Okie dokie here are my findings on a Sabertooth rev1 990fx w/ 1604 bios
Most tests were done with ripjawsx 2133 cl11 ram 4x4gb set, HX850 psu, Xigmatek Aegir Air cooler, on Windows 7.
Here are the main settings that were used in bios config.
D.O.C.P. Mode (not manual or auto)
nb as near to 2200 as possible (not auto)
ht ranging from 2160 to 2626, depending on how bad fsb mangles it.
Spreads all disabled
epu power saving off
cpu load line med
cpu/nb load line high
cpu current cap 100%
cpu/nb current cap 100%
cpu powerphase optimal
vrm spread disabled
cpu&nb voltage in manual mode
cpu 1.425 (sits at 1.404 and 1.356 on max load in IBT)
nb ht 1.25
Easily hold clocks regardless of multiplier vs ht vs nb vs nb @ 4.5ghz with about the same temp on idle and max.
4.5ghz @ 200fsb 22.5x 1.404v
4.5ghz @ 240fsb 19x 1.404v
4.5ghz @ 238fsb 2600ht 19x 1.404v
4.5ghz @ 245fsb 2450ht 18.5x 1.404v
4.5ghz @ 245fsb 2450ht 18.5x 1.404v mem@1966(983) 1.65v 10-10-10-28-2t
4.5ghz @ 200fsb 22.5x 1.404v mem@2133XMP 1.65v 11-11-11-30-2t
Now for the un-pretty grouping of cinebench 11.5 stuff between most of the achieved clocks. (note these were done without the microsoft bulldozer patches.)
Now for the mem benches using memmax2
these were all done @ 1600 because of a problem I found that you will see in the next set.
Now for the IMC/Board/BIOS problem I ran into, I don't know which to blame but one of them is mucking up any testing above 1600 memory speeds. (Both in 2 stick and 4 stick configs, I get the same numbers either way.)
And last but not least the HWinfo which was never changing through these tests.
I currently blame my reaaaly high ambient temp for making me need a liquid 240mm rig to go higher than 4.5 safely.