thermal problems with XFX RS XXX RX 570 OC+ card

[VirtualLarry]

I've had thermal problems with this card for any sort of heavy compute load, ever since I got it.

With the card singly in my R5 1600 (AIO LC), ATX case, not the best cooling, but it has a 120mm intake, and a 120mm blowhole exhaust, and the rear 120mm is the radiator in exhaust (push).

Running F@H on the GPU, with stock settings, or even with WattMan Power Limit at -20%, it still hits 89C and pins there.

Now I've switched to doing NiceHash (coin mining), and added my RX 470 4GB (also XFX) as the secondary "bottom" card. Both cards have backplates, so there's very, very little room between. Pretty-much none.

But for the life of me, I can't see to control the temps very well.

The bottom RX 470, running at 1201 Mhz, 54C, fan RPM 0, 60W power usage.

The top RX 570, running at 400-500 Mhz, 83C, fan 2200 RPM (set in WattMan from 2500-3000, not sure why it's slower), similar 60-70W.

Why the disparity? Is cooling a video card THAT HARD?

I'm seriously starting to think that the heatpipe / heatsink on this RX 570 is just faulty.

http://xfxforce.com/en-us/products/amd-radeon-rx-500-series/rx-570-rs-4gb-dd-triplex-rx-570p427d6

XFX Dual BIOS Mining

Equipped for gaming and mining

We know enthusiast like options, our RX 570 and 580 graphics cards are shipped with a Dual BIOS. If you want optimal gaming performance just install it and go. For cryptocurrency mining, simply shut the system down and flip the BIOS switch on the card. Our mining BIOS is already tuned to give increased mining hashrates without hassle.

Hmm, so I just need to flip the BIOS switch? Interesting. Let me try that.

 

[ao_ika_red]

Well, if it's about thermal dissipation related issue, the only way to go is RMA it because you can't check whether the thermal paste / heatpipe is working due to warranty sticker on the mounting bolts.

 

[VirtualLarry]

I removed, and re-installed the drivers (17.12.2 Adrenalin), after powering off and flipping the BIOS switch on both cards. That didn't work so well with the RX 470 (whole system was moving sluggish, something was clearly wrong), so I powered off and flipped the RX 470 switch back to it's original position, with the RX 570 in "mining" mode, which appears to have done virtually nothing for the temp issues, but it did clock the VRAM up from 1750 to 1850. But core clock is throttling around 400-500Mhz, rather than 1284, or even at least 1100Mhz. I did use WattMan to set the temp throttle limit down from 89C to 83C, seems more reasonable to me. But it's clear, that the heatsink / heatpipes on this card, are NOT adequate to cool the GPU.

IOW, since when is thermal throttling down to half performance or less, when choosing to use a supported feature (crossfire), acceptable?

Edit: Now it's throttling HARD. Down to 300Mhz core clock (lowest it will go, I think), and temps are climbing above the 83C max I set, now they're 87C. Fan is at 2200 RPM, 56-60W power usage.

I don't get it, CPUs that are 65W TDP, can be cooled by a smaller hunk of Al, with no heatpipes, with a small down-draft fan. You're telling me a 60W GPU, with dual fans, heatpipes, and lots of fins, can't cool that same TDP? WTF is going on here???

 

[ao_ika_red]

If RMA is out of option, I suggest to return into stock bios and dial back memory and core voltage. My 560 once called by gamernexus as having the poorest cooling solution but after undervolting it with some help from other member, now it runs peacefully at 1300MHz/1030mV/68°C/1200rpm.

edit: Have you checked heatpipe condition? There maybe a puncture or cutout that render its heat transfer capacity useless. I remember evga had this mfg defect on their 970 some years ago, it might happen to yours as well.

 

[PG]

I think there are two possible problems:

1 - card spacing - You can't have good cooling and temps if that top card can't suck in any air. Switch the cards around and I bet the "good" card will now be running really hot. What board do you have? Isn't there a better spacing you can use?

2 - replace the TIM if that doesn't mess up your warranty - It may just be that the heat isn't getting to the heatsink because of a bad TIM job.

 

[Puffnstuff]

Well, if it's about thermal dissipation related issue, the only way to go is RMA it because you can't check whether the thermal paste / heatpipe is working due to warranty sticker on the mounting bolts.

Doesn't XFX have a lifetime warranty? If so then RMA it.

 

[IntelUser2000]

I don't get it, CPUs that are 65W TDP, can be cooled by a smaller hunk of Al, with no heatpipes, with a small down-draft fan. You're telling me a 60W GPU, with dual fans, heatpipes, and lots of fins, can't cool that same TDP? WTF is going on here???

GPU heatsinks need to handle way more than 60W, for an RX 470/570.

The total is really the total consumption for the card, which is 120-150W. By the way, I doubt GPU heatsinks are as capable as it looks initially. They have a strict height requirement, so it doesn't block majority of your PCIe slots. That cuts down cooling capability, despite the fancy tech on it. And yes, it could be something as simple as needing a repaste. But warranty exists, then send it in.

 

[VirtualLarry]

Well, I noticed that the heatsink arrangement on my RX 570 has one LESS heatpipe than my RX 470 does, and it seems like the shorter one near the GPU, for some reason. Maybe that's why it doesn't seem to have the cooling capacity.

 

[3DVagabond]

Well, if it's about thermal dissipation related issue, the only way to go is RMA it because you can't check whether the thermal paste / heatpipe is working due to warranty sticker on the mounting bolts.

Unless AMD has changed their policy recently this is not true. At least in the US.

Doesn't XFX have a lifetime warranty? If so then RMA it.

Not any more. Haven't for a few years now.

Have you contacted XFX? If not, then do so. FWIU they offer excellent customer service. Their rep is also quite active online and seems to be very helpful.