Now that the initial hype of the GTX 1080 launch has died down, I thought it was time to do a preliminary editorial on the real Pascal flagship: the GTX 1080 Ti. We haven't published anything concrete on the Pascal flagship so far so this would serve as a baseline for future reports. The existence of the pascal chip powering this particular graphics card has been speculated upon but nothing had come to light so far. I am talking of course about the "Gamer" version of the professional GP100 core: the GP102.
*Not a real Nvidia logo. Feature image @Wccftech
The Nvidia GTX 1080 Ti Editorial - Powered by the GP102 GPU
Without further ado, I would like to formally acknowledge the existence of the GP102 core. It is worth pointing out that there have been no confirmed reports or leaks regarding this particular chip. No Zauba entries, no die shots yet. But I have heard enough from sources familiar with these matters to say with a reasonable amount of certainty that the core exists. Not to mention that it makes a lot of sense once you delve into it. This will be the core that will be present in not only the new TITAN card but the GTX 1080 Ti as well. Notice the fact however, that this post is tagged as an editorial, so everything mentioned here is based upon my personal insights as opposed to established facts (and therefore have a non-zero probability of being off the mark). Let us start with an introduction of the GP100 (or P100 as it is marketed) core.
The GTX 1080 Ti will be powered by the GP102, which would basically be the FP32 focused cousin of the GP100. On paper, the GP100 had a total of 60 SMs or Streaming Multiprocessors. Each SM had a 2:1 ratio of FP32 cores to FP64 cores. This effectively meant that you were looking at 3840 CUDA Cores on the FP32 side of things and 1920 on the FP64 side. For a total of 5760 cores, the GP100 clocks in at just 610mm². Of course, as the more tech savvy would remember, the P100 Accelerator didn't actually use the full GP100 die. It used one with 56 SMs for a total of 3584 FP32 Cores and 1792 FP64 Cores. The reasons for the lower amount of SM can only be attributed to bad yields (which are expected this early in a node and at this large a size) which is also one more reason why we should not expect the GP100 to power the GTX 1080 Ti (not a full one at any rate).
Of course, unlike the P100 accelerator, a gaming graphics card like the GTX 1080 Ti doesn't actually need the FP64 cores. So there is no point in wasting valuable die space in hosting DP units. It is with this that we can begin the discussion of what we can expect from the GP102. Keep in mind however, that accelerators like the P100 do not have ROPs - so we are not really looking at a linear pay-off. There is also the fact that the GP102 will almost certainly not be as big as the GP100. Our usual sources have been very tight lipped about the this particular die but they did state that it would be "exactly half way" between the GP104 and the GP100.
Since the GP100 and GP104 are 610mm² and 314mm² respectively, the GP102 should be in the ballpark of 462mm² to 478mm². A chip of that size should be able to host the full compliment of the GP100 FP32 cluster, or in other words 3840 FP32 CUDA cores. In any case, the amount of CUDA cores present on the die will be somewhere around this range (give or take a few SMs). According to our estimates, with that core count you are looking at power consumption within the range of 270W with GDDR5X. If Nvidia shifts to HBM2 for the GTX 1080 Ti, the consumption should fall within the power budget of 250W. According to our sources, HBM2 will start taping out in the third quarter of this year, so it is unclear whether Nvidia will opt to go with the same.
A GP102 powered GTX 1080 Ti (or a new TITAN branded card for that matter) at the ~470mm² die size will be able to conform to the 250W power limit and offer some pretty impressive performance numbers. With a core count of 3840, you are looking at 50% over the raw performance of the GTX 1080 (or roughly equivalent to a GTX 1080 SLI) which would make this a solid 4K/60 capable graphics card. I do not expect Nvidia to go for a full fledged die like the GP100 for various reasons (which I have detailed in my discussion below). Needless to say, the yields of a 610mm² large GPU will not be ideal enough for mainstream purposes among other reasons. So there you go folks, the new TITAN card or the GTX 1080 Ti (or whatever Nvidia will call it) will be powered by the GP102 core. As far as the when is concerned, we are still fuzzy on the details on that count, and anything is fair game. I have touched upon some of my opinions in the complete GTX 1080 Ti discussion below, for those interested:
Nvidia GTX 1080 Ti 'Expected' Specifications Comparision
|NVIDIA Graphics Card||Tesla P100||GTX 1080 Ti*||GTX 1080|
|Transistors||15.3 Billion||10.8 Billion||7.2 Billion|
|GPU Die Size||610 mm2||471mm²||314mm²|
|CUDA Cores Per SM||64||64||64|
|FP32 CUDA Cores (Total)||3584||3072-3584 (TBC)||2560|
|FP64 CUDA Cores / SM||32||TBD||2|
|FP64 CUDA Cores / GPU||1792||TBD||80|
|Memory Interface||4096-bit HBM2||GDDR5X||256-bit GDDR5X|
|Memory Size||16 / 32 GB HBM2||10 GB GDDR5X||8 GB GDDR5X|