Vega 64 Shutdown

[lessardsean13]

When my PC turns on, my Vega GPU cannot go under any major load, or the card will shut off (the PC remains on). Also, after a shutdown, it won't stay on for very long, and will shut off after a few minutes of use. I have to leave the system unplugged for several hours for the card to be useful for web browsing and YouTube videos again, but still no games. What is going on with the card? I have already tried different power supplies.

 

[MadOver]

How much are the Power Supplies? AMD requires 650w+ ...

 

[lessardsean13]

Mine is an 850W Gold PSU. Also, the GPU has two seperate power cables going to it from the supply. I wanted to add that the GPU Tachometer shuts off completely when the card shuts off, and there is no red LED.

 

[g-unit1111]

Mine is an 850W Gold PSU. Also, the GPU has two seperate power cables going to it from the supply. I wanted to add that the GPU Tachometer shuts off completely when the card shuts off, and there is no red LED.

That doesn't really help. It needs to be a specific brand / model. You can usually find this information on the side of the PSU.

 

[lessardsean13]

It is the Thermaltake Grand RGB 850W PSU (80+ Gold).

 

[g-unit1111]

It is the Thermaltake Grand RGB 850W PSU (80+ Gold).

That's a decent unit - it should be able to supply the power needed to power that particular GPU. Can you try removing the GPU and running on onboard video? Because if it posts without the GPU then it's definitely the GPU that needs to be RMA'd.

 

[lessardsean13]

There is no on board video, unfortunately, I have a Ryzen 1600X. Later today, I can test my spare Radeon HD 7450 I have on hand.

 

[mdd1963]

Although the PUS *should* handle it, you will know it is likely not the PSU only after substituting a known good one of sufficient wattage...(any PSU can be weak, and subject to brownouts under load; likewise, a good UPS can prevent brownouts as well if your home is weak in current supply, dimming lights when AC turns on, etc..)

 

[g-unit1111]

There is no on board video, unfortunately, I have a Ryzen 1600X. Later today, I can test my spare Radeon HD 7450 I have on hand.

Yeah I'd try it with the other GPU first. If you have another system you could try your GPU in that to make sure it works as well.

 

[lessardsean13]

The cheaper card worked fine, but now I want to know why the Vega card would be randomly shutting off. I had installed the Alphacool Eiswolf 240 on it recently, could it have been related to that?

 

[norcalsc]

The cheaper card worked fine, but now I want to know why the Vega card would be randomly shutting off. I had installed the Alphacool Eiswolf 240 on it recently, could it have been related to that?

YES, are you monitoring temps as it shuts down? It sound like maybe your pump isnt working or maybe improper install or possible damage? like its going into a thermal shut down because it cant cool itself. I would check all connections and make sure that all contacts points for gpu and vrm are correct. also see if that pump is working while your pc is powered on

 

[Karadjgne]

Hopefully you covered every prior contact like VRM's, vram etc with thermal tape and used decent paste on the gpu.

 

[lessardsean13]

I had installed the cooler and stuff correctly, and temps are fine, because it had been able to power up under low load for a while, and would crash under full load, and it now just crashes all the time. The cooling is fine, as Radeon Wattman reads the temps in the 20s.

Update: I had taken off the cooler and ran it bare, just for a minute or two, and it crashed the same way. There is no physical damage.

 

[Karadjgne]

There's only 1 sensor for temps, that's under the gpu. The processor. The entire other side of the card, includes the voltage regulatory circuitry, the vram and all the included mosfets etc. They have no temp sensor. So while the graphics processor might be a nice 20°, that's not to say unregulated components on the other end of the board aren't hitting 100°. Every single chipset, transistor, mofset, vram module etc that previously had some sort of connection to the heatsink, needs that same connection replaced. Miss just one, and the card goes thermally nuts and shuts down on you.

Reliance on software to tell you the temp of a non temp regulated component is a fools errand.

 

[lessardsean13]

The Eiswolf cooler contacts every part of the card, as it essentially is one giant heatsink (with the assistance of thermal pads of course). When installing it, instructions told me where to put every pad, so it is contacting every essential component. Also, there was an included backplate that also included thermal pads for the rear components as well.

Update: By any chance, is Arctic Silver 5 conductive? That is the paste I applied on the card.

 

[Karadjgne]

No, AS5 as such really isn't conductive, any more so than any other paste, it's just junk imho. It's a mediocre paste that doesn't do as well as other 'grease' type pastes. Noctua, grizzly kryonaut, arctic MX-4, gelid extreme would be better choices.

I'd check, then double check, make 110% sure visually that every essential component is covered by the pads. Visually check vrs the OEM heatsink, make sure one of the components didn't get folded over accidentally, or missed because it was folded over. Get an accounting of everything 1 by 1. Doesn't take much to take a custom pcb and change a few things, the XFX is slightly different than reference.

 

[lessardsean13]

I will check tomorrow afternoon. It is a Gigabyte reference PCB BTW.

 

[lessardsean13]

After checking the card, it didn't appear to be physically damaged, but it still doesn't work. I must have static shocked it or something, so I am sending it to Gigabyte in an RMA.

Final question: In the GPU box, there was a 2 6-pin female to 1 8-pin male connector. Would this provide more power to the card? Is it better than an 8-pin to 8-pin?

 

[maxalge]

After checking the card, it didn't appear to be physically damaged, but it still doesn't work. I must have static shocked it or something, so I am sending it to Gigabyte in an RMA.

Final question: In the GPU box, there was a 2 6-pin female to 1 8-pin male connector. Would this provide more power to the card? Is it better than an 8-pin to 8-pin?

no

same thing

6-pin is rated for 75w
( since you have dual 6-pin cable to single 8-pin) = 150w

8-pin ( or 6+2 ) native = 150w

 

[lessardsean13]

Okay, can any of you guys reason why the card works with the reference cooler? I had installed it for the RMA and did a test, and there are no issues, but the Eiswolf causes this problem.

 

[Karadjgne]

Prior to reinstall, was the back of the reference cooler pure flat plane? Or were there offsets that'd bring the heatsink closer to the pcb and components. If it's not pre flat, it's entirely possible that the Eiswolf is not making good contact with 1 or more components, so any thermal pad there is next to useless, allowing that 1 component to overheat and cause the issues? By rights, they should be identical under the heatsink, but that's not always the case.

 

[lessardsean13]

The reference cooler is diveted for the components such as VRMs, and the same can be said for the Eiswolf. Comparing the two coolers, they should both make proper contact, so it is not a cooling issue. However, it most likely was static shock, as I had discovered the Eiswolf's cooler was scratched, so the anodized aluminum was exposed. Luckily, the card still works and the cooler's pump is still useful, so I will return the cooling block and get another.