Overclock.net banner

[Gamers Nexus]Titan V benchmarked : An Async Future for nVidia

6K views 49 replies 33 participants last post by  AlphaC 
#1 ·

Some interesting stuff.

And yes, they know it isnt made for silly gaming but actually important stuff
tongue.gif
 
See less See more
1 1
#2 ·
The conclusion is the best part. This thing has so many cores, but needs clock speed to get those cores fully utilized. It also has some frametime concerns for gaming. Nevertheless, there will be many people buying this card to replace a titan xp or GTX 1080Ti. Not me
biggrin.gif
 
#4 ·
Quote:
Originally Posted by d3v0 View Post

The conclusion is the best part. This thing has so many cores, but needs clock speed to get those cores fully utilized. It also has some frametime concerns for gaming. Nevertheless, there will be many people buying this card to replace a titan xp or GTX 1080Ti. Not me
biggrin.gif
People will be buying these to replace Titan XPs or GTX 1080 Tis in compute workstations. Gamers should not be buying this card at all.
 
#5 ·
I don't know if you guys have noticed, but Nvidia is still touting the 1080 Ti as their flagship gaming card. Nvidia is not marketing the Titan V as a gaming card, so if that's all you do, you're making a huge mistake.
 
  • Rep+
Reactions: LAKEINTEL
#6 ·
Doesnt mean that 100s of people wont try it anyway
biggrin.gif


More dollars than sense, etc
 
#7 ·
Pure speculation on my part but if they are getting a fully enabled part out the door (retail not closed shipments) it would seem to indicate that the yield on this design is fairly high. Compared to the previous titan releasing with a cutdown chip to start then followed up with a fully enabled chip much later.
 
#8 ·
It's still in the air if NV will use the volta architecture for their next-gen geforce cards. Volta seems to be designed primarily for ai and machine learning tasks. Creating a smaller Volta chip seems counterproductive to me when you take away what makes Volta such a sought-after gpu in the first place. So, the rumors about Ampere being the next geforce architecture could be true. Volta is a datacenter-first architecture, and using R&D money towards Ampere instead of building a mainstream Volta makes more sense.

Anyway, at least Steve has confirmed all of our suspicions when it comes to gaming performance on this card. It performs well, but not nearly enough in some areas to justify the the spec advantage. Unfortunately, we can't even use this data to extrapolate what performance metrics are possible with the next-gen geforce cards due to NV most likely going with Ampere. In the end, we should just view this as the result of an endeavor to know more about something just for the sake of knowing, and to a lesser extent, as knowledge that Gamers Nexus has the finances available to purchase a Titan V
tongue.gif
 
#9 ·
So for the price of a Titan V you could buy 3 Vega 64, a huge power supply and a 2 day cruise on Carnival between Miami and Nassau including round-trip air-fare from anywhere in North America AND get 50% more performance than the Titan V....

Remember, we've been told that "Volta" is not gaming, that Ampere is gaming...

If I had spare money right now I'd bet a sweet $1k that the "70" Ampere card out-performs Titan V in gaming across the board. They dropped the V far too close to xmass money for it to be a good buy.
 
#10 ·
Quote:
Originally Posted by Mad Pistol View Post

I don't know if you guys have noticed, but Nvidia is still touting the 1080 Ti as their flagship gaming card. Nvidia is not marketing the Titan V as a gaming card, so if that's all you do, you're making a huge mistake.
Doesn't matter, it's been solidified in the consumer mind that Titan cards are ultra enthusiast level gaming hardware for those with money to blow. The Ti is obviously a just as good/better choice price/performance wise but that doesn't mean anything too those who the Titan has targeted for years now.

I don't get this "This Titan isn't a gaming one" thing when we all know good and well that's a moot point at this stage. The Titan REP is set in stone now, the number of times I've heard "You'd need a Titan to run that on Ultra" trumps any other card.
 
#11 ·
Quote:
Originally Posted by Echoa View Post

Doesn't matter, it's been solidified in the consumer mind that Titan cards are ultra enthusiast level gaming hardware for those with money to blow. The Ti is obviously a just as good/better choice price/performance wise but that doesn't mean anything too those who the Titan has targeted for years now.

I don't get this "This Titan isn't a gaming one" thing when we all know good and well that's a moot point at this stage. The Titan REP is set in stone now, the number of times I've heard "You'd need a Titan to run that on Ultra" trumps any other card.
except the 780ti beat the 700 titan
the 1080ti beat the Titan P...

There's the problem, it's a joke.
 
#13 ·
Quote:
Originally Posted by prjindigo View Post

except the 780ti beat the 700 titan
the 1080ti beat the Titan P...

There's the problem, it's a joke.
The 780 Ti lost to the Titan Black

The 1080 Ti loses to the updated, full Titan Xp.

The problem is, the Kepler Titans were prosumer cards. Sure they were awesome for games, but that was not their forte. Their real value was as HPC, CUDA, scientific and prosumer cards. Which was great, they made sense and were actually CHEAP for the target audience.

But Maxwell and Pascal titans are just gaming cards. Nothing more.

This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
 
#14 ·
Quote:
Originally Posted by Charcharo View Post

The problem is, the Kepler Titans were prosumer cards. Sure they were awesome for games, but that was not their forte. Their real value was as HPC, CUDA, scientific and prosumer cards. Which was great, they made sense and were actually CHEAP for the target audience.

But Maxwell and Pascal titans are just gaming cards. Nothing more.

This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
Exactly what I was going to say, 100% agree.
 
  • Rep+
Reactions: Charcharo
#15 ·
Quote:
Originally Posted by prjindigo View Post

except the 780ti beat the 700 titan
the 1080ti beat the Titan P...

There's the problem, it's a joke.
the 780 ti came 9 months after the original titan and it had a bigger core.
the 1080 ti came 8 months after the titan x with a similar core count but different configure. It was better at gaming but the titan had a slight edge in compute.

If anyone thought the top card will stay top card for more than a year, is delusional.
 
#16 ·
Quote:
Originally Posted by jarble View Post

Pure speculation on my part but if they are getting a fully enabled part out the door (retail not closed shipments) it would seem to indicate that the yield on this design is fairly high. Compared to the previous titan releasing with a cutdown chip to start then followed up with a fully enabled chip much later.
This is not a fully enabled part, it has 5120 SPs and a 3072-bit memory bus. GV100 has 5376 SPs and 4096-bit.

You can count on yields being crap with an 815mm^2 die, interposer and HBM2.
 
#17 ·
This just in. Titan VIP Q1, Titan VR fully enabled Q2...$4k and $6k respectably
 
#19 ·
Quote:
Originally Posted by prjindigo View Post

So for the price of a Titan V you could buy 3 Vega 64, a huge power supply and a 2 day cruise on Carnival between Miami and Nassau including round-trip air-fare from anywhere in North America AND get 50% more performance than the Titan V....
Would 3 Vega 64 greatly outperform Titan V on machine learning tasks other tasks that Titan V is designed for?
 
#20 ·
I think the key takeaway here, apart from the frame time troubles which are most likely due to immature drivers, is that the Volta architecture seems to be scaling well with DX12/Vulkan games with Async Compute enabled.
 
#21 ·
#22 ·
I compiled some of the results into an excel sheet if anyone's interested

BYOIYlt.png

https://www.gamersnexus.net/guides/3170-titan-v-gaming-benchmarks-async-future-is-bright-for-volta

The N/As are there because the cards that are tested are inconsistent from game to game; some of them don't include the Titan Xp overclocked and For Honor does not include the Titan V overclocked.

The Titan V is 17% faster on average than the Xp and 28% faster when comparing both cards overclocked. Though remember that the overclock results for some of the games where the Titan V does badly aren't all available.
 
#23 ·
Quote:
Originally Posted by Charcharo View Post

The 780 Ti lost to the Titan Black

The 1080 Ti loses to the updated, full Titan Xp.

The problem is, the Kepler Titans were prosumer cards. Sure they were awesome for games, but that was not their forte. Their real value was as HPC, CUDA, scientific and prosumer cards. Which was great, they made sense and were actually CHEAP for the target audience.

But Maxwell and Pascal titans are just gaming cards. Nothing more.

This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
Thanks to AMD providing competition a bit giving semi pro drivers to Vega Frontier, Nvidia did the same thing and Titans are actually work horses for cad programs and good values now.

https://www.youtube.com/watch?v=oVxVIkw0xd8&feature=youtu.be&t=6m12s

These are the newest results with the newest drivers vs the frontier edition, and the titan XP are actually good cad rendering cards now among other pro applications. If your gaming and doing the same productivity applications the frontier edition is good at(aside from fp16 stuff), a titan XP will actually serve you well and make that 1200 dollars if your using it for certain professional applications a good value, atleast better than the Vega frontier because of the big leads(50% in gaming and 1-60% faster performance in professional applications) for 200 dollars more.

Until GV or GA102 comes out, titan XP will likely hold their value pretty well because of this.
 
#24 ·
Quote:
Originally Posted by Usario View Post

This is not a fully enabled part, it has 5120 SPs and a 3072-bit memory bus. GV100 has 5376 SPs and 4096-bit.

You can count on yields being crap with an 815mm^2 die, interposer and HBM2.
No, you're wrong in several ways.

The first of which is that this is a damaged bin part not an under-enabled part. nV can charge 2x as much for fully functional parts. These cards have a bad HBM2 stack on them or they'd be 16gb.

The second is that each HBM2 stack is only 128 bits wide and cannot be addressed with larger than 64bit data, it has two 64bit channels into it not 1024.
The parallelism of the ram in the HBM2 stack is immaterial.

These cards are 384bit, are missing an HBM2 stack, may have damaged nVlink parts and are useless in arrays because of it.

I've described elsewhere how HBM and HBM2 stacks work but basically the "ram controller" for the stack is in the bottom and has access channels to the PCIe and the GPU, the GPU communicates with 2x64bit bi-directional channels per stack at very high speed.

Claiming that the HBM is 3072 or 4096 wide is like claiming that 4-channel boards with 8 slots populated with 8gb sticks of ram are 1024bit wide memory buses. There may be 1024 bits of memory port address in there but it's still only 4x64 bit channels to the CPU for 256.

Don't propagate marketing twaddle like "4096" and "x86"... they're lies.
 
#25 ·
Quote:
Originally Posted by prjindigo View Post

No, you're wrong in several ways.

The first of which is that this is a damaged bin part not an under-enabled part. nV can charge 2x as much for fully functional parts. These cards have a bad HBM2 stack on them or they'd be 16gb.

The second is that each HBM2 stack is only 128 bits wide and cannot be addressed with larger than 64bit data, it has two 64bit channels into it not 1024.
The parallelism of the ram in the HBM2 stack is immaterial.

These cards are 384bit, are missing an HBM2 stack, may have damaged nVlink parts and are useless in arrays because of it.

I've described elsewhere how HBM and HBM2 stacks work but basically the "ram controller" for the stack is in the bottom and has access channels to the PCIe and the GPU, the GPU communicates with 2x64bit bi-directional channels per stack at very high speed.

Claiming that the HBM is 3072 or 4096 wide is like claiming that 4-channel boards with 8 slots populated with 8gb sticks of ram are 1024bit wide memory buses. There may be 1024 bits of memory port address in there but it's still only 4x64 bit channels to the CPU for 256.

Don't propagate marketing twaddle like "4096" and "x86"... they're lies.
None of what you said is relevant at all to my post or the question I was answering. For the record, I didn't even say why it wasn't fully enabled, and I could not care less about your being flustered with how HBM is marketed.
 
#26 ·
Quote:
Originally Posted by Charcharo View Post

This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
Its the first card with the "TITAN" imprint on the side of the shroud instead of Geforce, isnt that clear enough?

From this point onward, Nvidia is clear that TV no longer belong to Geforce or even any part of it.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top