Question MemTest86 errors

Marky000

Distinguished
Sep 23, 2013
39
3
18,545
Hi people.

Have a system that freezes occasionally running Windows 10. Can run for hours or even a couple of days until it freezes completely, requiring hard reset.

I tested the hard drive using Western Digital DLdiag utility and it can back no errors. Will be replacing it with SSD anyway.

However, I have run several MemTest86 tests which reported some errors. Images attached.

View: https://i.imgur.com/FsBqBqk.jpg

View: https://i.imgur.com/rXSCFAz.jpg


When I test the sticks on their own, no issues. When I swap the 2 sticks around (2 dimm sockets), same error and same failing. address. Not sure why it is exact same failing address after swapping sticks.

When both sticks are installed, BIOS reports the following primary timings @1333MHz. Image attached:

View: https://i.imgur.com/5BXbHWV.jpg


My theory is that the high performance Corsair stick (CAS9 @1333) is forcing the Crucial stick (CAS11 @1600) to run at timings which may be too quick. Only theory I can come up with really.

I would have thought that the lesser stick would force the timings, however since the lesser stick is actually rated for the faster speed of 1600MHz, it may think it is capable of running at CAS9?

I have paired different speed RAM for many years but these sticks have a very big CAS difference so this to me seems to me the most likely expalantion.

I have been trying to figure this out for days so any comments or suggestions would be appreciated.

Short of removing one of the modules, could I maybe manually adjust the timings to say CAS 10 etc?

Thanks
 
Last edited:

Reaper_7799

Distinguished
Do like samir says...set loose timings for both, try again and go from there....ram can either work perfectly together with diff timings/speeds as it is supposed to downclock to the lowest denominator or if on auto a setting both should pass but sometimes it just doesn't dplay well at all, depending on board....The actual rated speeds for each memory can also cause that, not only timings...I'd bring it down manually to the lowest ones speed and then up the timings or you can try and run higher speed with the other lower timings....but neither might work, you may have to sacrifice it all and bring it down to low speeds and low timings. You'll just have to trial and error
 
Last edited:
  • Like
Reactions: SamirD
You do not have a matched ram kit.
Ram is sold in kits for a reason.
A motherboard must manage all the ram using the same specs of voltage, cas and speed.
The internal workings are designed for the capacity of the kit.
Ram from the same vendor and part number can be made up of differing manufacturing components over time.
Some motherboards, can be very sensitive to this.
This is more difficult when more sticks are involved.
You can sometimes compensate for errors by increasing the ram voltage in the motherboard bios.
 
You do not have a matched ram kit.
Ram is sold in kits for a reason.
A motherboard must manage all the ram using the same specs of voltage, cas and speed.
The internal workings are designed for the capacity of the kit.
Ram from the same vendor and part number can be made up of differing manufacturing components over time.
Some motherboards, can be very sensitive to this.
This is more difficult when more sticks are involved.
You can sometimes compensate for errors by increasing the ram voltage in the motherboard bios.
This is more for ddr4 as ddr3 wasn't as picky. I think just manually setting the timings will work fine for the OP.
 

Marky000

Distinguished
Sep 23, 2013
39
3
18,545
See this image to see what I have tried:

View: https://i.imgur.com/g6LXoC7.jpg


In BIOS, when memory frequency is set to 1333MHz, all the timings match across both channels. CL = 9-9-10-24-1 @1.5V. Those timings look logically ok for a 1600MHZ CL11 and 1333MHz CL9 stick but they fail in Memtest86.

The image you see above is when I set frequency to 1600MHz speed. In the past, this used to fail Memtest86 test but I hadn't noticed the different CAS Latency for each channel. So, what I have done is manually adjusted the CAS Latency to 11 and the DRAM Row Cycle time so that timings match across channels. The DRAM voltage at this frequency is 1.65V. Primary timings = 11-9-10-24-2

Running Memtest86 test now, No errors yet at 50% percent of a full pass. If it fails, I will set the primary timings to 11-11-12-28-2 because that is the timings of the 1600 stick when I install it in isolation. So, as it stands, it is actually a memory overclock.

Will advise how the test goes.
 
Last edited:
  • Like
Reactions: SamirD

Reaper_7799

Distinguished
GL man...memory OC or just matching timings/speed for mismathcing sticks can be finicky sometimes and a pain with cmos. This board actually tests mem oc 5 times, then will take me straight back to bios if it doesn't pass, which is nice. No playing with buttons or clerar cmos or anything.
 
See this image to see what I have tried:

View: https://i.imgur.com/g6LXoC7.jpg


In BIOS, when memory frequency is set to 1333MHz, all the timings match across both channels. CL = 9-9-10-24-1 @1.5V. Those timings look logically ok for a 1600MHZ CL11 and 1333MHz CL9 stick but they fail in Memtest86.

The image you see above is when I set frequency to 1600MHz speed. In the past, this used to fail Memtest86 test but I hadn't noticed the different CAS Latency for each channel. So, what I have done is manually adjusted the CAS Latency to 11 and the DRAM Row Cycle time so that timings match across channels. The DRAM voltage at this frequency is 1.65V. Primary timings = 11-9-10-24-2

Running Memtest86 test now, No errors yet at 50% percent of a full pass. If it fails, I will set the primary timings to 11-11-12-28-2 because that is the timings of the 1600 stick when I install it in isolation. So, as it stands, it is actually a memory overclock.

Will advise how the test goes.
Excellent work! I hope it sticks at these timings as this would be really nice to have both at 1600Mhz, even if the timing is looser.
 

Marky000

Distinguished
Sep 23, 2013
39
3
18,545
Excellent work! I hope it sticks at these timings as this would be really nice to have both at 1600Mhz, even if the timing is looser.

Unfortunately, test failed with all combinations, even 2 sticks at 1333 and loose timings. So, i ended up removing the 1600 stick and just keep the high performance 1333 stick at CL9.

I couldn't risk further issues as this client has had an unstable PC for months. I wonder if it might be a motherboard issue that somehow prefers just 1 DIMM installed.

At least I havent seen any more freezes but once in a while, I did see the PC just reset itself but I voltage tested all the wires on the PSU including the 12V rails. All voltages looked good.

So maybe a subtle MOBO issue that hopefully will be 95% stable moving forwards.
 
Unfortunately, test failed with all combinations, even 2 sticks at 1333 and loose timings. So, i ended up removing the 1600 stick and just keep the high performance 1333 stick at CL9.

I couldn't risk further issues as this client has had an unstable PC for months. I wonder if it might be a motherboard issue that somehow prefers just 1 DIMM installed.

At least I havent seen any more freezes but once in a while, I did see the PC just reset itself but I voltage tested all the wires on the PSU including the 12V rails. All voltages looked good.

So maybe a subtle MOBO issue that hopefully will be 95% stable moving forwards.
Interesting. Well, if you can get another 1333 or 1600 module, then you can try again. I don't think there's any issues with the motherboard or power supply, but if a more closely matched module also fails, then then there is a chance.
 

Marky000

Distinguished
Sep 23, 2013
39
3
18,545
Interesting. Well, if you can get another 1333 or 1600 module, then you can try again. I don't think there's any issues with the motherboard or power supply, but if a more closely matched module also fails, then then there is a chance.
I still find the whole thing a bit odd. The sticks weren't really so far apart in terms of speed/timings.

Correct me if I am wrong but the recommendation is to tighten CAS by an increment of 1 for every drop in frequency. I know I read that here somewhere although I think that person might have said that applies for every drop of 200MHz.

And when the sticks were installed on their own and forced to run at 1333, the auto timings were virtually identical.
 
I still find the whole thing a bit odd. The sticks weren't really so far apart in terms of speed/timings.

Correct me if I am wrong but the recommendation is to tighten CAS by an increment of 1 for every drop in frequency. I know I read that here somewhere although I think that person might have said that applies for every drop of 200MHz.

And when the sticks were installed on their own and forced to run at 1333, the auto timings were virtually identical.
I do too, but this is where the quality and design of the modules come into play. Thing was this did happen in the DDR3 days, but it just wasn't as common of an issue.