Friday, December 8th 2017

NVIDIA Announces TITAN V "Volta" Graphics Card

NVIDIA in a shock move, announced its new flagship graphics card, the TITAN V. This card implements the "Volta" GV100 graphics processor, the same one which drives the company's Tesla V100 HPC accelerator. The GV100 is a multi-chip module, with the GPU die and three HBM2 memory stacks sharing a package. The card features 12 GB of HBM2 memory across a 3072-bit wide memory interface. The GPU die has been built on the 12 nm FinFET+ process by TSMC. NVIDIA TITAN V maxes out the GV100 silicon, if not its memory interface, featuring a whopping 5,120 CUDA cores, 640 Tensor cores (specialized units that accelerate neural-net building/training). The CUDA cores are spread across 80 streaming multiprocessors (64 CUDA cores per SM), spread across 6 graphics processing clusters (GPCs). The TMU count is 320.

The GPU core is clocked at 1200 MHz, with a GPU Boost frequency of 1455 MHz, and an HBM2 memory clock of 850 MHz, translating into 652.8 GB/s memory bandwidth (1.70 Gbps stacks). The card draws power from a combination of 6-pin and 8-pin PCIe power connectors. Display outputs include three DP and one HDMI connectors. With a wallet-scorching price of USD $2,999, and available exclusively through NVIDIA store, the TITAN V is evidence that with Intel deciding to sell client-segment processors for $2,000, it was a matter of time before GPU makers seek out that price-band. At $3k, the GV100's margins are probably more than made up for.
Add your own comment

135 Comments on NVIDIA Announces TITAN V "Volta" Graphics Card

#101
R0H1T
bugNo, no, no, we have to get to the bottom of this. I mean, shall we let a company sell a niche product at whatever price they want? Where does this lead? /s
Oh sure, wait till we get Titan XV or the other abomination, Titan XVZ :rolleyes:
Posted on Reply
#102
Vayra86
cdawallLOL I am already trying to get one or four. Depends how electroneum works out for me
Buy em all, build an AI that mines faster, profit
bugNo, no, no, we have to get to the bottom of this. I mean, shall we let a company sell a niche product at whatever price they want? Where does this lead? /s
Wait, we have a say in this? Who do I write? :D
Posted on Reply
#103
rtwjunkie
PC Gaming Enthusiast
bugNo, no, no, we have to get to the bottom of this. I mean, shall we let a company sell a niche product at whatever price they want? Where does this lead? /s
Actually, yes. That's capitalism. If it doesn't sell, then they won't release future similar products at that price.
Posted on Reply
#104
bug
Vayra86Wait, we have a say in this? Who do I write? :D
Of course we do. 1 million rants on TPU forums and the price gets halved. /s
rtwjunkieActually, yes. That's capitalism. If it doesn't sell, then they won't release future similar products at that price.
Did you really miss my /s (aka "end sarcasm")?
Posted on Reply
#105
rtwjunkie
PC Gaming Enthusiast
bugOf course we do. 1 million rants on TPU forums and the price gets halved. /s



Did you really miss my /s (aka "end sarcasm")?
Sadly, it appears I did. :oops: My apologies.
Posted on Reply
#106
matar
So nVidia if we spend $3000 is there a guarantee that it can run Crysis 3 @ 4K 100% Maxed out with no less then 60FPS . Please answer nVidia crysis 3 was released in early 2013.
Posted on Reply
#107
bug
matarSo nVidia if we spend $3000 is there a guarantee that it can run Crysis 3 @ 4K 100% Maxed out with no less then 60FPS . Please answer nVidia crysis 3 was released in early 2013.
I don't even think that was a good game, who' still playing it after almost 5 years?
Posted on Reply
#108
Fluffmeister
okidnaWhat are you talking about? AMD have THIS card! The most beautiful card ever made!!!!!
If you already have a card with minuscule PCB and a humongous heatsink, it's a WIN, NO question asked!



It's funny isn't it, even big Volta has hit the market before custom Vega.

Poor Volta. :P
Posted on Reply
#109
xorbe
matarSo nVidia if we spend $3000 is there a guarantee that it can run Crysis 3 @ 4K 100% Maxed out with no less then 60FPS . Please answer nVidia crysis 3 was released in early 2013.
Wait for the consumer gaming version without the Tensor matrix compute cores. I'm still curious to see benchmarks though of course, like everyone else. I can't imagine too many takers for gaming.
Posted on Reply
#110
Xzibit
xorbeWait for the consumer gaming version without the fp64 matrix compute cores. I'm still curious to see benchmarks though of course, like everyone else. I can't imagine too many takers for gaming.
No FP64

Tensor cores are 4x4 matrix FP32 FP16 FP16 FP32

The big V100 only has 32 FP64 units same as P100. Not sure if this one has the same or not.
Posted on Reply
#111
xorbe
Hah, that's what I get from trying to simplifying from Tensor to fp64, swing and a miss! Thanks.
Posted on Reply
#113
Xzibit
FluffmeisterPretty sure it's a double precision monster:

www.anandtech.com/show/12135/nvidia-announces-nvidia-titan-v-video-card-gv100-for-3000-dollars

This is one Titan card haters can't say isn't aimed at compute workloads.

You can thank me later.
Hopefully it is. The P100 had 1/2 DP as well but that got cut down on the Titan variant as the chart shows. They cut the L cache a bit as well from the information so far.

The Quadro part is always more powerful so who knows how they are positioning the parts until full details come out.
Posted on Reply
#114
Fluffmeister
XzibitHopefully it is. The P100 had 1/2 DP as well but that got cut down on the Titan variant as the chart shows.
Well put your credit card away for now, I'm sure it will become clear in the coming days before you pull the trigger.
Posted on Reply
#115
Xzibit
FluffmeisterWell put your credit card away for now, I'm sure it will become clear in the coming days before you pull the trigger.
I was hoping you, xkm or rez will hook me up
Posted on Reply
#116
Fluffmeister
XzibitI was hoping you, xkm or rez will hook me up
Sadly, despite the common misconceptions, me and Nvidia aren't charities.
Posted on Reply
#117
R-T-B
evernessinceOh boy, if you think a $3,000 Nvidia video card is going to beat a gaggle of RX 580s in mining then you just speaking out your butt. There is no universe in which this Volta GPU will beat 15 RX 580s in mining, especially considering AMD has a large performance advantage in that category. One RX 580 alone gets around a GTX 1080s performance in mining. Also, da fuq does "marketing" have to do with video cards?

Please proof read your comments and study up on mining.
It depends on the coin actually. NVIDIA cards are much more competitive today.
Posted on Reply
#118
cdawall
where the hell are my stars
R-T-BIt depends on the coin actually. NVIDIA cards are much more competitive today.
If not better.
Posted on Reply
#119
Vya Domus
FluffmeisterPretty sure it's a double precision monster:
A compute monster in general ? Yes.

DP monster ? I'd say hardly anything to that degree ...
Posted on Reply
#120
jabbadap
Vya DomusA compute monster in general ? Yes.

DP monster ? I'd say hardly...
Tesla P100 sxm2 version had 5.3TFLops of fp64 power, if this card has unlocked fp64 it would be ~7.4TFlops of fp64.
XzibitHopefully it is. The P100 had 1/2 DP as well but that got cut down on the Titan variant as the chart shows. They cut the L cache a bit as well from the information so far.

The Quadro part is always more powerful so who knows how they are positioning the parts until full details come out.
Well unlike the pascal titans(gp102s full fp64 was 1/32), gv100 chip is capable for 1/2 fp64 so there is some hope at least.
XzibitNo FP64

Tensor cores are 4x4 matrix FP32 FP16 FP16 FP32

The big V100 only has 32 FP64 units same as P100. Not sure if this one has the same or not.
If one chip have 32 fp64 units it can do 64 single precision floats so 64 fp32. 1/2 fp64/fp32 ratio is the full fp64 hw.
Posted on Reply
#121
Fluffmeister
Vya DomusA compute monster in general ? Yes.

DP monster ? I'd say hardly anything to that degree ...
I'd say 7 TFlops was pretty good, but fair enough. /shrugs
Posted on Reply
#122
Assimilator
FluffmeisterIt's funny isn't it, even big Volta has hit the market before custom Vega.

Poor Volta. :p
This is now the second card (after 1070 Ti) that NVIDIA released but didn't need to.

AMD, on the other hand, absolutely needs custom Vega variants to be released but they're nowhere to be seen.

People bitching that NVIDIA is ripping customers off by charging $3,000 for a graphics card, would be much better served directing their ire at AMD for failing to provide anything resembling competition.
Posted on Reply
#123
Vya Domus
AssimilatorAMD for failing to provide anything resembling competition.
That has absolutely nothing to do with anything I'm afraid.

It's safe to say people should be intelligent enough too look at the feature set of this card and realize that it is not meant to go up against any other consumer product currently on the market , not even one from Nvidia's own product stack.
Posted on Reply
#125
T4C Fantasy
CPU & GPU DB Maintainer
maxed out silicon would be 84 SMs 5376 Cuda Cores
Posted on Reply
Add your own comment
Apr 23rd, 2024 21:13 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts