Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Yeah, I meant Volta.Did you mean to say Vega, or something else? If so, why do you believe that?
It'll still be better than Reddit r̶e̶v̶i̶e̶w̶s bragging.LinusTechTips bought one, and will be benching it soon. Better than nothing.
Exactly. All this points to amazing things when the GeForce version hits the market. GV102 March-April, maybe? My wallet will be ready.this isn't the same architecture to even assume that, need to see where the bottleneck is, I'm almost positive its bandwidth. Need to see different settings, different benchmarks, to see any type of assertion that is efficiency of the shaders that are the problem.
Any case pure graphics score, its 30% faster than a 1080ti......, as of right now in Firestrike.
Superposition, Titan V, matches (actually beats by 10%) a Titan 1080ti score that is overclocked with LN2 2.5 ghz, so yeah that is impressive looking at close to 70% more performance from that score.
http://www.guru3d.com/news-story/nvidia-titan-v-graphics-card-benchmarks.html
Also keep this in mind when making comments that are pretty obtuse, this chip is 800mm2, half of its die (DP units and Tensor cores) is NOT used in games but still powered up. They will take up a good chuck of power. With those out of the picture, one can speculate that the power this chip is really drawing for the parts it needs for a gaming chip, is close to 75% of the 250 watt TDP it has. That is pretty crazy when you look at the performance this thing is giving ya. That's 1080 power usage at 40% more performance over a 1080ti, we are looking at 70-100% per tier improvement.
I'm waiting for a 3rd party aftermarket cooler this time around even if it's months down the pipeline from the Founders Edition.
Exactly. All this points to amazing things when the GeForce version hits the market. GV102 March-April, maybe? My wallet will be ready.
The Titan V has fewer Tensor cores enabled and one whole HBM2 stack disabled when compared to the Tesla V100. It has the same number of CUDA cores AFAIK.Do you think it's gonna be faster than Titan V ?
Is Titan V fully unlocked, or are they holding something back for the next Volta based Titan "Whatever" ?
LinusTechTips bought one, and will be benching it soon. Better than nothing.
The Titan V has fewer Tensor cores enabled and one whole HBM2 stack disabled when compared to the Tesla V100. It has the same number of CUDA cores AFAIK.
But like razor1 said: half of the 800mm^2 GPU isn't being used in games, and the SPs are probably not optimized for the gaming stack. GV102 will probably be another 10-15% faster in games, which will make it the same performance jump from Maxwell Titan to Pascal Titan (60-70%).
They wont be making the consumer version 800mm, likely 500mm or less. Unless they are doing something new the Titan has always outperformed the TI part by at least a tiny amount, so I doubt the non titan version is going to be faster. That has been the way Nvidia has typically launched their products and I would expect the same here.
The irony of someone going by "Vega" being first in line to buy a $3000 nvidia card should not be overlooked here.Got mine in, time for some real testing!
Got mine in, time for some real testing!
If you paid attention to the launch, this is a Titan Line, they have shifted it off the Geforce line for this particular card. Most likely all Ti products will be under the Geforce branding from now on.
Can't look at it that way anymore.
Yeah, things with the 'Titan' name have gone from being HPC parts to upsized consumer parts, and now back to being HPC parts (which is what the GTXn80's were for the longest time).
Of course, that means that what GV102 actually gets in terms of size and allocation of resources is anyone's guess; they could put in logic to make it a good Quadro without going full Tesla, or it could just be a larger consumer part, and they could feasibly make it any size. 50% larger than GV104 will be? 100%? 20%? Throw a dart at a board .
She's got a purty mouth.
She's got a purty mouth.
Looks like my core tops out at 1900 MHz. Just constantly pinging at the power roof with +120% power limit. Time to work on memory and then onto benches.
Looks like my core tops out at 1900 MHz. Just constantly pinging at the power roof with +120% power limit. Time to work on memory and then onto benches.