AMD RX 580 - Disappointing performance (10.12.6)

[Jultsu]

So the Nvidia Web Drivers on 10.13.1 are newer than the ones on 10.12.6?

I'll try High Sierra next week on a fresh new SSD, if I see no improvement I guess I'll try to find a 980ti before I start breaking $hit up.

 

[Justcurious1]

Hi everyone.

Am I missing something?

I'm running the same RX 580. I don't think you're missing anything. I think it's a confluence of things. It's Windows DX 11 and 12, when they might come into play in test situations like Valley and Heaven. I'm not sure, I'm just assuming because they are a game engine test and MacOS has terrible OpenGL performance. I think it's the AMD drivers which I'm sure are better on Windows. Then there is Metal and where it rears its mysterious head. For instance, my 1080 gives very good OpenCL benchmarks in LuxMark vs the 580 but sucks big ones when using FCPX and Motion5. The RX 580 Bruce test was 13 seconds, which creamed the 1080 40+ seconds. What's going on there? I don't know. Apple does. I'm betting Metal support. I can edit 4k 30fps video with several effects and prerender turned off with the RX 580 with absolutely no issues. Smooth as silk. The 1080 can't match it. I got the same scores you did on your Cinebench bench but got a full 10fps more on the 1080p Valley bench. I have no idea why.

 

[LeoDuke]

@FaNIX

Hi bro,
got 10.12.6 + Sapphire RX 580 Nitro + 2 x DELL U2715H but I don't use W-Green.

Here some of my scores.









I think:
we can't compare RX 580 with GTX 1070 but RX is definitely not crap

 

[Jultsu]

RX580 is not a shitty card by any means, it's just the drivers on macOS which let it down. In fact they let it down so badly that even an ancient 4 year old card beats it by at least a healthy 37% margin (688 points @RX580 vs. 940 points @780Ti, I have my CPU OC'd at 4.2GHz as well and I had tesselation @ Extreme, yours was @ Normal)

Please try to update to 10.13 and see if performance increases?




And regarding my earlier post:

I'll try High Sierra next week on a fresh new SSD, if I see no improvement I guess I'll try to find a 980ti before I start breaking $hit up.

I did try to install 10.13, couldn't. Can't even get to the installer, crashes immediately after ACPI sleep states message. Looks like I'm stuck on 10.12 at least until 10.14 comes out, so I guess I'd better enjoy (and make do with) that Kepler performance..

 

[LeoDuke]

RX580 is not a shitty card by any means, it's just the drivers on macOS which let it down.

I agree with u !!! The drivers r not good
Windows 10 score is much better

In fact they let it down so badly that even an ancient 4 year old card beats it by at least a healthy 37% margin (688 points @RX580 vs. 940 points @780Ti, I have my CPU OC'd at 4.2GHz as well and I had tesselation @ Extreme, yours was @ Normal)

Yo 6 cores OC to 4,2GHz + Asrock X79 Extreme6 are very strong components

I use 7700K@4,2 but with iATX Board Gigabyte Z170N-WiFi, which is not so good

Wenn I have a little bit more time for it, I'll try to bench my RX580 with Asus Z170 Sabertooth from my Music-PC.
Maybe the score will be better...

I made bench with the same settings like u, here is the difference

I did try to install 10.13, couldn't. Can't even get to the installer, crashes immediately after ACPI sleep states message. Looks like I'm stuck on 10.12 at least until 10.14 comes out, so I guess I'd better enjoy (and make do with) that Kepler performance..

I've installed 10.13 with my RX580 easily, without any issues and it works fine but I installed it on an old 2,5" HDD just to let it know if it works. I think, it's nonsense to make any benchmark with this old HDD.
When I install 10.13 on SSD, I'll give u the benchmark feedback immediately.

P.S. Question to all owners of RX 580: guys, should we sell our strong and beautiful graphic card

 

[Jultsu]

Clock-for-clock your i7-7700 should be just a hair stronger, so CPU speed is irrelevant in this comparison. Heaven probably uses only 1 CPU core anyways. 632/940 points mean that RX580 is 33% slower, or in other words, the relic from 2013 (780ti) is 50% faster.

Disgustingly poor performance from the Radeon.

 

[LeoDuke]

Clock-for-clock your i7-7700 should be just a hair stronger, so CPU speed is irrelevant in this comparison.

My bench experience /Geekbench3-4, CinebenchR15, LuxMark, Unigine Heaven/ shows:

1. CPU - relevant and important
2. Board - relevant and important
3. RAM - irrelevant and unimportant

Heaven probably uses only 1 CPU core anyways.

I don't know - r u sure ?

632/940 points mean that RX580 is 33% slower, or in other words, the relic from 2013 (780ti) is 50% faster.

Exactly 32,77% in Unigine Heaven comparison but as a graphic card:

http://gpu.userbenchmark.com/Compare/Nvidia-GTX-780-Ti-vs-AMD-RX-580/2165vs3923

Disgustingly poor performance from the Radeon.

Not from the Radeon (!) but only from APPLE and Lilu + WhateverGreen maybe.

As I said: as soon as possible I'll install 10.13 on some SSD and then we can talk again about the RX 580.

I'm curious myself

 

[Stork]

The Sapphire RX 580 Pulse 8GB graphics works OOB with acceleration in Sierra 10.12.6 and High Sierra. This card is the one Apple sells to developers as part of it's developer eGPU program.

See my post here for the latest info on how well AMD graphics cards are supported in Sierra 10.12.6 and High Sierra.

 

[arianrassoul]

P.S. Question to all owners of RX 580: guys, should we sell our strong and beautiful graphic card [/QUOTE]

not if youre a fcpx user, im not having any issues with this card, am not a gamer only fcpx after effects and premiere

 

[rna23]

P.S. Question to all owners of RX 580: guys, should we sell our strong and beautiful graphic card

not if youre a fcpx user, im not having any issues with this card, am not a gamer only fcpx after effects and premiere[/QUOTE]

Hi, I'm wondering how fast You can export the BruceX 5K test? I had two r9 280X and I got to 20s. Now I purchased a RX 580 and I'm on 75 That is like a 980Ti, which I tried before. A single 280X gave me much better results - around 30s.

I don't know what is wrong ...

But my results in Luxmark look similar to other people - from my point of view very poor ... just between the 280x and 290x.