Memtest86 HAMMER TEST and G.Skill Trident X DDR3

Hi everyone!

I received all of the parts for my build and after assembly I wanted to test the RAM before overclocking the Devils Canyon CPU I got.

Every single stick of my G.Skill Trident X RAM 2133 9-11-11-31 fails the Hammer Test. I've read up on the test itself, and I have also been told by G.Skill "The Hammer test is only for DDR4 RAM, if it passes the other tests the RAM is fine". From what I've read this is not necessarily true; but also to my knowledge now I expect that any replacement Trident X RAM will have the same problem.

I know "how big of a problem" it is, but what I'm wondering is, has anyone else encountered this with G.Skill? Does Corsair Vengeance Pro or Dominator have the same issues? I'm trying to make an educated decision here on whether or not to switch brands. Ideally, I would have no memory issues.

Thanks!

Try using OCCT instead.

What does that have to do with memory?

I'm failing to see anything on their website remotely mentioning RAM.

Covering full memory is actually memory test so you can use OCCT, Prime95 or any other test which is using large amount of RAM. Personally I see that AIDA64 memory test is finding errors pretty quick and is testing memory controller too.

Memtest86+ in the latest version is working with DDR4 but for some reason multithreaded tests are not working on both my X99 rigs.
On DDR3 all should work fine and if it's crashing but all other tests are working fine then I think it's more software issue than memory. Memtest86+ has some stability issues on some motherboards that's why they added safe mode.

If you are testing stability then in all cases it's best to check it with couple of programs.

Woomack, thank you for the info! I'm currently running OCCT 4.4.1 and comparing the results to what I found at http://www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k/2.

I was skeptical when I got the RAM because one of the heat spreaders actually was peeling off just gently taking it out of the packaging. I had already ordered a replacement so I figure I will test it when it comes and if it doesn't fair any worse than the original set I will stick with it.

I'm testing the 4790k @ 4.7Ghz @ 1.3v with the Swiftech H220x and seem to be fairing far better than they are temperature wise. Currently my max temps are 81c on OCCT and 91c on Intel Burn Test. Curious if anyone has any feedback on OC'ing that processor.

I should probably re-post in a different section.

iterator said:

Hi everyone!
Every single stick of my G.Skill Trident X RAM 2133 9-11-11-31 fails the Hammer Test. I've read up on the test itself, and I have also been told by G.Skill "The Hammer test is only for DDR4 RAM, if it passes the other tests the RAM is fine". From what I've read this is not necessarily true; but also to my knowledge now I expect that any replacement Trident X RAM will have the same problem.

I know "how big of a problem" it is, but what I'm wondering is, has anyone else encountered this with G.Skill? Does Corsair Vengeance Pro or Dominator have the same issues? I'm trying to make an educated decision here on whether or not to switch brands. Ideally, I would have no memory issues.

Thanks!

Click to expand...

I have G.Skill Ares RAM and it also fails -- only a few bits, but still fails -- the hammer test. According to Wikipedia (https://en.wikipedia.org/wiki/Row_hammer), and despite what the G.Skill rep told you, this problem does affect DDR3. According to the Wikipedia article, "DDR4 memory employs hardware support for the so-called target row refresh (TRR), which prevents the row hammer effect from occurring without introducing any performance or power consumption impacts."

According to the passmark Memtest86 site, this is a design flaw in the memory and " it is likely that the RAM modules with the same model would still fail the Hammer test."

A thread on the passmark Memtest86 forum says that some Corsair and Kingston memory also have this problem, but that a particular model of memory, "HyperX Savage", from Kingston does not. (http://www.passmark.com/forum/showthread.php?5077-How-to-relate-to-errors-in-Hammer-Test-13) The write hypothesizes that HyperX Savage is a recently released model, so the flaw may have been fixed.

I'm trying to find other sources that list RAM modules not susceptible to the problem, but that's the only one I can find so far.

For what is worth my Kingston HyperX Fury 8+8 GB fails the row hammer test with ~400 errors in 4 passes

2 modules like:

8GB DDR3 PC3-14900
Kingston / KHX1866C10D3/8G
10-11-10-30 / 933 MHz / 1.5V

432 errors in 4 passes

Lowest Error Address 0x3547BAC (53MB)
Highest Error Address 0x43F4DFB44 (17396MB)
Bits in Error Mask 00000000FFFF5FC7
Bits in Error 27
Max Contiguous Errors 1

on a Asus Z97-A Intel i5.

Contacted Kingston and got a useless scripted reply.
Other than the bogus reply from G.Skill mentioned above, no manufacturer has said anything about it.

Now my dilemma is:

do I start a (possibly) long RMA process just to get an other module that will fail the same?

or

I just live with it? had I tested it 5 months ago I would be a very happy customer giving this RAM maximum stars everywhere.

mariolino said:

For what is worth my Kingston HyperX Fury 8+8 GB fails the row hammer test with ~400 errors in 4 passes

2 modules like:

8GB DDR3 PC3-14900
Kingston / KHX1866C10D3/8G
10-11-10-30 / 933 MHz / 1.5V

432 errors in 4 passes

Click to expand...

I got about the same with that same memory, and the error count didn't change much when I dropped to 1.35V. I've read that increasing the refresh rate can eliminate the errors. OTOH I tried 4 samples of my 4GB Samsung 1600 MHz memory, and they showed 0 hammer test errors. The chips on them are Samsung K4B2G0846D -HYK0

We all need to figure out where the setting is in the BIOS to raise the DRAM refresh interval.

It looks like virtually everybody is required to manually change the DRAM refresh interval. BIOS manufacturers apparently didn't release an update with this already!