Post your GPU scores - to prep for next gen cpu's I have 1080ti 3080ti 6900xt included

[Panos]

Ah gotcha.

Also i wondered, if it affects anyone if you disable fullscreen optimizations from right clicking the exe and go to compatibility and doing what i just said about fullscreen thing. Probs not but wondered. Or maybe running the game as administrator.

Tried them on windows. Nothing.

On Linux we do not have such options as the game is native Vulkan, doesn't go through the DXVK translation.

[Panos]

Quick dirty overclock in a 77 deg F room on the 3080ti. Can't wait until I get some time to really push this card. Looks like a great piece of silicon.

Code: Select all

3080 ti w/ a moderate overclock in X4 test = 247 fps

Seriously looks like Nvidia is the way to go with x4. I didn't even try on the overclock. Got 203 rd place equivalent in Unigine Super @ 13277 pts

Since you are using linux, have a look the link bellow. Added a vkBasalt configuration file to use RealSpace reshade in Linux too.

Your 3080Ti wouldn't bother with 8 fps drop

viewtopic.php?f=180&t=414524&p=5079293#p5079293

[Skeeter]

thats going off topic recommending reshade etc, lets keep it results only, and u still havent updated the topic on amd forums btw just a fyi.

[Panos]

thats going off topic recommending reshade etc, lets keep it results only, and u still havent updated the topic on amd forums btw just a fyi.

Why you do not add to it?

[Panos]

AMD posted new drivers today 21.9.1 activating SAM for 5000 series
With the 1st page settings there is no FPS drop when the ships appeared

Code: Select all

5700XT = 81 fps when the spire & ships appear, 83 fps otherwise

yet still 1080Ti a comparable GPU is still 40% faster.

Anyone from Egosoft has looked at this?

6700XT

Skeeter what I see is my 5700XT with 21.9.1 (added SAM support for 5000 series) performs exactly the same when the ships are appearing to your 6700XT!!!!!!!
That shouldn't be happening, your card is faster than mine in general by 40% (and the GTX1080Ti)

[Skeeter]

@timon37 @cbj

Now that 4.10 is done, can you guys look into this please?

[Panos]

@timon37 @cbj

Now that 4.10 is done, can you guys look into this please?

Imho considering we had no responces yet, Egosoft should help us to find some RTX 3080 at MSRP, I will buy it
It won't cost them anything compared to fixing the engine and we are the only ones complaining

[Panos]

Windows was identical for the 3080ti. But yeah... I'm very happy linux is gaining ground. Especially with the steam deck! Now maybe all game devs will really start taking linux serious.

*** Edit : Ooops I see what you mean after re-reading. Yes... Linux undoubtedly appears to be way smoother with regard to frame time excursions when cpu bound for me. Love Linux!

Have you activated SAM & NGCC on Linux for the AMD cards if you use MESA 21+?

You can force it on ALL cards (including FuryX, Vega etc) by adding to the environmental variables

Code: Select all

RADV_PERFTEST=sam,ngcc

For those having 6000 series, try to see if there is any difference by activating VRS by brute force

Code: Select all

RADV_FORCE_VRS=2x2
RADV_DEBUG=novrsflatshading

[spookywatcher]

This took a very long time. Hope you enjoy it.

I couldn't just sit and not investigate where the 6900xt was sh*tting the bed. I just had to know where it started to puke out. All fps scores were taken after each setting change and a game restart. I started with the absolute minimum settings (except for 1440p monitor). Everything was off or zero or minimum. Then I changed each MSAA. Then I added LOD and Effect Distance to 100%. Let's call that "Baseline". Then Changed each setting for each MSAA setting. Just adding the settings to the previous settings. Anyhow... You'll get the idea and then can see where the perf justs nerfs. I hope the formatting turns out correct so it's easy to read. And this was still done on 4.0 hf3 because GOG still doesn't have 4.10 out for linux.

3080ti - Linux bone stock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 304     283             283        283           283            280       280       280          280
2X  = 304     283             283        283           283            280       280       280          279
4X  = 304     283             283        283           280            280       278       263          232
8X  = 273     241             241        230           230            197       183       173          157

6900 xt - Linux (kernel drivers) 294 watts max power but no actual overclock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 298     275             275        275           275            275       275       275          248
2X  = 298     275             275        275           271            260       242       225          207
4X  = 298     275             275        238           238            195       178       175          155
8X  = 152     134             132        126           126            102        90        84           78

6900 xt - Win 10 294 watts max power but no actual overclock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 286     265             265        263           262            262       262       262          253
2X  = 286     265             265        262           262            262       262       258          215
4X  = 286     265             265        258           255            216       197       192          167
8X  = 165     144             144        135           130            113        97        91           86

[Panos]

This took a very long time. Hope you enjoy it.

I couldn't just sit and not investigate where the 6900xt was sh*tting the bed. I just had to know where it started to puke out. All fps scores were taken after each setting change and a game restart. I started with the absolute minimum settings (except for 1440p monitor). Everything was off or zero or minimum. Then I changed each MSAA. Then I added LOD and Effect Distance to 100%. Let's call that "Baseline". Then Changed each setting for each MSAA setting. Just adding the settings to the previous settings. Anyhow... You'll get the idea and then can see where the perf justs nerfs. I hope the formatting turns out correct so it's easy to read. And this was still done on 4.0 hf3 because GOG still doesn't have 4.10 out for linux.

3080ti - Linux bone stock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 304     283             283        283           283            280       280       280          280
2X  = 304     283             283        283           283            280       280       280          279
4X  = 304     283             283        283           280            280       278       263          232
8X  = 273     241             241        230           230            197       183       173          157

6900 xt - Linux (kernel drivers) 294 watts max power but no actual overclock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 298     275             275        275           275            275       275       275          248
2X  = 298     275             275        275           271            260       242       225          207
4X  = 298     275             275        238           238            195       178       175          155
8X  = 152     134             132        126           126            102        90        84           78

6900 xt - Win 10 294 watts max power but no actual overclock

Code: Select all

MSAA          LOD &         Texture     Shadows      Soft Shad        SSAO       Glow    Chrom +       Fog 
             Effect Dist     High        High          On             High      High    Distort        High
Off = 286     265             265        263           262            262       262       262          253
2X  = 286     265             265        262           262            262       262       258          215
4X  = 286     265             265        258           255            216       197       192          167
8X  = 165     144             144        135           130            113        97        91           86

Wow

That is a lot of data thank you.
My main question is why there is no impact on the 3080Ti except on maxed out settings.
Have you activated NGCC (culling) on Linux when using the 6900XT?
Do you get an orange warning on the 3080Ti or 6900XT when selecting MSAA 4x + and Fog?

Anyone from Egosoft has any idea what is going on?

[spookywatcher]

Wow

That is a lot of data thank you.
My main question is why there is no impact on the 3080Ti except on maxed out settings.
Have you activated NGCC (culling) on Linux when using the 6900XT?
Do you get an orange warning on the 3080Ti or 6900XT when selecting MSAA 4x + and Fog?

Anyone from Egosoft has any idea what is going on?

Nah.. I already pulled it back out and put the 3080ti back in. The 6900xt with the right graphics settings was much nicer this time around...but it still was WAY too stuttery still even with some tweaks and Freesync monitor. Plus I wanted to put back the 3080ti for my next set of benchmarks, this time it's going to be all about memory timings, mem speed in relation to cpu clock speed etc etc. This should be a good one.

[Falcrack]

I checked newegg for GTX 2060 prices, a logical upgrade to my GTX 1060. Lowest price is $654, with $45 shipping.

Nope, no way, not time for an upgrade.

[Panos]

I checked newegg for GTX 2060 prices, a logical upgrade to my GTX 1060. Lowest price is $654, with $45 shipping.

Nope, no way, not time for an upgrade.

Go to EVGA queue. You will get a 3060.

[Panos]

does disabling in steams settings under shaders them 2 shader options help with fps in any way?

I wonder if we can run the Vulkan Configurator to find out where there is the issue

https://gpuopen.com/learn/vulkan-best-practice-layer/

[Panos]

Installed new 21.10.1

No change. 5700XT perf is crippled still.

Any idea if/how can use the new AMD Vulkan profiler to assist Egosoft on finding out why the game perf is crippled on ALL AMD GPUs?

We are now on 2nd month and no respond by devs even if have reported this thread to perf issues discussion.

Tonight I will install new MESA and new AMDVLK with kernel 15.5rc4 and going to test all on Linux too.

[Skeeter]

Yes it would be helpful if ego posted here and said yes we see a issue and are working on it or something.

[Panos]

Yes it would be helpful if ego posted here and said yes we see a issue and are working on it or something.

Willing to help as tester for free. I have the time and the patience. The latter due to my job as tech analyst and developer for 25 years

But...

[spookywatcher]

anecdotal evidence...

As I have said before the 3080 ti (nvidia in general) is the way to go with X4 vs any AMD gpu for some reason. And full disclosure I have NOT tested what I'm about to write with the 6900xt yet. I've been really busy and scrambling with my job and the recent mandates so I haven't had the opportunity to test it yet with the 6900xt.

But I stumbled across something that I already knew but didn't think that it would be a factor. In Linux the cpu governor defaults to the powersave governor mode. And during testing the clocks on my 6700k show locked in at 4.8 ghz (which is my setting during summer time) even with the default powersave cpu governor active. In powersave mode the cpu will run @ 4.8 ghz and in the overlay I don't see any deviations from 4.8 ghz.

However I just tested changing the pstate to performance mode with the code, this causes all cores to run at max cpu freq all the time:

Code: Select all

echo "performance" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

And it appears that the stutter (1% and .01% lows) is much less in all areas. Again.. I have not tested this with the 6900xt yet but being X4 is a mostly cpu bound game I wondered if the 1% and .01% lows would improve if I set the cpu governor to performance. And the problem with space sims is... The 1% and .01% lows are VERY noticable. Not so with first person shooters. And after setting the governor to performance mode by the code mentioned above, all aspects of X4 seem much much smoother.

Now being in the cooler months, I felt comfortable in overclocking a bit more. So first I tried matching my "uncore" (meaning cpu cache etc) to match the cpu clock of 4.8 ghz. It has been at 4.8ghz cpu and 4.7 ghz uncore previously due to voltage required. So now it's 4.8 ghz and 4.8ghz uncore. And I got about 7 fps more in testing and couple with cpu governor of "performance" it's even better. When I clock the cpu to 5.0 ghz and the uncore to 4.8 ghz (I can't get uncore to 5.0 ghz) it's about 10 fps more. And in both cases the "stutter" was significantly better and even more so better when set the cpu governor was set to "performance" mode by the code above. But again...even without the extra overclocks (using my summer overclock of 4.8ghz cpu and 4.7 ghz uncore) it was noticeably smoother.

Hopefully I get some free time and can test with the 6900xt but I think that's more of an amd thing.

So just a heads up. If on linux... set your cpu governor to performance mode by the code above. It'll revert back to powersave after a reboot, so no worries there. There are "fixes" to have it be persistent after reboot, but I don't need that. It's just really easy to open terminal and set it via command line before a gaming session.

** ETA ** I don't know of any way to measure the 1% or .01% lows on Linux. So this is just a "feel" thing.

[Panos]

anecdotal evidence...

As I have said before the 3080 ti (nvidia in general) is the way to go with X4 vs any AMD gpu for some reason. And full disclosure I have NOT tested what I'm about to write with the 6900xt yet. I've been really busy and scrambling with my job and the recent mandates so I haven't had the opportunity to test it yet with the 6900xt.

But I stumbled across something that I already knew but didn't think that it would be a factor. In Linux the cpu governor defaults to the powersave governor mode. And during testing the clocks on my 6700k show locked in at 4.8 ghz (which is my setting during summer time) even with the default powersave cpu governor active. In powersave mode the cpu will run @ 4.8 ghz and in the overlay I don't see any deviations from 4.8 ghz.

However I just tested changing the pstate to performance mode with the code, this causes all cores to run at max cpu freq all the time:

Code: Select all

echo "performance" | sudo tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

And it appears that the stutter (1% and .01% lows) is much less in all areas. Again.. I have not tested this with the 6900xt yet but being X4 is a mostly cpu bound game I wondered if the 1% and .01% lows would improve if I set the cpu governor to performance. And the problem with space sims is... The 1% and .01% lows are VERY noticable. Not so with first person shooters. And after setting the governor to performance mode by the code mentioned above, all aspects of X4 seem much much smoother.

Now being in the cooler months, I felt comfortable in overclocking a bit more. So first I tried matching my "uncore" (meaning cpu cache etc) to match the cpu clock of 4.8 ghz. It has been at 4.8ghz cpu and 4.7 ghz uncore previously due to voltage required. So now it's 4.8 ghz and 4.8ghz uncore. And I got about 7 fps more in testing and couple with cpu governor of "performance" it's even better. When I clock the cpu to 5.0 ghz and the uncore to 4.8 ghz (I can't get uncore to 5.0 ghz) it's about 10 fps more. And in both cases the "stutter" was significantly better and even more so better when set the cpu governor was set to "performance" mode by the code above. But again...even without the extra overclocks (using my summer overclock of 4.8ghz cpu and 4.7 ghz uncore) it was noticeably smoother.

Hopefully I get some free time and can test with the 6900xt but I think that's more of an amd thing.

So just a heads up. If on linux... set your cpu governor to performance mode by the code above. It'll revert back to powersave after a reboot, so no worries there. There are "fixes" to have it be persistent after reboot, but I don't need that. It's just really easy to open terminal and set it via command line before a gaming session.

** ETA ** I don't know of any way to measure the 1% or .01% lows on Linux. So this is just a "feel" thing.

Cheers

Btw have you activated SAM and culling with the 6900XT on Linux in the environmental variables?

Code: Select all

RADV_PERFTEST=sam,nggc

Optional at the command line parameter for the game add radeonsi_enable_sam=true

To make sure it runs after the restart

Code: Select all

AMD_DEBUG=info glxinfo | grep vram

If

Code: Select all

 vram_vis_size

is equal to your

Code: Select all

vram_size

then sam is enabled and working.


FYI since you are doing a lot of testing and fiddling have a look here
https://gazoche.xyz/fps-overlays-on-linux.html

The Mesa Vulkan overlay section as is the one running in Vulkan and has a nice option to extract the log into file.

[Panos]

anecdotal evidence...

I was looking for something irrelevant and this came to my attention.

Look at Basemark Vulkan chart. Is matches the perf differences we see in X4 Foundations eg 5700XT vs 1080Ti

https://www.guru3d.com/articles_pages/r ... ew,23.html

However that doesn't reflect on games like RDR2 (is same on all other Vulkan games)

https://www.guru3d.com/articles-pages/a ... ew,10.html

So it seems, there is somewhere Egosoft might be able to ask help from AMD to improve the perf.