MP 1,1-5,1 5.1. Random Shutdown problems!

[PetteriKiller]

Edit: I changed the topic to match the following discussion better

Hi-ho folks!

I have a Frankenstein-Mac Pro 5.1 2009 which has 2 x Intel Xeon X5690 3.46GHz processors in it. (Mojave on NVMe PCI-e disk and the boot ROM naturally 144.0.0.0.0)

I did have some random and shudden shutdowns - which I now believe are connected to Radeon RX 580 for its poor ventilation because of the very large double NVMe -board right on the slot above of the GPU - but after installing the Macs Fan Control, I noticed that the temperature of the CPU A Core from PCECI runs around 15 celsius degrees higher than the temperature of the CPU B core from PCECI.

Why so and is it normal?

I changed the processors myself and added the thermal compound around equal amount on both.

Thanks for answers.

 

[tsialex]

I noticed that the temperature of the CPU A Core from PCECI runs around 15 celsius degrees higher than the temperature of the CPU B core from PCECI.

Why so and is it normal?

It's the airflow, CPU B gets the colder outside air, CPU A gets the hot air from CPU B.

The difference with my Mac Pro it's around 5 to 7ºC when idle, and 10 to 12ºC when in constant full use.

 

[PetteriKiller]

^So I can stop worrying.

Thanks You!

 

[w1z]

If you have no PCIe card occupying slot 3 move your GPU to slot 2 and NVMe PCIe card to slot 1 to remedy any excess heat with the GPU.

The different in temperatures between both CPUs is the not the reason you are having sudden/random shutdowns.

So check:

1. That you RX 580 is connected with dual 6-pin to 8-pin as it has been reported on here that this GPU can cause shutdowns when stressed
2. Check and report your northbridge tdiode temperature
3. That shutdowns are not software-related

 

[PetteriKiller]

1. That you RX 580 is connected with dual 6-pin to 8-pin as it has been reported on here that this GPU can cause shutdowns when stressed

I noticed! Yesterday I did not feel like working, and instead tried Geekbench 5 GPU Metal test and both times: random shutdown! I let the beast sleep for a while and changed the NVMe carrier to one slot away and voilá! no shutdowns whatsoever.

I do not like the two fan GPU idea in Mac Pro, even the RX 580 seems to be some kind best to for Mac Pro 5.1.

I wonder would is be wise to test another GPu, AMD W7100 or the RX Vega 56.

Anyway, testing is always funny.

Edit. AMD will fit for sure on Mac Pro 5.1, but how about the thickness of RX Vega 56 as RX 580 Pulse fits, not the Nitro version, since I cannot sacrifice two slots for GPU.

 

[PetteriKiller]

2. Check and report your northbridge tdiode temperature

Give a kind advise what is that and I´ll check after nap!

 

[w1z]

Give a kind advise what is that and I´ll check after nap!

Open Macs Fan Control, look for "Northbridge Tdiode" temperature and share it with us. If the temperature is very high then your northbridge heatsink's rivets have failed/broken off which could result in the unexpected/random shutdowns you are reporting and eventually damage to your processor board.

 

[PetteriKiller]

Thanks. When stressed (Premiere CC running) the temperatures were at their highest:

Northbridge Diode 78 celsius
Northbridge Heatsink 64 celsius

 

[w1z]

Thanks. When stressed (Premiere CC running) the temperatures were at their highest:

Northbridge Diode 78 celsius
Northbridge Heatsink 64 celsius

Those are normal temperatures. Are the shutdowns occuring when stressing your GPU?

Edit: sorry, just noticed that you made a post on your GPU. Is it powered by a single or dual 6-pin to 8pin cable?

Also, if you moved your NVMe from slot 2 to slot 3 or 4, your sequential read/write drive speeds, which are important for video production work, will be impacted (become slower). Best thing is to move the NVMe to slot 1 and GPU to Slot 2 if you are not using Slot 3 or increase the PCIe fan speeds to cool the GPU when installed in Slot 1.

 

[PetteriKiller]

^Please remember that my machine is originally 4.1. I cannot move the GPU anywhere else!

..by the way I noticed that the AMD Radeon RX 5700 and the RX 5700 XT are slightly thinner as their thickness is around 35 mm. as my Sapphire RX 580 is 40 mm. I'm not sure if the extra 5 mm. would make any difference. I forgot that the nVIDIAs are not the only GPUs that do not work with Mojave..

 

[w1z]

It doesn't matter if it's a 4,1 or 5,1. The GPU can be installed in slot 2, which is also x16 just like slot 1, at the cost of not being able to use slot 3.

Try it for yourself -

Slot 1 NVME --> Slot 2 GPU --> Slot 3 Empty (used by GPU due to size) --> Slot 4 Available for use

 

[bsbeamer]

You 100% can move the GPU into another slot, it just will block access.

Would suggest you check your RAM if it truly is a random shutdown issue. Also would suggest you test on a clean install to eliminate OS as source of issue vs. hardware.

 

[PetteriKiller]

I guess the problem is here, my precious and huge dual Syba SI-PEX40129:

I guess I do have to alternatives to try: thin GPU or smaller dual-carrier for NVMe´s. I do not like the idea of letting the Syba go..
[automerge]1570542396[/automerge]

It doesn't matter if it's a 4,1 or 5,1. The GPU can be installed in slot 2, which is also x16 just like slot 1, at the cost of not being able to use slot 3.

Try it for yourself -

Slot 1 NVME --> Slot 2 GPU --> Slot 3 Empty (used by GPU due to size) --> Slot 4 Available for use

I do not dare to test today, but tomorrow evening probably. Thanks anyway very much for Your advise!

 

[AlexMaximus]

I guess the problem is here, my precious and huge dual Syba SI-PEX40129:

I guess I do have to alternatives to try: thin GPU or smaller dual-carrier for NVMe´s. I do not like the idea of letting the Syba go..
[automerge]1570542396[/automerge]


I do not dare to test today, but tomorrow evening probably. Thanks anyway very much for Your advise!

I have the same IO Crest card and a modded Pixlas Vega 7 16GB card in my system. Heat is the main factor that kills your machine. Those shutdowns are a safety net, but it is clear that your Crest card blocks your GPU and overheats it in slot two as experienced. My Vega although sporting three fans would not handle the heat at all if I would put the Crest in slot two.
So out of experience a different GPU would not present a solution for you, it would be even worse. The single duct fans on Vega 56 & 64 are a lot worse than your dual 580 fans.
There are only three options here that I debate with some friends at the moment:

1. Be fine with the speed decrees to around 1500 MB/s with the Crest card in Slot 3, 4 or 5 and lose NO PCI slot.
2. Lose one slot in your Mac and put the Crest card in Slot #1 by using the GPU in Slot 2 and blocking 3 with that but have full SSD Speed that this switch is designed for.
3. Use an aftermarket water-cooling solution (i.e. Alphacool) or Liquidcoolded Vega64 LE & have full potential SSD speed, using all PCI slots, but the sacrifice at least one standard HDD SATA Bay + the Apple PCIe bay fan in the process.

This is actually the reason why I still use old AHCI blades in my Crest because in Slot3 an NVME would not be much faster.
Below you can see my Vega Evolution so far with this heating issue settling on slot 3 for now..

 

[h9826790]

Are the shutdowns occuring when stressing your GPU?

Is it powered by a single or dual 6-pin to 8pin cable?

OP, please answer these questions. It is crucial for abnormal shutdown issue.

 

[PetteriKiller]

^Oh, sorry: Yes, shutdowns occurred when set GPU was stressed.

And yes: powered with single cable..but the shutdown do not occur anymore, since I moved the IO Crest carrier further from the RX 580 Pulse.

 

[h9826790]

^Oh, sorry: Yes, shutdowns occurred when set GPU was stressed.

And yes: powered with single cable..but the shutdown do not occur anymore, since I moved the IO Crest carrier further from the RX 580 Pulse.

You should NOT power the RX580 with just single mini 6pin.

I suspect that give the RX580 some extra room to "breath" avoid the force shutdown is because your card now able to run few degrees cooler. And that will let the GPU draw little bit less power, which makes it incidentally just draw below the ~120W (mini 6pin) force shutdown protection limit.

You should use a DUAL mini 6pin to single 8pin cable to power the RX580. This is the proper way to power this card, and able to avoid unnecessary force shutdown even under extreme stress.

In fact, I believe if you run Furmark now, your cMP will still shutdown.

 

[MIKX]

TeamProfitcom Dual Mini 6 Pin to 8 Pin PCI Express Video Card Power Adapter Braided Sleeved Cable for Mac Pro Tower/Power Mac G5 15-inches

Features: Dual Mini 6 Pin to 8 Pin PCI-e cable is to connect a PCIe video graphics card with 8 pin port to Mac motherboard with mini 6 pin port; Specification: Cable Length: 15-inches Connector 1: standard GPU 8 Pin male ; Connector 2 : 2x mini 6 pin male ( not standard 6 pin male);Gauge:18AWG In...

www.amazon.com

 

[tsialex]

You should NOT power the RX580 with just single mini 6pin.

I suspect that give the RX580 some extra room to "breath" avoid the force shutdown is because your card now able to run few degrees cooler. And that will let the GPU draw little bit less power, which makes it incidentally just draw below the ~120W (mini 6pin) force shutdown protection limit.

You should use a DUAL mini 6pin to single 8pin cable to power the RX580. This is the proper way to power this card, and able to avoid unnecessary force shutdown even under extreme stress.

In fact, I believe if you run Furmark now, your cMP will still shutdown.

What is the RX 580 manufacturer and model @PetteriKiller?

If it's a Dell RX 580 with just a single mini-PCIe 6pin connector, it's a card known to cause shutdown problems with Mac Pros, since it's really a RX 480 that was factory modded to use RX 580 firmware and use more power from the PCIe slot than a Mac Pro can power.

 

[h9826790]

What is the RX 580 manufacturer and model @PetteriKiller?

If it's a Dell RX 580 with just a single mini-PCIe 6pin connector, it's a card known to cause shutdown problems with Mac Pros, since it's really a RX 480 that was factory modded to use RX 580 firmware and use more power from the PCIe slot than a Mac Pro can power.

I believe that's the Sapphire PULSE RX580

 

[tsialex]

I believe that's the Sapphire PULSE RX580
View attachment 867957

Missed that, thx.

A Sapphire Pulse RX 580 needs a dual mini PCIe 6-pin to single 8-pin cable and will shutdown the Mac Pro with only one mini PCIe 6-pin connected.

 

[PetteriKiller]

Thanks for advise!

Well, I will order asap the cable - which means I will get it in a week since I'm living & working here behind God´s back in the Finnish countryside.

I built the Frankenstein-Mac Pro around a year ago, and it has been working very-very well in a daily use of Photoshop CC & Premiere CC, unless the shutdowns which started suddenly september 17th this year - I still remember the panic when the first shutdown happened. That is the reason I started suspecting the processors, some update etc. first. Hope I did not make permanent harm for the RX 580, even if it´s cheap, even here in Finland.

2. Lose one slot in your Mac and put the Crest card in Slot #1 by using the GPU in Slot 2 and blocking 3 with that but have full SSD Speed that this switch is designed for.

I will do that. I have my archives in a two 2,5" SSDs RAID set-up on Sonnet card and there's no particular need for speed for my archives.

 

[PetteriKiller]

Are the shutdowns occuring when stressing your GPU?

"Stressed" means here that I was using Premiere CC with 4K video with shutdowns, and then I noticed that the new Geekbench 5 could not do the Open CL- or Metal -test for RX 580 because of an immediate shutdown.

 

[h9826790]

Thanks for advise!

Well, I will order asap the cable - which means I will get it in a week since I'm living & working here behind God´s back in the Finnish countryside.

I built the Frankenstein-Mac Pro around a year ago, and it has been working very-very well in a daily use of Photoshop CC & Premiere CC, unless the shutdowns which started suddenly september 17th this year - I still remember the panic when the first shutdown happened. That is the reason I started suspecting the processors, some update etc. first. Hope I did not make permanent harm for the RX 580, even if it´s cheap, even here in Finland.



I will do that. I have my archives in a two 2,5" SSDs RAID set-up on Sonnet card and there's no particular need for speed for my archives.

That force shutdown is a protection. It shouldn't hurt the hardware, but may cause data lost.

However, multiple frequent hard shutdown can't be good to the RX580 nor the cMP. You should fix this powering issue ASAP.

There are many reason why it didn't happen before.

e.g. as I said before, temperature, the GPU draw more when it run warmer. When the graphic card was new, the heatsink / fan was clean, it may run few degrees cooler then now. Or may be simply because the thermal paste's performance is degraded.

Or, the software now able to better utilise the hardware. So, even there is no hardware changes (including the cables), but the software upgrade / update may cause higher demand to the GPU, which just stressful enough to trigger the shutdown protection now.

 

[Rolo10]

Hi Guy, I'v same Mac (5.1, dual intel x5690, Shappire Pulse x580) with random SHUTDOWM when stressed run. -- Testing with Geekbrench 5 shutdown . Fans normally running ok, I've never ear it, I've dual cable, and I've two thing to pay attention a) over Sahppire I've a Lycom NVme, tap 50% fan b) Mac are connected and UPS APC... any tips could you give to avoid this shutdown?? thk. a lot