Different Results from different versions of MemTest86

[gregeeh]

Hi all,

Ryzen 1700
Asus B350 Plus M/B
16GB PC4-25600 DDR4 XMP 3200MHZ 14-14-14-31
CPU overclocked to 3700MHz
RAM clocked at 3200MHz

Using version 4.3.7 I always get errors on test 6.
Using version 8.1 I never get errors on any test.

Why??

TIA

Greg

[David (PassMark)]

Are you using multithreading? The single threaded test isn't nearly as effective and using all CPU cores.

What was the memory address of the error?

[gregeeh]

Are you using multithreading? The single threaded test isn't nearly as effective and using all CPU cores.

What was the memory address of the error?

Sorry, but do not understand the Multithreading question. I'm using the default settings of both versions. ie: Just let it boot up and do its thing.

Memory address for Test 6 in version 4.3.7 is 001A4F6AAB0 (6735.4MB) - 001AEF6AAD0 (6863.3MB).

Greg

[David (PassMark)]

When running V8, there is text on the screen, "CPUs Found", "CPUs running", "CPUs active".

How many are running for most of the tests? It should be 8 for that CPU. But some motherboards have UEFI BIOS bugs.

[gregeeh]

CPU's Found: 16
CPU's Started: 16
CPU's Active: 1

EDIT: I note my M/B, ASUS PRIME B350-PLUS, is on the list with bios bugs

[David (PassMark)]

You could trying forcing multi-threading from the config menu, ASUS might have fixed the bug in newer BIOS and not told us.

[gregeeh]

I forced multi-threading in V8 and now I'm getting 43 errors in Test 8. No multi-threading in V8, no errors.

V4.3.7 - Only errors in Test 6.

So are the errors real errors or errors caused by the bios? And which version do I believe? Do I leave multi-threading ON or OFF?

[David (PassMark)]

What firmware do you have installed? Maybe we can un-blacklist this board and have multi-threading on by default for this board.
If you aren't seeing crashes or lockups, leave multi-threading on.

From our point of view this is a good result. As we need V8 to be at least as effective as V4 in finding errors.

I think the errors would be real. But maybe caused by your overclocking. You would need to test to be sure.

[gregeeh]

What firmware do you have installed? .

4207 - The latest.

If you think the errors are real why don't I see the same errors on the different versions?

[David (PassMark)]

Some possibilities,

1. The system instabilities aren't due to bad RAM at all, but something else.
2. The errors are random or effected by some external factor. The RAM is running on the edge of failure (due to pushing the timings to the limit & overlocking) but only becomes unstable in combination with some external factors. For example the slight increase in temperatures or increase in EMI from having the entire system under load might push the RAM over the edge into a failure case. i.e. you have no safety margin. This is where BurnInTest can help by fully loading the system while testing RAM.
3. You got unlucky, and the weak RAM cells happen to be at addresses that are reserved by the system in UEFI mode, but not BIOS mode. UEFI BIOS reserves some RAM for the firmware and UEFI data. This RAM can not be tested (generally it is only maybe 2% of all the RAM, but sometimes you get unlucky). A example of this memory map can be seen here.
4. The differences in results are due to change we made to the tests. For example V8 runs as a 64bit application. V4 ran as a 32bit application and used PAE sliding address windows. We changed a number of the algorithms to be faster, but there is a class of error (e.g. bit fade) that only happen in slower tests. We added entirely new tests using new CPU instructions in V8 (compared to V4)