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

[CBJ]

is the menu background even GPU rendered and not some standard video?

It is not a video. It's a scene built, and rendered, using the game engine. It's actually quite a good test of the GPU, as the CPU load for rendering it is a lot lower than in-game.

[Tempest]

is the menu background even GPU rendered and not some standard video?

It is not a video. It's a scene built, and rendered, using the game engine. It's actually quite a good test of the GPU, as the CPU load for rendering it is a lot lower than in-game.

Thank you, for confirming CBJ.

i spose there aren't any "rules" for the CPU at all? (PBO off for AMD etc.)

i'll have to take a look at MSI Afterburner and revert it to default clocks.

[spookywatcher]

If you sell the 6900XT at MSRP let me know

Sorry...it's in a glass showcase on my "I Love Me Wall". Took me 6 months of trying to get that thing. My favorite monitor is quite old so if it ever pops a cap and I got to get a new monitor... the 6900xt is going in. And I refuse to believe it's performance. I've a little bit of time today... hope to be able to work on it and get it working well.

[spookywatcher]

With the 3080ti installed I am pumping out 150 watts more into my ambient and it is very noticeable here in the summer months.

In play, at same settings and cpu bound fps, I would expect the 3080ti to be able to run at lower power (maybe after some fiddling with the power limit or v/f curve).

I could. However how I set my settings is as follows: I lower my graphics until there is basically no GPU load. Then fly around and see what and where my normal high cpu bottleneck is. e.g. near a busy station in busy sector. Then let's say my fps is 120 there. I'll then raise my graphics settings until I reach about 98% in that situation looking at the sun in black hole sun sectors. Because putting that directly in the crosshairs really loads the gpu up. Or the asteroid belt...etc you get the idea. So basically I have max cpu performance and max gpu in balance. I could set the same settings as the 1080ti but then...what's the point? So I just put the 1080ti back in. I'll probably put the 3080ti back in when the summer heat wave ends for us in socal.

[spookywatcher]

@ Spookywatcher

whats the reason for this test? is the menu background even GPU rendered and not some standard video?

your other thread seemed way more elaborate : viewtopic.php?t=427248

i should probably get off my arse and test the 5800x + 1080Ti even though the CPU is somewhat "old news" soon

Reason was.. I was lucky enough to be able to get two of the top of the line GPU's to compare with my old 1080ti. And when I saw the 6900xt choke at those settings and after confirming with multiple OS installs that it indeed was 160 fps I thought others should know. I still want the 6900xt to be my daily driver and will continue to work at it. Maybe with it being put in a 5950x system it'll be able to stretch it's legs. Hoping someone with that kind of setup lets us know what they are getting. Would save me $1100 for the cpu mobo combo.

And yes... I'm almost 100% positive the opening menu scene is a rendered scene with very little cpu overhead. And the results were very repeatable and didn't rely on savegames etc.

I think your 5800x and 1080ti are about the most perfect combo. I'm still tempted to go get the 5950x. Just still researching mobos and trying to find which one would best suit me and my goals for it.

[spookywatcher]

is the menu background even GPU rendered and not some standard video?

It is not a video. It's a scene built, and rendered, using the game engine. It's actually quite a good test of the GPU, as the CPU load for rendering it is a lot lower than in-game.

Thank you, for confirming CBJ.

i spose there aren't any "rules" for the CPU at all? (PBO off for AMD etc.)

i'll have to take a look at MSI Afterburner and revert it to default clocks.

Yeah.. for the CPU test... Bang away. If you can pump out the clocks and IPC... I'd love to see it! I'm thinking of getting the 5950x since in that thread he was king at 6-8% faster than my golden boy cpu.

[spookywatcher]

And again... I encourage anyone else with other gpu's to post what they get with those settings, especially any ampere and rdna2 gpu's. Main reason is...maybe some has a chance of upgrading to another GPU via craigslist or FB marketplace. Then they could know what to expect in gains compared to their other GPU's.

I'm hoping to test my 1050ti and rx290 today and posting those.

[Tempest]

5800X - PBO ON/Auto
4x8GB 3800 - XMP ON - subtimings untouched
471.68 W10 21H1
1080Ti stock

Code: Select all

1080Ti (strix gaming oc) = 125 Spire enters right, 122 while touching the X logo , 144 last the 1/4 before going off-screen 

same results for 4.10 Beta 6

[spookywatcher]

I gave the 6900 xt almost every chance I could (aside from buying the 5950x + x570 mobo). The performance just isn't there. I did a fresh install of win 10 with fresh driver install. Boosted the power cap to 293 watts (which is max for amd reference board). The clocks were regularly smashing 2555 mhz and even saw 2565 mhz a bunch.

The fps was a little better. Instead of 164 before I now got 176 fps. So a little better with the more refined windows drivers. But compared to the 3080ti it just sh*ts the bed. It's almost as if ray tracing were turned on.

The only other possible thing I could think of is that the low performance might be because of my test system. It's an Asus z170 w/ 6700k @ 5.0 ghz + 4.8 ghz uncore( I boosted it for the retest ) + 32 gig dual rank dual channel trident z 3600 mhz cl 15. The 6900xt was installed in the top x16 slot for all the tests because it didn't block my other pcie slots.

Maybe the z170 platform just doesn't like the RDNA 2. My platform crushes it with the 3080ti (even installed in the x8 slot), but maybe it's just an RDNA 2 thing. So until either I pick up the 5950x with a x570 or someone else with a 6800xt or 6900xt and updated platform like I mentioned can test it... We'll just not know. Really hope someone with the updated platform and the 6900xt tests it out and reports back.

D@mn... I really wanted that 6900xt to be my daily driver since I'm a linux guy.

[Panos]

I gave the 6900 xt almost every chance I could (aside from buying the 5950x + x570 mobo). The performance just isn't there. I did a fresh install of win 10 with fresh driver install. Boosted the power cap to 293 watts (which is max for amd reference board). The clocks were regularly smashing 2555 mhz and even saw 2565 mhz a bunch.

The fps was a little better. Instead of 164 before I now got 176 fps. So a little better with the more refined windows drivers. But compared to the 3080ti it just sh*ts the bed. It's almost as if ray tracing were turned on.

The only other possible thing I could think of is that the low performance might be because of my test system. It's an Asus z170 w/ 6700k @ 5.0 ghz + 4.8 ghz uncore( I boosted it for the retest ) + 32 gig dual rank dual channel trident z 3600 mhz cl 15. The 6900xt was installed in the top x16 slot for all the tests because it didn't block my other pcie slots.

Maybe the z170 platform just doesn't like the RDNA 2. My platform crushes it with the 3080ti (even installed in the x8 slot), but maybe it's just an RDNA 2 thing. So until either I pick up the 5950x with a x570 or someone else with a 6800xt or 6900xt and updated platform like I mentioned can test it... We'll just not know. Really hope someone with the updated platform and the 6900xt tests it out and reports back.

D@mn... I really wanted that 6900xt to be my daily driver since I'm a linux guy.

I hope you are NOT using MSI AB.

[spookywatcher]

I gave the 6900 xt almost every chance I could (aside from buying the 5950x + x570 mobo). The performance just isn't there. I did a fresh install of win 10 with fresh driver install. Boosted the power cap to 293 watts (which is max for amd reference board). The clocks were regularly smashing 2555 mhz and even saw 2565 mhz a bunch.

The fps was a little better. Instead of 164 before I now got 176 fps. So a little better with the more refined windows drivers. But compared to the 3080ti it just sh*ts the bed. It's almost as if ray tracing were turned on.

The only other possible thing I could think of is that the low performance might be because of my test system. It's an Asus z170 w/ 6700k @ 5.0 ghz + 4.8 ghz uncore( I boosted it for the retest ) + 32 gig dual rank dual channel trident z 3600 mhz cl 15. The 6900xt was installed in the top x16 slot for all the tests because it didn't block my other pcie slots.

Maybe the z170 platform just doesn't like the RDNA 2. My platform crushes it with the 3080ti (even installed in the x8 slot), but maybe it's just an RDNA 2 thing. So until either I pick up the 5950x with a x570 or someone else with a 6800xt or 6900xt and updated platform like I mentioned can test it... We'll just not know. Really hope someone with the updated platform and the 6900xt tests it out and reports back.

D@mn... I really wanted that 6900xt to be my daily driver since I'm a linux guy.

I hope you are NOT using MSI AB.

Correct. I am NOT using afterburner. This was done using Radeon Software.

[Panos]

I gave the 6900 xt almost every chance I could (aside from buying the 5950x + x570 mobo). The performance just isn't there. I did a fresh install of win 10 with fresh driver install. Boosted the power cap to 293 watts (which is max for amd reference board). The clocks were regularly smashing 2555 mhz and even saw 2565 mhz a bunch.

The fps was a little better. Instead of 164 before I now got 176 fps. So a little better with the more refined windows drivers. But compared to the 3080ti it just sh*ts the bed. It's almost as if ray tracing were turned on.

The only other possible thing I could think of is that the low performance might be because of my test system. It's an Asus z170 w/ 6700k @ 5.0 ghz + 4.8 ghz uncore( I boosted it for the retest ) + 32 gig dual rank dual channel trident z 3600 mhz cl 15. The 6900xt was installed in the top x16 slot for all the tests because it didn't block my other pcie slots.

Maybe the z170 platform just doesn't like the RDNA 2. My platform crushes it with the 3080ti (even installed in the x8 slot), but maybe it's just an RDNA 2 thing. So until either I pick up the 5950x with a x570 or someone else with a 6800xt or 6900xt and updated platform like I mentioned can test it... We'll just not know. Really hope someone with the updated platform and the 6900xt tests it out and reports back.

D@mn... I really wanted that 6900xt to be my daily driver since I'm a linux guy.

I hope you are NOT using MSI AB.

Correct. I am NOT using afterburner. This was done using Radeon Software.

Had to do a run with your settings as I was intrigued.

Code: Select all

5700XT = 70 fps when the spire & ships appear, 80 fps otherwise

Definitely something is wrong with the AMD performance with the X4 then.
The 5700XT should be in par with the GTX1080Ti if not faster as in all other Vulkan & DX games, the 1080Ti shouldn't be 60% faster.

FYI the core was working at 2024-2046 and have 1800Mhz VRAM. So above standard 5700XT (is watercooled) which is around 1875-1950 and 1750Mhz ram.
I use 3900X which is clocking at 4650Mhz and 3800C14 ram. So no issue with CPU.


Btw if anyone reading this and has access to a Radeon VII which is comparable to the 5700XT please run the test to see if it is the vram bus or not.

It is not a video. It's a scene built, and rendered, using the game engine. It's actually quite a good test of the GPU, as the CPU load for rendering it is a lot lower than in-game.

CBJ thoughts on the above? Seems AMD cards under performing against comparable GPU
eg 1080Ti is over 60% faster than the 5700XT which is illogical, any other game vulkan/DX the 5700XT is in par of faster at 2560x1440
Same applies to the 6900XT spookywatcher has against the 3080Ti.

I will investigate further if some settings is the one affecting the perf.

[spookywatcher]

I hope you are NOT using MSI AB.

Correct. I am NOT using afterburner. This was done using Radeon Software.

Had to do a run with your settings as I was intrigued.

Code: Select all

5700XT = 70 fps when the spire & ships appear, 80 fps otherwise

Definitely something is wrong with the AMD performance with the X4 then.
The 5700XT should be in par with the GTX1080Ti if not faster as in all other Vulkan & DX games, the 1080Ti shouldn't be 60% faster.

FYI the core was working at 2024-2046 and have 1800Mhz VRAM. So above standard 5700XT (is watercooled) which is around 1875-1950 and 1750Mhz ram.
I use 3900X which is clocking at 4650Mhz and 3800C14 ram. So no issue with CPU.


Btw if anyone reading this and has access to a Radeon VII which is comparable to the 5700XT please run the test to see if it is the vram bus or not.

It is not a video. It's a scene built, and rendered, using the game engine. It's actually quite a good test of the GPU, as the CPU load for rendering it is a lot lower than in-game.

CBJ thoughts on the above? Seems AMD cards under performing against comparable GPU
eg 1080Ti is over 60% faster than the 5700XT which is illogical, any other game vulkan/DX the 5700XT is in par of faster at 2560x1440
Same applies to the 6900XT spookywatcher has against the 3080Ti.

I will investigate further if some settings is the one affecting the perf.

Thank You Panos for that! Yeah.. there is definitely something going on with x4 and AMD gpu's (at least the ones we've looked at). At this point there is no doubt that ampere is the better choice for x4. I just reinstalled the 3080ti and retested to make sure I didn't miss anything. And once again.. the 3080ti crushes it at 232-238 for me.

On side note... stutter was Significantly better on both the 1080ti and 3080ti vs the 6900xt. I'm not sure if that is a pure gsync monitor vs freesync monitor issue. But being next to a station or dense asteroid field at a complete stop and just turning in a circle with either gsync or freesync the gsync monitor was MUCH MUCH more smooth and enjoyable. I'll have to test the 3080ti on the freesync monitor next to see if it is as butter smooth as the gsync monitor.

It's crazy because the 6900xt scored (eventually after the power limit raise) 10405 in unigine super 1080p extreme vs the 3080ti scoring 11457. With those metrics the 6900xt should have scored about 210 fps in the opening menu scene.

I know RDNA 2 sh*ts itself at 4k compared to ampere in virtually all benchmarks. Wonder if that has anything to do with it. Maybe vram throughput speed? But that doesn't explain why the 1080ti is not too far off from the 6900xt in the test.

Anyhow... looking forward to anyone else contributing scores.

*** ETA *** the 3080ti paired with the freesync monitor was just as butter smooth as the pure gsync monitor.

[Panos]

Thank You Panos for that! Yeah.. there is definitely something going on with x4 and AMD gpu's (at least the ones we've looked at). At this point there is no doubt that ampere is the better choice for x4. I just reinstalled the 3080ti and retested to make sure I didn't miss anything. And once again.. the 3080ti crushes it at 232-238 for me.

On side note... stutter was Significantly better on both the 1080ti and 3080ti vs the 6900xt. I'm not sure if that is a pure gsync monitor vs freesync monitor issue. But being next to a station or dense asteroid field at a complete stop and just turning in a circle with either gsync or freesync the gsync monitor was MUCH MUCH more smooth and enjoyable. I'll have to test the 3080ti on the freesync monitor next to see if it is as butter smooth as the gsync monitor.

It's crazy because the 6900xt scored (eventually after the power limit raise) 10405 in unigine super 1080p extreme vs the 3080ti scoring 11457. With those metrics the 6900xt should have scored about 210 fps in the opening menu scene.

I know RDNA 2 sh*ts itself at 4k compared to ampere in virtually all benchmarks. Wonder if that has anything to do with it. Maybe vram throughput speed? But that doesn't explain why the 1080ti is not too far off from the 6900xt in the test.

Anyhow... looking forward to anyone else contributing scores.

*** ETA *** the 3080ti paired with the freesync monitor was just as butter smooth as the pure gsync monitor.

You are welcome.

There is some out of place performance issue. Nothing to do with the GPU. 6900XT is faster than the 3090 in many games at 2560x1440 regardless the API let alone the 3080Ti.
Even the 5700XT is faster or in par to 1080Ti in Vulkan games. So there is something to do with X4

[Skeeter]

95 with those settings, when spire comes it drops to 80.

Specs in sig.

Using 21.6.1 whql driver.

Btw anyone with a rdna2 card like 6xxx series, can you test something for me. Ultra more or less, glow high. 1440p. msaa 4x. Do that then when in a station like wharf for argon style ones, enable in the latest beta FSR ultra. Can you let me know if you see blinging flashing colored lights around as you move the cam, sorta stand where the big ships would land. Need to know if just me or others get artifacts with rdna 2 cards using msaa 4 and fsr ultra. msaa2 also causes it but no flashing lights but little black boxes near top of the screen when in managers office sometimes lookin at the orange walls near the top of the walls.

Cheers.

[Wanabe]

I am reporting in using an I7-7700K along with an AMD 6800, 2X 8GB CL16 3400
X4 Foundations version 4.0. Using the same settings as outlined in your original post, 1440P ETC.

Using a slightly out of date GPU driver on this test: 21.6.1, the current being 21.8.1. My GPU clocks to 2200MHZ core and 2000MHZ memory during the test. GPU load fluctuates between 96%-98%.

135 FPS for most of the scene, until the fighters and corvettes fly by and then it drops down to about 115 during their flyby and for a bit after, after-which it goes back upto 135FPS and later dips down to 125FPS for a short while.

[Panos]

I am reporting in using an I7-7700K along with an AMD 6800, 2X 8GB CL16 3400
X4 Foundations version 4.0. Using the same settings as outlined in your original post, 1440P ETC.

Using a slightly out of date GPU driver on this test: 21.6.1, the current being 21.8.1. My GPU clocks to 2200MHZ core and 2000MHZ memory during the test. GPU load fluctuates between 96%-98%.

135 FPS for most of the scene, until the fighters and corvettes fly by and then it drops down to about 115 during their flyby and for a bit after, after-which it goes back upto 135FPS and later dips down to 125FPS for a short while.

Thank you.

So seems there is performance issue with AMD cards with X4 Foundations by a lot.

[spookywatcher]

I am reporting in using an I7-7700K along with an AMD 6800, 2X 8GB CL16 3400
X4 Foundations version 4.0. Using the same settings as outlined in your original post, 1440P ETC.

Using a slightly out of date GPU driver on this test: 21.6.1, the current being 21.8.1. My GPU clocks to 2200MHZ core and 2000MHZ memory during the test. GPU load fluctuates between 96%-98%.

135 FPS for most of the scene, until the fighters and corvettes fly by and then it drops down to about 115 during their flyby and for a bit after, after-which it goes back upto 135FPS and later dips down to 125FPS for a short while.

a 6800 scoring just a smidge over a 1080 ti ? Crazy business. If this gets figured out and performance goes to inline where it "should" theoretically be... I'll be very happy and put that 6900xt back in.

[Skeeter]

Wonder what cbj would say about this discussion.

Maybe ego can reach out to amd and discuss optimization for their engine with them to try get some more performance out of amd cards vs nvidia ones.

Wonder if this would help ego for optimizing?

https://gpuopen.com/performance/
https://gpuopen.com/vulkan/?sessionGUID ... 78699a5502
https://gpuopen.com/rgp/ - really hope they can use this to deep dive into low level reporting to see the bottlenecks with this tool. Heck we can get it save a frame snapshot for you ego then send u it to analyse if you want.

Maybe one of the others might wana post their findings and results with a link to here on amd forums here as may help if a dev sees it there too?
https://community.amd.com/t5/graphics/b ... iscussions

[Panos]

Wonder what cbj would say about this discussion.

Maybe ego can reach out to amd and discuss optimization for their engine with them to try get some more performance out of amd cards vs nvidia ones.

Wonder if this would help ego for optimizing?

https://gpuopen.com/performance/

Maybe one of the others might wana post their findings and results with a link to here on amd forums here as may help if a dev sees it there too?
https://community.amd.com/t5/graphics/b ... iscussions

Yep. Clearly there is a huge perf issue with RDNA1 & 2, not only 2.

Added the post in the AMD forums.
https://community.amd.com/t5/graphics/r ... 604#M78542




Update. My post in AMD forums is marked as spam