RX 480 - Problems

If you think it might be a driver problem, see viewforum.php?f=79

Moderators: Site Moderators, FAHC Science Team

iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

gunnarre wrote:If the work unit threw a NaN fault once but not on a re-try then it indicates that your card is barely stable. FAH doesn't officially support over- and under-clocking GPUs, and will tax GPUs harder than games, so if you're doing any under-volting or under-clocking then I'd reverse that. Using the power limiting feature in the driver is a safer way to reduce power draw.
A 30% decrease in Power Limit immediately cause an: FS02:0x22:An exception occurred at step 999481: Particle coordinate is nan, error. Something seems of with the GPU after all.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
gunnarre
Posts: 567
Joined: Sun May 24, 2020 7:23 pm
Location: Norway

Re: RX 480 - Problems

Post by gunnarre »

iero wrote:The card is in its stock configuration.
The MSI RX 480 GAMING X 4 GB has a very slight factory overclock (3% according to the TechPowerUp GPU Database) and that might be enough to make it unstable when you give it a heavy workload like FAH. It might vary from project to project how much it taxes the card. Or it might be a function of the age of the card.
iero wrote:A 30% decrease in Power Limit immediately cause an: FS02:0x22:An exception occurred at step 999481: Particle coordinate is nan, error. Something seems of with the GPU after all.
What might be happening here is that a lower power limit causes the GPU to pause its work more often, and when it resumes working it has enough temperature budget to boost to a clock (on the GPU or the VRAM) that is unstable. This might not be noticable in graphics, but appears in compute. Paradoxically, this might get worse with better cooling, because the card would otherwise be temperature limited and not boost into an unstable clock.

Despite what I said about not underclocking, you might try reducing the boost clock to 1266 MHz and the memory clock to 2000 MHz, which are the factory settings for the reference model RX 480.
Image
Online: GTX 1660 Super, GTX 1080, GTX 1050 Ti 4G OC, RX580 + occasional CPU folding in the cold.
Offline: Radeon HD 7770, GTX 960, GTX 950
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

gunnarre wrote:
iero wrote:The card is in its stock configuration.
The MSI RX 480 GAMING X 4 GB has a very slight factory overclock (3% according to the TechPowerUp GPU Database) and that might be enough to make it unstable when you give it a heavy workload like FAH. It might vary from project to project how much it taxes the card. Or it might be a function of the age of the card.
iero wrote:A 30% decrease in Power Limit immediately cause an: FS02:0x22:An exception occurred at step 999481: Particle coordinate is nan, error. Something seems of with the GPU after all.
What might be happening here is that a lower power limit causes the GPU to pause its work more often, and when it resumes working it has enough temperature budget to boost to a clock (on the GPU or the VRAM) that is unstable. This might not be noticable in graphics, but appears in compute. Paradoxically, this might get worse with better cooling, because the card would otherwise be temperature limited and not boost into an unstable clock.

Despite what I said about not underclocking, you might try reducing the boost clock to 1266 MHz and the memory clock to 2000 MHz, which are the factory settings for the reference model RX 480.
I think I found the culprit, its the Mobo. I just saw that the GPU is running at x8 1.1 PCI-E.... I tried to resit the GPU, clean it with isopropyl alcohol, the contacts on the GPU too, cleared cmos, nothing. Still x8 1.1. -Warranty just expired too... My luck...
I think I'll try to lower the core and mem a bit, and if that fails, I'll just leave it stock, it seems to run at 90ish% usage most of the time, so I'll take it.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
gunnarre
Posts: 567
Joined: Sun May 24, 2020 7:23 pm
Location: Norway

Re: RX 480 - Problems

Post by gunnarre »

Unless you've tested the card in a different motherboard, or a different card on the motherboard, I'm not sure if you've isolated the issue yet. (e: The spiking and stopping behavior does sound like it might be PCIe bus bandwith related though.) If it is the motherboard, you might be able to run two cards in the other computer. Folding needs more PCIe bandwidth than crypto mining but less than gaming.
Image
Online: GTX 1660 Super, GTX 1080, GTX 1050 Ti 4G OC, RX580 + occasional CPU folding in the cold.
Offline: Radeon HD 7770, GTX 960, GTX 950
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

Both of my pc's have 1 pci-e x16 slot, so I would have to DDU my main pc to install the RX, and I'm not in the mood to be honest. I just did the same a few days ago to put the 3060 in place of the 5700 XT.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

gunnarre wrote:Unless you've tested the card in a different motherboard, or a different card on the motherboard, I'm not sure if you've isolated the issue yet. (e: The spiking and stopping behavior does sound like it might be PCIe bus bandwith related though.) If it is the motherboard, you might be able to run two cards in the other computer. Folding needs more PCIe bandwidth than crypto mining but less than gaming.
Would running Linux on my second system, the one with the potentially bad mobo, boost the PPD? Something that has to do with Linux being more bandwith forgiving? I do not know squat about Linux... :roll: :oops: :lol:
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

Image
This was 3 days ago and again today. I didn't do anything to fix it. it was running at x8 3.0. then x8 1.1. then x8 3.0. all on its own. WT f@@@ is going on... :oops: :shock:
Last edited by iero on Wed Mar 17, 2021 11:40 am, edited 1 time in total.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
gunnarre
Posts: 567
Joined: Sun May 24, 2020 7:23 pm
Location: Norway

Re: RX 480 - Problems

Post by gunnarre »

What is the particular model and revison of the motherboard. Which "BIOS"/UEFI and AGESA version?

Seems like the motherboard's PCIe lanes or the PCIe controller on the CPU or GPU are having hardware issues or firmware issues. Can you try a known good GPU in the bad motherboard? That should avoid having to mess with the OS of the computer that is working.
Image
Online: GTX 1660 Super, GTX 1080, GTX 1050 Ti 4G OC, RX580 + occasional CPU folding in the cold.
Offline: Radeon HD 7770, GTX 960, GTX 950
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

Its an Asrock A320M-HDV revision 3 P.1.10 . Latest BIOS on Asrock's website is 3.60.
Image
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

And back to x8 1.1. without me even being near the computer.
Image
Any ideas how to go about troubleshooting?
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
gunnarre
Posts: 567
Joined: Sun May 24, 2020 7:23 pm
Location: Norway

Re: RX 480 - Problems

Post by gunnarre »

Do this?
gunnarre wrote:Can you try a known good GPU in the bad motherboard?
We still don't really know if the GPU or the motherboard is the locus of the issue.

After that, try a different power supply.

Some caution here though - if the motherboard or PSU is very broken, you might harm the component you are testing. So maybe not test with the 3060.

If the PCIe slot itself is the problem, you might even try using a 1x-to-16x riser cable in the other PCIe slot. That will harm throughput, but even 1x 2.0 is better than a broken slot. Re-flashing the Bios might help, but if the CPU or chipset is broken, it might be time to toss the motherboard.
Image
Online: GTX 1660 Super, GTX 1080, GTX 1050 Ti 4G OC, RX580 + occasional CPU folding in the cold.
Offline: Radeon HD 7770, GTX 960, GTX 950
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: RX 480 - Problems

Post by bruce »

iero wrote:Would running Linux on my second system, the one with the potentially bad mobo, boost the PPD? Something that has to do with Linux being more bandwidth forgiving? I do not know squat about Linux... :roll: :oops: :lol:
Linux is a fine OS and Windows does require more bandwidth that Linux, but Linux tends to be more demanding of your need to learn about it :idea: :idea: . If you're comfortable with Windows, stay with Windows. Changing OSs for a small change in performance puts the demands on your brain :eugeek: .
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

Image
The above is from the mobo's manual. Does it mean that it runs in x8 3.0 if there is an nvme as a boot drive, or am I reading it wrong?
Cause if using a SATA drive, so that wouldn't explain the x8 either. Let alone the drop to 1.1. I have to get in there after all.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
gunnarre
Posts: 567
Joined: Sun May 24, 2020 7:23 pm
Location: Norway

Re: RX 480 - Problems

Post by gunnarre »

It means that a Raven Ridge CPU (like the 2200G) will only run in PCIe v3 x8 mode in the board. That should still be plenty for folding, and doesn't explain why your PCIe controller drops to PCIe v1.1 The manual does not say that the M.2 slot will slow down the PCIe slots. Unfortunately, the manual is lacking a chipset block diagram, which is often useful to understand any lane-sharing at a glance.

It might be the CPU, the motherboard, the M.2 drive, the GPU, the RAM or the PSU, so until you open it up and start swapping components, we're no closer to a solution. Is your SSD on the Storage QVL list? If it isn't then it doesn't necessarily mean that it's the SSD's fault, but you might try putting in a second hand laptop harddrive and boot from that instead. That's what I use in one of my folding machines (the one used to heat the wardrobe in winter). Edit: Or get an NGFF to 2.5 SATA SSD adapter, which lets you plug an M.2 SATA drive into a regular SATA port.

PS: Are you running your desktop on the integrated graphics or the GPU?
Image
Online: GTX 1660 Super, GTX 1080, GTX 1050 Ti 4G OC, RX580 + occasional CPU folding in the cold.
Offline: Radeon HD 7770, GTX 960, GTX 950
iero
Posts: 177
Joined: Tue Feb 09, 2021 10:40 am

Re: RX 480 - Problems

Post by iero »

gunnarre wrote:It means that a Raven Ridge CPU (like the 2200G) will only run in PCIe v3 x8 mode in the board.Unfortunately, the manual is lacking a chipset block diagram, which is often useful to understand any lane-sharing at a glance.
PS: Are you running your desktop on the integrated graphics or the GPU?
Image
You are indeed correct.
I'm using the GPU fro output.
I tried a known working GTX 1050, couldn't replicate the issue.
I think it is an issue with GPU-Z. the only times the x1.1, as far as I can tell, appears is then I start the program and the GPU is in an idle state.
If the GPU is folding during the boot up of the app, it always shows as x3.0.
Image
GPU only
RTX 3060 12GB Gigabyte Gaming OC [currently mining]
Folding since 14/02/2021
Post Reply