Microsoft is introducing a new feature in Windows to better deal with blue screens of death. In the release notes for Windows 11 Insider Preview Build 26220.6982 (Dev Channel), the company detailed that after a user experiences a blue screen, Windows will automatically perform a memory scan.
We’re introducing a new feature that helps improve system reliability. If your PC experiences a bugcheck (unexpected restart), you may see a notification when signing in suggesting a quick memory scan. If you choose to run it, the system will schedule a Windows Memory Diagnostic scan to run during your next reboot (taking 5 minutes or less on average) and then continue to Windows. If a memory issue is found and mitigated, you will see a notification post-reboot.
↫ Amanda Langowski at the Windows Blogs
In its current iteration, this memory scan will trigger after every single error code to collect as much data as possible, but Microsoft states it will refine and narrow the number of error codes in the future. In addition, this feature will not be available on Arm64 and systems with Administrator Protection and/or BitLocker without Secure Boot.
Let’s hope this feature won’t be a nuisance, but an actually useful feature that helps people uncover memory problems that otherwise remain undiagnosed.
I wish more consumer products supported ECC ram to detect and correct memory errors automatically. I’d prefer this hardware feature becoming standard over the recently discussed memory tagging feature they’re working on.
https://www.osnews.com/story/143601/intel-amd-to-bring-memory-tagging-to-x86-at-some-point/
DDR5 has on-die-ECC by default.
DDR5 has on-die ECC, which corrects single-bit errors within the chip to improve manufacturing yields and reliability, but it’s not the same as traditional ECC, which is a separate, more robust system.
Memory tagging has nothing to do with what ECC is doing.
0brad0,
Yes I know, I’m saying that “ECC for everyone” would interest me more than the tagging they are working on. It has nothing to do with the two being related other than the CPU needing to support them.
That’s still not how anything works.
0brad0,
What are you talking about? I never said the features are related or exclusive, if that was your criticism, then it doesn’t really apply since I am just ranking the importance to me: I’d rather see more hardware support ECC than tagging, it doesn’t mean they are related.
Alfman,
We need to take this to Intel.
Even when they have ECC support on “prosumer” CPUs, they would disable that on “consumer” versions. They would specifically create distinct markets. (Not too different than nvidia and their shenanigans. Or AMD as well. And AMD does this with ThreadRipper)
There is no reason for a high end CPU like i9 (or previous X99 chipset ones) not to have ECC.
GPUs do this as well. The “Workstation” counterparts not only have better cooling, they also offer ECC. Which is extremely important if you want to render professional video or similar.
sukru,
I agree. I see errors once in a while, like a game, browser crash, what have you, and with things like that we tend to assume it’s software. But without hardware telemetry data it’s really hard to determine if it’s a hardware or software fault particularly when the issue is intermittent. ECC can quickly identify faults and IMHO it should be standard. It may be technically “more important” on enterprise systems, but that’s not to say this is unimportant elsewhere.
After dealing with Windows for 30+ years and having more than enough blue screens memory was literally never the issue, not saying it can’t ever be the case but this seems like a waste of a users time most of the time.
0brad0,
Before last year I was in the same boat as you. However there’s a first time for everything and I definitively identified bad ram in my machine. The OS could boot and run for hours, there didn’t seem to be a rhyme or reason. Memtest did it’s job, though it had to run a considerable amount of time – something that would be unreasonable to do regularly IMHO.
Is there a similar utility for GPU ram? Searching just now I found these, but haven’t tried them yet.
This says it uses memtest patterns over opencl and cuda, but it’s old updates doesn’t give me confidence that it will work with modern hardware.
https://sourceforge.net/projects/cudagpumemtest/
This project performs a test using vulkan.
https://github.com/GpuZelenograd/memtest_vulkan