[Update] Nvidia Releases Volta-Based Titan 5 For Pc

Volta comes to desktop PCs equally the gilded Titan V for $3000.

Nvidia simply announced it's newest Titan graphics menu at the NIPS 2017 conference, the Titan V. NIPS stands for Neural Information Processing Systems, in addition to the focus is on AI in addition to deep learning. Not surprisingly, the Titan V volition create rattling good inwards those areas, in addition to spell it's ostensibly a menu that tin locomote used for gaming purposes, I don't retrieve that many gamers are going to locomote willing to purpose amongst a cool $3,000 for the privilege.
That's to a greater extent than than twice the toll of the electrical current champion, the Titan Xp—what could mayhap warrant such an expense? Simple: lots in addition to lots of compute power, cheers to the Volta GV100 architecture's Tensor cores, in addition to backed yesteryear 12GB of HBM2 memory. Here are the total specifications for the novel behemoth, amongst the Titan Xp in addition to GTX 1080 Ti provided for contrast.

Even ignoring the deep learning Tensor cores, this is a fauna of a graphics card. With 5120 active CUDA cores, this yesteryear far the highest performing graphics chip from Nvidia—and at that spot are yet iv disabled SMM clusters inwards this implementation of the GV100 architecture. Clockspeed is slightly lower than on the Titan Xp, in addition to quite a flake lower than the 1080 Ti, but the added cores to a greater extent than than brand upwardly for that.
In raw numbers, the GPU share of the Titan V tin create 14.9 TFLOPS of single-precision FP32 computations, which is 31 per centum to a greater extent than than the Titan Xp. Though create banking concern complaint that the reported boost clocks are conservative estimates—the 1080 Ti for lawsuit runs at closer to 1620MHz inwards close of my gaming benchmarks.
This is too the commencement pseudo-consumer graphics menu from Nvidia to utilize HBM2 memory, but entirely 3 of the potential iv HBM2 stacks on GV100 are active. With a retention clockspeed of 1.7Gbps in addition to a 3072-bit retention interface, that's practiced for 653GB/s of bandwidth.

But the existent ability of the Titan V, at to the lowest degree insofar equally deep learning in addition to AI calculations are concerned, comes from the 640 Tensor cores. There are 8 special-purpose Tensor cores per SMM cluster, in addition to each Tensor substance tin perform a 4x4 FMA (fused multiply add) functioning per cycle. That's 64 FLOPS for the multiply, in addition to some other 64 FLOPS for the addition, in addition to hence 128 FLOPS total. Multiply that yesteryear 640, in addition to it looks similar the Tensor cores run at a slightly lower clockspeed, equally Nvidia reports a total of 110 TFLOPS—that would locomote to a greater extent than or less 1340MHz, though Nvidia powerfulness locomote including other factors hither equally well.
Utilizing the Tensor cores volition require custom code—it's non something that you'll at i time create goodness from inwards games—so this is definitely non intended to locomote the ultimate gaming graphics menu for side yesteryear side year. In fact, I suspect we'll encounter something similar a GV102 substance at some betoken that completely omits the Tensor cores, perhaps fifty-fifty sticking amongst GDDR5X or GDDR6 retention instead of HBM2, but that it could locomote some fourth dimension earlier such a production materializes.
Architecturally, Nvidia has too stated inwards the yesteryear that Volta isn't simply Pascal amongst the improver of Tensor cores. The architecture has been changed inwards damage of thread scheduling in addition to execution, retention controllers, the pedagogy set, the substance layout, in addition to more. But despite the massive popular off size in addition to gobs of computational power, it's impressive that the Titan V remains a 250W part. That bodes good for the truthful consumer Volta parts that we'll eventually encounter some fourth dimension inwards 2018.
Or if you've got $3,000 burning a hole inwards your wallet, y'all tin buy the Titan V on Nvidia's site correct now. Hell, why non purchase several in addition to slap them inwards a arrangement for some serious computational power? Just don't excogitation on SLI, equally at that spot are no SLI connectors. Not that you'll necessitate those when you're edifice an AI to accept over the world.

Thanks PC GAMER