RTX 4080 & 7900 XTX/XT benchmarks DCS

[petsild]

22 hours ago, Th3ChosenOn3 said:

Thanks to @xoxen for providing his run.

Great now the benchmark is complete, thank you.
I bought a 4090 today, when I can fit the graphics into the case I can provide the data.

Edited December 23, 2022 by petsild

[BitMaster]

Haha !

Who would have thought that could be an issue ?

[exil]

On 12/23/2022 at 12:17 AM, ssg said:

Does anyone know if older radeons like 6900XT have horrible blur when looking sideways and moving head when limiting fps to for example 45 or 55 @90hz on VR? Maybe it is driver problem and could be fixed by AMD. 

I can get solid 45 or even 55 on most missions like Enigma on 7900XTX but due to microstuttering blur it is completely unplayable. 

Only setting I have found out that has no blur is when fps is same as refresh rate. Even if I get solid 90 and then limit it to 45 to avoid frame drops it starts to blur. So annoying. 45 fps would be just fine for me. 

 

Yes. It was on the 6800 and 6900 series as well and the reason why I went to team Green. By the way, this hasn't been fixed ever since. I somehow expected this behavior or was at least afraid it could happen again with the 7000 series. And now it's just confirmed. To be honest, I wouldn't hold my breath that this gets fixed anytime soon. 

[Dangerzone]

19 minutes ago, exil said:

Yes. It was on the 6800 and 6900 series as well and the reason why I went to team Green. By the way, this hasn't been fixed ever since. I somehow expected this behavior or was at least afraid it could happen again with the 7000 series. And now it's just confirmed. To be honest, I wouldn't hold my breath that this gets fixed anytime soon. 

That may have completely scuttled my thoughts of going to AMD. That's a very poor situation to be found in - a known problem, and what appears to be no interest to correct it at all with the previous generation, plus now as well. Very disappointing. I was seriously considering purchasing a 7900XTX in January. I guess my only hope now is that at least for most other gamers that don't need it, the AMD is attractive enough to force NVIDIA's prices down at some stage in the future. Either that, or DCS come out with MT/Vulkan where it's improvement is to a point where it gets rid of the nigglies. 

[exil]

6 minutes ago, Dangerzone said:

That may have completely scuttled my thoughts of going to AMD. That's a very poor situation to be found in - a known problem, and what appears to be no interest to correct it at all with the previous generation, plus now as well. Very disappointing. I was seriously considering purchasing a 7900XTX in January. I guess my only hope now is that at least for most other gamers that don't need it, the AMD is attractive enough to force NVIDIA's prices down at some stage in the future. Either that, or DCS come out with MT/Vulkan where it's improvement is to a point where it gets rid of the nigglies. 

Well, don't get me wrong. I'm not saying its a bad card in general. For anyone on 1080p or 1440p it might be very good. 

But for anyone in VR whose concern is also visuals and therefore needs motion reprojection it's really a bad card for the price if that stutter problem still persists. For that specific user I would not recommend buying AMD cards at all. I was really hoping I was wrong and they would have solved this issue with the 7000 series. In that case I would have probably sold my 4090 and bought a 7900XTX myself. But since they haven't, I don't expect it to be solved at all (VR is still just a niche market, and the need for motion reprojection even more). The AMD drivers division is not really big and I think they have other problems at the moment. 

[Dangerzone]

12 minutes ago, exil said:

Well, don't get me wrong. I'm not saying its a bad card in general. For anyone on 1080p or 1440p it might be very good. 

But for anyone in VR whose concern is also visuals and therefore needs motion reprojection it's really a bad card for the price if that stutter problem still persists.

And that's exactly what I'm looking for. VR.

[MK IX Hawk]

@petsildDo you have a VR headset? If you do would be great to get some performance comparison between 4080, 7900 XTX and 4090. Thanks. 

[Johnny Dioxin]

I'm using a Cooler Master MB511 case for mine - was really cheap at around £50 but it's great. Has fittings for 8 SSDs - 4 below the PSU/cable area, 2 on the back of the baseplate (what you attach the motherboard to) and two on top of the PSU enclosure. It has 3 fans on the front, two at the top and one at the back and my CPU & GPU temps dropped 10°C when I moved my PC into it.

It has all the RGB lightshow gubbins but I don't use that stuff (the 4080 has an RGB socket, too).

This was with my 2080Ti:

 

This with the 4080:

 

It's a hardened cellulose side panel. Only thing I don't like is that it doesn't have a USB-C socket, they are all USB A 3.1. As you can see I'm not massively strict on hiding all the cables - just the worst of it.

[ssg]

On 12/23/2022 at 6:45 PM, SkateZilla said:

ASW and STW

Sadly ASW on AMD just makes it different flavor of stuttering where as on Nvidia ASW is pretty smooth and usable.

Few posters have pointed out that AMD never fixed this issue for 6000-series so I guess I will return this card and see if 4080 ever drops to near 1200 euro here. 

Edited December 25, 2022 by ssg

[petsild]

15 hours ago, MK IX Hawk said:

@petsildDo you have a VR headset? If you do would be great to get some performance comparison between 4080, 7900 XTX and 4090. Thanks. 

I don't have a VR set yet and I don't plan to in 2023 either.

[petsild]

On 12/22/2022 at 7:37 PM, Th3ChosenOn3 said:

Thanks to @xoxen for providing his run.

Raw data of Plasma Torture Map on RTX 4090.

 

4x MSAA + x1,5 SSAA:

Average framerate : 97.2 FPS

Minimum framerate : 66.9 FPS

Maximum framerate : 122.3 FPS

1% low framerate : 53.3 FPS

0.1% low framerate : 44.4 FPS

Edit:
details settings as recommended by video and resolution 2560x1440.
My hw build now matches the signature.

Edited December 26, 2022 by petsild

[Bazz_Mulder]

On what resolution and settings preset?

[petsild]

41 minutes ago, Bazz_Mulder said:

On what resolution and settings preset?

I have added the data.

[SkateZilla]

waiting for 7900XTX Taichi Benchmarks/reviews, ASRock likes to go overboard with Taichi line.

[IanHx2]

waiting to find a sub-£500 card that can handle Reverb VR city low-level without stutter.....

[Andrei]

Jumped the gun, bought 4080 at last. Will do some VR comparison tests between 2070S and 4080 when it arrives

[Johnny Dioxin]

I've just done a flight in VR with my Pico 4 and the RTX 4080 - jaw-dropping smoothness. I used the same settings as with my 2080Ti, which always had either a solid 30 or 45fps, depending on headset used. Despite having my Pico 4 settings at 90Hz, DCS World ran at 72 - but OMG! Now I can finally up my settings. Not sure if the HD is processed yet - this is the onboard Pico headset recording - finally a VR headset manufacturer has included an easy onboard recording option - just a press of the screenshot button on the controller:

 

I recorded the test flight, which was just a take-off from Addis Ababa in the A-10A followed by a low-level flight around the city until my demise when I misjudged a couple of appartment blocks! Unfortunately, the commentary wasn't recorded. Nevertheless, fpsVR is visble for the whole flight - flatlining

Edited December 27, 2022 by Johnny Dioxin

[LucShep]

I understand the purpose of benchmarks, but I notice that most (all?) are testing with FC3 aircraft (Su-27, A-10A, etc), which have much lower detail in Cockpit 3D and far lower res textures. Same thing for the map, Caucasus is simpler and much better in performance than pretty much any other DLC map.

I'd say to try that with higher-detail modules, such as an F-14A/B Tomcat, F-16C Viper, AH-64 Apache (among others) and in, say, Syria or The Channel maps, to get a better representation of the real capability of the GPU in test. 

Edited December 27, 2022 by LucShep

[Johnny Dioxin]

Yes - it's a shame the commentary didn't record in my video as I was saying pretty much the same and I was only using the A-10A because I am redoing my profiles and that is top of the list! Next one will be in the A-10C II and F-16C, which I've also completed now.

Edited December 27, 2022 by Johnny Dioxin

[petsild]

57 minutes ago, LucShep said:

I understand the purpose of benchmarks, but I notice that most (all?) are testing with FC3 aircraft (Su-27, A-10A, etc), which have much lower detail in Cockpit 3D and far lower res textures. Same thing for the map, Caucasus is simpler and much better in performance than pretty much any other DLC map.

I'd say to try that with higher-detail modules, such as an F-14A/B Tomcat, F-16C Viper, AH-64 Apache (among others) and in, say, Syria or The Channel maps, to get a better representation of the real capability of the GPU in test. 

 

I currently find the Apache cockpit the most demanding, and it doesn't even have a functional radar, but it also depends on the number of ground units on the map.

For the benchmark, I would choose the free module route in combination with a challenging mission.

[Andrei]

2 hours ago, LucShep said:

I understand the purpose of benchmarks, but I notice that most (all?) are testing with FC3 aircraft (Su-27, A-10A, etc), which have much lower detail in Cockpit 3D and far lower res textures. Same thing for the map, Caucasus is simpler and much better in performance than pretty much any other DLC map.

I'd say to try that with higher-detail modules, such as an F-14A/B Tomcat, F-16C Viper, AH-64 Apache (among others) and in, say, Syria or The Channel maps, to get a better representation of the real capability of the GPU in test. 

 

Fair point. But gentlemen few post above did the F/A-18C on Marianas map, both and high and low level, which is fairly representative of the GPU performance. Granted, Hornet is not as heavy as Apache, but it's still a full fidelity module.

Some maps, and modules like Channel map are so poorly optimised, it doesn't make too much sense testing and benchmarking with - not overly representative of general performance. I guess not even 5090 with I9-15900K will make good framerates there

[LucShep]

35 minutes ago, Andrei said:

Fair point. But gentlemen few post above did the F/A-18C on Marianas map, both and high and low level, which is fairly representative of the GPU performance. Granted, Hornet is not as heavy as Apache, but it's still a full fidelity module.

Some maps, and modules like Channel map are so poorly optimised, it doesn't make too much sense testing and benchmarking with - not overly representative of general performance. I guess not even 5090 with I9-15900K will make good framerates there

That's a fair point, but then one can say the same about the "Plazma Torture Map test" used so often here in this thread - it'll show CPU limitations far, far sooner than anything GPU wise, won't it?

One would think that using 4K or VR, in MP and with supercarrier + F-14A/B or F/A-18C, or F-16C in Syria, or P-47 in Normandy, would be a far more sensible test for GPU and VRAM stress utilization related aspects (framerate, frametime, temps, wattage, bandwitdth, etc), for comparison purposes with these higher-end GPUs.
Those not having such modules may be able to Trial them, then use them for such a "benchmark".

Edited December 27, 2022 by LucShep

[dock999]

Hi to all gentlemen :

Have we drawn a definite conclusion yet? , that is , buying a 4080 card is adequate to play DCS in VR? or  you should buy 4090 instead to get a non- blurred experience?

 

By reading those posts on the OpenXR Guide for WMR Headsets, to a get non-blurred image in DCS VR , you need to reach FPS 55.

 

Welcome comments from the owners of 4080 and 4090.

Edited December 27, 2022 by dock999

[ODBU_Mongoose]

10 hours ago, dock999 said:

Hi to all gentlemen :

Have we drawn a definite conclusion yet? , that is , buying a 4080 card is adequate to play DCS in VR? or  you should buy 4090 instead to get a non- blurred experience?

 

By reading those posts on the OpenXR Guide for WMR Headsets, to a get non-blurred image in DCS VR , you need to reach FPS 55.

 

Welcome comments from the owners of 4080 and 4090.

 

I'd say a 4080 would do. Went from a 6800XT to 4080 and the difference is noticable.

FPS are between 45-55 fps on busy MP servers but the experience is just butter smooth. No blurred experience, no stutters. Silky smooth. Great frametimes.

Running on a 13600K, G2, 64GB 3200 CL16

[Thinder]

21 hours ago, Johnny Dioxin said:

I've just done a flight in VR with my Pico 4 and the RTX 4080 - jaw-dropping smoothness. I used the same settings as with my 2080Ti, which always had either a solid 30 or 45fps, depending on headset used. Despite having my Pico 4 settings at 90Hz, DCS World ran at 72 - but OMG! Now I can finally up my settings. Not sure if the HD is processed yet - this is the onboard Pico headset recording - finally a VR headset manufacturer has included an easy onboard recording option - just a press of the screenshot button on the controller:

 

I recorded the test flight, which was just a take-off from Addis Ababa in the A-10A followed by a low-level flight around the city until my demise when I misjudged a couple of appartment blocks! Unfortunately, the commentary wasn't recorded. Nevertheless, fpsVR is visble for the whole flight - flatlining

 

It's  gonna be PICO 4 and RX 7900 XTX MBA 24GB for me, I suppose we'll have similar VR experiences...