HCI Memory test advice

DarkHorizon472 · 27 Dec 2017 at 11:20

Testing memory with hci memory test, it looks like you need to run one per CPU thread.

For my Ryzen 1700 8 core / 16 thread I am running 16 mem tests of 1750mb at the same time. It has passed 150% and is still running with no errors so far, is that the right number to run & the right mb setting ?

Also what is the lowest % to give a useful test ?

Thanks.

kitfit1 · 27 Dec 2017 at 13:22

TBH i think that HCI test is very clumsy and unintuitive to use. Memtest64 is a much better solution for testing in Windows, you just choose maximum mem (which it is at by default anyway) and press Begin Test. The prog then pages out everything to the pagefile and uses all the memory that's left. You just leave it running.

DarkHorizon472 · 27 Dec 2017 at 14:05

Thanks, I will try that one as well.

Got to over 300% with no errors so stopped hci.

8 Pack · 28 Dec 2017 at 18:04

Around 20 mins as an initial guide is suggested by hci. I run for around 1000% for customers. After this maybe just use your pc to see if it's truly stable.

DarkHorizon472 · 28 Dec 2017 at 18:48

Thanks for the advice, after upping vcore a little it seems to be completely stable again now.

opethdisciple · 28 Dec 2017 at 21:03

I used HCI and it seems to have done the job for me.

Yes, 1 instance per thread. I test about 90% of available ram.

400%.

Doogles · 28 Dec 2017 at 22:54

kitfit1 said:
TBH i think that HCI test is very clumsy and unintuitive to use. Memtest64 is a much better solution for testing in Windows, you just choose maximum mem (which it is at by default anyway) and press Begin Test. The prog then pages out everything to the pagefile and uses all the memory that's left. You just leave it running.

Run both, I've heard cases of HCI not showing but Memtest86 failing, I guess it's like how I ran P64 for 4 hours and it was "stable" yet IBT crashed after 10 minutes.

kitfit1 · 29 Dec 2017 at 09:19

Doogles said:
Run both, I've heard cases of HCI not showing but Memtest86 failing, I guess it's like how I ran P64 for 4 hours and it was "stable" yet IBT crashed after 10 minutes.

Memtest86 is run from a usb stick booted from the bios, Memtest64 is run from within Windows. As it happens, none of them is foolproof and never have been. The problem with running any from within Windows is knowing exactly how much ram is available to test and how much of the OS can be moved to the Pagefile.
That's why i think Memtest86 run from a bootable USB stick is more likely to give an accurate result. It uses much more ram than anything run from within Windows and does the job within 13 hours with 32GB of ram.
I suggested Memtest64 though because it takes all the guesswork out of how much ram is free and how much ram is dumped onto the pagefile.

Doogles · 29 Dec 2017 at 15:09

kitfit1 said:
Memtest86 is run from a usb stick booted from the bios, Memtest64 is run from within Windows. As it happens, none of them is foolproof and never have been. The problem with running any from within Windows is knowing exactly how much ram is available to test and how much of the OS can be moved to the Pagefile.
That's why i think Memtest86 run from a bootable USB stick is more likely to give an accurate result. It uses much more ram than anything run from within Windows and does the job within 13 hours with 32GB of ram.
I suggested Memtest64 though because it takes all the guesswork out of how much ram is free and how much ram is dumped onto the pagefile.

That is why I run Memtest86 and HCI, it tests it in different environments, sure nothing is fall proof but it helps.

Rroff · 29 Dec 2017 at 15:24

If you go through all the patterns Memtest86 should be accurate in detecting faults - it does complete coverage of the RAM minus a tiny amount used by the system and the only variable really is if its run long enough to detect errors that only occur once the system is warmed up.

DarkHorizon472 · 29 Dec 2017 at 17:56

Thanks for the advice.

Silent_Scone · 29 Dec 2017 at 18:57

kitfit1 said:
TBH i think that HCI test is very clumsy and unintuitive to use. Memtest64 is a much better solution for testing in Windows, you just choose maximum mem (which it is at by default anyway) and press Begin Test. The prog then pages out everything to the pagefile and uses all the memory that's left. You just leave it running.

Someone made a tool on OCN that helps with assigning threads/amounts. Memtest Launcher Pro. Memtest64 is not anywhere near as stringent as HCI or Google Stress App test ran through BASH terminal.

http://www.overclock.net/t/1628751/...memory-stability-thread/1000_40#post_26419580

There is a new tool that is community driven that is proving to be worth trying out, too. I've found it to be far quicker at catching memory errors than both of the above thus far.

https://www.karhusoftware.com/ramtest/