Nvidia DX12 driver holding back Ryzen

[tommybhoy]

Not a big deal though is it, with Dx12 being a massive fail, all we've got is a couple of Dx11 games patched with it laters, and hardly any using Vulkan.

You must be in the 1% bracket of AMD users that say DX12 is a fail as the other 99% of them are delighted with the gains giving them faster performance than the direct and sometimes faster Nv equivalent.

Guessing you're under the impression there must be a fair few ****** off 290 users out there running DX12?

I don't care that Nvidia are crap in a couple of games compared to AMD

I don't care either, more delighted someones not crap at DX12 as my 1070 has been faster in near enough every title than anything AMD has had since last June, sure there's a small amount of vocal 1080+ users could be ****** that AMD can get close in a hand full of titles right enough leaving them unable to brag.

i know that when it gets here proper, and in a good number of games, when it matters, they won't be crap, just like the tessellation.

All them years AMD had a tessellator on their cards while Nvidia didn't, and then when the games started coming, Nvidia done a one, and BANG!, AMD were smashed, one hit!.

It's funny reading those slating something that increases performance because the other mob know's how it works due to the remarkable familiarity AMD has coding with all the new Api's.

The rest Iv'e already predicted will happen, Nv will be great at DX12 when they don't make AMD look better or it's time to stop looking weak.

 

[Freddie1980]

Loads still raging about his fury's purchase! Comedy gold I think we have our new lambchop

Is Lambchop barred from the graphics forum or something I've not seen him for ages.

That's some very interesting results it's videos like this is why I subscribe to his channel. Most of the tech channels are like 'here are our numbers and for a quick word from our sponsors' and don't bother with any of this deep investigation stuff.

 

[Nexus18]

Half hoping for his sake that Vega is a flop as if it's better than his incoming Ti he will have a breakdown .

Too late for that I think, he is a follower of the great Jen-Hsun Huang now All he sees now is green and anything red is associated with the rage of buying into crossfire

Is Lambchop barred from the graphics forum or something I've not seen him for ages.

That's some very interesting results it's videos like this is why I subscribe to his channel. Most of the tech channels are like 'here are our numbers and for a quick word from our sponsors' and don't bother with any of this deep investigation stuff.

No idea, hopefully he is though..... Doesn't matter anyway as he is on my ignore list

 

[smogsy]

I wouldn't doubt that, but seriously the Fury X vs Titan X Pascal numbers seems more than just what a driver can manage.

you could say the same the other way around

1) some are saying Nvidia are crippling AMD on purpose
2) i'm saying AMD have AMD+AMD optimizations

cant have one without the other.

personally drivers can make huge differences, from experiences 40%+ can be had from driver changes

 

[TheRealDeal]

you could say the same the other way around

1) some are saying Nvidia are crippling AMD on purpose
2) i'm saying AMD have AMD+AMD optimizations

cant have one without the other.

personally drivers can make huge differences, from experiences 40%+ can be had from driver changes

Sure they probably do have optimizations but i can't remember any of AMD's other Cpu's having this type of problem with Nvidia. If anything due to Nvidia's lower Dx11 over head there older Cpu's had a better time on Nvidia. This is very strange indeed but Ryzen is new and i won't be pointing the finger at Nvidia without more evidence.

 

[Rroff]

Sure they probably do have optimizations but i can't remember any of AMD's other Cpu's having this type of problem with Nvidia. If anything due to Nvidia's lower Dx11 over head there older Cpu's had a better time on Nvidia. This is very strange indeed but Ryzen is new and i won't be pointing the finger at Nvidia without more evidence.

Definitely something that needs more looking into before conclusions are made.

There are a few aspects that could be at play like that the Fury cards are very very fast at certain functionality that is normally not a big component of the overall performance envelope - but could come into play in unusual testing scenarios like low settings, etc. where nVidia might have sacrificed on those to put in more shader performance which counts more at higher settings, etc. etc. and again with regard to the way CPU assistance/scheduling works in the different drivers under different conditions.

 

[N19h7m4r3]

you could say the same the other way around

1) some are saying Nvidia are crippling AMD on purpose
2) i'm saying AMD have AMD+AMD optimizations

cant have one without the other.

personally drivers can make huge differences, from experiences 40%+ can be had from driver changes

Drivers shouldn't make such a huge difference, not 40%+, especially with DX12 where Driver's play very little role in performance, and the vast majority of optimisation and performance comes from the developers and hardware specific code paths.

 

[Killer7xx]

Funny how the situation is sort of reversed now, in DX11 AMD were known to have high CPU overhead and now with DX12 Nvidia has higher CPU overhead. The driver and software hacks they are doing to compensate for their lack of hardware to handle DX12 features like Async compute probably has a part to play as well. We know that for Pascal they are highly reliant on software scheduling to perform async compute or compute preemption as they like to call it, most likely this puts a decent workload into the CPU as well. This coupled with the fact that Nvidia's DX12 driver cannot multi-thread properly could result in a significant bottleneck.

Also I don't think AMD heavily optimized their drivers for Ryzen, it's just that Nvidia's DX12 drivers are heavily unoptimised in general and higher on overhead. If AMD realised that their drivers needed heavy optimisation, they would have also told Nvidia about it since both companies would have something to lose out on.

 

[Rroff]

especially with DX12 where Driver's play very little role in performance

Ironically as things stand the drivers come into play quite a bit in DX12 - for instance nVidia has a certain amount of software scheduling of stuff that AMD do in hardware (and vice versa there are some other things not DX12 specific AMD do in software that nVidia is offloading to their shader architecture) and so on.

 

[andybird123]

so is this another case of AMD over egging their marketing and now people are looking for scapegoats to explain why reality and marketing don't mix?

 

[N19h7m4r3]

so is this another case of AMD over egging their marketing and now people are looking for scapegoats to explain why reality and marketing don't mix?

Can you explain this comment and context please. I don't follow.

Ironically as things stand the drivers come into play quite a bit in DX12 - for instance nVidia has a certain amount of software scheduling of stuff that AMD do in hardware (and vice versa there are some other things not DX12 specific AMD do in software that nVidia is offloading to their shader architecture) and so on.

Just shows how badly it's going for NVIDIA in that regard really. That screenshot is from Their own Best practices presentation for DX12 from GDC last year.

 

[LoadsaMoney]

You must be in the 1% bracket of AMD users that say DX12 is a fail as the other 99% of them are delighted with the gains giving them faster performance than the direct and sometimes faster Nv equivalent.

Guessing you're under the impression there must be a fair few ****** off 290 users out there running DX12?

Its only a fail as it hasn't took off, when it was announced by MS, they said there would be about 25+ titles )or something like that), that fall, well, where the **** are they ?, as here we are, a couple of years later, and we're still waiting, all we've got is a couple or so Dx11 games, that have had it patched in laters, FAIL!, and i doubt any 290 owners are miffed at running Dx12, its a performance boost for them, they'll only be miffed at the lack of its usage id have thought.

 

[tommybhoy]

If you say so

 

[Killer7xx]

Its only a fail as it hasn't took off, when it was announced by MS, they said there would be about 25+ titles )or something like that), that fall, well, where the **** are they ?, as here we are, a couple of years later, and we're still waiting, all we've got is a couple or so Dx11 games, that have had it patched in laters, FAIL!, and i doubt any 290 owners are miffed at running Dx12, its a performance boost for them, they'll only be miffed at the lack of its usage id have thought.

Yes but sooner or later you can guarantee that most games will be either DX12 or Vulkan. DX9 died a slow death and yet DX11 adoption still happened.

 

[LoadsaMoney]

Yes but sooner or later you can guarantee that most games will be either DX12 or Vulkan.

and Nvidia won't be crap at them.

 

[AlamoX]

few weeks back, most ppl were convinced that Ryzen sucked for gamers, now almost all the problems have been identified, and all of them are quite simple to fixe, from Ram speed, drivers, and optimization.
now most games where Ryzen had 20-30% less performance, now it closed the gap considerably or even beating intel.
Ryzen is trending with competitive pro gamers and streamers, the reactivity in games that ppl called "Ryzen is smoother" turned out to be latency, where Ryzen packed so much stuff in the CPU itself getting rid of the bridges, lowered the latency, even the reactivity for the peripherals like mouse and keyboards connected through USB 3 speaks directly to the cpu.
and to be fair, Nvidia's stands to lose also with this, AMD probably didn't share much with Nviia before launch.

 

[~>Dg<~]

Rise of the Tomb Raider is an NVIDIA Gamework's title.

amd have worked closley with tomraider games for years cough cough tressfx anyone ?

 

[Goldblum]

Ironically AMD consistently showcased Ryzen with a Geforce 1080 at all the press events (Battlefield 4, Sniper Elite 4) leading up to the launch. You would like to think they would have been aware of this and have mention this in there review note to the press.

They did use 480 mGPU when demoing Sniper Elite at the launch event.

 

[smogsy]

Drivers shouldn't make such a huge difference, not 40%+, especially with DX12 where Driver's play very little role in performance, and the vast majority of optimisation and performance comes from the developers and hardware specific code paths.

why does everyone think i mean GPU drivers?

im talking about CPU optimisations in the GPU driver. the same that Nvidia did for intet on DX11

 

[Killer7xx]

and Nvidia won't be crap at them.

I'm sure they won't be crap but there's no guarantee that they will be ones on top like with DX11, AMD could make further strides with Vega and Navi while Nvidia have to keep playing catch up with drivers and hardware. It's a whole new era for AMD with Ryzen and the NCU GPU architecture, if they can hit the ground running with Vega they've got the momentum already for the next year or so until the launch of Volta. Nvidia will have to fight a much more competent and resourceful AMD than what they have been used to in the last few years.