what's the difference between memtest86+ V4 and memtest86 v6 when test DDR4？

[play008]

hi David，
Why there is no problem when DDR4 tested on memtest86+ V4,however many errors appear when tested on memtest86 V6Why there is no problem when DDR4 tested on memtest86+ V4,however many errors appear when tested on memtest86 V6，These errors occur when the hammer test.The same memory , platform,and the boot options of BIOS.

I have tow questions:
1.what's the difference between the two when test DDR4?
2.can you find which is the error IC from this picture?

[David (PassMark)]

See this page regarding the row hammer errors.

MemTest86+ V4 (and V5) were never tested with DDR4. It didn't exist when MemTest86+ was created.

MemTest86 V6 adds native 64bit support, DDR4 support, UEFI support, a new row hammer memory test and many other changes. So testing should be more effective in V6.

No, you can't determine the individual IC at fault. With row hammer errors being caused by a design flaw, it might be all the chips that are effected. I was lead to believe DDR4 was largely immune to row hammer errors. But this seems not to be totally true. This is the second report of row hammer errors in DDR4.

[play008]

hi David
I got it about the hammer error.
If it is not error hammer, can I determine the individual IC at fault?
If I would like to buy a paid version, can it be more accurate than the free version? Is there a error code description for the paid version?
If I buy a paid version, could you add a logo? For example, “passMark memtest86 V6-myname”
thanks a lot

[David (PassMark)]

can I determine the individual IC at fault?

Generally no.
The mapping algorithm between memory addresses and ICs isn't made public by Intel.

The differences between the paid and free editions can be found here,
http://www.memtest86.com/features.htm

Software customisation is possible, but it is a relatively expensive exercise compared to buying the generic off the shelf software.

[FuturePlus]

Hello Play008,

Can you look in your BIOS and tell us what the refresh rate is set to?

Hello David,

Can Passmark read the Refresh setting and print that out with the test? This is an important parameter with regards to Row Hammer failures.

[keith]

Reading the current refresh rate that is set on the system is chipset-specific, which means that we would need to implement code for every possible chipset. ie. There is no standard way of reading the refresh settings across all chipsets. At the moment, we do not have the resources for the amount of work required to implement and test the code to read the refresh rate for all possible chipsets.

[mseaborn]

Reading the current refresh rate that is set on the system is chipset-specific, which means that we would need to implement code for every possible chipset. ie. There is no standard way of reading the refresh settings across all chipsets. At the moment, we do not have the resources for the amount of work required to implement and test the code to read the refresh rate for all possible chipsets.

Note that it's possible to measure the refresh rate by timing memory accesses.

See my post on rowhammer-discuss:
https://groups.google.com/d/msg/rowh...Q/qFcDd4SEHUcJ

And see the code here:
https://github.com/google/rowhammer-...refresh_timing

This doesn't require any CPU/chipset-specific knowledge. It does require some signal processing to work out which memory access delays are due to refreshes, but since memtest has the machine to itself it shouldn't see too much noise in those delays.

[David (PassMark)]

it's possible to measure the refresh rate by timing memory accesses

Thanks for the suggestion.
We might have a look at it for a future release. As your code is at the moment it would need a significant rewrite. The code is written for Linux with file I/O, lots of floating point and various O/S API calls. None of this works in UEFI land, where there isn't a real O/S available.

[play008]

Hi David
Could you please give me your E-mail or another contact information online, in order to enquire something about the paid editions. Because sometimes I cannot access this forum.

[David (PassMark)]

All our contact details are on the Contact Us page.