Hey SPi freaks
Some time ago I was a bit sceptic, when I read reports about 20s+ disadvantage in benching 32M on Conroe setup @ 2x512MB vs 2x1024MB memory configurations. Yesterday, I finally had a chance to run a little comparison myself and just wanted to share results I got, in case anyone here is interested
I used my ASUS P5B Deluxe/E6600 config for this test. I benched 16M, 32M and dual 32M in absolutely identical conditions, except for just swapping memory or adding some. I used 2x512MB kit, then 2x1GB and finished off with 4x1GB. Below are the results:
2x512MB
2x1024MBCode:16M --> 6:54.172 32M --> 14:59.500 dual 32m --> 16:17.500 / 16:15.407
4x1024MBCode:16M --> 6:50.172 32M --> 14:50.266 dual 32m --> 15:57.735 / 15:56.375
Conclusions:Code:16M --> 6:49.047 32M --> 14:48.609 dual 32m --> 15:53.593 / 15:51.344
1. I stand corrected, benching SPi with 2x1GB is indeed beneficial, although gain is not as large as others have reported. However, this might be due to differences in platforms as mobos based on 965 chipset were not yet available at that time, only 975.
2. Dual 32M results show, that time gains from using 2x1GB over 2x512MB are larger compared to single 32M run.
2. Filling up all DIMM slots pretty much did not bring a noticeable gain, compared to 2x1GB config. I was hoping, rank interleaving with 4 modules would kick some a$$ but I guess it did not happen in my testing
... just for those who care
flame away...
Bookmarks