Login
Username:

Password:

Remember me



Lost Password?

Register now!

Sections

Who's Online
53 user(s) are online (32 user(s) are browsing Forums)

Members: 2
Guests: 51

LiveForIt, levellord, more...

Headlines

 
  Register To Post

X5000 - GfxBench2D score
Just popping in
Just popping in


See User information
Hi,

I see, that yesterday somebody posted new GfxBench2D score to hdrlaborg.nz page.

Fantastic 13,030.65 points! X5000 with Polaris10 card!
For the first time is X5000 faster then X1000.

I am glad to see, that here is driver speed improvent.
Thanks to all involved, especially Hans.

AmigaOS: Micro A1-C, AmigaOne XE, Pegasos II, Sam440ep-flex, AmigaOne X1000
MorphOS: Efika 5200b, Pegasos I, Pegasos II, Powerbook, Mac Mini, iMac, Powermac Quad
Go to top
Re: X5000 - GfxBench2D score
Just popping in
Just popping in


See User information
@sailor


I think we see noticable differences in that score.
1) RadeonRX driver version 2.10. Iirc, the latest public release is 2.8

2) It's a Gigabyte RX580 4 GB. Maybe this means that the RX580 is now supported for 3D and/or video acceleration as well?

3) When I compare the score with my own X5000/RX570 score, the difference seems mostly caused by improved memcopy performance: WritePixelArray Memory copy ~525MB/s -> ~1100MB/s ; ReadPixelArray ~25MB/s -> ~1000MB/s. Is this due to DMA/GART support?

The second best score is with a RX570. Although it's a different RX570, it's better for comparing with my setup. RadeonRX driver is 2.11 (i have 2.8) and pcigraphis.card is 54.1 (i have 53.18).

Somehow, the performance of everything except memcopy and "random score" is worse. This can be explained by a difference in GPU clock speed. The second one can be explained by the huge memcopy improvement as it is part of the random test.
The memcopy performance improvement is similar as for the RX580 card above: WritePixelArray Memory copy ~525MB/s -> ~1050MB/s ; ReadPixelArray ~25MB/s -> ~950MB/s.
The bottomline score improved from 8248 to 9171.



Edited by geennaam on 2022/1/28 10:50:01
Edited by geennaam on 2022/1/28 10:53:37
Go to top
Re: X5000 - GfxBench2D score
Just popping in
Just popping in


See User information
@geenaam

RX580 is already supported with 3D/Video acceleration since the last the public update.

Sinan - AmigaOS4 Beta-Tester
- AmigaOne X5000
- AmigaOne A1222
- Sam460ex
Go to top
Re: X5000 - GfxBench2D score
Just popping in
Just popping in


See User information
@SinanSam460

Ok. Didn't know that. I thought that only the video decoding issues were solved for the RX550/560/570.

So maybe 2.8 isn't the latest driver as well. Time to run updater.

Go to top
Re: X5000 - GfxBench2D score
Home away from home
Home away from home


See User information
@geennaam
Quote:

1) RadeonRX driver version 2.10. Iirc, the latest public release is 2.8


Yeah, the latest public one is 2.8. All after till 2.11 in beta for now.

Quote:

) It's a Gigabyte RX580 4 GB. Maybe this means that the RX580 is now supported for 3D and/or video acceleration as well?


They supported by 3d and video acceleration for some time already (and that support in public already too).

Quote:

3) When I compare the score with my own X5000/RX570 score, the difference seems mostly caused by improved memcopy performance: WritePixelArray Memory copy ~525MB/s -> ~1100MB/s ; ReadPixelArray ~25MB/s -> ~1000MB/s. Is this due to DMA/GART support?


That is because just a few weeks ago Frederik add DMA acceleration support to the graphics.library, meaning that WritePixelArray() and ReadPixelArray() are now DMA accelerated (the same as it was for x1000 before, and that was the reason why x1000 results were superior before).

But that is not driver's DMA, but just for graphics.library. So once Hans will deal with everything in terms of DMA in drivers, things should be faster in terms of pure MemCopy functions from ram to VRAM and from VRAM to ram.

Quote:

The second best score is with a RX570. Although it's a different RX570, it's better for comparing with my setup. RadeonRX driver is 2.11 (i have 2.8) and pcigraphis.card is 54.1 (i have 53.18).


If you mean that one : https://mail.hdrlab.org.nz/benchmark/g ... 2d/OS/AmigaOS/Result/2346

Then it's my yesterday's upload.

Quote:

Somehow, the performance of everything except memcopy and "random score" is worse. This can be explained by a difference in GPU clock speed. The second one can be explained by the huge memcopy improvement as it is part of the random test.
The memcopy performance improvement is similar as for the RX580 card above: WritePixelArray Memory copy ~525MB/s -> ~1050MB/s ; ReadPixelArray ~25MB/s -> ~950MB/s.
The bottom line score improved from 8248 to 9171.


For myself, i can't explain why i have in some parts that slower speed in comparison with another upload where one scored 13.000.000.

But I also have in ragemem test from sysmon strange 50% decrease in video bus reading speed, which I can't explain for the moment.

And my card is Saphire RadeonRX 560 2GB

Quote:

Ok. Didn't know that. I thought that only the video decoding issues were solved for the RX550/560/570.


As far as I know, for now (since some time) every RX card which works in 2D works in 3D and have video acceleration working. And if i remember right that all already in public.

Quote:

So maybe 2.8 isn't the latest driver as well. Time to run updater.


2.8 is the last one, but maybe you also need to update everything else, like warp3dnova to the latest public one, etc.

Go to top

  Register To Post

 




Currently Active Users Viewing This Thread: 1 ( 0 members and 1 Anonymous Users )




Powered by XOOPS 2.0 © 2001-2016 The XOOPS Project