Gary, what you're seeing on that HD7850 host sounds like a clear bug. Either the memory usage is misreported (but you're not seeing swapping either) or someone is handling memory badly. I don't know which Linux you're using - could this be a problem? Maybe some experimental kernel or driver?
You're probably using a 64 bit OS, aren't you? In this case dress space shouldn't be an issue at all.
You checked the RAM timings & clocks. There are actually many more timings than usually shown in a BIOS, but they mostly don't matter much (that's why they're not shown). So for any differences in minor timings it would be very hard to cause such significant performance differences. And larger modules tend to have worse timings (although this may not be true if the 2x2 GB kit is much older).
The interanl organization of the modules could be different, e.g. the 4 GB modules could do more interleaving over more banks (or was it "ranks"?). they could also differ in the command rate. Yet, I wouldn't expect any of this to cause such a large difference in GPU performance.
Free shows: free -m -s
)
Free shows:
free -m -s 2
[pre] total used free shared buffers cached
Mem: 16003 6028 9974 69 272 1387
-/+ buffers/cache: 4368 11634
Swap: 16335 0 16335
total used free shared buffers cached
Mem: 16003 6028 9974 69 272 1387
-/+ buffers/cache: 4368 11634
Swap: 16335 0 16335[/pre]
(using 'pre' tag)
RE: Free shows: free -m -s
)
Thanks. Another mystery solved.
Gary, what you're seeing on
)
Gary, what you're seeing on that HD7850 host sounds like a clear bug. Either the memory usage is misreported (but you're not seeing swapping either) or someone is handling memory badly. I don't know which Linux you're using - could this be a problem? Maybe some experimental kernel or driver?
You're probably using a 64 bit OS, aren't you? In this case dress space shouldn't be an issue at all.
You checked the RAM timings & clocks. There are actually many more timings than usually shown in a BIOS, but they mostly don't matter much (that's why they're not shown). So for any differences in minor timings it would be very hard to cause such significant performance differences. And larger modules tend to have worse timings (although this may not be true if the 2x2 GB kit is much older).
The interanl organization of the modules could be different, e.g. the 4 GB modules could do more interleaving over more banks (or was it "ranks"?). they could also differ in the command rate. Yet, I wouldn't expect any of this to cause such a large difference in GPU performance.
MrS
Scanning for our furry friends since Jan 2002