Memtest Strange Error only when Graphics Card is plugged in...

[temjam99]

Hi,

So, I've recently built a new rig and put it together last week, and after getting a few freezes in gpu heavy apps like battlefield 4 and unigine heaven, I decided to thoroughly check over my rig through varying tests. I immediately performed an extensive Memtest which turned up errors in Test 13 (Hammer Test). I ran only this test a couple more times just to be certain and noted that it failed in the exact same spot on each pass. Usually this:

>Test: 13 Addr: 35E4462B4 Expected: FFFFFFFF Actual: FFDFFFFF CPU: 0
and this...
>Test: 13 Addr: 26C0473B8 Expected: 00000000 Actual: 00200000 CPU: 0

Very Rarely there would also be another error in a different addres but this has only happened once despite now performing over 5 days of straight testing so I've chalked that down to absolute random chance.

Now, the strange things is, these errors go away when I unplug my graphics card, and in that case no errors are found. I was confused by the whole process before this, but am left baffled as to how simply plugging in a graphics device could be affecting the RAM sticks during a memtest, which as far as i know doesnt utilize anything to do with the GPU.

My RIG is as follows....
Intel I7-4790k @4.00ghz (Corsair H110iGT AIO Liquid Cooler)
Corsair VengeancePro DDR3 2133mhz 16GB XMP RAM
Palit Geforce GTX 980 Super Jetstream
MSI Z97 Gaming 7 Motherboard
Crucial m100 256gb SSD
EVGA 850 G2 GOLD PSU
NZXT phantom 530

I did notice someone else had a similar set up in another thread on these forums and replaced his RAM with another brand and his errors went away. I am just a little curious as to what could be the cause for the malfunction only when my 980 is inserted and not when it is not present....I don't want to go buying new RAM if there is another cause?

Any help would be hugely appreciated so thanks in advance.

[David (PassMark)]

1 or 2 errors isn't that much in the hammer test.
It could be just random.

If you had a 50% chance of getting an error per run, then probability of getting.
Fail - Fail - Fail - Pass (without the GPU) is 1 in 16. So not totally conclusive it is the GPU making a difference.

If it was conclusive show to be the GPU causing the RAM errors, then some possibilities are,
A) The GPU pulling enough power from the PSU that the the voltage levels to the RAM are slightly effected.
B) There is EMI from the video card, spraying out enough additional interference to cause the bit flip.

There are 100,000s of hits in Google for Battefield 4 freezing. Enough that I think there must be some other causes of a freeze.

I found this for example, "A bug affecting the PC version of Battlefield 4 is currently causing the game to freeze for some users, resulting in a sound loop. Developer DICE acknowledged the issue on the Battlelog and said it is currently working on a fix."

[temjam99]

I realise there are thousands of possibilites for Battlefield 4 as an individual game, but my system was stalling whenever ANY game or application pushed the system at all. As a result of this, I assumed there was a problem with the graphics card but couldnt always replicate the error in certain GPU benchmarks and stresstests so I began looking for other possible reasons.

I ended up running MemTest on the suggestion of a friend, who had a similar problem and fixed it by replacing his RAM. However, in my case I have now swapped out the GPU for an older GPU (amd 6950 instead of gtx 980) and am still getting the errors in Memtest. I'm running one more extended MemTest overnight tonight covering every CPU core on the HammerTest and shall see what it finds in the morning.

To summarise what still appears to be some sort of a memory problem...
1) When two sticks of RAM are inserted along with a GPU (of any kind or make) the memtest fails on HammerTest everytime (even if just testing one cpu core).
2.) When one stick of RAM is removed, the other will not fail a test, no matter what slot it is in.
3) when the GPU is removed altogether, both sticks of RAM will pass the full Memtest without error also....

My head is spinning trying to come up with a cause and effect here, but at the moment all i can come up with is to continually try every different combination of GPU slot and RAM slot until potentially coming up with more data. If anyone has any guesses as to what might be causing such repeatable errors please let me know.

Thanks in advance.

[David (PassMark)]

.... but my system was stalling

Can you describe was a stall is (in your system). e.g. is it just a short 2 second pause.
Or is it a unrecoverable lockup in a single application / game?
Or is it a unrecoverable lockup of the entire machine, forcing a power cycle to recover it?

Do you still get the 'stalling' with the new video card?

Due to the small number of errors, which are only in test #13, you still need to consider the possibility that the problem is totally unrelated to the RAM. It might be the RAM, but there are so many other reasons for pauses / freezes / lockups as well.

[temjam99]

The system freezes totally and requires holding the power button to turn off and back on again. While the freezing may not be caused by the RAM, it seems the RAM is still defective and I prefer to buy products with the peace of mind that they are working 100% (especially when it seems other products in the same price range *hyperx savage in other thread* seem to be capable of 100% passes).

The fact remains that I am getting some RAM errors and I am not happy with them, the only difference in my case is for some reason these errors only show up when I have a GPU inserted. It is this that I am trying to problem solve...

Could it be a mboard error or PSU error...or is it likely that the GPU affects the data cell arrangement of the RAM when it is plugged in and so it is STILL the RAM which is just being forced to reveal its own errors when a GPU is plugged in.

As per another thread on these forums, i may try and get hold of some HyperX Savage memory and see if that solves the problem. Of course I would like to know how my current memory can be affected by simply plugging in a GPU, as I didn't think that would be possible before yesterday, but hey, learn something everyday...

[temjam99]

Have now tested the GPU's (both 6950 and gtx980) in two seperate slots on my motherboard and the same errors appear no matter which card is in either slot. Something is happening when a GPU of any kind is inserted into my motherboard in any slot, that is affecting the RAM during a memtest, but again only in the Hammer Test.

[David (PassMark)]

The fact remains that I am getting some RAM errors and I am not happy with them

Yes, understandable. PCs can behave pretty strangely even when the there isn't a hardware fault.

Can you borrow some ram (of a different type) from another machine to see if the behavior changes.

You could also try increasing the voltage to the RAM from BIOS. e.g 1.55V instead of 1.5V
Or try decreasing the clock speed.