Jump to content
Corsair Community

RAM failing memtest is not a qualifier for RMA with Corsair?


bondisdead

Recommended Posts

The following is the info I supplied with a ticket with corsair today:

 

I have the following RAM:

P/N: CMK16GX4M2B3000C15W

 

I am using it in a MSI X399 Gaming Pro Carbon AC motherboard, with a AMD Ryzen Threadripper 1950x CPU. I have two kits, and populated both kits in the slots specified by the motherboard manual (A2/B2/C2/D2). I initially used the A-XMP profile for 3000MHz. When running Passmark memtest86 v7.4, I would notice about four errors on Test #10. I kept dropping frequency, thinking a Ryzen compatibility issue, but even at default 2133MHz, I would get a handful of errors on Test #10. I then decide to test the two kits separately, installing in slots B2/D2, and running at default 2133MHz. One kit fails on Test #10 as above, while the other passes. So it looks like this one kit is bad. I would like a replacement, so I can run in quad-channel. Would like to setup advanced RMA.

 

And this is the response from Corsair tech support:

 

Synthetic benchmark failure wouldn't be cause for an exchange. Art

 

Am I missing something? Failing memtest is not "bad enough" to get an RMA? What qualifies for an RMA, the RAM catching fire?!

Link to comment
Share on other sites

Hi,

 

Would you like me to clarify more of my response? I did say much more than what was quoted on here.

 

Please do. This was just the initial response from Corsair (you?).

 

You'd get approved if you had issues in everyday use of the RAM and not in synthetic benchmarks solely. Prime95 is a synthetic test, not normal use of the kit.

 

If you have errors in a memtest but no issues in regular use then there's no problem to fix.

 

Essentially, it sounds like Corsair will not honor warranty on RAM that fails "synthetic" benchmarks like Prime95 and memtest86. I would like to see this posted on the warranty page or some FAQ, as it will certainly affect me as a future buyer, and perhaps others.

 

Whenever I build a system, I run a standard suite of tests to insure stability. This includes memtest86 (note the use of *test* not *benchmark*), Prime95, IntelBurnTest, Intel Processor Diagnostic Test (assuming Intel system), Valley, Heaven, Cinebench, etc. Prime95 is puzzling. This is running a bunch of FFTs, which is pure math. If this fails with one set of RAM at stock / non-overclocked speeds, but not another, how is this not a memory error? I recently built several systems for a client with 64GB of RAM each. He was running complex math solvers, which max'd out the use of RAM. I don't want him to get bad results for his applications! A memory failure wouldn't crash the system, but provide inaccurate results, which he would probably never detect.

 

Using your explanation of these "synthetic" tests, if I am building a system for a customer using Corsair RAM, and encounter errors in memtest86, then I would not give that system to a customer. I am not going to tell him/her: "Oh, it fails some synthetic tests, but those are not real...don't worry!". Then when I get the system back in a few weeks/months, I will have to spend the time debugging, replacing RAM and re-testing. Meanwhile, I will have lost credibility with the customer, thus potentially losing future sales and referrals.

 

So yes, please explain the justification for refusing RMA replacement, and post this for everyone to see on a public place. Alternately, please show me where it is already posted.

Link to comment
Share on other sites

  • Corsair Employee

There's just no issue with the kit if all you're having is a synthetic benchmark problem, I'd like this to be succinct so it's clear.

 

If you had any other performance issues in every day use then that'd definitely be a reason to get an exchange.

 

It's not posted anywhere, as this is on a case by case basis, I don't believe anyone that has gotten this response before has disagreed with the assessment to be honest.

 

-Art

Link to comment
Share on other sites

You just said the magic set of words. Case by case basis. I would like to see other users here on the forums give their input as well, as I just find that failing memtest and Prime95 not a benchmark, but a stress test. If your RAM cannot pass stress test, it is not very robust. I have subsequently moved the same RAM to a totally different platorm, an ASRock X399 Taichi / AMD Ryzen 7 1700X, set the speed to 2933MHz via A-XMP. I ran memtest86, and sure enough Test #10 fails. Set the speed to 2133MHz and Test #10 fails. Put in an identical kit (CMK16GX4M2B3000C15W), set speed to 2933MHz via A-XMP and it passes.

 

With that analysis, you are saying, very succintly, there is *no problem* with this RAM? You would be very comfortably putting this RAM in the guidance system of the next airplane you fly in? Because there is nothing wrong with the RAM.

 

Corsair: Hello, Corsair Tech Support here, how can I help you?

Dell: Hi, this is Joe from Dell computer test department. We are encountering some failures when testing the RAM in our systems.

Corsair: So are you using a memory test program?

Dell: Yes, of course. What else would we use to test the RAM in our systems?

Corsair: Sorry, but memtest86 is a synthetic test. We don't support the results of a synthetic test. There is nothing wrong with the RAM.

Dell: I see...

 

Case by case basis. I somehow don't see Dell or whomever being brushed off when providing the results of the industry standard memtest86.

Link to comment
Share on other sites

  • Corsair Employee

Failing a test for any non-ECC kit would be inconsequential of other issues, you're good as I've mentioned before.

 

I would not recommend, ever, consumer grade memory for use on air traffic control systems, or anything else related to safety; that requires commercial grade set ups.

 

I see the made-up Dell interaction, but I'm failing to see the issue. You're not having an issue with the memory besides the failing of the test/synthetic benchmark. Once again, no problem with that as long as there are no other issues.

 

-Art

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...