Ryzen's poor performance with Nvidia GPU's. Foul play? Did Nvidia know?

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
As usual with Ryzen, things tend to get very emotional and overblown. A few games, mostly DX9, seem to not perform optimally with Nvidia cards at the moment. Doesn't mean there's a big controversy or something is "broken" and needs "fixing". It just means Nvidia, like everyone else, will have to profile and optimize for this brand new architecture. I'd expect these few performance anomalies to get fixed and general performance to improve in the next few driver updates.

It's not just DX9, that would be understandable though a bit surprising. It's just another example of how bad Nvidia's drivers have been running on Ryzen since release. DX12 is just one big mess that probably will continue with Vulkan. DX9 being funky I can live with. This DX12 mess is both terrible in the sense that this where we are going in games and also that the problem is the exact opposite of what is happening. DX12's big thing is making MT easier in fact some basic CPU work is split with little effort. So the DX12 on Nvidia benches that tank on 1600, 1700, and 6900 but run better on a 7700 when compared to the DX11 versions of these games is absolutely infuriating. We know its not the games because Radeon's still scale fine in these games. If this was a DX11 issue and DX12 was fine that would be one thing. But the fact that thread scaling doesn't work on DX12 is not good, the fact that it only happens on Nvidia takes the problem to another level.
 
Reactions: DarthKyrie

Rifter

Lifer
Oct 9, 1999
11,522
751
126
I have a hard time believing Nvidia wouldn't have seen this issue already when testing Intel's HEDT for years.

They knew about it for sure, they just ignored it. Then got completely blindsided by AMD bringing HEDT performance for mainstream prices and now its biting them in the ass hard, very hard if they dont fix this by Vega launch.

I just hope Nvidia will have this fixed in a few months and that it'll benefit all 6+ core systems.

They dont have alot of time, they have until vega launches, if its not fixed by then they will start to loose sales.
 

moonbogg

Lifer
Jan 8, 2011
10,637
3,095
136
Yeah if its not fixed before Vega reviews come out, it will be an absolute slaughter on those certain games. Vega will already be fast and with this Ryzen/Nvidia bug it will make the Vega/Ryzen combo look that much better. Fires under everyone's asses! I love this.
 
Reactions: DarthKyrie

JimmiG

Platinum Member
Feb 24, 2005
2,024
112
106
It's not just DX9, that would be understandable though a bit surprising. It's just another example of how bad Nvidia's drivers have been running on Ryzen since release. DX12 is just one big mess that probably will continue with Vulkan. DX9 being funky I can live with. This DX12 mess is both terrible in the sense that this where we are going in games and also that the problem is the exact opposite of what is happening. DX12's big thing is making MT easier in fact some basic CPU work is split with little effort. So the DX12 on Nvidia benches that tank on 1600, 1700, and 6900 but run better on a 7700 when compared to the DX11 versions of these games is absolutely infuriating. We know its not the games because Radeon's still scale fine in these games. If this was a DX11 issue and DX12 was fine that would be one thing. But the fact that thread scaling doesn't work on DX12 is not good, the fact that it only happens on Nvidia takes the problem to another level.

Mostly DX9 games. Too few DX12 games out there to draw any conclusions. As we saw with AotS, simple optimizations can completely change the playing field. ROTTR is probably another example of a game that needs a Ryzen patch (it's just GPU bound on the 1060 and Fury).

It's no secret that Nvidia have been doing some driver trickery (that's a technical term) with DX12 to compensate for the less flexible Pascal architecture, moving some tasks from the GPU to the CPU (scheduling etc.). It's also not surprising that this "trickery" was made for Intel quad-core CPUs, as they were by far the most common up until now. With more than 4 cores in use (for the driver trickery bit), it might simply be that the overhead from coordinating all those threads outweighs the performance benefit.
 
Reactions: w3rd

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Mostly DX9 games. Too few DX12 games out there to draw any conclusions. As we saw with AotS, simple optimizations can completely change the playing field. ROTTR is probably another example of a game that needs a Ryzen patch (it's just GPU bound on the 1060 and Fury).

It's no secret that Nvidia have been doing some driver trickery (that's a technical term) with DX12 to compensate for the less flexible Pascal architecture, moving some tasks from the GPU to the CPU (scheduling etc.). It's also not surprising that this "trickery" was made for Intel quad-core CPUs, as they were by far the most common up until now. With more than 4 cores in use (for the driver trickery bit), it might simply be that the overhead from coordinating all those threads outweighs the performance benefit.
Sure there are ways for games to leverage the extra CPU's and that is what AotS is doing, RTS's have tons of room for multithreading AI and unit movements and such. But that still doesn't resolve the main issue. That is in these 4-5 games that have DX11 and either DX12/Vulkan. DX11 scales on both AMD and Nvidia setups and any game going to the trouble of supporting DX12 is going to be more threaded than most DX11 games. They also scale properly on AMD setups in DX12, just not on Nvidia setups above 4c8t. Sure a game developer can use more cores for other work to better utilize the cores but there is inherent support that Nvidia has killed in the name of optimizing the driver for 7700 to mitigate its CPU overhead in DX12. I don't think Nvidia is purposely tanking Ryzen. But it's a trend I don't to be a part of if the DX12 penalty wasn't bad enough, I wouldn't want to pay $400+ on GPU that in future games will only run at its premium in single CPU implementation. Upping the res mitigates this and I am running a 4k monitor right now. But its still a messy business that really makes me want Vega to be high end with a halfway decent price/perf. Nvidia has time to straighten it out but it still leaves a bad taste in the mouth.

I really feel for 6900 owners, I don't think they ever realized what was going on, because if you spend that much on a CPU your probably running a decent monitor, but there are benchmarks proving that this wasn't always the case for them so at one point they just suddenly lost upper end frame rates on a driver update.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,362
5,026
136
nVidia knows, and they've known since Intel HEDT CPUs launched. They just can't do anything about it without completely rearchitecturing their driver which is optimized for DX11 and extracting the most performance out of their hardware using a software scheduler.

Edit: Poll choices are rather extreme. I voted yes they knew, but I don't think you can attribute it to malice. More of because "fixing" it would require so many resources it's not worth it to them (yet).
 
Last edited:

Rifter

Lifer
Oct 9, 1999
11,522
751
126
nVidia knows, and they've known since Intel HEDT CPUs launched. They just can't do anything about it without completely rearchitecturing their driver which is optimized for DX11 and extracting the most performance out of their hardware using a software scheduler.

Edit: Poll choices are rather extreme. I voted yes they knew, but I don't think you can attribute it to malice. More of because "fixing" it would require so many resources it's not worth it to them (yet).

Me too, voted top option but i dont believe they are doing this for reasons of malice, i mean its not like they havent done shady things before and this couldnt be malice its just in this exact instance i think they just got blindsided by AMD's performance with Ryzen and didnt expect to need to have good scaling to 6+ cores with their drivers.

The only thing thats baffling to me is that they havent fixed it yet.
 

DaveSimmons

Elite Member
Aug 12, 2001
40,730
670
126
The only thing thats baffling to me is that they havent fixed it yet.

It isn't just a bug fix, it probably requires a new design. That takes more than a few weeks: design, implementation, testing, iteration on them to improve performance, QA and fixes, validation with Microsoft, release.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
The main issue is that Nvidia has problems with more than 4 cores under DX12 and it's not Ryzen specific.
The fact that they have a problem is bad enough but it's even worse that at least half of reviewers don't know it and they keep pushing DX12 results for Broadwell-E and Ryzen.

I did notice an odd result in AT's Ryzen 5 review

More results for this title http://www.anandtech.com/show/11244...x-vs-core-i5-review-twelve-threads-vs-four/14

This is DX11 and how do we explain it? Maybe in some titles Nvidia can optimize specifically for Ryzen , beyond fixing their many cores DX12 issue.
 
Reactions: T1beriu

moonbogg

Lifer
Jan 8, 2011
10,637
3,095
136
It is Ryzen specific. Show me a benchmark where an Intel 6 core suffers like that. At worst, it performs like a quad core of the same architecture. These Ryzen CPU's are performing terrible, much worse than you would expect. Performance doesn't only fail to scale with more than 4 cores, Ryzen actually goes way backwards as far as performance is concerned with an Nvidia GPU. The situation is not the same for Intel CPUs. This problem is so obvious it just sticks out like crazy in these reviews.
 

Yakk

Golden Member
May 28, 2016
1,574
275
81
Yes, I have to assume (yes assume, there I said it) nvidia's R&D department definitely did do some testing on all new hardware, including Ryzen.

I'm curious if nvidia chose not to optimize their drivers for Ryzen cause pleading ignorance is worst. Or if the undertaking is just too big at this point because of all the driver level hacks on a per game basis.

Interesting developement to be sure.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
It is Ryzen specific. Show me a benchmark where an Intel 6 core suffers like that. At worst, it performs like a quad core of the same architecture. These Ryzen CPU's are performing terrible, much worse than you would expect. Performance doesn't only fail to scale with more than 4 cores, Ryzen actually goes way backwards as far as performance is concerned with an Nvidia GPU. The situation is not the same for Intel CPUs. This problem is so obvious it just sticks out like crazy in these reviews.

https://www.computerbase.de/2017-03.../#diagramm-battlefield-1-dx12-multiplayer-fps

Look at the 6900k vs. the 7700k in the reviews. Also specifically look at the games that support DX11 and DX12. I like the BF1 MP bench because we know that BF well MT for MP. It goes 6900, 6950, 1800x, 7700k in DX11. In DX12 the 6950 and 1800x tank, the 6900 goes down in performance, and the 7700k is the only CPU to show real growth when moving to DX12. Below is the those 4 games in that test with DX11 and DX12 numbers. Once you see the pattern, it becomes pretty obvious when other reviews are seeing the same issue even if they aren't including a 6900 or 5960 in their comparisons.

BF1 DX11 720p
6900k 143.8
1800X 122.4
7700k 116.4
BF1 DX12 720p
6900k 122.4 down 14.9%
1800X 90.7 down 25.9%
7700k 127.6 up 9.6%

Deus Ex DX11 720p
6900k 106.7
1800X 80.5
7700k 87.1
Deus EX DX12 720p
6900k 83 down 22.2%
1800X 63.6 down 21%
7700k 83.6 down 4%

Rise of the Tomb Raider DX11 720p
6900k 165.7
1800X 135.7
7700k 152
Rise of the Tomb Raider DX12 720p
6900k 172.5 up 4.1%
1800X 117.5 down 13.4%
7700k 168.2 up 10.65%

Total War Warhammer DX11 720p
6900k 45.5
1800X 40.3
7700k 43.3
Total War Warhammer DX12 720p
6900k 34.6 down 24%
1800X 30.7 down 23.8%
7700k 42.4 down 2.1%
 
Reactions: KompuKare and w3rd

imported_jjj

Senior member
Feb 14, 2009
660
430
136
It is Ryzen specific. Show me a benchmark where an Intel 6 core suffers like that. At worst, it performs like a quad core of the same architecture. These Ryzen CPU's are performing terrible, much worse than you would expect. Performance doesn't only fail to scale with more than 4 cores, Ryzen actually goes way backwards as far as performance is concerned with an Nvidia GPU. The situation is not the same for Intel CPUs. This problem is so obvious it just sticks out like crazy in these reviews.

http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...and-discussion.2499879/page-215#post-38823668
Ryzen loses a bit more than the 6900k but there is ZERO evidence that those extra issues are Nvidia issues not Ryzen or game specific issues.
We know for example that some games could not count the cores and threads right for Ryzen and that's not on Nvidia.
Nvidia has no reason to harm Ryzen as it would result in better Radeon sales and share losses for Nvidia.It would only get much worse when Vega arrives.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...and-discussion.2499879/page-215#post-38823668
Ryzen loses a bit more than the 6900k but there is ZERO evidence that those extra issues are Nvidia issues not Ryzen or game specific issues.
We know for example that some games could not count the cores and threads right for Ryzen and that's not on Nvidia.
Nvidia has no reason to harm Ryzen as it would result in better Radeon sales and share losses for Nvidia.It would only get much worse when Vega arrives.

Imported. I didn't even realize that the numbers from the computerbase was yours. I have copy pasted those into other forums so much that I completely forgot where I originally got them from. Anyways, this is me crediting your stuff late.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Imported. I didn't even realize that the numbers from the computerbase was yours. I have copy pasted those into other forums so much that I completely forgot where I originally got them from. Anyways, this is me crediting your stuff late.

Thanks but it's ok, only noticed that you've already posted them after i posted, didn't do it to take credit.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Thanks but it's ok, only noticed that you've already posted them after i posted, didn't do it to take credit.
No I understand. I was just clicking on the link you posted and saw the exact same formatting and thought to myself, oh yeah that's where I got them from. I mean they are Computerbase numbers, but I am to lazy to actually type them out myself. It was more me thanking you for putting the effort in originally it makes it really easy to basically give a quick summary when pointing people to the Computerbase review.
 
Reactions: imported_jjj

imported_jjj

Senior member
Feb 14, 2009
660
430
136
No I understand. I was just clicking on the link you posted and saw the exact same formatting and thought to myself, oh yeah that's where I got them from. I mean they are Computerbase numbers, but I am to lazy to actually type them out myself. It was more me thanking you for putting the effort in originally it makes it really easy to basically give a quick summary when pointing people to the Computerbase review.

In another post i've complied some data from Techspot but only for Ryzen and the 7700k http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...and-discussion.2499879/page-223#post-38831695
Only 4 games but there is Radeon data too. For Radeon, Total War is odd but maybe the latest patch wasn't applied.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
nVidia designs for what reviewers are using today. With today being the operative word. They'd rather sell you something for tomorrow, tomorrow.
And it's not a terrible choice but it's also a dichotomy that I am completely opposed to. I am a feature guy. I get what does what I need now but also has things I might use in the future. Obviously there will still be 7600 purchasers, 7700 purchases and so on. But I am not swapping out parts all the time I need equipment that will run what I need today but also tommorow.

That's why this DX12 thing is troubling. I don't mind getting Nvidia cards. If it's the fastest and best approach for now and work well in the future I am fine with it. But not at the cost of Everytime there is a driver update wondering if this one finally fixes performance on my CPU or later if this is the update that will break performance on my CPU just because it doesn't have the market penetration to interest Nvidia.
 

AMDisTheBEST

Senior member
Dec 17, 2015
682
90
61
Maybe the devs don't know how to code dx12 yet. They need some time to learn the new technology and sharpen their skills.
 
Reactions: DarthKyrie

unseenmorbidity

Golden Member
Nov 27, 2016
1,395
967
96
nVidia designs for what reviewers are using today. With today being the operative word. They'd rather sell you something for tomorrow, tomorrow.
Except they try to ensure that tomorrow doesn't come. It's much more profitable to sell you the same gpu over and over.
 
Reactions: DarthKyrie

tamz_msc

Diamond Member
Jan 5, 2017
3,865
3,729
136
Considering how NVIDIA's drivers have progressed over the years, I doubt that they're going to open the can of worms that their driver software is at the moment to properly code for a new CPU architecture.

Anecdotally speaking, it all started with the 'Game Ready' moniker that came into effect a few years ago. I distinctly remember a time when it took almost a month before NVIDIA claimed that their drivers had optimizations for recently released games, and they clearly specified the percentage improvements in the said game tested against a particular card. Now they are in a race to deliver drivers every other week coincident with game releases, and that seems to be their priority at the moment. This change negatively affected quality, a cursory glance over at the NVIDIA forums shows that with every other release that intends to fix something, it also breaks something else completely unrelated. They are in a perpetual tug of war collecting feedback about issues and delivering on them, whether successfully or unsuccessfully.

Yesterday only there was another post about the infamous TDR issue - something that has been happening from 2008!

In the light of all this, it will be months before they get to fix their drivers for proper usage of multi-core CPUs beyond four cores.

How long did it take for them to release 'DX12 optimized' drivers for Pascal? That should give you a clue.
 
Reactions: DarthKyrie
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/    |