Support NeoGAF

[DemonCleaner]

So my “worst case NVIDIA fixes it” is off by quite a bit. NVIDIA goes to hardware scheduler in future gen is the fix?

i don't think so. the software scheduler with it's tile-based rendering approach was/is pure genius and probably the main reason nvidia had much better TFLOP efficiency vs GCN.
without it nvidia would have real problems to feed that monstrous amounts of cudacores they have in Ampere and we are expecting for their next archs.


as i see it the downsides are somewhat niché without any big impact for most use-cases

 

[Armorian]

In some dx11 games this is reversed...

But yeah for Vulcan and DX12 I noticed higher driver overhead for nvidia few years ago in benchmarks but so far no one was talking about this much...

 

[Md Ray]

i was expecting a lot more of different games/engine and cpu in the test to draw this conclusion

Some additional information: If you want to learn more about this topic I highly recommended watching this video by NerdTech:

You can also follow him on Twitter for some great insights into how PC hardware works: https://twitter.com/nerdtechgasm As for the video, please note I also tried testing with the Windows 10 Hardware-Accelerated GPU Scheduling feature enabled and it didn't change the results beyond the margin of error. This GeForce overhead issue wasn’t just seen in Watch Dogs Legion and Horizon Zero Dawn, as far as we can tell this issue will be seen in all DX12 and Vulkan games when CPU limited, likely all DX11 games as well. We’ve tested many more titles such as Rainbow Six Siege, Assassin’s Creed Valhalla, Cyberpunk 2077, Shadow of the Tomb Raider, and more.

 

[Md Ray]

You can't count one game and declare that the entire driver set for the Nvidia graphics boards are at fault. You would have to give multiple examples of games exhibiting the same result. Perhaps it's the game.

See the pinned comment in the video.

This GeForce overhead issue wasn’t just seen in Watch Dogs Legion and Horizon Zero Dawn, as far as we can tell this issue will be seen in all DX12 and Vulkan games when CPU limited, likely all DX11 games as well. We’ve tested many more titles such as Rainbow Six Siege, Assassin’s Creed Valhalla, Cyberpunk 2077, Shadow of the Tomb Raider, and more.

 

[M1chl]

In some dx11 games this is reversed...

But yeah for Vulcan and DX12 I noticed higher driver overhead for nvidia few years ago in benchmarks but so far no one was talking about this much...

AMD at it again j/k

And meh afterall could be Windows handling DX12 or Vulkan

Also we worked with nVidia back then, heavily. Theri dev support is :chef_kiss:

 

[Md Ray]

Can we really says it's a global problem with 2 games and one cpu to support the theory ?

Two games? Again, they've tested more. See their pinned comment.

And why not question amd when the only intel cpu in mainly one test is "only" a i3-10100 and it perform best than 2600x ?

What? NVIDIA driver overhead issue exhibits even with low-end Intel CPU (i3=10100) here, so it's not a Ryzen-only problem.

Also, 2600X released in 2018 whereas the i3-10100 released in 2020, of course, the i3 is going to beat it due to IPC and other improvements.

so i know what fanboy will decide here with that small bad sample:

HUB said:
"This GeForce overhead issue wasn’t just seen in Watch Dogs Legion and Horizon Zero Dawn, as far as we can tell this issue will be seen in all DX12 and Vulkan games when CPU limited, likely all DX11 games as well. We’ve tested many more titles such as Rainbow Six Siege, Assassin’s Creed Valhalla, Cyberpunk 2077, Shadow of the Tomb Raider, and more."

 

[Md Ray]

Hardware Unboxed at it again...

And then when Nvidia next tells them to fuck off due to their disingenuous takes they will start crying and plead innocence.

C'mon now. What do you have to say about the data they've gathered? It's all fake and HUB is conspiring against NVIDIA? Please.

 

[Md Ray]

Is it the same with Intel? Maybe AMD is pulling some shady shit with their CPUs?

Yes, it's the same even with a low-end Intel CPU. It's not a Ryzen-only issue.

They do great monitor reviews. I must of missed something because I didn't know these guys were disliked as well haha.

Needs to be tested with Intel CPUs as well to confirm.

So you didn't watch the video.

 

[Md Ray]

I bet he didn't do a single intel benchmark

He did.

 

[llien]

...but previews are always taken with a large dose of salt.

It is fine to mislead viewers in a preview, I'll remember that, thanks.

For the haters; the worst that can happen is NVIDIA looks into it and identifies / fixes an overhead or other driver quirk and gamers benefit from a better product in the future.

For the buthurt ones: it is more nuanced than that. NV's code could be optimized into outer space, but just be doing more than AMD's (bringing other benefits, such as faster perf on GPU overall, when not CPU constrained)

You can't count one game and declare that the entire driver set for the Nvidia graphics boards are at fault. You would have to give multiple examples of games exhibiting the same result. Perhaps it's the game.

Agreed.

 

[Md Ray]

The fact he didn’t use any Intel CPU to drawn better conclusions is out of my mind like he is trying to find the result he wants lol

Expected better from you. He did use an Intel CPU.

Sorry but people on here don't watch the vid and immediately jump to conclusions. Makes you people look like a .

 

[sackings]

This has been known for years IIRC. The people buying the latest and greatest from NVIDIA are always using a top-end cpu, and for people on the lower end they either dont know or dont care.

 

[Md Ray]

You can't count one game and declare that the entire driver set for the Nvidia graphics boards are at fault. You would have to give multiple examples of games exhibiting the same result. Perhaps it's the game.

Also, the driver issue exhibits even with Intel CPU. Pointing this out because I saw your reaction when someone said "I bet he didn't test with Intel CPU".

 

[BattleScar]

People forget that its easier to upgrade a GPU than it is to upend your entire system and get a new motherboard/CPU combo. AM4 and Ryzen has mercifully changed that, but still people are more often willing to just drop in a GPU upgrade as its still significantly easier to do.

 

[FireFly]

It is fine to mislead viewers in a preview, I'll remember that, thanks.

I never said it was fine. My point is that DF can still be considered a perfectly reliable site for reviews, which is why people consistently use them. (Your original query)

(Personally I think not checking performance on every quality setting on Doom Eternal was probably an oversight, but we're never going to agree on this)

 

[Epic Sax CEO]

AMD at it again j/k

And meh afterall could be Windows handling DX12 or Vulkan

Also we worked with nVidia back then, heavily. Theri dev support is :chef_kiss:

You're wrong, it's not "again", is more like "yet".
Back on CCN times the hardware scheduler could feed all the cores properly and the DX11 driver had higher overhead than Nvidia's. This was discussed for years and was a huge negative for AMD, why they didn't or couldn't fix the issue, reason why they created Mantle that later morphed into Vulkan and pushed Microsoft to release DX12.
AMD was thinking how it could survive in the future, that's why they bet their future on Zen and waited until Navi to "fix" the problems with CCN. But people still remember those times and believe that AMD still have those problems, they're mistaken. Meanwhile Nvidia's drivers worsened it's overhead, something that was a advantage in the past became and disadvantage and people didn't noticed. This hurt AMD in the past and may hurt Nvidia soon, that's why it needs to be pointed and discussed to motivate Nvidia to put the effort into fixing this.

 

[martino]

Ok it's more clear for me with this video and yes i should have watched to the end (but not moving from 2 games after nearly 15 min ,i stopped. my mistake !)
Indeed it could become a problem and it doesn't seems fixable for released card too
Let's see if dev trends will favor AMD here

 

[MightySquirrel]

I'd dare wager that very few are choosing to play on a 1080p monitor and plus medium settings at that on their 3090s.

Still Hub could have at least tested Intel vs Amd and find out is it just Amd issue or with intel as well.

 

[M1chl]

You're wrong, it's not "again", is more like "yet".
Back on CCN times the hardware scheduler could feed all the cores properly and the DX11 driver had higher overhead than Nvidia's. This was discussed for years and was a huge negative for AMD, why they didn't or couldn't fix the issue, reason why they created Mantle that later morphed into Vulkan and pushed Microsoft to release DX12.
AMD was thinking how it could survive in the future, that's why they bet their future on Zen and waited until Navi to "fix" the problems with CCN. But people still remember those times and believe that AMD still have those problems, they're mistaken. Meanwhile Nvidia's drivers worsened it's overhead, something that was a advantage in the past became and disadvantage and people didn't noticed. This hurt AMD in the past and may hurt Nvidia soon, that's why it needs to be pointed and discussed to motivate Nvidia to put the effort into fixing this.

It's also j/k : )

 

[Insane Metal]

Does this even make sense? Who in the world would be running a 3090 with a 1600X? What?

 

[onunnuno]

Is it the same with Intel? Maybe AMD is pulling some shady shit with their CPUs?

Of course. AMD implemented hardware changes in their CPU silicon so it affects their competitor.

Or nvidia is using weird compiler flags, or nvidia is too reliant on software scheduling of draw calls. but no, what makes sense in neogaf is AMD physically using expensive silicon to "block" nvidia LOL

 

[Md Ray]

Still Hub could have at least tested Intel vs Amd and find out is it just Amd issue or with intel as well.

Another one who is commenting before watching the full vid.

 

[Md Ray]

Does this even make sense? Who in the world would be running a 3090 with a 1600X? What?

Another one missing the point of the video. A lot of people in the world are running 3600X + 3070. But 5700 XT is outperforming 3070 when CPU-bound with that CPU. Why? Does THAT make sense to you?

 

[Insane Metal]

Another one missing the point of the video. A lot of people in the world running 3600X + 3070. But 5700 XT is outperforming 3070 when CPU-bound with that CPU. Why? Does THAT make sense to you?

Oh.

Ok then sorry, I'm dumb lol

 

[MeatSafeMurderer]

In some dx11 games this is reversed...

But yeah for Vulcan and DX12 I noticed higher driver overhead for nvidia few years ago in benchmarks but so far no one was talking about this much...

Not really surprising. Everyone knows AMD's driver overhead in OpenGL / DX11 is atrocious, it's part of the reason they fell behind comparable NVIDIA cards so consistently for years. Equally NVIDIA's support for DX12 and Vulkan has been lacking for years, this isn't news at all (although the apparent extent to which this is so is...concerning). The amount of people stanning for NVIDIA ITT is concerning though...and most of them didn't even bother to watch the video. This is clearly indicative of an issue. No, most people aren't going to pair a 3090 with 1600X or i3 but it doesn't make it any less valid...and this will have an impact on CPU utilization on higher end CPUs too, potentially impacting 1% and .1% lows.

 

[Armorian]

Oh.

Ok then sorry, I'm dumb lol

Your CPU will get insufficient faster on nvidia hardware... Higher CPU overhead means that and that is true for all current top processors. I hope this will be fixed in the future, nv clearly got lazy with drivers

 

[martino]

Your CPU will get insufficient faster on nvidia hardware... Higher CPU overhead means that and that is true for all current top processors. I hope this will be fixed in the future, nv clearly got lazy with drivers

driver will not fix this if the scheduler is still software like pointed in another video in this thread

 

[MightySquirrel]

people still remember those times and believe that AMD still have those problems, they're mistaken. Meanwhile Nvidia's drivers worsened it's overhead, something that was a advantage in the past became and disadvantage and people didn't noticed.

On the same page..

109 v 65 fps on what suppose to be very similar perf cards, but because of terrible Amd driver overhead you have the latest 6800xt looking like an entry card in comparison in many dx11games still.

 

[Md Ray]

Ok then sorry, I'm dumb lol

It's ok, mate. Nah, you're not.

 

[Bo_Hazem]

Very interesting!

 

[RobRSG]

What a stupid test. 3090 + 1600X is a terrible, terrible benchmark for anything.

 

[ethomaz]

Expected better from you. He did use an Intel CPU.

Sorry but people on here don't watch the vid and immediately jump to conclusions. Makes you people look like a .

That is a issue with his channel... he lose the trust to be watched.

Indeee he looks like a

BTW the test is weird no matter how you try to defend it.

 

[VFXVeteran]

You're wrong, it's not "again", is more like "yet".
Back on CCN times the hardware scheduler could feed all the cores properly and the DX11 driver had higher overhead than Nvidia's. This was discussed for years and was a huge negative for AMD, why they didn't or couldn't fix the issue, reason why they created Mantle that later morphed into Vulkan and pushed Microsoft to release DX12.
AMD was thinking how it could survive in the future, that's why they bet their future on Zen and waited until Navi to "fix" the problems with CCN. But people still remember those times and believe that AMD still have those problems, they're mistaken. Meanwhile Nvidia's drivers worsened it's overhead, something that was a advantage in the past became and disadvantage and people didn't noticed. This hurt AMD in the past and may hurt Nvidia soon, that's why it needs to be pointed and discussed to motivate Nvidia to put the effort into fixing this.

Well, all you have said is of course true, BUT this won't affect most people with PCs. Most people are generally running in a GPU-limited case most of the time. All that's needed is to crank the resolution to 1440p and you are now GPU-limited. Going forward, I think it'll be something that Nvidia "may" get to but let's face it - there are way more important things to master than 1080p efficiency.

 

[DemonCleaner]

In some dx11 games this is reversed...

But yeah for Vulcan and DX12 I noticed higher driver overhead for nvidia few years ago in benchmarks but so far no one was talking about this much...

Are that your Benchmarks? A bit of context would be nice.

Anyway, you cant claim thats a driver overhead issue if you dont controll scaling with weaker cpus.

What you have shown is much more likely some sort of engine stall or occupancy issue like we have seen in so many dx11 games on amd gpus over the years.

 

[Marlenus]

I knew the owner of the video/article before open the thread.

That tells you a lot.

The fact he didn’t use any Intel CPU to drawn better conclusions is out of my mind like he is trying to find the result he wants lol

He tried a 10100 which is very close to a 7700K and is something where people who built a system 3 years ago may me looking to buy a 3070 and this data shows if you are CPU limited in DX12 titles you are better off with Radeons.

The inverse of this was true with a lot of DX11 games where the NV driver was better than AMDs.

 

[Soulblighter31]

Are that your Benchmarks? A bit of context would be nice.

Anyway, you cant claim thats a driver overhead issue if you dont controll scaling with weaker cpus.

What you have shown is much more likely some sort of engine stall or occupancy issue like we have seen in so many dx11 games on amd gpus over the years.

 

[martino]

also let's hope DX11 is not the future....
it's time for it to retire now

 

[Marlenus]

That is a issue with his channel... he lose the trust to be watched.

Indeee he looks like a

BTW the test is weird no matter how you try to defend it.

The test is valid and now DX12 is the predominant API it is relevant in the same way that a few years ago the AMD DX11 issues were relevant.

NV need to fix it because a 5600XT should nkt be faster in any situation than a 3090.

Plenty of people keep a CPU for 5-10 years and have 1 or 2 GPU upgrades in that time. Plenty of 8700K and 7700K owners will be looking for a new GPU to keep their rig ticking for another 12 or so months and for people with these CPUs this information is valuable.

I have a 2200G and it is well worth knowing that I am better off plumping for an AMD gpu should I decide to add one.

 

[Soulblighter31]

The test is valid and now DX12 is the predominant API it is relevant in the same way that a few years ago the AMD DX11 issues were relevant.

we dont have any sort of data to make any type of claims though. 2 games on 4 cpus from the same manufacturer means nothing. He claims he tested more games ? Oh really ? Present the data then. Its literally 2 random games that we have here. There no situation if we dont get more data

 

[Marlenus]

we dont have any sort of data to make any type of claims though. 2 games on 4 cpus from the same manufacturer means nothing. He claims he tested more games ? Oh really ? Present the data then. Its literally 2 random games that we have here. There no situation if we dont get more data

He tested more than 2 games but for the sake of brevity he focused on 2 and showed a bit of a 3rd. He tested the Intel 10100 as well which is a 7700K level part and were very popular several years ago. He said the data is available to floatplane members or patreon members but I bet it gets released in a few days.

It looks to be a DX12 issue although Vulkan is worth testing too. With DX11 it is usually the other way around since NV would hand tune the driver for specific games to reduce load on the main thread that was issuing draw calls. Of course DX11 is now old hat so while it was relevant a few years ago it is not so much anymore.

Not sure I understand the NV cope in here. Pretty embarrassing, they just make GPUs, they don't need, and ought not be, worshipped.

 

[Epic Sax CEO]

On the same page..

109 v 65 fps on what suppose to be very similar perf cards, but because of terrible Amd driver overhead you have the latest 6800xt looking like an entry card in comparison in many dx11games still.

I'm expressed myself a bit bad leaving one a part of my thoughts.
It's not a problem with "current" APIs. Isn't rare for a game to launch today without DX12 or Vulkan support? AMD gave up fixing the DX11 drivers because it was almost impossible to fix and they were working on changing their architecture, meanwhile people just assumed that Nvidia would have no problems with the new APIs because if everything was working so well for them with DX11 shouldn't be a reason for things to worse. This is how people were wrong. AMD made as many bets on the GPU side as the did on the CPU side, and the bets are paying off.

 

[Werewolfgrandma]

Yes, it's the same even with a low-end Intel CPU. It's not a Ryzen-only issue.


So you didn't watch the video.

I did sorta. I watched the generation ryzen one and figured this was just that data. Seems they did a bit more as well.

 

[Rikkori]

 

[MightySquirrel]

I'm expressed myself a bit bad leaving one a part of my thoughts.
It's not a problem with "current" APIs. Isn't rare for a game to launch today without DX12 or Vulkan support? AMD gave up fixing the DX11 drivers because it was almost impossible to fix and they were working on changing their architecture, meanwhile people just assumed that Nvidia would have no problems with the new APIs because if everything was working so well for them with DX11 shouldn't be a reason for things to worse. This is how people were wrong. AMD made as many bets on the GPU side as the did on the CPU side, and the bets are paying off.

That's fair. But the way I'm looking at this data is that even on an older amd1600x you can still get ~60 fps. That wasn't always the case especially in dx9 and evidently still not in some dx11 games on amd cards. Those having 3090s most likely play at 4K or at the very least Wqhd and most likely have decent not low end CPUs so they will never run into any issues with this overhead. It's a good find by Hub, but it's far from same situation.

 

[Kenpachii]

He did.

Should probably mention that in the OP then.

 

[Marlenus]

Should probably mention that in the OP then.

Or you could just watch the video. It has timestamps in the description with a clearly titled chapter called 'Is this just a ryzen problem?'.

 

[Techies]

There is no problem, except for bad CPU/GPU pairing.

It will always be a give and take situation, Nvidia has a greater feature set so that by itself would take more resources and the more you can offload on the CPU, the more power is available to the GPU. In the case of high end GPUs, you would pair it would high end CPUs and optimize in such a way that you you make maximum use of the CPU, cap it out in a way that every bit of performance is available on the GPU to push the max fps.

 

[lukilladog]

Good thing I don’t game on a weak cpu

Bad thing is that the nvidia drivers might not be allowing your strong cpu to flex its muscles as much as it could.

 

[MeatSafeMurderer]

Bad thing is that the nvidia drivers might not be allowing your strong cpu to flex its muscles as much as it could.

Also the thing that people are ignoring is that the bottlenecking also affects lower tier NVIDIA cards, it's not just the 3090, it's the whole stack, and even rears its ugly head at 1440p...in a generation or two xx70 will be matching or very close to the 3090, and people running xx70 cards are much more likely to be using older / more budget friendly CPUs. The cope in here is real.

 

[zeomax]

This is problematic when next-gen CPU-heavy games start to come out in the near future.

When will this be? When the next gen games starting to use 8 cores?