Quick Power Testing
I am still doing some more analysis on this, but the power draw of the Titan V appears to be quite interesting. As I pointed out on the first page, the clock speeds start in the 1700+ MHz range, but quickly drop to the 1550-1600 MHz range under sustained usage. That is a bigger variance than normal and our power testing shows an interesting profile as well.
Click to Enlarge
Please excuse the crude screenshot from our monitor tools, but what you see above is the card starting out at ~250 watts of power draw. When the card hits the default temperature limit of 84C, we see the power draw decrease to the 210-220 watts range, and we matched that up with the change in clock speed.
So, why does this happen? There are a couple of interesting options. First, we could have a great GPU that is able to maintain its clock speeds with very low leakage, and we are getting better than expected performance. However, it could also be that our GPU is standard fare, and instead we are seeing NVIDIA's decision to either a.) limit the noise profile of the card and not spin the cooler any faster or b.) over-rate the TDP of the card.
More to come on this later – we have questions in at NVIDIA and more testing to do!
First and foremost, I want everyone to remember that no matter the scores we showed or talked about, I do not recommend that anyone buy a $3000 graphics card for a gaming system. Please. Listen to the words that I am typing to you. Do. Not.
Now that’s out of the way – holy hell does the Titan V offer impressive gaming performance for a product not aimed at this market. Comparing it to the current top of the line gaming/not-gaming card you can purchase, the Titan Xp, our testing shows that the GV100 in this implementation is offering as much as 38% better gaming performance, with my reasonable average hovering in the 18-20% range. Hellblade was the best example of performance improvement, seeing a jump of 38% at 2560×1440 and 29% at 4K. Even Sniper Elite 4 and The Witcher 3 showed improvements in the 20% range, with no indications of concerns for frame time variance, etc.
If we lower the price of entry to comparison even further, say down to the GTX 1080, the Titan V is able to produce average frame rates that reach 80% better in some areas. Hellblade actually runs 97% faster on the Titan V than the GTX 1080!
For AMD, the best option it can offer is the Vega 64 liquid cooled option, and as expected, it performs between the GTX 1080 and the GTX 1080 Ti in most cases. But just like with the GTX cards we have in this review, there is no true price comparison option for the Radeon brand to the Titan V.
It’s clear from our testing that had NVIDIA released a Volta-based GPU with the GV100 in the $1000-1500 range, it would have a justifiable case for targeting high-end enthusiast gamers. But, instead, the Titan V is three grand, putting it well outside the range of gamers with the exception of the crazy few that have more money than they know what to do with. (Or maybe, early Bitcoin investors.)
For the rest of us, the Titan V points to an interesting future for NVIDIA and the GeForce product family. Though I have my doubts on us seeing a GV100 derivative part under the GeForce name, I think the rumored Ampere chip might be the goodness we see with Volta today, minus the unnecessary double-precision and Tensor Core hardware that only serves compute workloads, raises costs, and lowers fabrication yields.
We will be following up this piece very soon with a look at the Titan V against the Titan Xp and the Vega architecture in compute-based workloads. If you are looking for a Titan V review that will tell you about the capabilities it offers for GPGPU, scientific computing, ray tracing, and more, check back soon for part 2!