What exactly is Test 6 testing?

[Roxxors]

So I got this CL36 8000mhz memory running at 8407. No problem. 20 Memtest passes and no errors. But when I bump it up to 8500mhz...
This is from multiple passes...
Test 1-5, no errors...
Test 6, 166 errors...
Test 7-13 (skipping 10), no errors...

I know it's says: Block Move, 64 Byte Blocks, but what error are you testing for? Boundaries? Edge Flipping?

It's only Test 6 that is failing and I need to know what exactly is it testing for so I can fix it...

[Roxxors]

Soooo... Since I like danger, I went back to 8500mhz and ran MT again, same thing Test 6, 230 errors, no errors on any other tests. Soooo, I just did it, I booted Windows and ran all my other tests and benches... It all works fine... No problems... I am now sub 8.5ns...

I would still like to know how to make those MT errors on Test 6 go away, but is has no effect on the system operation in the real world...

[David (PassMark)]

I would still like to know how to make those MT errors on Test 6 go away

Running your RAM at it's specified speed would be the obvious answer

but is has no effect on the system operation in the real world

RAM errors cause random effects. How sure are you that a few numbers weren't changed in your Excel spreadsheet. Or cat.jpg file had a few pixels of the wrong colour? Or that flight control system didn't just direct an aircraft slightly off course.

Test descriptions can be found here.
https://www.memtest86.com/tech_indiv...est-descr.html

Test 6 tends to be one of the more effective tests at finding errors. It moves a lot of data around very quickly.

[Roxxors]

"Running your RAM at it's specified speed would be the obvious answer "

Yea, but who does stuff like that?

"RAM errors cause random effects. How sure are you that a few numbers weren't changed in your Excel spreadsheet. Or cat.jpg file had a few pixels of the wrong colour? Or that flight control system didn't just direct an aircraft slightly off course."

Absolutely sure. I test it all the time. I know you are the PR man, MemTest must have a reason to sell. I get that, but I built and run this box every day and play with it all the time and I know what is stable and not stable... Memtest is ok for what it does, but it's hardly bulletproof and I have other tools to use just to doublecheck things...

[David (PassMark)]

I know you are the PR man

Engineer actually.
We don't have any PR people.

[Roxxors]

Engineer actually.
We don't have any PR people.

My apologies. Where I come from they hire people from marketing just to promote the product on the websites...If someone says something not good about the product, the post is either "lost" or played down...
Just used to it...

[Roxxors]

I still need to know what test 6 is testing for. I read the description, nice, but useless. I'm not asking what it does. I am trying to know why it's even there... Is it looking for edge bleeding?

[David (PassMark)]

As best we know the original version was written in 2000 (25 years ago) by Robert Redelmeier.
We assume it was added to an old version of MemTest86 around the same time (it was before our time).
It was one of a suite of very small programs to "load x86 CPUs as heavily as possible for the purposes of system testing". We assume that this is why it got added.

Chris Brady re-wrote it using the same concept of moving memory blocks. Then we wrote our own version (for 64bit, UEFI support, ARM support, threading & debug logging).

It is also one of the tests that is written in hand coded assembler code.

So as best we know, it got added as it seemed like a good idea at the time. The original code is gone, but the concept was good and has lived on, as it has proved to be an very effective memory test.