to render 3D images, that was the entire purpose behind MMX, 3DNOW and SSE. External solutions are so much faster they make the SIMD FPU look like a slide show in comparison.

BTW using the SIMD FPU coprocessor to render images is known as software rendering and has been available for decades. It pales in comparison to hardware assisted rendering, which is really just a fancy way to say that the code required for the calculations have been embedded inside powerful external SIMD FPU coprocessors with access to incredibly high speed dedicated memory. We don't use the FPU to render 3D images because dedicated SIMD renders are so much faster. We see a similar relationship between APU's and dGPU's. The APU's can do the work but the dGPU's do it much much faster. So the question then becomes, is the APU fast enough for current software? How fast is current software growing and how is it's requirements expanding vs current technology growth? As someone above mentioned, we are a long long way away from R9 290X + 4790K performance inside a 300mm^2 100W chip, and by the time we get that we will have external GPU's that are 4~5x faster then the R9 290x and who knows what kinds of CPU's we will be using.

 

[blackkstar]

Is there some sort of way to use multiple dies in a single package? Ergo you make a big GPU and a big CPU and put the two together?

I do not think a 90% GPU APU would be very good. AMD's own documentation states that both CPU and GPU are required for HSA to work. 90% GPU APU is throwing out almost all latency computer for throughput.

Not everything will be HSA. As I have said in the past, something like a 90% GPU APU would leave you with something that absolutely sucked at traditional CPU tasks. And no matter what you do, you're going to have legacy software.