why does mesa report cards as 570/580/590
they are all different GPUs. which is it then?... thats kinda stupid imo

Last edited by Xpander on 24 Jul 2020 at 5:42 pm UTC

why does mesa report cards as 570/580/590
they are all different GPUs. which is it then?... thats kinda stupid imo

because is the same silicon chipset just with more compute cores, stream processors enable, high grade chipset become 580/590, low quality chipsets become 570, same as CPUs that low quality chips end being low tier, they are just the same but with some deffective cores that get disabled. In the end is the same Polaris 10 chip.

That wasn't my point..I know what it is. But why it reports all those names..It could surely have the correct name by checking how many stream processors are enabled or whats the powerlimit or whatnot

why does mesa report cards as 570/580/590
they are all different GPUs. which is it then?... thats kinda stupid imo

because is the same silicon chipset just with more compute cores, stream processors enable, high grade chipset become 580/590, low quality chipsets become 570, same as CPUs that low quality chips end being low tier, they are just the same but with some deffective cores that get disabled. In the end is the same Polaris 10 chip.

That wasn't my point..I know what it is. But why it reports all those names..It could surely have the correct name by checking how many stream processors are enabled or whats the powerlimit or whatnot

Or just by reading the device string reported from the host OS.

Just received my new 5600 xt, Gigabyte Windforce version. Slight OC on the Gpu. Slight OC on the Cpu @3.8GHZ (Ryzen 2600)
it reports as a 5700xt, but it's a 5600xt.

this is also a 5600XT but it shows 5700XT:

!link

No OC on my side.

I read about issues with it's vbios and the journey to flash it here and part two here but I still have to read into the details.

It feels like it's underperforming.

UPDATE:

I checked the bios with amdvbflash and the "latest" I find on techpowerup is XFX.RX5600XT.6144.200327.rom

    Product Name is :    NAVI10 A1/A2 D1990301 XLE 6GB 300e/875m 
    Device ID is    :    731F
    Bios Version    :    017.001.000.068.000000
    Bios P/N is     :    113-170WCNAVIXLE6
    Bios SSID       :    5710
    Bios SVID       :    1682
    Bios Date is    :    03/27/20 21:25 

while the vBios I extracted from the GPU is already

    Product Name is :    NAVI10 A1/A2 D18903 XLE 6GB 300e/875m 
    Device ID is    :    731F
    Bios Version    :    017.001.000.068.000000
    Bios P/N is     :    113-160WDNAVIXLE6
    Bios SSID       :    5710
    Bios SVID       :    1682
    Bios Date is    :    05/14/20 02:16 

So I guess I'm uptodate as long as I can not find any more recent vBios for this GPU.

Last edited by beko on 30 Aug 2020 at 2:46 pm UTC

seems Ok to me without OC I get 3400 or so...better than me. XFX is not precisely one of the best brands for AMD, likewise Gigabyte (the one I have). I think the best ones are Shappire, MSI and ASUS. Probably those ones can hit 3600 points without OC. Anyway it's just a trick because the cards performing better have higher clocks, so the OC I did would be like normal clocks for a Shappire card imo. In this cards the clocks are lower because the thermal solution is not upto the task.
Too bad that this cards cannot be bios modded, like polares did, if I understand correctly now you might brick your card if you flash a modded bios.

I read that it won't boot if you flash it without a valid signature. This is what the -rsa command is for on the flash util.

I researched somewhat and it looks like my precise model is XFX Radeon RX 5600 XT THICC III Pro 14Gbps 6GB GDDR6 and XFX has currently no more recent vBios so I'm good on this.

Interesting is that it has a vBios switch. The switch position closer to the power connector is "performance" mode while there is also a quiet mode. I wonder.. mine is currently on performance and in idle I have 15W consumption and 0 rpm on the fan. That sounds awesome to me :D

I'll get the OP updated today, have neglected it a bit

Also just realised that I never added my Vega 7 to it

Edit , Maybe not. Don't have write access to the spreadsheet

Here's mine, its only a year + late

!link

Last edited by pete910 on 31 Aug 2020 at 9:09 am UTC

Edit , Maybe not. Don't have write access to the spreadsheet

PM your email to woox2k or me

Last edited by Xpander on 31 Aug 2020 at 9:53 am UTC

It's my turn,

With My RX 5600 XT, I tested with and without OC, with kernel 5.4 and 5.8.
+ comparison on Windows OpenGL and DirectX

!link
!link
!link
!link
!link
!link

Last edited by Vinouch on 6 Sep 2020 at 1:14 am UTC

New machine, same gfx card as before (score 3172). No OC.

!superposition benchmark

Last edited by damarrin on 6 Sep 2020 at 7:59 am UTC

New machine, same gfx card as before (score 3172). No OC.

so your score improved because of the cpu, mobo and ram?

Hard to say I guess? Different OS, newer kernel, driver, even newer benchmark version possibly. Quite a few variables there.

Gawd I Hate spreadsheets

Updated! Only showing top 50 as it's getting a little big.

Gawd I Hate spreadsheets

Updated! Only showing top 50 as it's getting a little big.

thanks!!

you have some errors Beko's and mine are both 5600xt not 5700xt. position 18th and 21st.

Fixed!

!link

"Gigabyte TU102" is an RTX 2080 Ti.

Last edited by CatKiller on 6 Sep 2020 at 1:20 pm UTC

New machine, same gfx card as before (score 3172). No OC.

that 980ti is doing extremely good, considering that Liam's old 980ti did 2900pts or so (on 2017), seems that Nvidia drivers have improved a lot. Or maybe Ryzen cpus squeeze more performance out of them.

Different version so cant say if driver or not really , Need to run Ver 1.0 and Ver 1.1 on same machine to see if there's a difference to be sure .

Last edited by pete910 on 7 Sep 2020 at 7:25 pm UTC

OMG !

It perform better with wine Vulkan/DXVK !

DXVK :

!link

Same machine, on Linux :

!link

Last edited by Vinouch on 10 Sep 2020 at 2:20 pm UTC

This is also on nvidia. Its the directX version of the benchmark that just performs better than openGL. And you dont loose much with DXVK translation. OpenGL isnt just up to scratch for this heavy thing. Probably also a lot less optimized.

It perform better on Linux DXVK than on Windows native Directx 11
On my windows 7 directx 11 : 3659
On my Linux DXVK : 3801
https://www.gamingonlinux.com/forum/topic/2683/post_id=29621
So, the better score I got is on Linux DXVK

Last edited by Vinouch on 10 Sep 2020 at 7:53 pm UTC

Yeah cause AMD windows divers just suck in dx11 and windows bloat will probably kill some perf also, depends on your system ofc, how clean it is etc


edit: you can also test with DXVK on windows to see if windows is the issue or AMD DX11 drivers just suck :)
just drop the d3d11.dll and dxgi.dll into the superposition folder, next to executable and try

Last edited by Xpander on 11 Sep 2020 at 8:01 am UTC

you can also test with DXVK on windows... just drop the d3d11.dll and dxgi.dll into the superposition folder, next to executable and try

I'm interesting. Where to get these two files? Chance to work on dx11 games?

you can also test with DXVK on windows... just drop the d3d11.dll and dxgi.dll into the superposition folder, next to executable and try

I'm interesting. Where to get these two files? Chance to work on dx11 games?

from here: https://github.com/doitsujin/dxvk/releases
just make sure you use the correct ones. for 64bit games the 64bit ones.. though i dont think there are many 32bit games these days anyway.

Also windows is not officialy supported DXVK, some games might not work like that.