Thursday, December 7th 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

#26
okidna
P4-630Oh boy.....AMD what is your answer?
What 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!



Posted on Reply
#27
Rahmat Sofyan
that price . . wowww .

whats the main difference between this Titan Volta vs Tesla Volta ?

now I know that RX Vega price much cheaper from it should be and the stock very limit, HBM2 was the problem ?

or indeed volta arch way much better than pascal arch or all radeon and nvidia arch ?
Posted on Reply
#28
dj-electric
It concerns me a little how nobody is actually talking about the target demographic for this card, the deep learning market with its deep pockets
Posted on Reply
#29
medi01
P4-630Oh boy.....AMD what is your answer?
Got rid of Raja.
MarkoszThis thing has freakin Tensor cores, meant for deep neural networks.
That's a different card.
nVidia will likely make sure it's not usable for "pro" purposes (as usual).
Posted on Reply
#30
okidna
medi01That's a different card.
nVidia will likely make sure it's not usable for "pro" purposes (as usual).
Not that much different according to their claim :
Posted on Reply
#31
silentbogo
Even with that price tag this card will sell like hot sh#tcakes.
Remember, there are no Volta-based Quadro cards yet, and Tesla V100 is trending at around $8000 (as a part of a server bundle, a lot more by itself).... :roll:
Posted on Reply
#32
Vya Domus
Honestly the prices seems a little more justified this time around. The GPU die is massive and I am convinced it still costs a fortune to make since it's fully enabled, 12 GB of HBM2 isn't cheap either.
ViperXTRdoes it have async compute yet? :V
There isn't really a clear cut definition of a GPU being asynchronous compute capable or not. It's more like a gradient and Pascal was a lot better than Maxwell at it.
Posted on Reply
#33
bug
Price aside (Titan were never priced reasonably, as far the the average consumer is concerned), has anyone else noticed this needs less power than Vega64?
Posted on Reply
#34
P4-630
bughas anyone else noticed this needs less power than Vega64?
Yeah that was a reason I said:
P4-630Just 6+8 pin power!

Oh boy.....AMD what is your answer?
Posted on Reply
#35
birdie
To all the retards who think this card is overpriced: this is not a gaming card for fuck's sake. This is a professional card for AI/FP computations. It's a miracle it costs just 30% of its full featured brother.

Have people collectively gone mad or full-retarded and whenever they see anything priced above $500 from NVIDIA they think they are milking us? No one fucking forces you to buy anything from NVIDIA.

Fuck my life.
Posted on Reply
#36
Vya Domus
birdieFuck my life.
Clearly , since it's obvious you haven't got anything better to do than calling people retards for expressing their opinions.

Get out of here troll.
Posted on Reply
#37
bug
birdieTo all the retards who think this card is overpriced: this is not a gaming card for fuck's sake. This is a professional card for AI/FP computations. It's a miracle it costs just 30% of its full featured brother.

Have people collectively gone mad or full-retarded and whenever they see anything priced above $500 from NVIDIA they think they are milking us? No one fucking forces you to buy anything from NVIDIA.

Fuck my life.
It's just a childish reaction, over $500 price/performance ratio is not a factor. Just like price/performance ratio is not a factor when you buy a $100,000+ car.
That's why pointing out that it's overpriced is really just stating the obvious (underscoring in the process that the person doesn't have anything more useful to say).

Edit: This price also goes hand in hand with Mr. Hwang's statement that they're not bringing Volta to consumers yet because it's still too expensive to make. Though the usual grain of salt still applies.
Posted on Reply
#38
Frick
Fishfaced Nincompoop
@cdawall Benchmarks please. You know you want to.
Posted on Reply
#39
Manu_PT
So the gtx 2080 will cost 1000€? Nice to know. Time to keep my current gpu for indies, mobas, rts and retro gaming, and invest on a console for triple a titles.

Nah Im not in the mood to spend so much money on 16gb ddr4, a 256gb (lol) ssd and a 4k uncompromised 60fps gpu to play console Ubisoft/EA ports bruteforced to better fidelity on the same console assets.. Thanks anyway. Getting too old for this shit.

Rip PC aaa gaming. Murders: Intel and Nvidia.

Most used GPUs on steamstats: gtx 750ti and 960. Most played games: League, Dota, CS, TF2, pubg, minecraft, world of tanks, warframe, heartstone and stardew valley. Go figure.
Posted on Reply
#40
B-Real
cucker tarlsonHoly smokes.
If I had $3K to drop, I wouldn't think twice. Wait for the reviews, this thing is gonna perform in TXp SLI range.

People who bought two of those pimpire Titans must be furious, nvidia stikes back with a fully enabled TXV a few weeks later.
Lower clock speeds. I wouldn't bet a single penny it performs around TXp SLI.
Posted on Reply
#41
lordofdusk95
OK with that that specs above how fast this card compared to titan Xp or 1080ti in gaming? do we get like 30% performance increase with those specs? I just want to know, so that I can have a better view with the future gaming/consumer volta.
Posted on Reply
#42
Vya Domus
Manu_PTMost used GPUs on steamstats: gtx 750ti and 960. Most played games: League, Dota, CS, TF2, pubg, minecraft, world of tanks, warframe, heartstone and stardew valley. Go figure.
That's no one's fault, people simply prefer cheap , repetitive gameplay. And inexpensive, slower GPUs are fine for that.
Posted on Reply
#43
bug
fizhsmileOK with that that specs above how fast this card compared to titan Xp or 1080ti in gaming? do we get like 30% performance increase with those specs? I just want to know, so that I can have a better view with the future gaming/consumer volta.
That's the million dollar question. Nobody knows the answer, but we'll know as soon as these puppies get benchmarked. We still don't know the availability for these, but the optimist in me hopes the announcement was made now because Nvidia thought they'd make a buck or two over the holidays season.
Posted on Reply
#44
the54thvoid
Super Moderator
Dj-ElectriCIt concerns me a little how nobody is actually talking about the target demographic for this card, the deep learning market with its deep pockets
But Volta is an AI design, the tensor cores will be useless for consumers not doing that work.
Me, post 18.

Yeah, people buying this for gaming will be mighty disappointed if they expect it to be anything vastly superior to Titan Xp or GTX1080ti. Then again, if folks have the monies, why not.

I wonder if anyone will make a water block for it?
Posted on Reply
#45
birdie
Vya DomusClearly , since it's obvious you haven't got anything better to do than calling people retards for expressing their opinions.

Get out of here troll.
There are meaningful informed opinions, and there's hatred/rending the air/fanboyism/envy. Too many comments here are the latter. It's akin to blaming Ferrari for producing and selling the LaFerrari. The difference is Titan V is a useful product which exists to improve the world, while the LaFerrari exists only to show off.
Posted on Reply
#46
Manu_PT
Vya DomusThat's no one's fault, people simply prefer cheap , repetitive gameplay. And inexpensive, slower GPUs are fine for that.
No, people prefer to play games that run well on their systems that they could afford. No one likes to play assassins creed at 1080p 30fps with stutters everywhere at Low/medium settings with a gtx 750ti.

If they could afford something better they would and they might stop playing Dota or League of Legends 24/7 to try different games. But with the current prices on PC hardware it is impossible to most people. Thing is getting too expensive. More than the 1990s. Yes I´m that old. I´m an IBM PC1 warrior, way before the Commodore stuff and 486. Alley Cat and Ricky Dangerous days. I´ve never seen this industry so bad as it is right now, price wise of course.
Posted on Reply
#47
Tomorrow
xkm1948Ok I will take those old Titan Xp and Xs off their hands. $300 about right? :D
Same here. Remember people. Keep the environment and sharing is caring :D
Upgrayedd$3K?? ... Double the price of last gen. Is it the HBM2, can't be. Is it the Tensor cores? Will a Non-Tensor version be significantly cheaper?
Likely the massive GPU itself along with the cutting edge 12nm manufacturing process. I doubt HBM2 adds significant cost. Vega Frontier Edition has 16GB HBM2 and costs less than 800$.
Liviu CojocaruI guess this means that Volta (2080 or 1180) will be around 1000$ ... or maybe not. I would like to see some benchmarks for this card :)
Well the pessimist in me predicts that Volta xx80 will cost what 1080 Ti costs today and xx80 Ti will cost what Titan Xp costs today. I hope i'm wrong tho and we won't see a significant price increase.
punaniAny tests out as of now ? Even if totally silly and i would never get/afford one, still would love to see how it compares
fizhsmileOK with that that specs above how fast this card compared to titan Xp or 1080ti in gaming? do we get like 30% performance increase with those specs? I just want to know, so that I can have a better view with the future gaming/consumer volta.
bugThat's the million dollar question. Nobody knows the answer, but we'll know as soon as these puppies get benchmarked. We still don't know the availability for these, but the optimist in me hopes the announcement was made now because Nvidia thought they'd make a buck or two over the holidays season.
30% seems like a best case scenario considering it has 30% more CUDA cores than 1080 Ti (5120 vs 3584). Tho HBM2 bandwidth helps too. On the other hand the ~500Mhz deficit hurts the final perfromance assuming you can hit roughly 2Ghz boost on 1080Ti and Titan V has 1455Mhz boost.

What i'm really intrested regarding GV102 is wheter or not they will increase the amount of CUDA cores (and by how much) by cutting out the Tensor cores and FP64 capability since cutting these out will free up space. Or will they cut these out and further reduce the number of CUDA cores to something like 4500 range?
Posted on Reply
#48
Assimilator
B-RealLower clock speeds. I wouldn't bet a single penny it performs around TXp SLI.
You're right - it'll probably perform better.
Posted on Reply
#49
Vya Domus
AssimilatorYou're right - it'll probably perform better.
Well , it's good to be optimistic.

I personally hope it will play games at triple-4K and 144hz. :laugh:
Posted on Reply
#50
bug
Tomorrow30% seems like a best case scenario considering it has 30% more CUDA cores than 1080 Ti (5120 vs 3584). Tho HBM2 bandwidth helps too. On the other hand the ~500Mhz deficit hurts the final perfromance assuming you can hit roughly 2Ghz boost on 1080Ti and Titan V has 1455Mhz boost.

What i'm really intrested regarding GV102 is wheter or not they will increase the amount of CUDA cores (and by how much) by cutting out the Tensor cores and FP64 capability since cutting these out will free up space. Or will they cut these out and further reduce the number of CUDA cores to something like 4500 range?
Traditionally, lesser chips never had more CUDA cores. What would be interesting is if Nvidia somehow kept those tensor cores around because they found out a way to actually put them to good use in games. Unlikely, I know, but can hope...
Posted on Reply
Add your own comment
Nov 14th, 2026 10:01 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts