BruceX benchmark with RX 580 slower in High Sierra 10.13.3

[Prince134]

Hi All,
In my machine, Mac Pro 5.1 2010, with RX 580, the BruceX test in High Sierra 10.13.2 ran at around 20 sec. Now the same test under 10.13.3 runs at 32 sec. This is 50% lower than before. I think this was due to the Meltdown/ Spectre patch in 10.3.3. Tested in beta 10.13.4 is the same lower result. My FCP is 10.4. Can Any one confirm this?

 

[Teddy Leung]

Me too -> test under 10.13.3 runs at 32 sec

 

[h9826790]

Can anyone with other AMD card can confirm this slow down is applicable as we’ll?

If yes, may be the patch.

If no, most likely are driver related.

 

[devon807]

Ran the BruceX test on 10.13.4 Beta and it did it in 20 seconds. Have not tested on 10.13.3

 

[beaker7]

There are widespread OpenCL issues in 10.13.3

 

[Prince134]

Ran the BruceX test on 10.13.4 Beta and it did it in 20 seconds. Have not tested on 10.13.3

Strange, I tested both 10.3.3 and 10.3.4 beta. It gave me the same 32 sec. Now don't know why you got 20 sec?

 

[devon807]

Odd. BruceX is consistently running 20 Sec on 10.13.4. What version of FCPX are you using?

 

[Prince134]

Odd. BruceX is consistently running 20 Sec on 10.13.4. What version of FCPX are you using?

It is 10.4. Teddy Leung got the same speed as mine as he posted after me. I got 20 sec before I upgraded to 10.13.3, and I further upgraded to 10.13.4 beta, same.

 

[devon807]

It is 10.4. Teddy Leung got the same speed as mine as he posted after me. I got 20 sec before I upgraded to 10.13.3, and I further upgraded to 10.13.4 beta, same.

I am also on FCPX 10.4. This is very peculiar. You are getting 30+ seconds on the beta?

 

[Prince134]

I am also on FCPX 10.4. This is very peculiar. You are getting 30+ seconds on the beta?

Correct! I am running x5690, RX 580 8GB, BruceX Prores 444X. Wondering what makes our difference! Did you run it Nvidia and AMD both cards together?

 

[devon807]

I ran the test on ProRes 422. Only RX 580

 

[theitsage]

Hi All,
In my machine, Mac Pro 5.1 2010, with RX 580, the BruceX test in High Sierra 10.13.2 ran at around 20 sec. Now the same test under 10.13.3 runs at 32 sec. This is 50% lower than before. I think this was due to the Meltdown/ Spectre patch in 10.3.3. Tested in beta 10.13.4 is the same lower result. My FCP is 10.4. Can Any one confirm this?

FCPX 10.4 is most likely the culprit. This version refuses to use these AMD cards in eGPU mode. Try running FCPX 10.3.4 and let us know.

 

[devon807]

Tested again, this time with 10.13.3, and got 21.6 seconds. ProRes 422.

--EDIT-- Tested again with ProRes 444 XQ 22 Seconds.

 

[Prince134]

Tested again, this time with 10.13.3, and got 21.6 seconds. ProRes 422.View attachment 749374
--EDIT-- Tested again with ProRes 444 XQ 22 Seconds.

The only thing different from yours is that I ran on PCIE SM951. Did you run it on SATA drive?
[doublepost=1517456569][/doublepost]

FCPX 10.4 is most likely the culprit. This version refuses to use these AMD cards in eGPU mode. Try running FCPX 10.3.4 and let us know.

I ran it on board slot one not eGPU mode. My FCPX has upgraded to 10.4 already. Teddy got the same result, so I am not the only one.

 

[devon807]

The only thing different from yours is that I ran on PCIE SM951. Did you run it on SATA drive?

I also ran it on my SM951. How are you powering your RX 580?

 

[Prince134]

I also ran it on my SM951. How are you powering your RX 580?

I run with EVGA Power Link. Shouldn't be an issue.

 

[devon807]

This is odd. Im not sure of any other discrepancies. We have similar setups, but you and Teddy have X5690's which i'm sure make no difference.

 

[itdk92]

Will test tomorrow

 

[Prince134]

Things about what cause it can be more interesting!

Everyone know APPLE did the trick by preventing from Nvidia getting its better OpenCL performance over AMD (for years). But now it hit AMD card user if the OS not wise enough by thinking you are using Nvidia. My same setup go back to 20 sec in the BruceX test after I did a clean install of 10.13.3 with solely AMD Rx 580 on my other SSD. Then I tested with my existing installs of the Mac OS 10.13.4 beta. Surely it ran 32 sec. From here I suspect that because the 10.13.4 beta was upgraded from the installs that with Nvidia web driver and was using gtx 1080ti. Last night I uninstalled Nvidia's web driver. Boom! 20 sec as it should be with RX 580.

I think Teddy has a Nvidia web driver installed as I had. He can uninstall it and confirm my finding. In short if it's true, your RX 580 will only run FCP at level of Nvidia if you have Nvidia web driver present even though you are now solely running on an RX 580! This is the same idea as APPLE under clocking it's iPhones for previous generations after new phone debut. I know they have excuse about the battery gate (unexpected shot down for some iPhones). But what about handicaps Nvidia in a similar way??

Previously we know in FCP, Nvidia always underperforms vs AMD, for years we don't know why since Nvidia has better GPU. Now they incidentally murdered AMD card by 50% as long as Nvidia web driver present? Need more confirmation from other users, at least this is what I found.

 

[devon807]

This is crazy! I was also testing a GTX 970/980 and had the web drivers installed. An interesting theory indeed. I can test it with and without the drivers.

 

[devon807]

I tested the theory, and am getting mixed results. With the drivers it scored 19sec @ProRes 422 and without it got 20-22sec. This is odd. I have not tried a clean install, only uninstall/reinstall.

 

[Teddy Leung]

I just del all render file and output master again, 37Sec, with render file output use 34sec.

i am connecting 2 dell 4K Monitor now.

OS 10.13.3
FCPX 10.3.4

 

[Prince134]

I just del all render file and output master again, 37Sec, with render file output use 34sec.

i am connecting 2 dell 4K Monitor now.

OS 10.13.3
FCPX 10.3.4

Do you have Nvidia driver in your System Preferences Pane?

 

[Teddy Leung]

Do you have Nvidia driver in your System Preferences Pane?

Never used Nvidia Card in my mac

 

[hajwer]

I do also have a problem with my BruceX performance in Final Cut Pro.

Yesterday I got a pretty cheap Mac Pro 4,1 (flashed it to 5,1) with 8 Cores and tested it against my 3770K Hackintosh. Both machines were equipped with the same GTX1080, Final Cut 10.4 Test Library was stored both times on an internal SSD. So I ran the BruceX test an noticed a huge difference in the results (see below). And I have no idea, where this difference is coming from, because BruceX is completely GPU-related? So maybe someone else has.

3770K@4,3GHz, 16GB DDR3-1333, macOS 10.12.4 = 32 Seconds
MacPro4,1@ 8x 2,26GHz, 20GB DDR3-1066, macOS 10.12.6 = 60 Seconds

The only difference these two machines have is the single core CPU performance, but CPU idled about 99% during the run. First I ran 10.13.3 on the Mac Pro, got also 60 Seconds and thought the difference is because of nVidia messed up the new drivers. So I installed Sierra and got the same slow result.

Any thoughts? :/