I get these irregular errors on memtest vulkan that happen every ~5 seconds to ~5 minutes.
first of all I have considered it may be something wrong with the program since it seemingly doesn't affect performance, but OCing the VRAM makes it happen slightly more often. so I am thinking it might be that the errors are negligible and that ECC takes care of it. essentially that the performance boost outweighs the loss from error correction
I actually keep getting performance increases in e.g. Steel Nomad up to 2810 mhz fast timing, but at these speeds when the VRAM heats up the GB/s checked and written drops over time (without showing any different errors in ~5 minutes). I imagine in those scenarios that it'd crash after a while. with better temps (like from lowering pl) I can run 2750 mhz fast timing without this degradation
regardless of any of this I am erroring even on default VRAM settings and it's concerning… I also don't think it used to behave like this when I first got it as I was running memtest vulkan then and found 2750 mhz fast timing to not error at all. could it be a driver issue (25.3.2), or should I look into RMA? or does it not matter at all? I haven't found the erroring to affect anything, but I am worried about it getting worse
2 Comments
The errors will affect gaming the longer you play a game and could be the reason driver crashes 2 hours into a game, as a example.
Also when your doing the memtest, do you have adrenalin open to view metrics, or the overlay with graph. If not do so and pay attention to total board power and clock speed. Does it spike down and rebound very quickly when the memtest speed dips?
Typically you don’t want the speed to differ at all. [like this ](https://imgur.com/a/Ltiyiuv)
If your getting errors at fully stock everything I’d rma as its not normal and memory should not error out stock.
What stress test is this? Is it publicly available?