I rather think it's the particle systems - especially those, that get loaded upon the processor once it reaches six cores. Talk about 99th percentile on a 10-core-model of sub 20 fps.
No DX12 Software is Suitable for Benchmarking *spawn*
- Thread starter trinibwoy
- Start date
DX12 drivers are not allowed to be further multi-threaded than their clients, I believe. No hidden driver-time.
Alessio1989
Regular
Trying to do some multi-threading driver magic on an runtime that leave multi-threading on the shoulders of developer could be a very, very bad idea.
On DX12 (and vulkan) you control synchronization over CPU-to-GPU, GPU-to-CPU and intra-GPU.
In the last case, the drive could ignore some barriers if the GPU architecture do not need them at all in some cases, but generally, D3D12 and Vulkan do not allow those sort of driver "optimization" IHVs generally do with older version of DirectX and OpenGL. This is also a big advantage on the driver development side: less complex code and more common cases. Of course shader optimization is another story.
On DX12 (and vulkan) you control synchronization over CPU-to-GPU, GPU-to-CPU and intra-GPU.
In the last case, the drive could ignore some barriers if the GPU architecture do not need them at all in some cases, but generally, D3D12 and Vulkan do not allow those sort of driver "optimization" IHVs generally do with older version of DirectX and OpenGL. This is also a big advantage on the driver development side: less complex code and more common cases. Of course shader optimization is another story.
I don't quite understand, I'm afraid. Do you mean that if a given application feeds the driver with, say, four queues, that is the maximum number of threads the driver can use on top of that?
You should just PM Razor1 and both get your bias double-confirmed at the same time, while saving the rest of the forum from your constant bullshit.
FYI, I already benched TR at low settings and I know what is going on here. At 1080p Very High the 1060 is 2 fps ahead (58 vs 56) but at 1080p Lowest the RX 480 is 35 fps ahead (125 fps vs 90). Nvidia's DX12 driver is horrible in this game and bad in others, however it's being masked with GPU bottlenecks in most cases. It's simply a reversal of AMD's DX11 situation.
And no it's not just in Geothermal Valley, it's throughout the whole game. Syria masks the worst of it due to lacking long draw distances, but even there the drop in DX12 performance with Nvidia and Ryzen is clear.
How anyone can watch HUB's video and not realise that when Titan X ties with a 1070, yet Crossfire 480's thrash both isn't a clear driver issue is beyond me though. This "DX12 Ryzen Crossfire" theory is probably the most laughable thing I've read yet.
Last edited:
Is this the same guy who's been benching Starcraft II for years even with it's 26% faster performance between the 7700K and 7600K, or showing KL twice as fast as Broadwell-E?
Anyone can run macros on a couple of PC's and chart the end results. Understanding what is being seen is another matter altogether.
Jawed
Legend
This is with Ryzen?
What about with 7700K?
The problem in this game may be like the compiler problem that was discovered in Ashes of the Singularity. A different compiler resulted in a huge jump in performance on Ryzen.
It seems extremely unlikely that it's deliberately coded to perform badly or that NVidia is being negligent.
Which is still, to be quite frank, shocking. Excuses are made that AMD's architecture hates all the babying that D3D11 does, slowing it down. Well, there may be an element that NVidia's architecture is more conservative, favouring D3D11 and perhaps in certain corner cases, unable to gain the best advantage from D3D12. But it's getting harder to see real evidence that that's the case, seemingly as NVidia masters the techniques required to make its GPUs mesh with D3D12 at a low level.
AMD's D3D11 driver quality seems more thoroughly woeful - salvaged in many recent games by the presence of a D3D12 option.
NVidia seems to gain performance with the D3D12 option in this game, when compared with D3D11. So at the very least NVidia is doing good work there, even if they're lagging AMD in this one case.
The real problem comes when reviewers don't call out a huge difference as more than just an outlier.
No, it means the driver allocates no own threads, but it can execute multi-threaded when you do call it that way. The core functions are thread-safe (you thread in any way you want), and commands-lists are affinity locked (you can use 1 and the same thread only, but you can record many command-lists concurrently). Even if there are opportunities for a threading-"AI", it is not allowed.
The problem is here in RED, in your review (well video or what ever you want to call it) you stated the same thing, but you tested NO other game than this game. its a driver bug in this specific game, good for you found a bug, but that doesn't mean its there in other games, please if you want to test your theories, TEST them, don't just blurt out things you DON'T know nor have evidence for.
Hardwareunboxed (techspot) looked at BF1 DX12 (they have also tested other games with Ryzen in their Ryzen review with DX12 and nV hardware) and didn't see what you came up with at all but came across the same problem you did in the specific game and the GF's are being underutilized because of well we don't know outside of the Ryzen CPU being hit hard.
You don't know how to benchmark first off, cause well quite frankly you don't. How else can I put it. You are a rumor mongering youtuber, who has been way off in numerous videos. Sorry credibility lacking very much when you make mistakes because of lack of knowledge and then you talk crap about other reviewers that have been in the industry for many more years. There is a certain thing called integrity in journalism, something that you lack because of all the videos you have done, are mostly off base. You don't even have integrity of your peers, how the hell do you think a person watching your videos should take that?
You were told in your comments in that video you should NOT use multi GPU yet you didn't have the understanding of what that commentor was getting, he then reminded you AGAIN when you posted in the comments of HardwareUnboxed video, and why HardwareUnboxed came to a much different conclusion than you.
Why are your viewers knowing more about the hardware test bed that you are testing than yourself? Ask yourself that. And if they are valid, which those comments where valid, why you argue against them?
You didn't realize mGPU has a much different code path and definitely driver aspects?
These are things that you should KNOW right off the bat that will get called out if you do the things you did. Worse yet you went on speculating based off of what, errors upon errors of bad testing, and unfounded problems on a platform that is in diapers for now.
Last edited:
RX 480 is a 5.8 TFlop, GTX 1060 is 4.4 TFlop. Maybe it is just an effect of lower geometry detail at Low settings in game? At Very High both cards bottleneck at geometry processing (which 1060 has slightly better), but on Low Radeon shows much stronger pixel processing power.
Anyway - Is it really a miracle (and nVidia's driver flaw) that a 5.8 TFlop card has beaten a 4.4 TFlop card? And it's only sometimes.
DavidGraham
Veteran
I won't allow this kind of talk in the forums here, If you don't know how to comment with the sufficient respect to other members (and stick to technical arguments alone), then you can crawl back to whatever AMD tainted Youtube hole that spawned you. Have fun there alone with your hordes of AMD fanatics.
Which doesn't mean squat, different limitations exist at the lowest settings, CPU isn't likely among them. AMD GPUs gain the most there (as in the case of Deus EX MD and The Division). Once you get back to Ultra/Very High, these gains are back to "usual" levels again. Or even become non existent.
You can't jump to that conclusion without doing the necessary investigations to exclude other factors, the HardwareUnboxed video clearly showed this is not happening @1080p. It clearly showed the FuryX not doing Ryzen any favors. PCGamesHardware testing shows the phenomenon only happening in two games at the extreme ends of the CPU limitation spectrum (@720p). And frankly this is some how expected, the 1080Ti needs more CPU power to extract it's performance. You don't strangle it with a Core i3 and expect it to deliver mind boggling fps.
Trying to exalt Ryzen out of some performance irregularities should be accompanied by the appropriate levels of technical scrutiny, not by some amateur's, half assed, half cooked, armchair theories that mask themselves as deeply informed when in reality they are nothing more than a pile of rubble. Until further testing and validation that is.
Except that it happens with several GPUs. You said it yourself: the single 480 was close to the OC'ed 1070, which means it is also close to the FuryX, both are close to the TitanX, this is CPU limitation 101.
Interesting, his findings with 3600Mhz ram + Ryzen + Nvidia are much better than the comparison posted some days ago. AMD is also pulling ahead in several CPU limited scenarios (Ryzen or Intel, doesn't matter) in Dx12, so it might just be a limitation with Nvidia drivers, or maybe he is still GPU bound at 720p
Last edited:
Don't talk to me about "respect" when you're using phrases like...
"If your methods are flawed then your conclusions are also flawed."
"That's his theory alright. Only that it is a bunch of crap."
"I will wait for an independent trust worthy source".
Make sure you and your mate here spread the same crap over multiple forums as well. Oh wait you already did.
So Fury X has different bottlenecks, maybe due to lacking the primitive discard accelerator? We've seen it plenty where Fury X can barely scrape past 10% faster than the 390X.
Concentrate on HUB's and DF's results with Titan X, where they can't get above 80fps yet my Crossfire 480's hit 90fps or HUB's 295X2 wasting the Titan X.
Oh so now it needs further testing and validation? That's not what you were saying 5 posts ago.
Yes it's a "CPU limitation" at max settings when a 1070 matches a Titan X even though two much slower AMD GPUs are fine, "different kinds of bottleneck" at low settings when the 480 is 40% ahead of the 1060 and all the other results like The Division, BF1, Deus Ex and more incoming are simply "flawed testing".
Everything except the Nvidia driver. We get it David.
How exactly is the driver "horrible" when the performance depends on the effort by the developer to optimize their games? The API-Overhead test from Futuremark shows that nVidia's driver runs fine.
If you have the time it'd be good to test with higher clocked memory on Ryzen, it seems to perform much better with those. As a separate video, it doesn't have to be about Nvidia specifically.
DavidGraham
Veteran
So your methods are flawed and more trusty source are insults? How old are you? 5?
And yes you are not a trusty source, and you will continue not to be, heck your channel is just a collection of AMD ramblings repeated like a broken record.
Yep, finally you are getting the hang of it. Maybe you are not a lost cause after all.
At medium/low settings AMD cards achieve the same large gains with Intel CPUs as well. So it's not Ryzen related.
If all you know about of bottlenecks in the tech world is bad drivers than I suggest you stay clear of any sane technical discussions here and go back to having 5 years old discussions with your subscribers. Geez.
I just can't get my head around why we are seeing I7-7700k vs Ryzen benchmarks when investigating how nvidia driver works in DX12 surroundings on both platforms. Comparision should be against Intel 8C/16T counterparts. Only good comparision I've seen in RotR was Digitalfoundrys benchmarks, which clearly showed nvidia doesn't have problem under DX12 with lower clockspeeds, but higher core count.
RotR starts at 4:04
Deffinately no multithreading issues on Intel platform under DX12.
RotR starts at 4:04
Deffinately no multithreading issues on Intel platform under DX12.
DavidGraham
Veteran
I strongly believe in that too, but if all you have to show for your argument is personal attacks, then a stand must be taken.
Testing with a 1060/480 at medium settings is probably the cause for this. Medium is lower than even console settings. This reduces GPU and CPU workloads. Which isn't good for testing anything related to this matter.Interesting, his findings with 3600Mhz ram + Ryzen + Nvidia are much better than the comparison posted some days ago. AMD is also pulling ahead in several CPU limited scenarios (Ryzen or Intel, doesn't matter) in Dx12, so it might just be a limitation with Nvidia drivers, or maybe he is still GPU bound at 720p
