Pascal’s Architecture: What Follows Maxwell

With the launch of a new generation of GPUs we’ll start things off where we always do: the architecture.

Discrete GPUs occupy an interesting space when it comes to the relationship between architecture and manufacturing processes. Whereas CPUs have architecture and manufacturing process decoupled – leading to Intel’s aptly named (former) tick-tock design methodology – GPUs have aligned architectures with manufacturing processes, with a new architecture premiering alongside a new process. Or rather, GPU traditionally did. Maxwell threw a necessary spanner into all of this, and in its own way Pascal follows this break from tradition.

As the follow-up to their Kepler architecture, with Maxwell NVIDIA introduced a significantly altered architecture, one that broke a lot of assumptions Kepler earlier made and in the process vaulted NVIDIA far forward on energy efficiency. What made Maxwell especially important from a development perspective is that it came not on a new manufacturing process, but rather on the same 28nm process used for Kepler two years earlier, and this is something NVIDIA had never done before. With the 20nm planar process proving unsuitable for GPUs and only barely suitable for SoCs – the leakage from planar transistors this small was just too high – NVIDIA had to go forward with 28nm for another two years. It would come down to their architecture team to make the best of the situation and come up with a way to bring a generational increase in performance without the traditional process node shrink.

Now in 2016 we finally have new manufacturing nodes with the 14nm/16nm FinFET processes, giving GPU manufacturers a long-awaited (and much needed) opportunity to bring down power consumption and reduce chip size through improved manufacturing technology. The fact that it has taken an extra two years to get here, and what NVIDIA did in the interim with Maxwell, has opened up a lot of questions about what would follow for NVIDIA. The GPU development process is not so binary or straightforward that NVIDIA designed Maxwell solely because they were going to be stuck on the 28nm process – NVIDIA would have done Maxwell either way – but it certainly was good timing to have such a major architectural update fall when it did.

So how does NVIDIA follow-up on Maxwell then? The answer comes in Pascal, NVIDIA’s first architecture for the FinFET generation. Designed to be built on TSMC’s 16nm process, Pascal is the latest and the greatest, and like every architecture before it is intended to further push the envelope on GPU performance, and ultimately push the envelope on the true bottleneck for GPU performance, energy efficiency.

HPC vs. Consumer: Divergence

Pascal is an architecture that I’m not sure has any real parallel on a historical basis. And a big part of that is because to different groups within NVIDIA, Pascal means different things and brings different things, despite the shared architecture. On the one side is the consumer market, which is looking for a faster still successor to what Maxwell delivered in 2014 and 2015. Meanwhile on the high performance compute side, Pascal is the long-awaited update to the Kepler architecture (Maxwell never had an HPC part), combining the lessons of Maxwell with the specific needs of the HPC market.

The result is that there’s an interesting divergence going on between the HPC side and its GP100 GPU, and the consumer side and the GP104 GPU underlying GTX 1080. Even as far back as Fermi there was a distinct line separating HPC-class GPUs (GF100) from consumer/general compute GPUs (GF104), but with Pascal this divergence is wider than ever before. Ultimately the HPC market and GP100 is beyond the scope of this article and I’ll pick it up in detail another time, but because NVIDIA announced GP100 before GP104, it does require a bit of addressing to help sort out what’s going on and what NVIDA’s design goals were with GP104.

Pascal as an overarching architecture contains a number of new features, however not all of those features are present in all SKUs. If you were to draw a Venn diagram of Pascal, what you would find is that the largest collection of features are found in GP100, whereas GP104, like the previous Maxwell architecture before it, is stripped down for speed and efficiency. As a result while GP100 has some notable feature/design elements for HPC – things such faster FP64 & FP16 performance, ECC, and significantly greater amounts of shared memory and register file capacity per CUDA core – these elements aren’t present in GP104 (and presumably, future Pascal consumer-focused GPUs).

Ultimately what we’re seeing in this divergence is a greater level of customization between NVIDIA’s HPC and consumer markets. The HPC side of NVIDIA is finally growing up, and it’s growing fast. The long term plan at NVIDIA has been to push GPU technology beyond consumer and professional graphics, and while it has taken years longer than NVIDIA originally wanted, thanks in big part to success in the deep learning market, NVIDIA is finally achieving their goals.

This means that although GP100 is a fully graphics capable GPU, it doesn’t necessarily have to be put into video cards to make sense for NVIDIA to manufacture, and as a result NVIDIA can make it even more compute focused than prior-generation parts like GK110 and GF110. And that in turn means that although this divergence is driven by the needs of the HPC market – what features need to be added to make a GPU more suitable for HPC use cases – from the perspective of the consumer market there is a tendency to perceive that consumer parts are falling behind. Especially with how GP100 and GP104’s SMs are differently partitioned.

This is a subject I’ll revisit in much greater detail in the future when we focus on GP100. But for now, especially for the dozen of you who’ve emailed over the past month asking about why the two are so different, the short answer is that the market needs for HPC are different from graphics, and the difference in how GP100 and GP104 are partitioned reflect this. GP100 and GP104 are both unequivocally Pascal, but GP100 gets smaller SM partitions in order to increase the number of registers and the amount of shared memory available per CUDA core. Shared memory and register contention on graphics workloads isn’t nearly as great as with HPC tasks – pixel shader threads are relatively short and independent from each other – which means that while the increased ratios benefit HPC workloads, for graphics the gains would be minimal. And the costs to power and die space would, in turn, far outweigh any benefits.

The GTX 1080 & GTX 1070 Review GP104: The Heart of GTX 1080
Comments Locked


View All Comments

  • Robalov - Tuesday, July 26, 2016 - link

    Feels like it took 2 years longer than normal for this review :D
  • extide - Wednesday, July 27, 2016 - link

    The venn diagram is wrong -- for GP104 it says 1:64 speed for FP16 -- it is actually 1:1 for FP16 (ie same speed as FP32) (NOTE: GP100 has 2:1 FP16 -- meaning FP16 is twice as fast as FP32)
  • extide - Wednesday, July 27, 2016 - link

    EDIT: I might be incorrect about this actually as I have seen information claiming both .. weird.
  • mxthunder - Friday, July 29, 2016 - link

    its really driving me nuts that a 780 was used instead of a 780ti.
  • yhselp - Monday, August 8, 2016 - link

    Have I understood correctly that Pascal offers a 20% increase in memory bandwidth from delta color compression over Maxwell? As in a total average of 45% over Kepler just from color compression?
  • flexy - Sunday, September 4, 2016 - link

    Sorry, late comment. I just read about GPU Boost 3.0 and this is AWESOME. What they did, is expose what previously was only doable with bios modding - eg assigning the CLK bins different voltages. The problem with overclocking Kepler/Maxwell was NOT so much that you got stuck with the "lowest" overclock as the article says, but that simply adding a FIXED amount of clocks across the entire range of clocks, as you would do with Afterburner etc. where you simply add, say +120 to the core. What happened here is that you may be "stable" at the max overclock (CLK bin), but since you added more CLKs to EVERY clock bin, the assigned voltages (in the BIOS) for each bin might not be sufficient. Say you have CLK bin 63 which is set to 1304Mhz in a stock bios. Now you use Afterburner and add 150 Mhz, now all of a sudden this bin amounts to 1454Mhz BUT STILL at the same voltage as before, which is too low for 1454Mhz. You had to manually edit the table in the BIOS to shift clocks around, especially since not all Maxwell cards allowed adding voltage via software.
  • Ether.86 - Tuesday, November 1, 2016 - link

    Astonishing review. That's the way Anandtech should be not like the mobile section which sucks...
  • Warsun - Tuesday, January 17, 2017 - link

    Yeah looking at the bottom here.The GTX 1070 is on the same level as a single 480 4GB card.So that graph is wrong.
    Remember this is from GPU-Z based on hardware specs.No amount of configurations in the Drivers changes this.They either screwed up i am calling shenanigans.
  • marceloamaral - Thursday, April 13, 2017 - link

    Nice Ryan Smith! But, my question is, is it truly possible to share the GPU with different workloads in the P100? I've read in the NVIDIA manual that "The GPU has a time sliced scheduler to schedule work from work queues belonging to different CUDA contexts. Work launched to the compute engine from work queues belonging to different CUDA contexts cannot execute concurrently."
  • marceloamaral - Thursday, April 13, 2017 - link

    Nice Ryan Smith! But, my question is, is it truly possible to share the GPU with different workloads in the P100? I've read in the NVIDIA manual that "The GPU has a time sliced scheduler to schedule work from work queues belonging to different CUDA contexts. Work launched to the compute engine from work queues belonging to different CUDA contexts cannot execute concurrently."

Log in

Don't have an account? Sign up now