RTX 4080 & 7900 XTX/XT benchmarks DCS

[Caysuyu]

13 minutes ago, HansPeter1981 said:

There is a chance you are the first one running DCS with the 7900XTX. No pressure . AMD graphic card owners in the DCS community will be pressing the refresh Button all day... to wait for your feedback... 

Don't want to add pressure neither but... You are our only hope @Th3ChosenOn3 (pun intended) 

[AlpineGTA]

Yes, I have a XFX Merc 319 AMD RX 6900XT, so it will be interesting to see.

[Th3ChosenOn3]

I've done a quick benchmark with the Plazma Torture map using the RX 7900 XTX. I'm getting double the average framerate compared to the GTX 1080, and going from being GPU bottlenecked, to CPU bottlenecked. Overall quite impressed, although I'm not sure how it compares to competing cards.

When I get some time tomorrow, I'll do a few more benchmarks as were suggested here.

Note: Timestamps added on the video.

 

Edited December 15, 2022 by Th3ChosenOn3
Timestamps

[79Au]

Huh. I also have a GTX 1080 and 5800X CPU. But the problem is, the 7900XTX is more than twice as expensive, so we still get the same fps/$ as in 2017... Yawn. I'm not an eye candy guy. F4.0 has better SAMs, better ATC, better AI, dynamic campaign, and it runs on a Pentium III 1GHz and 64MB GPU. (No, we don't need a nasa computer or multithreading™ before the DCS core can be improved)

 

[HansPeter1981]

Interesting. I would say considering the GPU was never at 100% it did well...
I tried to run the same test with my 6800XT and 5800X3D.
DCS  Open Beta 2.8.0.33006 (Hope your Video was on the same Version?)

Most of the time my GPU was at 99%.
Same Settings and resolution just to compare.
I tried to fly same Altitude, might have been slightly off to the right. If I play the trk file she always pitches up and down so I flew it live with ATT hold

I took a screen shot and compared DATA for the "same moment"

I have 85 FPS at 99% 
Your 7900XTX has 82 at 67% - In theory that would be 121 at 99%
Your 1080TI has 43 at 99%

So in theory based on those results the 7900XTX performs:
1.4 times in comparison to the 6800XT (70% of 7900XTX)
2.8 times in comparison to the 1080TI (35% of 7900XTX)
at 1440p.

It would be interesting to know if the stack is the same at 4k or even better in comparison.
 

Edited December 16, 2022 by HansPeter1981

[Lurker]

16 hours ago, SkateZilla said:

The issue is,

You can seat them correctly, but the clipping mechanism on them don't always hold them in. Especially with side to side torque is applied.

If you are applying torque to the cables, after you've plugged them in, then you need to recheck the seating. This is standard for any build I've ever done, regardless of what Im doing. Could Nvidia have made this a bit more foolproof? Sure. Are most people idiots who rush through their work without a second thought? Absolutely. 

Edited December 16, 2022 by Lurker

[xoxen]

9 hours ago, Th3ChosenOn3 said:

I've done a quick benchmark with the Plazma Torture map using the RX 7900 XTX. I'm getting double the average framerate compared to the GTX 1080, and going from being GPU bottlenecked, to CPU bottlenecked. Overall quite impressed, although I'm not sure how it compares to competing cards.

When I get some time tomorrow, I'll do a few more benchmarks as were suggested here.

Note: Timestamps added on the video.

 

 

Thanks a lot for sharing, very interesting. As you have the same CPU than I have I will see if I can simulate the same resolution on my wide screen display. Should be feasable. Then we can theoretically compare the 7900 XTX with the 4080. Actually I´m not at home but at the weekend I should find the time. Never did a video so most probably I will post screenshots.

[m005106]

12 hours ago, Th3ChosenOn3 said:

I've done a quick benchmark with the Plazma Torture map using the RX 7900 XTX. I'm getting double the average framerate compared to the GTX 1080, and going from being GPU bottlenecked, to CPU bottlenecked. Overall quite impressed, although I'm not sure how it compares to competing cards.

When I get some time tomorrow, I'll do a few more benchmarks as were suggested here.

Note: Timestamps added on the video.

 

 

Congrats on the upgrade.

I noticed some micro stuttering after feet dry. Is the driver WHQL? I noticed stutters on my 6800xt while running optional non WHQL drivers.

[Th3ChosenOn3]

26 minutes ago, m005106 said:

Congrats on the upgrade.

I noticed some micro stuttering after feet dry. Is the driver WHQL? I noticed stutters on my 6800xt while running optional non WHQL drivers.

Thank you. 

I'm on the Windows 10 WHQL drivers. I don't remember experiencing micro stutters while running the benchmark, but I wasn't heavily focused on the screen.

I did notice glitchy smoke in the benchmark, and clearing my FXO/Metashaders folders today seems to have gotten rid of them. I've also enabled Adaptive Sync, and had a very smooth experience for the two hours I've been able to play today, but I'm not sure if I should keep it on for the benchmarks as I'm not sure if it would affect FPS.

30 minutes ago, skywalker22 said:

What is this app called? In top left corner.

MSI Afterburner with Rivatuner Statistics.

Edited December 16, 2022 by Th3ChosenOn3

[Th3ChosenOn3]

3 hours ago, xoxen said:

Thanks a lot for sharing, very interesting. As you have the same CPU than I have I will see if I can simulate the same resolution on my wide screen display. Should be feasable. Then we can theoretically compare the 7900 XTX with the 4080. Actually I´m not at home but at the weekend I should find the time. Never did a video so most probably I will post screenshots.

Yeah, it would be great to see a benchmark with your 4080, as we have similar systems.

If you change your resolution to 2560x1440 and enable fullscreen, the image will get stretch over the entire screen (What happened to me) and you'll be able to do the benchmark with no issues on your ultrawide.

[m005106]

@Th3ChosenOn3 Yeah, it might have been the youtube compression making it look like stutter.

Well it seems like I might make the jump to 7900xtx after all. Thanks for the benchmark!

[xoxen]

11 minutes ago, Th3ChosenOn3 said:

Yeah, it would be great to see a benchmark with your 4080, as we have similar systems.

If you change your resolution to 2560x1440 and enable fullscreen, the image will get stretch over the entire screen (What happened to me) and you'll be able to do the benchmark with no issues on your ultrawide.

I can even split the screen of the ultrawide into two halfs, in that case I should not even have the streched issue. I´m curious and most probably will do it tonight already. But congrats to your new card and the good results. A good buy from my point of view.

 

If you can sent me your track file I could run the identical benchmark.

Edited December 16, 2022 by xoxen

[Andrei]

Thanks @Th3ChosenOn3 for the benchmark. Really good stuff.

@xoxen looking forward to your run with 4080. That might be the final validation I need to choose for one or another card  I would expect them to be quite close though.

[Th3ChosenOn3]

On 12/15/2022 at 10:28 AM, frachy said:

Can you have a test of the DCS Default "AH 64D, Instant action, runway start" and "F16C, Instant action, Take off" with DCS Default "High" Image quality settings? So I can compare to my 3080 12G.

Wasn't sure which resolution you wanted me to use, so I used my native 3440x1440 resolution with the High Preset. I also didn't really know what to do, so I flew around for a bit. Hope it helps with your comparison.

14 minutes ago, skywalker22 said:

Just a bit of a help with MSI Afterburner. I set all I want to see, but I cannot see any of the Framerate parameters, except Framerate (which shows current fps). I cannot see Min, Avg, Max, 1% and 0.1% in the app. And I have set the to ASD. Anyone know why?

You'll have to setup Benchmark key binds, and then press Begin Recording to have them show up.

[frachy]

29分钟前，skywalker22说：

Works, thx @Th3ChosenOn3.

Was just wondering, why I don't have more FPS, since GPU and CPU are far from their max, at least based on these numbers.

Currently DCS use only 1 core of CPU (except some audio processing ), your CPU's 1 core is at 100% now. Upgrade CPU will solve this, even an 12400F.

Edited December 16, 2022 by frachy

[Dovivan]

Just my experienc. Upgrade from 6900XT to 4080 (Gigabyte AERO). Playing VR (G2) on OpenXR with 100% render and 80% NIS/FSR. Now i can change settings from visibility Hight to Extreme and shadows from FLat to Hight in my testing scene. (freeflight above marianas the biggest city with Hornet and 3 layers clouds). (2X MSAA) still testing but looks like good improve. But i needed upgrade win11. I had lot stuttering after instal new GPU in Mixed Reality. After update (some cum. updade 22H2) looks fine no stuttering with this settings. Today open beta patch

Edited December 16, 2022 by Dovivan

[frachy]

1小时前，Th3ChosenOn3说：

Wasn't sure which resolution you wanted me to use, so I used my native 3440x1440 resolution with the High Preset. I also didn't really know what to do, so I flew around for a bit. Hope it helps with your comparison.

 

You'll have to setup Benchmark key binds, and then press Begin Recording to have them show up.

Thank you. I've test it with 3080 12G overclocked about 100Mhz, about the same performance as 3080TI. at 2560x1440 the GPU Load and FPS is almost the same as your video, and 3440/2560=1.34. at 3840*2160 the FPS is about 60%-70% of your video but the screen pixels are 1.67 times too. Seems different resolution's comparison is hard...

I've compared my 3080 12G oced with others 4090, the score is about 1/1.7. 

So roughly guess, I thought (irresponsibly)  3080Ti vs 7900XTX vs 4090 is about 1 / 1.35 / 1.7 ，which is about the same as other games benchmark. These numbers are not accurate but only give a rough idea.

Edited December 16, 2022 by frachy

[Th3ChosenOn3]

6 hours ago, xoxen said:

If you can sent me your track file I could run the identical benchmark.

I forgot to save a track file, but I've created another for you to use.

 

20 minutes ago, frachy said:

Thank you. I've test it with 3080 12G overclocked about 100Mhz, about the same performance as 3080TI. at 2560x1440 the GPU Load and FPS is almost the same as your video, and 3440/2560=1.34. at 3840*2160 the FPS is about 60%-70% of your video but the screen pixels are 1.67 times too. Seems different resolution's comparison is hard...

I've compared my 3080 12G oced with others 4090, the score is about 1/1.7. 

So roughly guess, I thought (irresponsibly)  3080Ti vs 7900XTX vs 4090 is about 1 / 1.35 / 1.7 ，which is about the same as other games benchmark. These numbers are not accurate but only give a rough idea.

 

I can run the same mission again at 3840*2160 to give a better estimation of GPU performance if it would be helpful?

Plazma Torture Map - Level Flight.trk

[frachy]

1分钟前，Th3ChosenOn3说：

I forgot to save a track file, but I've created another for you to use.

 

I can run the same mission again at 3840*2160 to give a better estimation of GPU performance if it would be helpful?

Plazma Torture Map - Level Flight.trk 623.2KB · 0次下载

Yes that's definitely helpful.

[SkateZilla]

 

[CybrSlydr]

If this is true, this is one of the biggest OOFs I've seen in a long time.

https://wccftech.com/rumors-allege-amd-shipped-unfinished-rdna-3-navi-31-a0-gpu-silicon-in-radeon-rx-7900-series/

Rumors Allege AMD Shipped Unfinished RDNA 3 Navi 31 A0 GPU Silicon In Radeon RX 7900 Series

As discovered by Kepler_L2, it looks like early RDNA 3 silicon had a non-working shader prefetch HW. This was featured in three chips, the GFX1100 (Navi 31), GFX1102 (Navi 33), & GFX1103 (the APU lineup consisting of Phoenix chips). Now based on the latest GitHub submission. Kepler is also suggesting that Navi 32 GPUs are actually based on the 'GFX1103' IP whereas Navi 33 chips feature the 'GFX1102' IP. The same issues are found on every other chip besides Navi 32 which is going to be featured in several mainstream discrete GPUs on desktop and mobility platforms by early 2023.

...

But that's not all, other major features such as the VOPD instructions featured on the RDNA 3 GPUs claimed to offer a big improvement in performance but in reality, they only managed to deliver a 4 percent improvement over RDNA 2 in ray tracing titles. The RDNA 3 silicon was designed to handle dual Wave32 instructions for twice the floating point performance by utilizing 64 multi-precision & multi-purpose ALUs implemented across 2 SIMD32 units. Talking to an AMD representative, folks over at HardwareTimes were able to confirm that AMD is currently in the process of fine-tuning the performance here and expects optimizations along the way:

[LucShep]

38 minutes ago, CybrSlydr said:

If this is true, this is one of the biggest OOFs I've seen in a long time.

https://wccftech.com/rumors-allege-amd-shipped-unfinished-rdna-3-navi-31-a0-gpu-silicon-in-radeon-rx-7900-series/

Rumors Allege AMD Shipped Unfinished RDNA 3 Navi 31 A0 GPU Silicon In Radeon RX 7900 Series

As discovered by Kepler_L2, it looks like early RDNA 3 silicon had a non-working shader prefetch HW. This was featured in three chips, the GFX1100 (Navi 31), GFX1102 (Navi 33), & GFX1103 (the APU lineup consisting of Phoenix chips). Now based on the latest GitHub submission. Kepler is also suggesting that Navi 32 GPUs are actually based on the 'GFX1103' IP whereas Navi 33 chips feature the 'GFX1102' IP. The same issues are found on every other chip besides Navi 32 which is going to be featured in several mainstream discrete GPUs on desktop and mobility platforms by early 2023.

...

But that's not all, other major features such as the VOPD instructions featured on the RDNA 3 GPUs claimed to offer a big improvement in performance but in reality, they only managed to deliver a 4 percent improvement over RDNA 2 in ray tracing titles. The RDNA 3 silicon was designed to handle dual Wave32 instructions for twice the floating point performance by utilizing 64 multi-precision & multi-purpose ALUs implemented across 2 SIMD32 units. Talking to an AMD representative, folks over at HardwareTimes were able to confirm that AMD is currently in the process of fine-tuning the performance here and expects optimizations along the way:

WOW  
If true, that's a severe mishap by AMD, and some pretty heavy discounts on the RX7900 series are in order... which would actually make it better!  LOL

Meanwhile, an unlikely turn of events has just happened (maybe due to the underwhelming reception of the AMD RX7900 series)... 
NVIDIA GeForce RTX 4080 takes up best selling GPU spot at Newegg:
https://wccftech.com/nvidia-geforce-rtx-4080-becomes-neweggs-best-seller-rtx-4090-takes-3rd-spot/

These are strange times, when a overpriced $1600+ GPU that was destined to rot in the shelves just a week ago, becomes the most sold GPU of them all.   

Edited December 16, 2022 by LucShep

[CybrSlydr]

38 minutes ago, LucShep said:

WOW  
If true, that's a severe mishap by AMD, and some pretty heavy discounts on the RX7900 series are in order... which would actually make it better!  LOL

Meanwhile, an unlikely turn of events has just happened (maybe due to the underwhelming reception of the AMD RX7900 series)... 
NVIDIA GeForce RTX 4080 takes up best selling GPU spot at Newegg:
https://wccftech.com/nvidia-geforce-rtx-4080-becomes-neweggs-best-seller-rtx-4090-takes-3rd-spot/

These are strange times, when a overpriced $1600+ GPU that was destined to rot in the shelves just a week ago, becomes the most sold GPU of them all.   

 

Yeah, but how many of those buying the 4080 are scalpers trying to take advantage of the 7900XTX whimper?

[Th3ChosenOn3]

3 hours ago, frachy said:

Yes that's definitely helpful.

Seems I won't be able to test a resolution above 3440x1440. Unfortunately I can't seem to get the game to behave properly with a resolution of 3840*2160 on my 21:9 aspect ratio monitor.

Sorry.

[xoxen]

This is my first video and regarding the settings I followed the ones mentioned before:

For me it looks like the 4080 is slower than the 7900XTX. I opened the track file and ran it. Comparing different scenes show sometimes more, but overall less FPS. Not sure how to judge this for now. Actually my CPU is undervolted, but it runs with the same 4850 MHz than in the other system.

Edit: I carefully compared the videos. It is not a clear picture. The 7900 starts strong, has a weaker phase but is stronger when flying over the river later on. There the 4080 dips more down, even into the 50s. My conclusion is that both have it strengths. Hard to say.

Edit 2: For the video I used the Nvidia driver 526.98 WHQL, I just installed the newer version 527.56 WHQL and it performs better. I will create a video tomorrow as I´m tired and I´m ready to go to bed.

Edited December 16, 2022 by xoxen
Additional information