Memtest errors on new memory

Putting my new system through its paces and thought I had a stable OC but when I changed from the small fft prime 95 test to blend I started getting errors. Eventually took cpu back down to stock and am still getting errors.

Ran memtest and it found a whole bunch of errors (60+ and still going on first pass).

Obvious answer is that the memory is bad. My only doubt is that the kit I bought was a 4x8gb kit that is technically a dual channel kit. I'm sure I'm not the first person to do this, has anyone else experienced this out should I just rma the memory and hope that is the issue.

For what it's worth this memory obviously isn't on the approved list from asus, but I believe some slower speed memory utilizing the same chips is so I don't think that is the issue...

tachi1247, I'm no expert on the X99 platform so you may get better answers when someone that has hands on experience with them chimes in. I'm pretty sure for that setup you need to use the 125 mhz strap to run the memory at those speeds on your board. You may also need to add a bit of Dram voltage as well as VCCSA voltage to help stabilize the Cpu IMC. You can try lowering the Dram speed to something like 2600 and see if you're still getting errors. If not I'd bet it's lack of voltage that is causing the issue, not that the sticks are bad.

I'm running them at the 1.35v recommended for the 3000mhz speed. I'm also running them underclocked as I have yet to be able to get it to run at 3000 reliably. Currently they're at 2800, but I'll drop the speeds again when the test finishes.

For what it is worth I came across a lot of people in here while researching the OC that say their cpu runs better on the 100 strap and they also had high speed memory so I would assume that it should work even though it does indeed default to the 125 strap in the xmp profile. I guess I'll try running at the 125 strap as well.

Post some screen shots of Cpu-z, Cpu, Spd and memory tabs. Maybe something will stick out that could be the culprit of the instability.

will post screen shots when I'm done with the mem testing.

stopped the memory test at 2800 on the 100 strap as it was getting destroyed by the hammer test and was taking forever. had 200+ errors when I stopped it.

Switched it over to the 125 strap and tried again at 2666 and made it past the early tests where I was getting errors at 2800 with no issues. Stopped the test and changed the mem to run at it's advertised speed of 3000 on the 125 strap and started the test again. made it through the first 12 tests without errors on the first pass and it's doing the hammer test now and so far so good.

Sounds like it is as you suggested and the memory really doesn't like the 100 strap. Assuming it makes it through the test error free at 3000, what speed would you expect that I can run it at on the 100 strap without getting errors? 2400? 2666?

Edit: Made it entire through the first pass at 3000 on the 125 strap without any errors. Guess that would certainly indicate that the memory doesn't like to run on the 100 strap at high speeds. Once it finishes all of the passes I'll try it again at 2666 on the 100 strap.

Why not just leave it be with the 125 strap at 3000Mhz, if it's error free?

The cpu won't overclock well on the 125 strap. It wants at least 0.06 volts more for the same speed and maybe more. I gave up trying to get it stable on the 125 strap when I realized how much extra voltage it was requiring. 4.1 @ 1.23v with 2800 was fine on the 100 strap (p95 small fft ran over night but fails blend within hours). 4.125 @ 1.29 was still crashing the XTU benchmark within 10 seconds on the 125 strap so I didn't even get to p95.

Since the memory speed doesn't seem to affect system performance that much I'd rather sacrifice a little memory speed on the 100 strap to run a lower voltage 24/7 for my OC.

tachi1247 said:

The cpu won't overclock well on the 125 strap. It wants at least 0.06 volts more for the same speed and maybe more. I gave up trying to get it stable on the 125 strap when I realized how much extra voltage it was requiring. 4.1 @ 1.23v with 2800 was fine on the 100 strap (p95 small fft ran over night but fails blend within hours). 4.125 @ 1.29 was still crashing the XTU benchmark within 10 seconds on the 125 strap so I didn't even get to p95.

Since the memory speed doesn't seem to affect system performance that much I'd rather sacrifice a little memory speed on the 100 strap to run a lower voltage 24/7 for my OC.

Click to expand...

Understood and I do agree with you that you're better off getting the Cpu Oced, especially at lower voltage then trying to run the memory higher and the Cpu lower. I wish I had more answers for you about what you need to do to get the memory to run at it's rated speed with the 100 Mhz strap. Maybe tomorrow you'll get more answers from the members who have more experience with the X99 setup.

On ASUS board 100MHz bclk is working fine with memory @3000 so you can simply enable XMP but manually change CPU and memory ratios ( so x30 for memory ). When you reach high enough CPU clock then you can also check x32 memory ratio as your memory can probably make it.

Woomack said:

On ASUS board 100MHz bclk is working fine with memory @3000 so you can simply enable XMP but manually change CPU and memory ratios ( so x30 for memory ). When you reach high enough CPU clock then you can also check x32 memory ratio as your memory can probably make it.

Click to expand...

Maybe it's just the memory I bought, but it doesn't work at 2800 or above on the 100 strap. I can set it in the bios and it boots into Windows, but I get errors in p95 blend and when I ran memtest @ 2800 I had 70 something errors before getting to the hammer test where it went over 200 before I stopped it. It ran all 4 passes error free @ 3000 on the 125 strap.

I'm just changing the multiplier on the cpu and adjusting the ram speed under the xmp profile in the bios. Manually setting vcore and dram voltage. Everything else on auto. If there are any suggestions you can make to get it stable they would be appreciated.

OK so still playing with this. I set the cpu back at 4.1 @ 1.235v on the 100 strap after not being able to get it to run xtu benchmark at 4.2 even up to 1.26v with the memory at 2133. at 4.1 & 2666 it ran prime blend for 24 hours without any errors.

Set the memory back to 2800 and added 0.01 to the system agent offset. ran 12 instances of hci memtest (to use all my free ram) in windows and it kicked out 7 errors in one of the tests after a few minutes. I restarted the computer and bumped the system agent offset up to 0.02v. when I got back into windows only 24gb of ram were detected. went back and reset system agent to 0.015. no go. reset it back to 0.1 no go. still 24gb of ram. went back to 0. same issue. reset the memory back to jdec spec and voltage, booted up and it found all the memory. set it back to 2666 and running the hci memtest again. no errors so far.

So any advice on what the next step should be? From my research, the consensus is that I shouldn't have any issues running this ram at 2800 or 3000 (or more), but I must be doing something wrong or not have something set correctly. Or could I just have a dud chip with a weak IMC?

Any suggestions??

Ignore the hammer test man. It's flawed. Everything shows errors in it. I believe there is a thread here about it.

Mr.Scott said:

Ignore the hammer test man. It's flawed. Everything shows errors in it. I believe there is a thread here about it.

Click to expand...

Not concerned about hammer test. Was getting errors before that test at 2800 on the 100 strap. Ran fine at 3000 on the 125 strap without any errors .

My main issue is I can't get the ram to run faster than 2666 on the 1000 strap. It causes freezes and/or errors within minutes or hours.