Всем привет.

I have two identical cards Asus DUAL-RX580-O8G (Hynix memory) with exactly the same symptoms.

1. All voltages are correct
2. All resistances are correct
3. Both cards show the boot screen and I can enter the bios, all is normal
4. dmgg.py says the RAM chips are good, zero errors (python3 dmgg.py d0000000 8 8)
but
5. Both cards are detected
6. Windows does not initialize cards with drivers, I get a picture of "noise" like random data written to the video memory and computer reboots. These are not typical artifacts but rather white noise on the whole screen and then restart.
This "white noise" looks like a monitor which lost synchronization between clock and data and is similar to the analog noise, small pixels in random colors.
7. VRAMs work properly, there is 12V PWM on each coil input.

When both cards are a secondary cards then I get a blue screen when I try to use GPU-Z on the faulty card. They show up in device manager as good (no exclamation mark) but when I want to access them they crash computer.
Both behave exactly the same, the same failure.

Should I say both GPUs are dead?
Do you have a tip what to do next?

Hello,
You can try to flash the bios at the beginning

https://www.techpowerup.com/vgabios/194 ... 192-170417

Thank you very much for pointing me to this driver. There is a little bit better, they are unstable as they were but when I finally manage to start windows I have this kind of artifacts. This time the Windows did not start properly but it was this. What do you think?

i think you need to check ram with "tserver". Maybe some memory chips with errors.

can you send me a picture via private message? because i can't open it here.

The first card:

---

Posted after 1 minute 10 seconds:
The second card:

After test use command

pico memfa.log

and there you will found information about faulty channel.

according to the photo, replace the damaged VRAM.

On 300MHz/300MHz nothing is wrong, tests always pass so on a 1000MHz/1000MHz. When I initially increased to the -mem=2000 nothing happened so I increased also GPU and occasionally there are single bits in red. I have such strange symptoms for the first time and two cards with almost the same behavior.
Could it be a memory controller? Or rather memory itself? I don't have Hynix chips now anyway, I must order them from China.
When I set memory to 2200 the Linux hangs and I have to power off everything.

Personally i don't think that problem will be with ram controller, if all your test are passed correct then problem might be with GPU.
But from your previous messages i seen that both tests was failed, so there should be some information about error

Can you send me a picture of artefacts via private message?
because i dont have 15 posts to check your pictures.

I have sent these jpegs to PM.
I analyzed these cards again. They are very strange twins.

1. Number one
- Linux cli works
- boardtest=memfa C0 VRAM has 1 eror ".......... ......... ...1.... ........." always the same
- HDMI connected directly to the card -> artifacts in memfa
- boardtest=quickmfg goes to the certain point and stays there
- No need to power cycle in order to issue another command

2. Number two
- Linux cli works
- boardtest=memfa always PASS, memory is good even at 2000MHz
- HDMI connected directly to the card gives PASS in memfa, this test goes perfectly
- boardtest=quickmfg goes to the certain point and hangs the card
- This card requires power cycle after boardtest=quickmfg, only 5V is present and small LDO gives 3.3V all other voltages are gone.
- I discovered that when I want to increase system clock from 300 to a higher value it immediately disables all voltages but I can increase memory clock without problems and run memfa test. This card does not react on commands for setting the fan speed 0,1,2, and 3. No reaction.

They both look good under thermal cam, no point heating or border heating of the GPU. All VRMs of the GPU are functional, 7-8% of duty cycle on each phase 0.900V.

Hi, maybe you are found some solution? I have this card with 4GB Elpida vram. And she has the same problems. I check fases with oscilloscope and all ok on it. Only 3D test give CRC errors. I think problem with GPU. I rebolling it, but problem not solved.

Not yet, I am waiting for Hynix DDR5 in order to repair one. This one with faulty RAM starts on one GPU VRM second from the top. This faulty with good RAM starts on all VRMs together
So now there is a comedy I don't know what behavior is good

победили?

Hello
Look, this is really from the memories:
when entering windows through another gpu everything is flawless. As soon as you use gpu-z, it will read your entire GPU. but if an error occurs only in one bit, it will cause confusion on both boards. then the pc freezes immediately. It's from memory.

What software do you use to test the memories?
I use linux for all of them up to the RX 580.
always impeccable.