[WCCFtech] AMD and NVIDIA DX12 big picture mode

[Headfoot]

It's how software development works... they do the final optimization pass near the end after all the other systems are in place so you can profile them all and do fixes/optimizations. You can only optimize to a certain degree when the thing is half done. There's no conspiracy here. Its just getting closer to launch

 

[ShintaiDK]

I must be missing something because it's clear as day that AMD gained as much as NV did from newer drivers. Sorry to burst the focus group's bubble, but there's no NV driver magic here as usual? Just propaganda?

DX12 Medium.
980TI gained 11%
Fury X gained 6.3%

DX12 Epic:
980TI gained 5.5%
Fury X gained 4.8%.

Same goes for GTX970 vs 390. Bigger gains for NVidia.

The next driver may show the opposite tho. AMD gaining more than NVidia.

 

[NTMBK]

I must be missing something because it's clear as day that AMD gained as much as NV did from newer drivers. Sorry to burst the focus group's bubble, but there's no NV driver magic here as usual? Just propaganda?

Thankfully the "DX11 slower than DX12" problem has gone away 🙂 I was a bit worried by that initially. But yes, AMD has also gained. It's early days for DX12 drivers, DX11 has had years to get optimized!

 

[railven]

I must be missing something because it's clear as day that AMD gained as much as NV did from newer drivers. Sorry to burst the focus group's bubble, but there's no NV driver magic here as usual? Just propaganda?

Guess WCCF is back to shill site status. The AMD check must have bounced. Haha.

EDIT:

Thankfully the "DX11 slower than DX12" problem has gone away 🙂 I was a bit worried by that initially. But yes, AMD has also gained. It's early days for DX12 drivers, DX11 has had years to get optimized!

The DX11 side of that game is lacking some work. They saw 3% drop in medium settings. Battlefront shows AMD can do DX11 right.

 

[96Firebird]

Are people learning not to jump to conclusions from BETA benchmarks yet?

 

[Headfoot]

Are people learning not to jump to conclusions from BETA benchmarks yet?

No, haha. So long as there are cheerleaders here and their team is winning a round, they'll jump to conclusions

 

[railven]

Are people learning not to jump to conclusions from BETA benchmarks yet?

Well, my post kind of poked at that. 😀 Haha.

But, we need to draw every single possible data from alpha/beta/concept_art otherwise we'll look like fools!

 

[Leadbox]

I'm obviously missing something here, the cards lie exactly the same as before in terms of performance and here it's only tested at 1080p.
Where's all the smugness coming from?

 

[gamervivek]

The don't use AA at the behest of nvidia.

One comparison I got off from overclockers.uk site.

Average FR 31.6, GPU Fury X @1100/550, Normal FR 42.5, Medium FR 30.9, Heavy FR 25.7, CPU 3770k @4.3

http://forums.overclockers.co.uk/sho...0&postcount=32

Now,

Average FR 43.4, GPU Fury X @1120/560, Normal FR 49.3, Medium FR 42.1, Heavy FR 39.9, CPU 3770k @4.2

http://forums.overclockers.co.uk/sho...&postcount=250


There is a setting difference, namely time slice. Not sure if it's helping performance or degrading it.

And Oxide dev. made clear that async compute was an afterthought in the game and they implemented it themselves to play around with it instead of AMD's sponsorship. Typical AMD I might add.

 

[Azix]

Are people learning not to jump to conclusions from BETA benchmarks yet?

If you can't use these benchmarks as a measure you really can't use any. "BETA" just means its not out ,it does not preclude the code being solid and representative. The devs already said its as good as launch code so that's what matters.

 

[96Firebird]

I guess at least one didn't learn... :|

 

[Dygaza]

The don't use AA at the behest of nvidia.

One comparison I got off from overclockers.uk site.

Average FR 31.6, GPU Fury X @1100/550, Normal FR 42.5, Medium FR 30.9, Heavy FR 25.7, CPU 3770k @4.3

http://forums.overclockers.co.uk/sho...0&postcount=32

Now,

Average FR 43.4, GPU Fury X @1120/560, Normal FR 49.3, Medium FR 42.1, Heavy FR 39.9, CPU 3770k @4.2

http://forums.overclockers.co.uk/sho...&postcount=250


There is a setting difference, namely time slice. Not sure if it's helping performance or degrading it.

Noticed you were referring to my benchmarks done on oc.uk page. Keep in mind that there is other differences between those 2 shots, than just driver version.

Big difference when it comes to cpu performance was upgrade of ram speed, which clearly shows that DDR1600 ain't enough when you start pushing cpu to the limits. This gave huge cpu performance ugprade (from 68.8 cpu fps to 84.7).

http://forums.overclockers.co.uk/showpost.php?p=28542438&postcount=858

Even bigger difference in cpu performance was the aots patch 0.55.xxxxx which gave huge improvement for cpu. This performance uplit had nothing to do with drivers from AMD or Nvidia.

Also benchmark has evolved since first benchmarks, so direct comparision is hard to make between drivers.

 

[tential]

If you can't use these benchmarks as a measure you really can't use any. "BETA" just means its not out ,it does not preclude the code being solid and representative. The devs already said its as good as launch code so that's what matters.

So what conclusions are you drawing now that the 980ti is faster than fury x and nvidia was able to improve performance more than amd with new drivers?

 

[tential]

I guess at least one didn't learn... :|

I'm concluding now that if you threw away your 980ti, pray to God it's still where you left it.
And that nvidia is best at improving it's drivers over time.

Gotta make sweeping conclusions off 1 beta bench after all!

 

[Leadbox]

So what conclusions are you drawing now that the 980ti is faster than fury x and nvidia was able to improve performance more than amd with new drivers?

What benchmark are you looking at and what has changed from the benchmark b4? Wasn't the 980ti slightly faster than the fury x at 1080p in AOTS when first tested?

 

[littleg]

What benchmark are you looking at and what has changed from the benchmark b4? Wasn't the 980ti slightly faster than the fury x at 1080p in AOTS when first tested?

No, the 260 was beating the 980ti until those superheroes in green flew to the rescue. Get with the narrative, man, you'll never make it into a 'focus group' otherwise.

 

[railven]

No, the 260 was beating the 980ti until those superheroes in green flew to the rescue. Get with the narrative, man, you'll never make it into a 'focus group' otherwise.

It was the 250X, get your narrative right. Amateurs.

 

[Azix]

I guess at least one didn't learn... :|

Because you're wrong. If you want to say one benchmark, fine. If you just want to pretend "BETA" means anything but beta

A pre-release of software that is given out to a large group of users to try under real conditions. Beta versions have gone through alpha testing inhouse and are generally fairly close in look, feel and function to the final product; however, design changes often occur as a result.

If it was an alpha, sure.

So what conclusions are you drawing now that the 980ti is faster than fury x and nvidia was able to improve performance more than amd with new drivers?

wasn't drawing any. It's a given nvidia will be slower at this point and didn't particularly care what kind of tricks they pulled off so far after the benchmark launched. If this is what will be expected on dx12 games we can dispense with the delusion that nvidia has better driver support. Also I realized it was a benchmark. Limited scope means easier to optimize for. Nobody is going through a whole game to optimize performance, but a benchmark with one map and known variables is easier (actually similar with fable benchmark - probably easier). So, I just didn't care.

I also thought it was too limited and lacked information. No idea what clocks the cards are running at, they only used 1080p, they had no antialiasing. All in all not on par with the slew of ashes benchmarks that started this.

I was more concerned with people acting like ashes doesn't matter. Even fable matters. Both give an idea of what to expect and I most definitely would not risk a 980ti at this point (if you don't OC anyway). Not that I would buy a Fury X either, but it stands to suck far worse than kepler (relatively). The direction we are headed is obvious. I am the type that likes what I buy to remain strong, not the buy this year throw away next type. If I think it will lose next year, its garbage to me.

 

[tential]

wasn't drawing any. It's a given nvidia will be slower at this point and didn't particularly care what kind of tricks they pulled off so far after the benchmark launched. If this is what will be expected on dx12 games we can dispense with the delusion that nvidia has better driver support. Also I realized it was a benchmark. Limited scope means easier to optimize for. Nobody is going through a whole game to optimize performance, but a benchmark with one map and known variables is easier (actually similar with fable benchmark - probably easier). So, I just didn't care.

I also thought it was too limited and lacked information. No idea what clocks the cards are running at, they only used 1080p, they had no antialiasing. All in all not on par with the slew of ashes benchmarks that started this.

I was more concerned with people acting like ashes doesn't matter. Even fable matters. Both give an idea of what to expect and I most definitely would not risk a 980ti at this point (if you don't OC anyway). Not that I would buy a Fury X either, but it stands to suck far worse than kepler (relatively). The direction we are headed is obvious. I am the type that likes what I buy to remain strong, not the buy this year throw away next type. If I think it will lose next year, its garbage to me.

Please explain the bold.
Just to refresh everyone's memory, the gtx 980ti was the fastest in this benchmark, before factoring in oc.

 

[Silverforce11]

Please explain the bold.
Just to refresh everyone's memory, the gtx 980ti was the fastest in this benchmark, before factoring in oc.

1080p.

 

[Arachnotronic]

Cool. So both us NVIDIA users and our fellow AMD users will be able to enjoy DX12 titles at great framerates 🙂

 

[Keysplayr]

I must be missing something because it's clear as day that AMD gained as much as NV did from newer drivers. Sorry to burst the focus group's bubble, but there's no NV driver magic here as usual? Just propaganda?

How is Venus these days?

 

[Silverforce11]

Gonna post this here because some people are confused with the interpretation.

This is the new 1080p data:

Compare to the old 1080p data:

The order of the GPUs haven't changed. Fury X ~= 980Ti (decimal points!!), other AMD GPU > NV GPU.

The most interesting result was at higher resolutions, where NV GPUs ran slower in DX12 than DX11 mode.

Old 1440p data (other sites found worse DX12 vs DX11 for NV even at 1440p):

Old 4K data:

Sadly, Computerbase.de did not re-bench beyond 1080p.

 

[Keysplayr]

Old and new data was included in the graphs. But thanks.

 

[LTC8K6]

"Reports of my death were an exaggeration."