Unlike the previous Titan it looks like it will be a cut down Pro Tesla card with the full floating point capabilities. And not based on the gaming variants we have been accustomed too. I guess previously this would have been marketed as a quadro card. It's not aimed at most of us so we will have to wait for the real consumer level volta cards in the future (which will likely perform significantly better at regular tasks too).
It will be nice to see reviews.
I expect superb performance, desire, rage, jealousy, awe and wonder.
Pretty much this:
https://www.anandtech.com/show/1213...dia-titan-v-video-card-gv100-for-3000-dollars
Moving on and diving into the numbers, Titan V features 80 streaming multiprocessors (SMs) and 5120 CUDA cores, the same amount as its Tesla V100 siblings. The differences come with the memory and ROPs. In what's clearly a salvage part for NVIDIA, one of the card's 4 memory partitions has been cut, leaving Titan V with 12GB of HBM2 attached via a 3072-bit memory bus. As each memory controller is associated with a ROP partition and 768 KB of L2 cache, this in turn brings L2 down to 4.5 MB, as well as cutting down the ROP count.
In terms of clockspeeds, the HBM2 has been downclocked slightly to 1.7GHz, while the 1455MHz boost clock actually matches the 300W SXM2 variant of the Tesla V100, though that accelerator is passively cooled. Notably, the number of tensor cores have not been touched, though the official 110 DL TFLOPS rating is lower than the 1370MHz PCIe Tesla V100, as it would appear that NVIDIA is using a clockspeed lower than their boost clock in these calculations.
As mentioned earlier, NVIDIA is unsurprisingly pushing this as a compute accelerator card, especially considering that Titan V features tensor cores and keeps the TITAN branding as opposed to GeForce TITAN. But there are those of us who know better than to assume people won’t drop $3000 to use the latest Titan card for gaming, and while gaming is not the primary (or even secondary) focus of the card, you also won't see NVIDIA denying it. In that sense the Titan V is going to be treated as a jack-of-all-trades card by the company.
To that end, no gaming performance information has been disclosed, but NVIDIA has confirmed that the card uses the standard GeForce driver stack. Now whether those drivers have actually been optimized for the GV100 is another matter entirely; Volta is a new architecture, markedly so at times. Speaklng solely off the cuff here, for graphics workloads the card has more resources than the Titan Xp in almost every meaningful metric, but it's also a smaller difference on paper than you might think.
As for NVIDIA's intended market of compute and AI users, the Titan V will be supported by NVIDIA GPU Cloud, which includes a number of deep learning frameworks and HPC-related tools.
The card is being pushed as a compute card,and more importantly,to make sure they keep the card within TDP for more consumer workloads have cut the memory bandwidth and dropped the clockspeeds,and gone with a "wide and slow" approach to drop power.
Will it be faster than a GP102 in gaming,probably,but its massive in comparison,so I expect its more down to brute force.
To keep the power requirements reasonable,they have thrown another 6 billion transistors at the problem over big Pascal and changed the internal organisation of the GPU.
Now look at the consumer Titan Xp in comparison in terms of memory bandwidth,TFLOPs,etc. Remember HBM2,saves power and GPU die area.
For a 73% increase in die area,and a 76% increase in transistors,FP32 performance has gone up by a whole 15% in FP32 performance and that is on a "new node".
This is clearly a card and GPU made for real work not gaming,as it has massive FP16 and FP64 throughput.
Volta is made on TSMC 12NM which is an improved version of TSMC 16NM,which is lower leakage
https://www.electronicsweekly.com/news/business/tsmc-introduces-12nm-half-node-2016-11/
Regarding gaming Volta cards,I am not so sure,due to this:
https://www.tweaktown.com/news/59816/nvidias-next-gen-geforce-teased-ampere-unveil-2018/index.html
Ampere is probably a new FP32 line which will make up the basis of the sucessor to Pascal,but on the improved node.
I expect we will see a greater increase in FP32,but also the die sizes will be smaller.
It fits with Nvidia now making more dedicated graphics card lines.
