Sign in with
Sign up | Sign in
Your question

Ganged Memory Stable, Unganged Unstable

Last response: in Motherboards
Share
November 4, 2011 8:57:01 PM

More of a comment than a question.

I am running Windows 7 64 bit on an Asus 990FX Sabertooth MB, a Phenom II X4 970BE processor and 4 sticks of Gskill Ripjaw DDR3 (total 16gb memory). When running with the Memory in unganged mode Prime95 stops with an error after 1 or 2 tests. In Ganged mode Prime95 runs indefinitely (or at least over 12 hours).

I understand the difference between the two modes (1x128bit vs. 2x64bit) but why would one be Prime95 Stable and the other not?
November 24, 2011 2:12:31 PM

Finally I found someone with exactly the same issue. I have the same motherboard, 965BE processor, but corsair memory. I just discovered this relationship today, but I also do not understand the reason why this is true? Anyone out there know?
m
0
l
November 24, 2011 6:46:16 PM

are you running this machine at all default and stock clock speeds/voltages? what do you have your ram's voltage set to?
m
0
l
Related resources
November 24, 2011 7:26:58 PM

Frizzo said:
are you running this machine at all default and stock clock speeds/voltages? what do you have your ram's voltage set to?


Hello,
Prime 95 will give the following failure almost immediately on all four cores if I select "Round off checking" and my memory is in unganged mode:
FATAL ERROR: Rounding was 0.5, expected less than 0.4

This will happen at default motherboard settings (no overclock) and when overclocked. I tried a lot of different combinations, but the results are pretty predicatable with "ganged" and "unganged" making the difference.

The memory runs at 1.50 volts at all speeds except 1866 where it sets itself to 1.65 volts per the XMP-1866 profile.

I think it is interesting that the other fellow had exactly the same results using a different brand of memory.

I have not tried to isolate the individual memory sticks, but I probably should?

Thanks for lending a hand.

m
0
l
November 24, 2011 10:24:29 PM

yea although it probably won't matter, but you can try 1 stick. what is the stock memory speed? 1600? try running it at 1600 or whatever the stock speed is at 1.65v and then try prime95 again.
m
0
l
November 25, 2011 2:42:32 PM

Well, my ganged/unganged theory fell apart yesterday as my memory started failing Prime95 in ganged mode. I tested each stick in its respective slot with no errors. I did notice that one of the slots didn’t seem to lock the memory down that well and I’m wondering if I had a stick not fully seated before. In any case, I decided to swap the sticks from the B slots to the A slots.
Now I can NOT get Prime 95 to fail in ganged or unganged, stock speeds or overclocked. I guess I fixed it! Ha Ha
m
0
l
December 1, 2011 11:23:11 PM

Frizzo said:
are you running this machine at all default and stock clock speeds/voltages? what do you have your ram's voltage set to?



Yes. Everything is at default. 1.5v and 9,9,9,24 timings. 1600 speed.
m
0
l
December 1, 2011 11:27:22 PM

bgoyins said:
Well, my ganged/unganged theory fell apart yesterday as my memory started failing Prime95 in ganged mode. I tested each stick in its respective slot with no errors. I did notice that one of the slots didn’t seem to lock the memory down that well and I’m wondering if I had a stick not fully seated before. In any case, I decided to swap the sticks from the B slots to the A slots.
Now I can NOT get Prime 95 to fail in ganged or unganged, stock speeds or overclocked. I guess I fixed it! Ha Ha


When I booted my build the first time it crashed quite often. I finally figured out that the problem was with my ASUS Sabertooth 990FX motherboard automatic settings for my memory. It wanted to run the memory at 1.65v when it was rated for 1.5. When I manually set the voltage (and timings) in the BIOS the crashes stopped. Of course, then I found the Prime95 problem I first mentioned above.

BTW, since first posting this my machine has been stable with no BSOD crashes. I recently updated the BIOS to 0813. Maybe I will change back to unganged memory and see it the Prime95 problem still exists.
m
0
l
!