Discussion in 'Videocards - AMD Radeon Drivers Section' started by Aizu, Jan 10, 2023.
No i believe it also happens to custom cards
yea reference only
How to kill cards, nope not with drivers, now how to kill one its easy, open card put termal paste and just torque the card out of spec, done its dead. This is how the normal user kills a card and cracks the die. So if its a mining card probbaly to do this faster they used a eletric tool. Drivers dont kill cards in this day , they will overheat and trotle if the fans just stop, and then they will make the system shutdown.
I can risk damaging my x570-E chipset by running an old vision of openrgb cos it causes the chipset fan to pause that has no fan control.
So tell me more about how drivers cannot cause issues that potential will always be there.
Chipsets arnt graphics cards though. Pretty sure a chipset cannot throttle, reduce currents or shut down to protect itself.
Yeah yeah you cant kill AMD cards cos they are perfect, Nvidia makes a mistake you donkey, AMD makes a mistake its oke we all make mistakes
Could it be certain GPUs have coolers mounted in a way that would only be an issue with a notable change in operating temps? What if a driver changes how long a GPU stays in boost frequencies, or how long it stays at X voltage? What if temperatures are spiking more now because of more-aggressive ramps that weren't accounted for during initial testing?
The driver controls frequency and temps enough to affect game performance notably; it's not ideal by-default on Linux for RDNA2, but apparently is fine on Windows which would have higher and more consistent temperatures. Same card firmware, just different drivers with different heuristics(?)
Spoiler: Made-up example
Maybe 68xx/69xx performance was causing issues with power balancing and doing long throttling periods leaving high heatsink temps and overall lower performance, so now it shifts frequencies more-aggressively in order to maintain temps lower, so now 60% GPU load has 90% max frequency now instead of like 70% of max frequency. Now the GPU stays cooler, and most games are unaffected, except ones with sharp GPU load differences (100% during game with frequent frame-capped load screens), which now because of the new hints, cause the GPU to downclock harder during the lower GPU scenes instead of staying at the higher clocks longer like before, and then causing the temperature to shift sharply along with that.
I'm just guessing But it seems to be able to be tested for anyone determined enough to monitor frequencies, power usage, and temps with different loads and different drivers.
I'm also pretty late to this discussion and I'm not aware if this was debunked already or not
Yeah because you compare motherboard chipsets to the GPU chip, that is completely irrelevant
VBIOS behaviour honestly
Yeah thats an assumption i suffer brand loyalty, ive got as many AMD cards as i do NV cards, i dont care who makes then same as Intel and AMD CPUs, id say all the SSDs i use are all different brands as are the MOBOs i use. Its nothing to do with the name of the manufacturer, its the likely hood that with what we currently know, it does seem a little unlikely that this is a global issue ie drivers, when we have only seen this reported very recently and only from a single source. The NV thing with the power socket was reported globally as were the AMD vapour chamber issue. But just a single repair shop reporting this in such a tight timeframe, doesnt sound plausible.
The lack of widespread followups is loud.
Im far from an expert in any of this but im confidant it would take something pretty significant to produce the damage that has been reported and shown. Ive been on these drivers since they came out and i run my 6900XT up to 320W and all my temps whether idle, at lower performance 200W/2300mhz or flat out 320W 2600mhz (under water) remain similar to all previous drivers *relative to the room ambient. Even if the driver stopped the fans then the resultant overheating, throttling and shut down wouldn't produce that level of structural failure.
Im not convinced, those dies looked like it was more than just 110C, and i doubt 110C would cause any issues, these chips are built and designed to work at a % level above the recommended max temps.
Because GPUs have bios hard thermal protection built in that the drivers cannot alter. At 110c they throttle regardless if the fans are running or not they will continue to throttle down to the lowest point to stay under 110c which is spec.
I don't have any issues with my graphics card (reference) with last driver 22.11.2 WHQL
I mean probably but the bios is hard locked to 110c hard throttle down to not exceed 110c regardless. Only way you are exceeding that and not hard stop throttling is with a modded bios
Guess that means if i take off my cooler the card will survive like Jay tested in past with Nvidia gpu.
Follow up video:
So as expected, the "Radeon driver bad" theory was a nothing burger. the people that were saying the drivers are to blame must take a good look in the mirror and please stop talking about stuff that you have no idea about.
Honestly to me this all was just a big scam. Now he comes up with conlusions like "could be crypto miners, only a few were the original owners and the previous owner stored them wrongly". And with that Info he was waiting for a week keeping up these wrong claims that it could be the driver..