MP 1,1-5,1 Issue with RX 580 in 5,1.

[thewoolysheep08]

Installed a Sapphire RX 580 Pulse 4gb into my 5,1 using a dual mini 6-pin to 8-pin. Im running Martin Lo's OC 0.8.0 on 10.15.7.

The issue i am experiencing is when exporting from adobe media encoder with hardware acceleration. After starting the export, all my screens will go black and the GPU fans will kick up to 100%. Im assuming this is the automatic power restriction preventing the card from drawing too much power.

Im confused since this 580 with only 1 8-pin shouldn't be drawing more than 225 watts which the 5,1 can provide, however moments before the screens go black i am observing my 580 reaching a power usage upwards of 225 sometimes even reaching 300 before it shuts off.

I have already reflashed my GPU's BIOS to a stock one incase it was caused by a mining bios (this gpu was bought second-hand).
My current suspicion is that my cheap dual mini 6-pin to 8-pin cable is defective and only one side is working which would explain why it shuts off around 200 watts, but it doesn't explain why a card with a max tdp of 225 is reaching 300 watts.

If you have any ideas of what this might be caused by, i'd be very thankful.

 

[Matty_TypeR]

The thing is a lot of RX 580 cards can power spike well over 225W if your back plain is sensitive to that it will just shut down the Mac. your issue is a bit different in it doesn't shut down completely and GPU fan's go to 100% but reaching 300w is the reason it shuts down.

The Pixlas Mod would eliminate the shut downs and any sensitivity of the back plain. Its not advisable to keep pushing it as complete failure of the back plain could be next.

 

[thewoolysheep08]

I sorta figured this was going to be the answer. I was trying to avoid the pixlas mod since the parts are a bit difficult to get where i live. Im still quite confused though since my previous mac pro had an rx 580 with an identical power cord and had zero issues. Perhaps because it was a different brand.

 

[thewoolysheep08]

Also one thing i think is worth mentioning is that when the screens go black and the gpu fans kick up, i can still hear system audio and can still talk with people in a voice call. It is almost like the machine is still running just without the GPU. Is this typical behaviour of the protection shutdown or could this be a symptom of a failing graphics card?

 

[h9826790]

My current suspicion is that my cheap dual mini 6-pin to 8-pin cable is defective and only one side is working which would explain why it shuts off around 200 watts, but it doesn't explain why a card with a max tdp of 225 is reaching 300 watts.

You can ignore that 300W reading, which is a known inaccurate number.

You can read my post from here

Sapphire PULSE RX580 8GB VBIOS Study

*** For info, this mod is NOT compatible with the OpenCore boot screen mod *** This thread is about how to mod the Sapphire PULSE RX580 8GB card's ROM image. If you worry about bricking the card, or void the warranty etc. STOP HERE! For those who has MVC flashed RX580. Do NOT perform this mod...

forums.macrumors.com

From the first screen capture (the Furmark one), on the RHS, you can see the IOReg report 393W. Aparently that's way beyond a normal RX580's power draw.

But if you check the power draw from Booster A, B, and the slot. Then the card was just drawing about 225W at that moment.

Since the SMC should report the current correctly, which can be easily convert to power delivery. And the IOreg's reported power draw is way off from the normal range. Therefore, we may safely conclude that the IOreg power draw is inaccurate.

For HWMonitor or iStat, these software only read the GPU power draw from IOReg. Which means, the same errornous number.

If you really want to know the power draw, capture the power draw from Booster A, B, and slot. Then calculate the total power draw by yourself.

 

[thewoolysheep08]

What do you think might be the issue then? If the card is not exceeding the safe limits, do I simply have a defective GPU?

 

[h9826790]

What do you think might be the issue then? If the card is not exceeding the safe limits, do I simply have a defective GPU?

Possible power cable issue / fault card / not seating correctly / dust in the PCIe slot / GPU overheat.

But since you can reproduce the issue, and it seems only happen when start to draw high power. Therefore, most likely power cable issue, or GPU overheat.

For GPU overheat, it's easy to test. Because the GPU should crash when you run Furmark, Luxmark, Unigine Heaven, etc. You may try if you see the same crash when you run those benchmark. And try to capture the temperature just before crash.

Of course, you can simply just open up the card, clean the dust, re-paste it, and see if that help anything.

 

[thewoolysheep08]

I doubt it is GPU overheat since i have monitored temperatures in iStat and only gets to around 70C max.
I doubt it is dust in PCI slot since i recently did a full clean of the machine and cleared all dust.

Do you think if i were to install the Pixla's mod, it would resolve this issue?

 

[h9826790]

I doubt it is GPU overheat since i have monitored temperatures in iStat and only gets to around 70C max.
I doubt it is dust in PCI slot since i recently did a full clean of the machine and cleared all dust.

Do you think if i were to install the Pixla's mod, it would resolve this issue?

Can't tell, should make no difference on a healthy cMP for RX580. As you can see from my link up there, the RX580 can run Furmark with dual mini 6pin to 8pin. Video exporting shouldn't demand more than that.

But if the Boost A, B itself has power delivery issue. Then Pixlas mod may help.

 

[h9826790]

Can you try DV / iMovie export? I wonder if that's the problem of AME.

 

[thewoolysheep08]

I can try an iMovie export. What program were you using to monitor wattage from boosters?

 

[h9826790]

I can try an iMovie export. What program were you using to monitor wattage from boosters?

It's a legacy software that not avail anymore

Hardware Monitor: Description

Product description for Hardware Monitor

www.bresink.com

 

[jscipione]

Installed a Sapphire RX 580 Pulse 4gb into my 5,1 using a dual mini 6-pin to 8-pin. Im running Martin Lo's OC 0.8.0 on 10.15.7.

The issue i am experiencing is when exporting from adobe media encoder with hardware acceleration. After starting the export, all my screens will go black and the GPU fans will kick up to 100%. Im assuming this is the automatic power restriction preventing the card from drawing too much power.

If you have any ideas of what this might be caused by, i'd be very thankful.

It’s possible that you have a defective card or cable, but before we can assume that, I would like you to perform a simple test to ensure that GPU acceleration is working correctly.

Open a video file using Handbrake. Under Video tab select Video Encode: H.264 (VideoToolbox) and press Start.

If your hardware acceleration is not configured correctly the encode will fail right away. If not, allow it to encode the video and see if the system locks up.

 

[thewoolysheep08]

Im having trouble even getting the Furmark benchmark to start. Im going to try ungine heaven benchmark and see what happens there.

 

[thewoolysheep08]

It’s possible that you have a defective card or cable, but before we can assume that, I would like you to perform a simple test to ensure that GPU acceleration is working correctly.

Open a video file using Handbrake. Under Video tab select Video Encode: H.264 (VideoToolbox) and press Start.

If your hardware acceleration is not configured correctly the encode will fail right away. If not, allow it to encode the video and see if the system locks up.

Was able to encode a video of me running Heaven benchmark through handbrake. Used your settings however the video card was not used at all, instead it used CPU.

This video illustrates what i am experiencing.\

 

[thewoolysheep08]

This actually leads me to believe that this is an issue to do with thermals not the cable, seeing as the benchmark can run for so long without crashing. It would also line up with the fact that the fans are ramping upto full speed and the machine is still running without graphics which doesnt make sense if it were the safety power cutoff.
Let me know what you think.

 

[thewoolysheep08]

Tomorrow i will benchmark this card in a windows PC and see if the same issue arrises. Then i can figure out whether it is a thermal issue or an issue with my cMP. If it is an issue with the cMP, ill install pixlas mod.

 

[h9826790]

That looks like a thermal issue to me.

Your GPU temperature increase from 54C to 75C within 3 seconds, then the core clock speed drops to 567MHz. This is a sign that the GPU was in heavy thermal throttling to avoid overheat.

As you can see from my ROM study thread. The GPU was staying above 1200MHz even approaching the end of the Furmark Benchmark.

Hopefully there is no real damage has been done to the GPU yet, and re-paste can fix the issue (if you want to, liquid metal can help quite a bit to cool down the RX580 PULSE. But you really need to know what you are doing, liquid metal can kill the card if not apply correctly and carefully)

 

[thewoolysheep08]

Repasted the GPU today. Appears to have fixed it.
The gpu can now easily reach 1200mhz and stays within 65c, before it couldnt even sustain 400mhz without going to 80c.
Can't believe i overlooked such a basic issue
Thanks for your help Martin.

 

[rodedwards]

I had similar problem with RX580 and Mac 5,1. When processing 8k video it would suddenly switch off. I thought it was a power drain issue, but when i checked temps with Mac Fan Control, it turned out my Northbridge Diode temps were over 110C.

I took Mac to repairer and apparently it was a CPU Diode overheating problem. My Processor Heat Sink clip had broken so the heat sink didn’t work and the thermal paste was five years old and all dried up.

Got it fixed and problem solved. Hope its also fixed yours.

 

[thewoolysheep08]

I have since fixed the issue but thanks for the suggestion. I think my northbridge is okay for the moment since it never gets above 65c, plus its pretty common for it to run hot.

 

[TzunamiOSX]

I have since fixed the issue but thanks for the suggestion. I think my northbridge is okay for the moment since it never gets above 65c, plus its pretty common for it to run hot.

So it was the thermal paste of GPU?