Can't get past HCI Memtest (new build)?

[SimpleMan67]

I just put this new build together. It seems to be running great. However, I tried running HCI Memtest last night and made it to about 150% and got an error.

So I ran LinX and made it through 20 cycles no problem. I then ran Memtest86+ for over nine hours without an error. I then ran Prime95 Large FFT for over eight hours without an error. I then tried running HCI Memtest again and as soon as I started the 8th instance I got an error. I had each instance set to test 1500MB

My core temps haven't went over 66C.

The only thing I can come up with is my motherboard doesn't like having all four DIMM slots full.

I was going to pull the sticks out of A1 and B1 and see if I can get through HCI Memtest.

The only change I've made to the BIOS is I set Ai Overclock Tuner to X.M.P. so that my memory would run at 1600MHz. Everything else is at the default setting.

So what do you guys think? Is it bad RAM or could it be something else?

ASUS Maximus IV GENE-Z
i7-2600K (Noctua NH-U12P SE2)
EVGA SuperClocked GeForce GTX 580 1536MB
Corsair Vengeance LP 16GB (4x4GB) CML16GX3M4A1600C9
Intel SSD 320 Series 160GB
Corsair AX1200
Seagate Barracuda XT 2TB

Corsair 650D
Windows 7 64-bit

 

[fastamdman]

Don't trust HCI memtest as it runs in windows. Memtest86+ is a better way of running memtest.

If you are LinX stable with avx or sp1, aka 100+gflops, then don't worry about it. I would run LinX on max/all for AT LEAST an hour, if not more. As long as you are linx/ibt stable that is all that matters.

Don't worry about HCI memtest, it's a poor program choice for stability.

Edit: something else you can try is upping the vtt slightly, instead of 1.050 try 1.1 or 1.175, don't go more then 1.2 though.

Either way if you can do like 6 hours of LinX you are 100% stable *on all/max setting that is*

 

[SimpleMan67]

Don't trust HCI memtest as it runs in windows. Memtest86+ is a better way of running memtest.

If you are LinX stable with avx or sp1, aka 100+gflops, then don't worry about it. I would run LinX on max/all for AT LEAST an hour, if not more. As long as you are linx/ibt stable that is all that matters.

Don't worry about HCI memtest, it's a poor program choice for stability.

Edit: something else you can try is upping the vtt slightly, instead of 1.050 try 1.1 or 1.175, don't go more then 1.2 though.

Either way if you can do like 6 hours of LinX you are 100% stable *on all/max setting that is*

Thanks for the reply! I'll give your suggestions a try.

After I posted this I pulled the two sticks out of A1 and B1 and fired up HCI Memtest again (800MB per instance) and so far I'm at 132% with no errors.

 

[fastamdman]

Is your 2600k overclocked or stock?

 

[AdamK47]

HCI Memtest is single threaded, so turbo boost may play into the instability.

 

[Idontcare]

Don't trust HCI memtest as it runs in windows. Memtest86+ is a better way of running memtest.

If you are LinX stable with avx or sp1, aka 100+gflops, then don't worry about it. I would run LinX on max/all for AT LEAST an hour, if not more. As long as you are linx/ibt stable that is all that matters.

Don't worry about HCI memtest, it's a poor program choice for stability.

Edit: something else you can try is upping the vtt slightly, instead of 1.050 try 1.1 or 1.175, don't go more then 1.2 though.

Either way if you can do like 6 hours of LinX you are 100% stable *on all/max setting that is*

Terrible advice, certifiably.

My GSkill sticks passed memtest+ (24hrs) but would not pass HCI memtest.

Gskill confirmed my sticks were bad. HCI was right, memtest+ failed to detect this issue.

Memtest might be easier, and more to your liking, but dismissing the results of HCI memtest is simply not prudent. Telling others to dismiss them is simply terrible advice.

 

[SimpleMan67]

With just two 4GB sticks in A2 and B2, I was able to run HCI Memtest past 1000% coverage without any errors.

So now what? Should I test the other two sticks just to make sure they're both good?

If those sticks are good, then what does it mean if I can't get through HCI Memtest with all four DIMMs full?

 

[thewhat]

Test the other two the same way.

The problem could only be with all four slots full. It's normal, more slots = more work for the memory controller etc.

I also got errors with HCI Memtest while everything else showed no error. Slightly increasing RAM voltage helped.
Besides increasing voltage, you could also try loosening the timings.

 

[fastamdman]

The biggest question is whether or not you are overclocked. If so, HCI memtest is NOT a valid test to run where as memtest is. No matter how you look at it "idontcare" if he can pass multiple loops of LinX or IBT with the avx pack, hes going to be stable. Whether or not he needs a slight bump in voltage is all that comes into play. The sticks themselves are just fine, he either needs slightly loser timings or a simple voltage bump.

Idontcare - if his sticks were "bad" as yours were supposedly bad, he would be having errors in the real memtest aka memtest86+ and or in linx/ibt/prime95.

If you are really hung up on this application though and you want to pass it for peace of mind, simply bump up the vtt voltage. Either way my advice is 100% sound, if you pass LinX and the real memtest, you are stable plain and simple.

 

[AdamK47]

The way I see it is if it passes LinX or Prime95, but fails HCI Memtest, then it's still not stable. HCI Memtest is simply detectings errors more readily than the other two apps could not. I know I wouldn't trust an overclock if it failed. I would be changing timings and/or voltages.

 

[Plimogz]

Don't trust HCI memtest as it runs in windows.

So if ever my games crash, by your logic, I should just ignore that event because they run in windows? Sure is annoying, though...

Programs which readily detect errors and abnormal behaviour which others let go by unnoticed are more useful, not less, and certainly shouldn't be dismissed.

Edit: something else you can try is upping the vtt slightly, instead of 1.050 try 1.1 or 1.175, don't go more then 1.2 though.

this sounds about right, from what I've read.

 

[Kenmitch]

With just two 4GB sticks in A2 and B2, I was able to run HCI Memtest past 1000% coverage without any errors.

So now what? Should I test the other two sticks just to make sure they're both good?

If those sticks are good, then what does it mean if I can't get through HCI Memtest with all four DIMMs full?

Test the other sticks the same way. If they also pass then put the other set back into the rig. If it was me I think I'd try bumping up the memory voltage a little bit if it still fails. Maybe try 1.53-1.55v's and see if it helps.