MP 7,1 W5700X Release

[flowrider]

Would it be better to sell the 580X

Radeon Pro 580X MPX Module - Price

That's a good question. There should be a few of us with Pro 580Xs that are in the closet. Mine has been there for two months now. I don't really believe there is a market. But, who knows. Lou

forums.macrumors.com

Lou

 

[UmarOMC]

Sorry it's a bit vague, I've had odd crashes and graphical glitches. Nothing really deal-breaking, but annoying enough that I prefer to use the older, more reliable GPU. I really hope 10.15.5 sorts it out – in which case I can save £1,000!

I'm in the same boat. I've got a PowerColor RedDevil 5700XT that gives me a boot screen and runs fairly rock solid…except when it doesn't. I'm not going to lie, I use the MacPro 7,1 for everything, but a lot of gaming as well. Seemingly randomly, after some time in Dying Light, Borderlands 3 or even a couple of times after the kids play Roblox (until we have space for me to set up the other computer…) it's caused the GPU Driver to completely hang the finder and all interactions therein. I get around it by pressing the logout keyboard combo and restarting the computer (immediately logging back in will leave me continuing to experience the hung Finder).

 

[Peterjk]

I think a good test here would be to *only* connect the 5K2K to the Mac Pro and see if the lag still exists. If it doesn’t, then a second 580X should work (put the 5K2K on one, and your other two on the HDMI ports of the other). If the GPU is having issues with the 5K2K by itself, then a beefier card may be your better option.

Thanks @Krevnik. I just tried that, and while it was better when only the one monitor (LG 5k2k) was attached and running, I am aiming for something even more smooth. So my plan as of right now, is to get the W5700X (approx. 5th of May) try it out and keep it if it's better and if need be leave the 580X in the machine as well to drive the two less to demanding and less important screen in my room.

 

[mattspace]

Thanks @Krevnik. I just tried that, and while it was better when only the one monitor (LG 5k2k) was attached and running, I am aiming for something even more smooth.

That sounds like something else is wrong - I've read the 580's hardware should theoretically be able to drive 5 x 4k (even 5k in DP1.4) displays. If interactions in GUI elements are laggy, surly there's something more fundamental going on.

Do you have iStat or similar installed to see if the GPU is actually being loaded, or if it's just low-load faithfully displaying a lag that's coming from something else?

 

[Krevnik]

That sounds like something else is wrong - I've read the 580's hardware should theoretically be able to drive 5 x 4k (even 5k in DP1.4) displays. If interactions in GUI elements are laggy, surly there's something more fundamental going on.

Do you have iStat or similar installed to see if the GPU is actually being loaded, or if it's just low-load faithfully displaying a lag that's coming from something else?

That’s why I bought up the scaling. Running this at “like 3440 x 1440” so that elements match what you’d expect to see on a 34” UW gets you two things:

First, the framebuffer being rendered into is 6880x2880, getting it very close to XDR territory at a little over 19 Mpx in size.

Second, the framebuffer has to be scaled by the GPU down to 5K2K. This is what I’d suspect is the source of the lag.

Maybe another test is to see if setting the 5K2K to a display setting of “like 2560x1080” improves the lag or no.

 

[thomast0001]

Hey hey! The W5700X I ordered on the 15th just arrived on my doorstep!

I need to wait a bit for it to acclimate before I install it.

 

[Aboo]

Hey hey! The W5700X I ordered on the 15th just arrived on my doorstep!

I need to wait a bit for it to acclimate before I install it.

Wow lucky you! Are you located in the US?

 

[thomast0001]

Wow lucky you! Are you located in the US?

Yes! On the West Coast, as close to China as I can get (short of Hawaii ). For the record, shipment details are:

2020/04/22 China Label Created
2020/04/22 Shanghai, China Shipped
2020/04/23 Incheon, Korea Arrived/Departed
2020/04/23 Anchorage, AK Arrived/Departed
2020/04/24 Louisville, KY Arrived/Departed
> Oakland, CA Arrived/Departed
> Sunnyvale, CA Arrived/Out for delivery

So let me tell you, that package *booked* across the Pacific! Once it hit the US, it took one *day* to fly around the country to my doorstep. Pretty impressive.

 

[erroneous]

Hey hey! The W5700X I ordered on the 15th just arrived on my doorstep!

Congrats! Any chance of fitting pics and Geekbench numbers? Pretty please?

 

[thomast0001]

Congrats! Any chance of fitting pics and Geekbench numbers? Pretty please?

I’ve attached pics of the outer box (with same kind of pull tab as the Mac Pro outer box), inner box (which was tape sealed), and then with the inner box open. The inner box opens with a cool unfolding action when you push open the two sides.

[edit] As for more details, including benchmarks, that’ll have to wait a bit till after it acclimates.

 

[erroneous]

I’ve attached pics of the outer box (with same kind of pull tab as the Mac Pro outer box), inner box (which was tape sealed), and then with the inner box open. The inner box opens with a cool unfolding action when you push open the two sides.

Wow, that's pretty. Thanks for taking the time to post, appreciate it.

 

[thomast0001]

Wow, that's pretty. Thanks for taking the time to post, appreciate it.

No problem! Glad to be of community service! ? When I have a chance (soon) I’ll post more. In the mean time I need to get back to my job, which is making it possible for me to make these purchases. ?

 

[codehead1]

As for more details, including benchmarks, that’ll have to wait a bit till after it acclimates.

Why does it need to acclimate? Is it unusually hot, cold, or damp?

 

[daveedjackson]

That's super. Mine dispatched 22nd. Alas because it's coming to the UK I don't have any tracking info until it reaches The Netherlands. So currently my shipping just says waiting.
[automerge]1587771834[/automerge]
There used to a tool not apple, which would track and give you all the movement into about your order, but think it went offline.

 

[Peterjk]

Maybe another test is to see if setting the 5K2K to a display setting of “like 2560x1080” improves the lag or no.

OK. I will test that tomorrow. FYI: I am running the LG 5k2k screen at 3840x1620.

 

[thomast0001]

Following up with W5700X installation pictures. Then benchmarks.

 

[thomast0001]

W5700X geekbench. I used the free version. First the 580X scores on my system for comparison (10.15.4, 12 core, 32GB memory).

CPU: single-core 1174, multi-core 11773
Compute: OpenCL 25113, Metal 27845

Now for the W5700X.

Compute: OpenCL 66985, Metal 22501

Wha.....?!!!!

Repeat scores were 13341 and 14206, respectively. Suffice it to say, I was not pleased.

However, there was an outstanding 10.15.4 supplemental update. I went through the lengthy installation process (which blanked the screen and rebooted multiple times). When the system finally settled down again I reran the benchmarks three times in a row.

Compute OpenCL: 65569,65120,65882
Compute Metal: 70695,73279,72831

MUCH better! ??

However, now performance on repeat runs is in the toilet again. Then I noticed backupd running. Whoops! A time machine backup is in progress! So take the performance dive with a grain of salt. I’ll follow up once the time machine backup completes.

 

[thomast0001]

Wow. This is super odd. The time machine backup and supplemental update have nothing to do with the severe performance degradation I’m seeing with the W5700X. It has something to do with running the geekbench Metal Compute benchmark while Safari is running. Note that I have a couple of stopped videos in Safari.

If I run the OpenCL test with Safari running then performance is fine, even with multiple runs. But as soon as I run the Metal test with Safari running, performance dies. And it stays dead even if I quit Safari. And even the OpenCL performance stays dead! However, a reboot fixes the problem.

So something bad is going on under some circumstances. Ah. It’s the stopped videos! Wow! I closed both video windows and now performance is stable for the Metal test, even with Safari running.

So kids, beware of using Metal if Safari has a window that contains a video, even if it’s stopped! Performance will dive, and may require a reboot to get it back.

[edit] Oops. Even with the video windows closed, performance is still bad if the Metal test is run with Safari running. It seemed to behave, but now it isn’t.

 

[skippermonkey]

Interesting. I have a 5700XT (in conjunction with two RX580s). Performance is generally fine, but if I have Hearthstone running at the same time as Safari, Hearthstone's usually super-smooth animation gets jittery. I've also had complete machine crashes too, with graphically intensive processes. I hope the 10.15.5 driver update will fix a lot of the issues with the 5700 series (the Mac Pro's PCIe utility doesn't even recognise the card).

UPDATE…
Geekbench 5

CPU, single-core: 1170; multicore, 18000 (24-core)

Sapphire RX580
OpenCL benchmark (avge): around 28000 (this is up around 42000 when not driving the display!)
Metal (avge): around 47500

5700XT
OpenCL benchmark (avge): around 62000
Metal (avge): around 53000 (when not driving the display!)
(Weirdly these both dropped when the 5700XT wasn't driving the display; 59500 and 47000 respectively)

Metal performance in the W5700 looks consistently higher (when it's behaving itself). But, hell, these figures are all over the place between the two systems. Frankly, I've no idea what's going on with most of these scores.

I opened a YouTube video and paused it, and the scores were largely unaffected (within the usual variance). I then ran the Metal benchmark for the 5700XT with a YouTube video playing in Safari at the same time. No huge difference in score – still around 53000. Even with two videos running the score only dropped slightly.

 

[daveedjackson]

Wow. This is super odd. The time machine backup and supplemental update have nothing to do with the severe performance degradation I’m seeing with the W5700X. It has something to do with running the geekbench Metal Compute benchmark while Safari is running. Note that I have a couple of stopped videos in Safari.

If I run the OpenCL test with Safari running then performance is fine, even with multiple runs. But as soon as I run the Metal test with Safari running, performance dies. And it stays dead even if I quit Safari. And even the OpenCL performance stays dead! However, a reboot fixes the problem.

So something bad is going on under some circumstances. Ah. It’s the stopped videos! Wow! I closed both video windows and now performance is stable for the Metal test, even with Safari running.

So kids, beware of using Metal if Safari has a window that contains a video, even if it’s stopped! Performance will dive, and may require a reboot to get it back.

[edit] Oops. Even with the video windows closed, performance is still bad if the Metal test is run with Safari running. It seemed to behave, but now it isn’t.

This is super strange.... a call to Apple? I do wonder how much testing they do before releasing something. You’d think that would be some of the first tests they do

 

[thomast0001]

This is super strange.... a call to Apple? I do wonder how much testing they do before releasing something. You’d think that would be some of the first tests they do

I submitted a report to Apple almost immediately. We’ll see if I ever hear back from them. The problem feels like a GPU allocated memory problem, although that’s just a guess.

The problem is particularly annoying because Geekbench wants to start the browser at the end of execution to upload results. I have to remember to quit the browser before attempting another run.

 

[daveedjackson]

Slightly worried what real world impact this will have. If I’m using this and have a video playing which often happens. I can’t afford my performance to be impacted because of that.

 

[blackie36]

That has to be a software / driver issue? I've ordered two which are due to arrive on Wednesday - fingers crossed this isn't a deal breaker. You'd have to think Apple QA couldn't let that slip through?

 

[bsbeamer]

To everyone testing: run with multiple monitors connected and as many ports as possible in use with GB5. Headless adapters work great for this if multiple monitors are not available. Noticed this under utilization issue when testing with RX5700XT in eGPU. Reported in MBP forum.

MBP16,1 with RX 5700 XT via eGPU (Reports, Issues & Benchmarks)

Been running some tests with an MBP16,1 configured with 2.4GHz, 5500M 8GB, 64GB RAM, 1TB SSD. Currently on 10.15.2. Have a Sonnet eGFX Breakaway Box 650 GPU-650WOC-TB3, which is the OC model that provides 100W power delivery (newest model). Been running a Sapphire Pulse RX 580 8GB in that...

forums.macrumors.com

 

[Peterjk]

So I've performed a few tests with my 580X and 3 monitors. One LG 5k2k, one 4k television and one reg. HD 1920x1080.

With ONLY the LG 5k2k attached to the 580X my screen seems to have equally smooth animation regardless of the resolution up til 5120 x 2160. But as soon as I start up with the other two screens (4k and HD), attached and on, the sluggishness begins. Also with the LG in smaller resolutions.