Predict: The Next Generation Console Tech

Status
Not open for further replies.
CPUs have taken a very conservative core count increase since 2005 (dual cores were available in 2005 and most new systems are dual or quad core) and their flops are still lagging behind the peak of 2005 consoles.

Strictly speaking Xenon was equalled (in peak GFLOP terms) in January 2007 with the launch of the Q6600. Both CPU's are theoretically capable of pushing 76.8 GFLOPs.

Cell wasn't exceeded for another 4 years until Sandybridge launched in January 2011 with AVX. At that time a 2500K was capable of 211.2 GFLOPS peak vs 204.8 in the PS3 implementation of Cell.

The highest end Sandybridge-E today can push 316.8 GFLOPS.
 
Thanks pjb, I stand corrected. The lack progressive core count and flops, with a focus for IPC as well as adding GPU, have had the PC CPUs lag IMO. The number of 2 core and 4 core CPUs is dominant; I guess the problem is chicken/egg. You need good examples of needing many cores, but until software asks for it Intel/AMD have little prompting them to offer such. But we won't get software pushing such until there is hardware to target it.
 
Yes it does seem CPU performance has been significantly slower to progress than GPU. But then I guess that's down to the fact that CPU code isn't particularly parallelizable and thus it can't just scale with moores law like GPU's. CPU's have to get smarter rather than wider and that's seemingly a lot more difficult.

Hell, if CPU code were parallelizable then we'd probably just have 1000 core Pentiums now ;)
 
Yes it does seem CPU performance has been significantly slower to progress than GPU. But then I guess that's down to the fact that CPU code isn't particularly parallelizable and thus it can't just scale with moores law like GPU's. CPU's have to get smarter rather than wider and that's seemingly a lot more difficult.

Hell, if CPU code were parallelizable then we'd probably just have 1000 core Pentiums now ;)

While I don't doubt making gaming engines and whatnot is very difficult across many threads we already are seeing developers turn the corner--maybe not as much on the PC but definitely on consoles (ironic, eh?) There really is no running away from using a job model across many cores. This was known way back in 2004 when the consoles were being discussed: serial CPU performance hate hit a wall. The pace of serial performance gains has been at a crawl for almost a decade. It is sad when a new process and design and inflated silicone footprints bring excitement for 15% gains in IPC.

The multicore future has been here for a long time. Like it or not. As someone else mentioned memory is also a big part of this (look how slow that has moved).

So I say CPUs better get wider, faster, or we are going to continue be stuck in the glut. But I think Intel wants this. They have won over the serial performance war. AMD committed to APUs when they bought ATI so there really is no pressure. But the only way to find best practices, develop new languages and tools that are multicore friendly, and start advancing on this front is getting the hardware out...

Or for the consoles to push forward with core-heavy designs and start pushing software in a direction that utilizes the hardware. Once there is software that can utilize those kinds of resources there will be more pressure for Intel to offer such. As it stands they have no interest in providing an 8 or 12 core desktop chip when it *undervalues their lucrative server market.* But this is the same market that has shipped tens of millions, maybe hundreds, of quad core PCs with with IGP class GPUs.

I am not confident it will work out of the box, but the concept of AMD's HSA should also help longterm--using the shaders on an APU as a giant SIMD (or however they are going to market it) will offer a chance to offload some major work and things that can benefit from a much wider chip design to get a big enough bump in a resource pool (looking at 4x bump in flops in the first HSA models over the peak flops of a high end CPU, much more from the pedestrian models) should also encourage this. Of course the PC market is so fragmented with 90% of sales at the much lower end it will take time. A lot of time. Unless someone in the console business decides there is more to be done with CPUs. Or, the fruit may be so high hanging and the ROI for consumers so low they just say meh, why should we lose our shirt when the CPU industry just doesn't care?
 
hardware is getting out, just relatively slowly. we have GCN, Kepler, Maxwell that attack the problem from the GPU side and Intel's Larrabee follow-up from the CPU side.

it's not the big thing right now, because it's hard not only from the software programming side. you can buy "tile processors" from a few vendors where 64 small cores or some other numbers are arranged in a grid and pass data to each other.
so it says, "64 cores in 20 watts!" or something.
but as you may imagine, the cores in the middle of the grid are starved for data.

so rather than go through this, and design a PCB or order an evaluation board, and hire PHDs to program it, you can just put a sandy bridge on a $50 motherboard with everything (ethernet, sata, serial interface, usb etc.)
 
If MS or Sony were looking at NV's and AMD's current architecture for GPGPU it seems AMD has a pretty wide lead right now with GCN over Kepler. Kaotik posted this in another thread and noted, ""suurempi on parempi" means bigger bar is better, while "pienempi on parempi" means smalle bar is better".

So, say, if Sony's PS4 had an APU like the A8-3850 and a discreet GPU and wanted to offer up the APU as an evolutionary concept to Cell (you can offload some physics as well as post process and pre-process tasks to the CPU's GPU) it would seem right now AMD has the architecture most relevant. Oh, and they also have a CPU, relatively a pretty good one per core, to go along with such.
 
Saw this on GAF concerning the Wii U:

Originally Posted by IdeaMan:
A bit of info concerning the CPU


After discussing with a member here about the specificities of the Wii U CPU, and searching on it for a particularity that could constitute an hindrance for some of middleware companies if they don't optimize their product for, i was able to find some vague infos, i can't go too much into details, and it's from a totally different source that the ones i was referring too before. Nothing groundbreaking, but it's a confirmation at least.

- The instructions set of the CPU, their "routine", is definitively even more advanced and targeted for gaming that the VMX128 included in the Xbox360 Xenon.
- Expect a noticeable lowering of the caches latency (fewer cycles), i assume L1 & L2.

It definitively reinforce the fact that Wii U CPU is more modern, efficient, and more powerful overall than the Xbox360 one. You can even ditch the scenario where the Wii U could be a Xenon+10% + not so much powerful GPU, with 1GB of ram, the CPU really seems to constitute an advancement.

http://www.neogaf.com/forum/showthread.php?t=469930&page=32
 
If MS or Sony were looking at NV's and AMD's current architecture for GPGPU it seems AMD has a pretty wide lead right now with GCN over Kepler. Kaotik posted this in another thread and noted, ""suurempi on parempi" means bigger bar is better, while "pienempi on parempi" means smalle bar is better".

that's interesting ; nvidia might still have an edge with software and the fact you write in CUDA's latest version.

but there's a problem with these benchmarks, you have to rewrite the applications for the GTX 680 for them to make sense.
it's not as with SMP CPUs where you can hit a compile switch and have it be mostly optimized.
 
So, say, if Sony's PS4 had an APU like the A8-3850 and a discreet GPU and wanted to offer up the APU as an evolutionary concept to Cell (you can offload some physics as well as post process and pre-process tasks to the CPU's GPU) it would seem right now AMD has the architecture most relevant. Oh, and they also have a CPU, relatively a pretty good one per core, to go along with such.

APU + GPU with some sort of asymmetrical CrossFire is something what came to my mind when I first heard about APU in PS4. One can use GPU part in APU for GPGPU stuff or for pure GPU stuff cross-firing with discreet GPU.

I also thought GPU part in APU can be used to emulate SPU's but that is unlikely as someone in this forum said.
 
1TFLOPs is a lot on a GPU? I guess it is perspective. Blaz, I am going to disagree for these reasons.

From the perspective of a 4x increase in flops from 2005 to 2013 in a potential console -- 8 years? No, I think that this is the opposite of incredible. That is pathetic considering (a) the chips have moved from a 90nm process to a 28nm process (a 10x increase smallest feature density), (b) frequencies at a set TDP have increased, and (c) shaders have proportionately increased faster than other units, hence flops have ballooned. How is a 4x increase in FLOPs incredible when the process allows at the same TDP, safely, in the 10-15x range, if not more?

From the market perspective 1TFLOP GPUs is going to be a 5 year old affair come 2013. 4850, 4870, and 4890 (256mm^2 on 55nm) all broke 1TFLOPs in the Summer of 2008 an the 4770 (137mm^2 on 40nm) was in spitting distance at 40nm (960GFLOPs). How is it incredible to get performance in 2013 that the PC had 5 years earlier in 2008?

From a market placement perspective I don't see how it is incredible either. 7770 at $159 MSRP is well over 1GFLOPs and the $109 MSRP 7750 is at spitting distance. These are the lowest end South Island series you can get (tyhe 7670 is an OEM product) or the 6670 from North Island breaks 1TFLOPs. Basically these are your below-mid range GPUs... in 2011. Oh, wait, this is a re-badge of the 5750 -- from 2009 which was almost 1/3rd of the single GPU enthusiest FLOPs at the time. So moving forward 4 years to 2013 I don't see how what was midrange in 2009 and has dropped to low end in 2011 can in 2013 be anything but, "packed in bottom of the barrel" in 2013. How are GPUs that cost $109 MSRP in early 2012 (probably cost less than $70 for the GPU, Memory, PCB, fan, output, etc when you consider the retailer cut, assembler and distributor cut, and then AMD's cut) going to be incredible almost 2 years later in 2013?

From looking what AMD is packing into APUs -- which is a GPU that has to share space with HOT CPUs on-- it has a constrained footprint budget and TDP, we are hearing APUs in 2013 will be hitting 800GFLOPs. Again, how is 1TFLOPs in 2013 impressive when a GPU sharing space and power limitations with 4 ("8") to 6 ("10") AMD CPUs at 800FLOPs incredible?

The 7970 in early 2012 hit 3.7GFLOPs. How is 18+ months later 1 GFLOPs going to be incredible when single chip PC GPUs are going to be marching toward 5GFLOPs?

From the pure technological perspective of where we were and where we are, yeah, it is impressive. Especially when you see that desktop CPUs have taken a very conservative core count increase since 2005 (dual cores were available in 2005 and most new systems are dual or quad core) and their flops are still lagging behind the peak of 2005 consoles. That said, they are also packing in GPUs now to go along with discreet GPUs. The technology is cool but in big picture perspective what I find incredible is that we think what amounts to entry level hardware with budgets in mm^2 WELL BELOW the past generation is nothing but a big step BACKWARDS.

That is what I think is incredible :p

Excellent post. I could not express better, because as much as we imagine the efficiency of the APU on the most favorable scenario these initial reports of alleged SDKs with 1/1.25Tflop levels are disappointing from all points of view (developers, gamer and even competitors will not have much leverage and accelerate their technology ...).

(We're talking 1/1.25Tflop shaders for 2013/2014! Even with sustained maximum efficiency would be imagined ... this is very few for consoles that should last at least 5 years)
 
According to Timothy Lotte (designer of FXAA et al), he expects next gen to be about 6 times as powerful as 360 is. This is just a tad faster than todays very high end laptop GPUs. I find this quite a small leap, considering a 7+ year leap in technology.
 
According to Timothy Lotte (designer of FXAA et al), he expects next gen to be about 6 times as powerful as 360 is. This is just a tad faster than todays very high end laptop GPUs. I find this quite a small leap, considering a 7+ year leap in technology.
Because of the power wall and issues with process shrinks, we can't expect the technology curve to carry on to the exponential degree we are used to.
 
Six times the performance of Xenos is roughly a 7850 (240 GFLOPS vs. 1.8 TFLOPS) . Sounds reasonable to me.

Cheers

Indeed. TDP sounds reasonable too given it will likely be at least partially based on sea islands if it debuts holiday 2014. Peak 7850 TDP is around 100W, leaving 80 to 90W for CPU and the remainder for other peripherals in a 200W budget.
 
I was more thinking of "6x across the board". Not something as meaningless as flops. But I might've been off anyways, as I was thinking of my 6870 as "at least 6x the power".
 
Status
Not open for further replies.
Back
Top