Nvidia Titan V

TaintedSquirrel

[H]F Junkie
Joined
Aug 5, 2013
Messages
12,678
b9gIoaA.png


https://www.nvidia.com/en-us/titan/titan-v/

Available now on the Nvidia store. $2999.99.

6 Graphics Processing Clusters
80 Streaming Multiprocessors
5120 CUDA Cores (single precision)
320 Texture Units
1200 MHz Base Clock (MHz)
1455 MHz Boost Clock (MHz)
850 MHz Memory Clock
1.7 Gbps Memory Data Rate
4608K L2 Cache Size
12288 MB HBM2 Total Video Memory
3072-bit Memory Interface
652.8 GB/s Total Memory Bandwidth
384 GigaTexels/sec Texture Rate (Bilinear)
12 nm Fabrication Process
21.1 Billion Transistor Count
3 x DisplayPort, 1 x HDMI Connectors
Dual Slot Form Factor
One 6-pin, One 8-pin Power Connectors
600 Watts Recommended Power Supply
250 Watts Thermal Design Power (TDP)
91° C Thermal Threshold


 
Last edited:
Some are saying it's not actually a gaming card? Not sure what's going on. No real specs anywhere yet.
 
Wow, guess my prediction end of 2017, early 2018 wasn't far off lol.

Looks like its the full Volta (V100) core! Impressive.
 
Wow, guess my prediction end of 2017, early 2018 wasn't far off lol.

Looks like its the full Volta (V100) core! Impressive.

Yup I called Volta Titan by the end of 2017 and most people said it would never happen lol. Now if it’s $3000 then I’ll likely pass and wait on the consumer release in 6 months but this should be great for those that want the latest and greatest no matter the cost.
 
Yup I called Volta Titan by the end of 2017 and most people said it would never happen lol. Now if it’s $3000 then I’ll likely pass and wait on the consumer release in 6 months but this should be great for those that want the latest and greatest no matter the cost.

Oh I'm getting this card if its the full V100, just for that reason lol. Cause its the biggest chip ever made that actually works!
 
6 Graphics Processing Clusters

80 Streaming Multiprocessors

5120 CUDA Cores (single precision)

320 Texture Units

1200 MHzBase Clock (MHz)

1455 MHzBoost Clock (MHz)

850 MHzMemory Clock

1.7 GbpsMemory Data Rate

4608K L2 Cache Size

12288 MB HBM2Total Video Memory

3072-bitMemory Interface

652.8 GB/sTotal Memory Bandwidth

384 GigaTexels/secTexture Rate (Bilinear)

12 nmFabrication Process

21.1 BillionTransistor Count

3 x DisplayPort, 1 x HDMIConnectors

Dual SlotForm Factor

One 6-pin, One 8-pinPower Connectors

600 WattsRecommended Power Supply

250 WattsThermal Design Power (TDP)1

91° CThermal Threshold2


Listed on their website and ready to order for a paltry price of $2999.00. Limit 2 per customer.
 
Last edited:
Yup I called Volta Titan by the end of 2017 and most people said it would never happen lol. Now if it’s $3000 then I’ll likely pass and wait on the consumer release in 6 months but this should be great for those that want the latest and greatest no matter the cost.


i wouldn't really call this a "titan" launch.. this is just a semi-professional card with a titan badge on it similar to what AMD did with the vega frontier edition hence the 3k dollar price tag.. now they can do exactly what AMD did and say consumer Volta released in 2017 even though it really didn't. just as AMD said Vega released when they said it would but really didn't.
 
i wouldn't really call this a "titan" launch.. this is just a semi-professional card with a titan badge on it similar to what AMD did with the vega frontier edition hence the 3k dollar price tag.. now they can do exactly what AMD did and say consumer Volta released in 2017 even though it really didn't. just as AMD said Vega released when they said it would but really didn't.

It’s a consumer release hence you can purchase one right now. I will say they finally did Titan right again by making it a prosumer release. GeForce Volta/Ampere whatever is definitely at least 6 months away from availability. They could announce something in March but I doubt we’d see it till May/June.
 
6000 usd works out to a tidy 7700 canadian pesos.

jesus.

edit: oh yeah tax. lol

8700 dollars.

too rich for my blood.
 
Curious to see what the Ti version will be like (hopefully a sum well under $1000 USD).

3 keys is a lot to spend (I've already spent $1k on my new monitor, plus another $1.2k on my new rig).

Still, can't deny that I'm tempted to get one, simply to see how much performance I'd be able to get out of it in games (even if it isn't gaming-oriented).
 
I still think nvidia will skip volta for gaming as there is no use for tensor cores in games and it doesn't seem that much faster than pascal in single precision.
 
I still think nvidia will skip volta for gaming as there is no use for tensor cores in games and it doesn't seem that much faster than pascal in single precision.


that's probably where Ampere steps in.
 
honestly there is room for AI in gaming and get rid of the dumb enemies

so you may get the tensor core for consumer

im thinking that ampere and ohm and watt what maybe what's AFTER volta - since we didn't get a roadmap last year - and AMPERE maybe next gen HPC

what was posted a few weeks back maybe be inaccurate about ampere being consumer and we are all speculating too much
 
Does this mean that we'll be getting mainstream X80 and X70 Volta cards in the next 6 months? hehe
 
honestly there is room for AI in gaming and get rid of the dumb enemies

so you may get the tensor core for consumer

im thinking that ampere and ohm and watt what maybe what's AFTER volta - since we didn't get a roadmap last year - and AMPERE maybe next gen HPC

what was posted a few weeks back maybe be inaccurate about ampere being consumer and we are all speculating too much

Gonna have to wait and see, even if games benefited from tensor cores, it will be a while before it is being program into games.
 
honestly there is room for AI in gaming and get rid of the dumb enemies

so you may get the tensor core for consumer

im thinking that ampere and ohm and watt what maybe what's AFTER volta - since we didn't get a roadmap last year - and AMPERE maybe next gen HPC

what was posted a few weeks back maybe be inaccurate about ampere being consumer and we are all speculating too much


Was looking into that, the AI for games is totally different, can't really do it the same way AI is done DL. Now if they make specific libraries for it for games, that would be pretty cool. But don't think they will, making realistic AI for games, kinda defeats the purpose of gaming. A computer will "evolve" to be better then human anyways at specific tasks, and that isn't what ya want to be gaming against, just won't be fun ;)
 
I was surprised at this.

That's a fully-enabled core, with a single failed HBM stack to castrate performance a bit.

ATLEAST HALF THE PRICE of the big boy, but probably ships without the pro drivers?
 
Last edited:
Was looking into that, the AI for games is totally different, can't really do it the same way AI is done DL. Now if they make specific libraries for it for games, that would be pretty cool. But don't think they will, making realistic AI for games, kinda defeats the purpose of gaming. A computer will "evolve" to be better then human anyways at specific tasks, and that isn't what ya want to be gaming against, just won't be fun ;)

It could be a nice thing for MP games by having offline AI that can think like humans.
 
true, but again, head shot after head shot lol. It will become the perfect enemy, the hardest part about making AI currently is making it so it has errors, so we can actually win and its gotta be believable that it looks realistic.

The computer knows everything that is going on in the game, it knows fire now and hit here, now if we as the opponent move, then it will miss. its going to come down to who moves faster and most unpredictable. And we know human behavior is never unpredictable, it will eventually learn how a person plays.
 
I still think nvidia will skip volta for gaming as there is no use for tensor cores in games and it doesn't seem that much faster than pascal in single precision.

Depends on how effective the architecture is in Vulkan/DX12, that's where the current line up fall short vs the competition.
 
you don't need an sli bridge for multi card cuda/computing support since they work independently.. these cards aren't meant for gaming.


Besidea mgpu compute does not need SLI - all messaging occurs thru PCIe

sli games also don't need that much processing power as GPU's evolve - they are legacy games and a single card outperforms any SLI older gen anyways at any game


future games will use MGPU compute
 
Would be fun for [H] to pick one of these puppies up just for shits and grins to throw it into the mix for the next gaming benchmark review...
 
Fully enabled core but with 3/4 the memory bandwidth. That might slow performance a bit there.

That screams a failed HBM2 stack.

And the significantly lower core clock seems tom indicate reject cores there as well. We save only the highest-clocking units for the Supercomputers :(

I expect gaming performance about 20% faster than a Titan X Pascal. Hardly worth the money.
 
Last edited:
Fully enabled core but with 3/4 the memory bandwidth. That might slow performance a bit there.

That screams a failed HBM2 stack,.

And the significantly lower core clock seems tom indicate reject cores there as well.

I expect gaming performance about 20% faster than a Titan X Pascal. Hardly worth the money.

You are being very very very conservative with the performance.
 
You are being very very very conservative with the performance.

Never mind, I misread the clock speed of the v100 Press Release. For some reason I thought it was a couple hundred faster than this beauty..

Titan V will have the SAME boost clock as the fastest version of v100 Tesla, so the only castration will be the memory bandwidth.

You are RIGHT, this will be at least 40% faster :D
 
Back
Top