Could AMD driver curse be back with mysterious RX 6000 failures?
>
Some of AMD’s RX 6000 graphics cards are giving up completely, with a bunch of broken GPUs suddenly brought to our attention by a German repair shop.
That shop, KrisFix, pointed out this problem YouTube (opens in new tab) (through WCftech (opens in new tab)with Buildzoid on Twitter as well), and the visual evidence provided is indeed concerning.
RX 6800/6900 series cards seem to have core failures for some reason: https://t.co/NScb6rQbkrJanuary 10, 2023
We are not talking less about cracked GPUs, and if that happens to the chip, the entire graphics card is of course immediately written off.
KrisFix informs us that in the past three weeks he received 61 AMD graphics cards for repair – that’s the RX 6900 and 6800 models – and that 48 of those cards had the exact same problem, a cracked GPU.
That is almost 80% of all those models that are sent in for repair, a surprisingly high percentage it must be said; and all those he further observes had a shorted SOC rail, a shorted memory rail, and a shorted memory controller rail (indicative of a GPU catastrophe). Normally he would expect about 10% of the submitted graphics cards to suffer from something as bad as a cracked chip – not 80%.
However, there was one thing in common with the problem boards: KrisFix notes that they all used the same graphics driver, the latest from AMD.
We’ve reached out to AMD for comment on what might be going on here, and we’ll update this story if we get a response.
Analysis: Is mining chaos a possibility?
As KrisFix points out, it’s hard to say with any certainty what’s going on here, but given the commonality with the driver, he theorizes that the latest Adrenalin version (released a month ago) might make performance a bit more difficult, and maybe some RX 6000 GPUs can’t handle this.
At the same time, he makes it very clear that he doesn’t have enough information to make that judgment just yet, and it’s pretty clear that we can’t draw any definitive conclusions as to whether the graphics driver is really at fault here.
At this point we only have word from a German repair service, with KrisFix itself admitting that we need to be very careful and that we need more evidence. Indeed, if there really was a problem with RX 6800 and 6900 GPUs, surely we’d be hearing a lot more about this now, a month after the driver came out?
On the other hand, you can argue about the number of people who regularly update their graphics driver. There are certainly a lot of people who stick with the same GPU driver for a while and certainly don’t accept every version released.
Another possibility is that the problem is limited to certain graphics cards, not all of them. Here’s another suggestion that comes into play: that the affected models are ex-mining GPUs. When mining operations threw in the towel last year, they sold the graphics cards in their farms, and these are GPUs that are literally lashed almost to death 24/7.
So it’s not hard to think that a good number of gamers have picked up second-hand stock of these ex-mining AMD GPUs – possibly unaware that they are ex-mining models, as sellers often hide this for obvious reasons – and these have become problematic.
Maybe the problem is with those ex-mining cards and the new driver. If the latter does indeed push for more performance – which anecdotally some owners have said they get slight frame-rate improvements – perhaps this is tipping some GPUs, which were almost dead anyway, over the edge. Of course, all this is just speculation.
It’s also worth noting that there are scattered reports of coil whine after updating to the latest AMD driver and of higher hotspot temperatures, again suggesting that the GPU is being pushed a bit harder. But we really have to wait and see if more reports of this problem with RX 6000 cards come in now that the alleged problem is out in the open.
If there does turn out to be a problem with the graphics driver itself, it won’t look good for AMD as 2023 is barely underway and Team Red has already had a GPU crisis with its new RX 7900 XTX flagship and a cooling problem. A second crisis this early in the year would of course be quite embarrassing.