NVIDIA Volta Rumor Thread

Page 17 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,551
2,538
136
Nice, that is within striking distance of 2x the performance of a stock 1080. Liquid cooled and it should surpass 2x of a stock 1080.

I'd rather see it benched again without being CPU bound. That test isn't too much quicker than a 1080ti, but I think the graph makes it obvious the CPU is holding it back, at least in this benchmark.



Edit: Maybe have your Titan V friend run it at 150% resolution like this? It should remove most or all of the CPU bottleneck.
 
Last edited:
Reactions: ZGR

Hail The Brain Slug

Diamond Member
Oct 10, 2005
3,551
2,538
136
Why is the core clock so low after a node shrink?

I can't find exact details now but I thought the "12nm" process they were using was just an optimized version of 16FF. It didn't result in any notable shrink in die size per core count.

Not only that, the die is gargantuan at 815mm^2, so I am not surprised the clockspeeds are similar to GP100.
 

Glo.

Diamond Member
Apr 25, 2015
5,834
4,842
136
95 to 130 FPS in 4K Overwatch, Epic settings?

It is quite a lot faster than GTX 1080 Ti...
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Why is the core clock so low after a node shrink?

Its a monstrous 'moon shot' style GPU specalised for compute. As large as physically possible on a relatively new node, a pile of specialised compute features etc etc.

The eventual consumer ones will be much more stripped down and then we'll see.
 

USER8000

Golden Member
Jun 23, 2012
1,542
780
136

Look at the results for Deus Ex:Mankind Divided and Overwatch.

A GTX1080TI will get roughly 71 to 73FPS at the same settings and in the comments they say the GTX1080TI gets 60FPS in the same section tested.

This seems more like the Titan V is 10% to 20% faster than a GTX1080TI.

Since this is on an improved 16NM process node and has a 73% bigger area,it seems the gains are limited and most of the improvements have been geared towards non-gaming software.

Time to wait for gaming Volta methinks!
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
OUCH!

Imma guess drivers aren't quite ready for volta...

I'm not very surprised at the performance being lower than expected. First, the memory bandwidth gains are smaller than compute. The gains there are 20% over Titan XP and 34% over 1080 Ti.

The performance gains are lot less than expected in DL as well: https://www.xcelerit.com/computing-...hmarks-deep-learning-nvidia-p100-vs-v100-gpu/

GV100 has 25% more BW than GP100.

Why is the core clock so low after a node shrink?

Everything is nearly bound by thermals and power use nowadays. They increased the SP count by 33%, meaning thermal headroom is used up entirely by the increased SPs. Since clock speed increases power usage further, they couldn't do it.

Architects likely focused on increasing efficiency and clocks when it was process-constrained back in the 40nm days. That also means less gains to be had on newer nodes.

For those interested in mining performance for the sake of knowing it, 64MH/s on Ethereum. Not that it matters because of the cost.
 

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
Look at the results for Deus Ex:Mankind Divided and Overwatch.

A GTX1080TI will get roughly 71 to 73FPS at the same settings and in the comments they say the GTX1080TI gets 60FPS in the same section tested.

This seems more like the Titan V is 10% to 20% faster than a GTX1080TI.

Since this is on an improved 16NM process node and has a 73% bigger area,it seems the gains are limited and most of the improvements have been geared towards non-gaming software.

Time to wait for gaming Volta methinks!
Deus ex dx12 all max 79Fps with 1080TI oc.Titan volta is 8% faster

Edit:Heaven 1440p extreme:
TITAN Volta 109.6Fps
1080TI OC 103.2Fps
This time only 5% faster.
 
Last edited:

imported_bman

Senior member
Jul 29, 2007
262
54
101
We should get some OCed results for games with cards arriving this week. I would expect the Titan V running at 1455MHz boost clocks to be not much faster than a 1080ti running at 2050MHz. With the Titan V liquid cooled it should be beating an overclocked 1080ti by 40%-50%, maybe even more.

Also the Titan V broke 10,000 on Superposition



What is interesting is if one looks at the detailed results and looks at the clocks:


I am not sure how Superposition record GPU clocks, but it looks to me that the Titan V's boost is capping out around 1844 MHz and averaging around 1820 MHz. Maybe it can spike up to 2010 MHz and that is why it records the max as 2010 MHz.

Compare this to the other high ranking results which remain at their max clock throughout the benchmark. (3rd Place Titan XP for example):



Even assuming the max clock as the average clock for the Titan V it still beats the XP by 1.44x clock for clock. Using an average clock of 1820 MHz it is beating it by 1.6x clock for clock. FP32 core count is only 1.33x higher on the Titan V vs Titan XP, so it looks to have a 10%-20% uplift from uarch (memory bandwidth looks to be nearly the same per FP32 core on these results).

Numbers for the Titan V vs 1080ti looks to be 1.55x clock for clock assuming 2010 MHz average for Titan V, 1.72x assuming 1820 MHz average. (As expected 1.55/1.44 = 1.72/1.6 = 3840/3584, Titan XP scales from the 1080ti with the FP32 core count).
 
Last edited:
Reactions: Bouowmx

Nomanor

Member
Jun 5, 2009
104
3
76
>17 pages

Can someone post a short summary pls?

Is this card good for gaming, if money is not an issue?
Is it better than 2 x Titans XP SLI ?
Is it fully unlocked, or are they holding something back for the next Volta based Titan "Whatever" ?
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Lets see. Pass/check above for Qn 2.

For Qn 1, pedantically yes I suppose. For the moment it is definitely the fastest thing out there.

For Qn3 - Nothing held back for what it is designed for but..... It really wasn't designed for gaming.

They're selling enough compute/deep learning stuff to do cards just for that now. Its a grotesque waste to buy this to purely game on it - about 1/2 the chip isn't used at all!

For that it will be much saner to wait for the purely gaming focused version that will 99.99% appear next Spring. Much cheaper and very likely a non trivial chunk faster with it due to specialising.
(Even if they don't add anymore resources it'll have more power to play with.).
 

imported_bman

Senior member
Jul 29, 2007
262
54
101
Some new benches from a user on Neogaf with the Titan V:
Assassin's Creed Origins 4k max settings stock:

Oc of 150Mhz to core and 200Mhz to Memory with no voltage change:


Shadow of War 4k max settings:
 
Last edited:

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
Some new benches from a user on Neogaf with the Titan V at stock:
Assassin's Creed Origins 4k max settings:

Shadow of War 4k max settings:
Its really not much faster than 1080TI oc.Question is how much performance gain can TITAN have from OC.815mm2 die will be TDP limited no matter what.
Btw 1080TI have 60fps in shadow of war 4k max.
 

imported_bman

Senior member
Jul 29, 2007
262
54
101
Its really not much faster than 1080TI oc.Question is how much performance gain can TITAN have from OC.815mm2 die will be TDP limited no matter what.
Btw 1080TI have 60fps in shadow of war 4k max.

1080Ti FE at stock is 52fps in the Shadow of War 4k benchmark using max settings from looking at Youtube.
 

imported_bman

Senior member
Jul 29, 2007
262
54
101
But 1080TI FE runs at 1700mhz.That big 815mm2 die wont be able oc like 1080TI.

Most of that die will not be used while gaming, relative TDP hit will mostly be a function of the FP32 ratio of the Titan V vs 1080Ti, minus the slight efficiency gain from moving to 12nm. There are also board power saving from using HBM, I also think the die nets some power savings from HBM since the memory controller is operating at a much lower frequency. So I don't think it is dire for the Titan V, it has already held ~1820 MHz using the default cooler while running Superposition. I don't think it will have a hard time hitting the upper end of Pascal's clocks (2000 MHz) with better cooling, after 2100 MHz all bets are off. I am guessing that the card will OCed will be able to hit above 1.6x-1.8x of a stock 1080Ti.
 

Head1985

Golden Member
Jul 8, 2014
1,867
699
136
Most of that die will not be used while gaming, relative TDP hit will mostly be a function of the FP32 ratio of the Titan V vs 1080Ti, minus the slight efficiency gain from moving to 12nm. There are also board power saving from using HBM, I also think the die nets some power savings from HBM since the memory controller is operating at a much lower frequency. So I don't think it is dire for the Titan V, it has already held ~1820 MHz using the default cooler while running Superposition. I don't think it will have a hard time hitting the upper end of Pascal's clocks (2000 MHz) with better cooling, after 2100 MHz all bets are off. I am guessing that the card will OCed will be able to hit above 1.6x-1.8x of a stock 1080Ti.
Nah.TITAN boost at default around 1700mhz just like 1080TI.But dont oc that good.Some users reporting 1900Mhz with power limit max out whole time.Should be around 20% faster OC TITAN volta vs OC 1080TI(when both oc)
 
Last edited:

imported_bman

Senior member
Jul 29, 2007
262
54
101
Benches are up https://www.gamersnexus.net/guides/...marks-async-future-is-bright-for-volta/page-2

OCed performance varies from 1.3x-1.75x of a 1080 Ti reference (outside of ATOS which bottlenecks), dependent on the game.

Nah.TITAN boost at default around 1700mhz just like 1080TI.But dont oc that good.Some users reporting 1900Mhz with power limit max out whole time.Should be around 20% faster OC TITAN volta vs OC 1080TI(when both oc)

We will see once someone hybrid mods the card, hybrid should get it to 2GHz, hybrid plus shunt mod should take it over. Gamersnexus plans to do those mods this week.
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
Well, yes. Asides from using the Titan to pick off early adopters at a premium they haven't really significantly moved the price tiers of their gaming cards for a while.

Compute stuff has always cost $$$$$.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |