NVidia and AMD driver performance review [HardOCP]

[Carfax83]

HardOCP has two nice reviews demonstrating driver performance evolution for both NVidia and AMD spanning two architectures; Maxwell and Pascal for NVidia, and Fiji and Polaris for AMD.

NVidia results.
AMD results

Interestingly, HardOCP tests both RotTR and Deus Ex MD in DX11 mode, as they claim that the performance is better than DX12. This just goes to show that DX12 currently is hit or miss, for both AMD and NVidia.

 

[thilanliyan]

This is interesting:
"There was a lot of concern, us included, about the AMD Radeon R9 Fury X having only 4GB of VRAM for "future" games when it was released. Playing those "future" games now, you wouldn’t even know that was a problem. When drivers have been properly optimized the AMD Radeon R9 Fury X can hold its own today."

And this:
"In this review, and the previous one, we have seen a clear pattern of AMD having a more consistent progression of performance on its GPUs from driver to driver over the course of the video card's lifetime compared to NVIDIA."

I think this is the general consensus even on these forums?

 

[AtenRa]

Interestingly, HardOCP tests both RotTR and Deus Ex MD in DX11 mode, as they claim that the performance is better than DX12. This just goes to show that DX12 currently is hit or miss, for both AMD and NVidia.

Vast majority of Gamers doesnt have a 4.7GHz Core i7 Skylake

Im sure if you have a slower CPU like Core i3/5 or Phenom II/FX you will gain a lot with DX-12 over DX-11 in the vast majority of games.

For reference,

Core 2 Quad 9450 @ 3.2GHz + RX 480 8GB GTR Black

Deus Ex Mankind Divided 1080p medium preset (build in benchmark)

DX-11
Average = 39,6fps
Minimum = 27,1fps

DX-12
Average = 45,5fps
Minimum =32,7fps

 

[MajinCry]

Vast majority of Gamers doesnt have a 4.7GHz Core i7 Skylake

Im sure if you have a slower CPU like Core i3/5 or Phenom II/FX you will gain a lot with DX-12 over DX-11 in the vast majority of games.

For reference,

Core 2 Quad 9450 @ 3.2GHz + RX 480 8GB GTR Black

Deus Ex Mankind Divided 1080p medium preset (build in benchmark)

DX-11
Average = 39,6fps
Minimum = 27,1fps

DX-12
Average = 45,5fps
Minimum =32,7fps

I can't find the forum thread, but I distinctly remember Phenom II users were getting double their framerates on Battlefield 4 in 64 man multiplayer, when using Mantle. Did find this user's benchmark, tho'.

http://forums.pcper.com/showthread....Mantle-In-Battlefield-4&p=4560051#post4560051

 

[AtenRa]

I can't find the forum thread, but I distinctly remember Phenom II users were getting double their framerates on Battlefield 4 in 64 man multiplayer, when using Mantle. Did find this user's benchmark, tho'.

http://forums.pcper.com/showthread....Mantle-In-Battlefield-4&p=4560051#post4560051

yea reminds me of Thief on Mantle i did back in 2014

http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/thief-4-cpu-benchmarks.2370857/page-6#post-36301071

 

[Mercennarius]

So conclusion is AMD drivers improve performance over time while Nvidia's really just focus on bug fixes...pretty much what most of us have seen first hand. A 290X has seen a GOOD 10%+ improvement in performance since release from drivers alone.

 

[Carfax83]

This is interesting:
"There was a lot of concern, us included, about the AMD Radeon R9 Fury X having only 4GB of VRAM for "future" games when it was released. Playing those "future" games now, you wouldn’t even know that was a problem. When drivers have been properly optimized the AMD Radeon R9 Fury X can hold its own today."

This conclusion is dubious without actual frametime analysis.

And this:
"In this review, and the previous one, we have seen a clear pattern of AMD having a more consistent progression of performance on its GPUs from driver to driver over the course of the video card's lifetime compared to NVIDIA."

I think this is the general consensus even on these forums?

This is easily explained by the fact that AMD just has more room to grow. Their drivers have historically always been handicapped when it comes to CPU overhead and multithreading compared to NVidia under DX11. NVidia on the other hand has drivers that are more or less topped out when it comes to CPU optimization.

Under DX12 and Vulkan, NVidia has similar performance gains.

 

[Carfax83]

Vast majority of Gamers doesnt have a 4.7GHz Core i7 Skylake

Im sure if you have a slower CPU like Core i3/5 or Phenom II/FX you will gain a lot with DX-12 over DX-11 in the vast majority of games.

For reference,

Core 2 Quad 9450 @ 3.2GHz + RX 480 8GB GTR Black

Deus Ex Mankind Divided 1080p medium preset (build in benchmark)

DX-11
Average = 39,6fps
Minimum = 27,1fps

DX-12
Average = 45,5fps
Minimum =32,7fps

I wasn't impugning DX12, as I know the worth of a low level API. I'm just saying that the problem of developer incompetence/inexperience with DX12, and the lack of proper optimization on the engine side for the API affects BOTH vendors and not just NVidia. Games that are designed primarily for DX11 with DX12 tacked on, typically have very spotty performance.

 

[Carfax83]

So conclusion is AMD drivers improve performance over time while Nvidia's really just focus on bug fixes...pretty much what most of us have seen first hand. A 290X has seen a GOOD 10%+ improvement in performance since release from drivers alone.

If this is your conclusion, then

 

[tviceman]

I'm interested to see how Fallout 4 fairs across all cards, especially 4gb ones, with the HD texture pack. On my OC'd GTX 1070 at 1440p, I can max all settings except Godrays (set to medium) to maintain a consistent 60fps nearly all the time.

 

[Piroko]

This conclusion is dubious without actual frametime analysis.

With all the attention [H] has been giving to frametimes and stutters in the past what exactly makes you think that this is the one time they didn't notice stuttering?

 

[Carfax83]

With all the attention [H] has been giving to frametimes and stutters in the past what exactly makes you think that this is the one time they didn't notice stuttering?

Whether they noticed it or not is irrelevant because everyone has a different sensitivity when it comes to perceiving micro stutters.. To make an accurate assessment of whether a 4GB framebuffer is sufficient for playing the latest games, requires frametime analysis. Without frametime analysis, it's all conjecture.

 

[casiofx]

This is very interesting for 980Ti evangelists.

This leads us into some thoughts we have about DX12 on the Maxwell architecture (GTX 980/Ti) series of video cards. According to our testing in Hitman DX12 and Gears of War 4 DX12 the GeForce GTX 980 Ti did not receive or benefit from performance upgrades via drivers. Whereas, the GeForce GTX 1080 (Pascal GPU) did. We also saw improvements to DX12 games on both AMD GPUs.

http://www.hardocp.com/article/2017/02/08/nvidia_video_card_driver_performance_review/13

 

[tg2708]

This is very interesting for 980Ti evangelists.



http://www.hardocp.com/article/2017/02/08/nvidia_video_card_driver_performance_review/13

Card performs well with any improvements. From the benchmarks I have seen the cards that have gained advantage from using dx12 as of the present still don't clearly outperform and in some cases does not outperform said card.

 

[Carfax83]

This is very interesting for 980Ti evangelists.



http://www.hardocp.com/article/2017/02/08/nvidia_video_card_driver_performance_review/13

The GTX 980 Ti already outperforms the Fury X in Gears of War 4, and as for Hitman, whatever is affecting Maxwell is likely on the engine side and not the API or drivers. That engine heavily favors GCN, and whatever changes NVidia made to Pascal seems to benefited the Pascal lineup greatly compared to Maxwell.

 

[SlickR12345]

HardOCP has two nice reviews demonstrating driver performance evolution for both NVidia and AMD spanning two architectures; Maxwell and P

Vast majority of Gamers doesnt have a 4.7GHz Core i7 Skylake

Im sure if you have a slower CPU like Core i3/5 or Phenom II/FX you will gain a lot with DX-12 over DX-11 in the vast majority of games.

For reference,

Core 2 Quad 9450 @ 3.2GHz + RX 480 8GB GTR Black

Deus Ex Mankind Divided 1080p medium preset (build in benchmark)

DX-11
Average = 39,6fps
Minimum = 27,1fps

DX-12
Average = 45,5fps
Minimum =32,7fps

Don't use build in benchmarks they are fake! Computerbase and hardocp have proven this. Total War, Deus Ex and COD infinite warfare have fake internal benchmarks.

In total war and deus ex both AMD and Nvidia cards lose performance under DX12.

Its useless to compare on internal benchmarks most of the time.

 

[AtenRa]

Don't use build in benchmarks they are fake! Computerbase and hardocp have proven this. Total War, Deus Ex and COD infinite warfare have fake internal benchmarks.

In total war and deus ex both AMD and Nvidia cards lose performance under DX12.

Its useless to compare on internal benchmarks most of the time.

They may not directly translate in to actual Gaming performance but I have just demonstrated RX480 gained performance under DX-12 when a slow CPU was used even in the build in benchmark.

 

[Bacon1]

deus ex both AMD and Nvidia cards lose performance under DX12.

Thats not true, I gain performance @ 3440x1440 with a Fury and i7 3770k in Deus Ex MD using DX12.

 

[Bacon1]

Whether they noticed it or not is irrelevant because everyone has a different sensitivity when it comes to perceiving micro stutters.. To make an accurate assessment of whether a 4GB framebuffer is sufficient for playing the latest games, requires frametime analysis. Without frametime analysis, it's all conjecture.

You should know that [H] uses PresentMon and frametime graphs. They don't publish them usually especially for a long review like this. But they would point out any stuttering.

 

[amenx]

I wish he would have thrown in a kepler card in the testing as that would likely have shown more interesting results.

 

[SlickR12345]

Thats not true, I gain performance @ 3440x1440 with a Fury and i7 3770k in Deus Ex MD using DX12.

Yes it is, your supposed result that no one can see and you have no proof of doesn't change that. I trust computerbase.de and hardocp more than some random guy on the internet(you)!!!!!!!!!!!!!!!!!!!!!!

 

[Krteq]

Yes it is, your supposed result that no one can see and you have no proof of doesn't change that. I trust computerbase.de and hardocp more than some random guy on the internet(you)!!!!!!!!!!!!!!!!!!!!!!

What? There is a gain in DX12 for GCN based cards.

For example (side by side video comparison):

...and more:
https://www.youtube.com/watch?v=M36CnLMHEU0
https://www.youtube.com/watch?v=fKheeDNQswM

 

[Shmee]

Please remember to stay on topic, and stay civil.

 

[richaron]

In total war and deus ex both AMD and Nvidia cards lose performance under DX12.

^ This is just wrong.

It's a line of thinking which comes from reading a few reviews and gathering information (good) but coming to the wrong conclusions in the end (bad). The main problem is usually you only see results from systems with very powerful CPUs. And it's a mistake to think those results are representative of all systems.

This article for example, whilst providing useful and interesting data, provides just a portion of the information we need to fully understand the driver improvements.

 

[[DHT]Osiris]

It's possible that AMD spends more time refining the drivers due to having less capacity to roll up 'New Hotness' in the form of hardware. NV is probably more than willing to press their fabs and design teams for new chips rather than refine existing drivers (and lose those sweet $$$'s).

I mean, I'd rather see AMD's model adopted on both ends if that is the case. We really don't need a new chip drip every 6 months.