RX 5700 XT on macOS Catalina 10.15.2
 
Notifications
Clear all

RX 5700 XT on macOS Catalina 10.15.2  

 of  4
  RSS

aleigh
(@aleigh)
Active Member
Joined: 2 years ago
 

And so here I am with a trashcan and the purge-wrangler TB2 enabler, and the solution works well, is stable, at least on 10.15.1, and doesn't really bench any differently than a TB3 mac despite TB2 being slower on paper. That's the reality of it, despite whatever the marketing or optics were, it is a really viable technical solution and means the difference between life and death for a machine otherwise equipped with D500s. Funny enough I only bought the trash can because they kept dragging their feet on the tower and I needed something. 

 

The joke about consumers being confused about TB2 EGPU support really is a joke, because TB is so confusing to begin with. Passive v. active cables, looks like USB-C but regular USB-C cables only half work (power & USB hub), etc. Or just USB-C to begin with, what a mess. Heh. But glad they saved us mac pro users from ourselves... It seems more confusing to not support it when they sell first-party TB2 adapters. I am not even being cynical, not choosing to think that they are trying to cripple the machine so people will buy an entire new platform, I just think they don't understand long-tail enterprise support. I saw the same thing with x-serve & x-raid gear, you know, 3 years out, and can't buy a part for something that should have a 10+ year service life. It's a culture thing. Now we're getting way off topic but this is why I had such trouble with the rumors of them making a car, I couldn't imagine how they would support something for more than a few years. 

Trashcan Mac Pro, MacBook Pro (13-inch, 2017, Two Thunderbolt 3 ports), Core X, XFX 5700XT, XFX 580


ReplyQuote
beGPU
(@begpu)
Active Member
Joined: 2 years ago
 

GB5 tests, best in set reported.  RX 5700 XT in eGPU on MBP16,1 in 10.15.2.  Something is wrong with these drivers.

25813 = 5500M dGPU with eGPU connected but not utilized for test
51591 = RX 5700 XT driving monitor via DisplayPort (27" ACD) and HDMI (Samsung 43" 4K TV)
27226 = RX 5700 XT driving monitor via DisplayPort (27" ACD)
27572 = RX 5700 XT driving monitor via HDMI (Samsung 43" 4K TV)
32905 = RX 5700 XT with no displays connected (headless)

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

Mini i5 liked
ReplyQuote
craftsman
(@craftsman)
Estimable Member
Joined: 3 years ago
 
Posted by: @begpu

GB5 tests, best in set reported.  RX 5700 XT in eGPU on MBP16,1 in 10.15.2.  Something is wrong with these drivers.

25813 = 5500M dGPU with eGPU connected but not utilized for test
51591 = RX 5700 XT driving monitor via DisplayPort (27" ACD) and HDMI (Samsung 43" 4K TV)
27226 = RX 5700 XT driving monitor via DisplayPort (27" ACD)
27572 = RX 5700 XT driving monitor via HDMI (Samsung 43" 4K TV)
32905 = RX 5700 XT with no displays connected (headless)

Myself and others reported the same massive swings and inconsistency all year. Prior to that others noticed scores were dropping without any changes to the app or their Mac. I don't think its the drivers because Luxmark and GFXBench Metal have stayed consistent. Can you perform other more real world tests with the above configs?

Note that Geekbench delivers the results from their server. You can't perform a test and get results without a connection.

2019 16" MacBook Pro (RP5500M) [9th,8C,H] + RX 5700 XT @ 32Gbps-TB3 (Razer Core X Chroma) + macOS 10.15.1 [build link]  

ReplyQuote
beGPU
(@begpu)
Active Member
Joined: 2 years ago
 

MBP16,1 always has a display active unless you go into clamshell mode.  Clamshell mode was not used for these tests.  

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

ReplyQuote
craftsman
(@craftsman)
Estimable Member
Joined: 3 years ago
 

@begpu

Here's mine.

33590 = 5700XT Headless 
33059 = 5700XT + Eizo Display Port
27260 = 5700XT + Samsung 4K TV HDMI
49863 = 5700XT + Eizo DP + Samsung 4K TV HDMI

Same stupid results. The computer sends a configuration and benchmark results string to the server and the results come back inconsistent.

 

2019 16" MacBook Pro (RP5500M) [9th,8C,H] + RX 5700 XT @ 32Gbps-TB3 (Razer Core X Chroma) + macOS 10.15.1 [build link]  

ReplyQuote
craftsman
(@craftsman)
Estimable Member
Joined: 3 years ago
 

Geekbench 5.1 is out. 

The 5700XT Anniversary Edition now gets +/- 42,000 on Metal and OpenCL. The Metal result is terrible and slower than a 580. The OpenCL result is about 40% slower than on Windows. Geekbench for Mac goes in the bin.

The 9900K PC still consistently getting over 1400 single core and 10,000 multicore. I perfected the overclocking and thermals on this system. Gigabyte Designare motherboard has been a star. In some renders it performs about as well as some 12 and 16 core systems that have lower clock speeds. If I ever need a Threadripper I'll put it in a Gigabyte Designare with TRX40 socket.

Screenshot 2019 12 24 at 08.43.08
Screenshot 2019 12 24 at 08.41.54
gb51cpu
2019 16" MacBook Pro (RP5500M) [9th,8C,H] + RX 5700 XT @ 32Gbps-TB3 (Razer Core X Chroma) + macOS 10.15.1 [build link]  

Mini i5 liked
ReplyQuote
beGPU
(@begpu)
Active Member
Joined: 2 years ago
 

@craftsman

I've been able to test scores 70,000+ with Sapphire Pulse RX 5700 XT 8GB in an eGPU enclosure ONLY when multiple monitor configurations in specific resolutions are set correctly. 70361 and 70890 are the two METAL tests with MBP16,1 and DP1, DP2, and HDMI connected at UHD/4K resolutions. Any other less and scores are around 42000-45000 generally speaking, but can drop to around 35000 when JUST an HDMI is connected.  This immediately signals an issue with the drivers in macOS Catalina 10.15.2... in addition to a terrible way to test METAL scores with GB5 (5.1.0 503054)

 

I have not been able to get the third DP (DP3) to work correctly with this card.  Results in an immediate system freeze, whether used in isolation or in combination.  Hard restart required, or automatically does it.  Will freeze immediately again, unless disconnected.  

 

For reference, the 5500M 8GB dGPU of MBP16,1 gets around 28,000-30,000 in all testing so far.

To do: Create my signature with system and expected eGPU configuration information to give context to my posts. I have no builds.

.

Mini i5 liked
ReplyQuote
craftsman
(@craftsman)
Estimable Member
Joined: 3 years ago
 
Posted by: @begpu

@craftsman

I've been able to test scores 70,000+ with Sapphire Pulse RX 5700 XT 8GB in an eGPU enclosure ONLY when multiple monitor configurations in specific resolutions are set correctly. 70361 and 70890 are the two METAL tests with MBP16,1 and DP1, DP2, and HDMI connected at UHD/4K resolutions. Any other less and scores are around 42000-45000 generally speaking, but can drop to around 35000 when JUST an HDMI is connected.  This immediately signals an issue with the drivers in macOS Catalina 10.15.2... in addition to a terrible way to test METAL scores with GB5 (5.1.0 503054)

 

I have not been able to get the third DP (DP3) to work correctly with this card.  Results in an immediate system freeze, whether used in isolation or in combination.  Hard restart required, or automatically does it.  Will freeze immediately again, unless disconnected.  

 

For reference, the 5500M 8GB dGPU of MBP16,1 gets around 28,000-30,000 in all testing so far.

Thanks. I don't think it is worth relying on these scores no matter how high or low they go.

Sad to hear about your DP3 issue. I was only using the same DisplayPort 3 next to the HDMI port on the 50th Anniversary Edition. 

2019 16" MacBook Pro (RP5500M) [9th,8C,H] + RX 5700 XT @ 32Gbps-TB3 (Razer Core X Chroma) + macOS 10.15.1 [build link]  

ReplyQuote
aleigh
(@aleigh)
Active Member
Joined: 2 years ago
 

I did just get around to trying 10.15.2 yesterday on a TB2 RX 5700 XT setup w/ a mac pro, despite everyone saying there were problems, as I should have listened, because there were problems. System kernel panic'd with the eGPU plugged in. Unplugged and the update was able to complete. With pure-wrangler everything uninstalled (verified from latest script), the system would kernel panic if I plugged the eGPU in. Ended up restoring to 10.15.1 via Time Machine which was thankfully painless, just a big time waste. It's giving me PTSD of owning a Windows PC. 

Trashcan Mac Pro, MacBook Pro (13-inch, 2017, Two Thunderbolt 3 ports), Core X, XFX 5700XT, XFX 580


ReplyQuote
mac_editor
(@mac_editor)
Famed Member Moderator
Joined: 5 years ago
 

@aleigh

It’s a known issue with Navi on TB1/2: https://github.com/mayankk2308/purge-wrangler/issues/44

Just set up using the script without plugging the eGPU in. This is also mentioned in the Beginner’s Guide footnotes. 

"Desultory reading is delightful, but to be beneficial, our reading must be carefully directed." — Seneca

Author: kryptonitepurge-wranglertbt-flashpurge-nvdaset-eGPU
Insights Into macOS Video Editing Performance
Launching Apps on Specific (e)GPUs on macOS

Master Threads:
2014 15-inch MacBook Pro 750M
2018 15-inch MacBook Pro

 
2019 13" MacBook Pro [8th,4C,U] + RX Vega 64 @ 32Gbps-TB3 (Mantiz Venus) + macOS 10.14.6 & Win10 [build link]  


itsage liked
ReplyQuote
 of  4