So check this out. As many know, most of the time 4 sticks of ddr2-1066 in an AMD (or Core2*) mobo cannot run at 1066. The memory controller and such just cannot handle all the load. This was true of mine, I have been running my 4 gb sticks of Gskill 5-5-5-15-2.1vdimm 1066 @800mhz 4-4-4-12 for almost a year, as it would give a rounding error almost immediately in Prime95 mixed torture test every time at 1066. Mem is always at 2.1v, I am not comfortable running ddr2 any higher than this.
Well after a new bios flash (this could have been the cause of the new functional speed too, just flashed the F6 bios to my ma790gp-ud4h in sig), and bumping both the CPU_NB and NB voltages up a minor +0.05v (this could also be it, I was unaware either of these voltages had an effect on memory overclocks), now they can do 1066 just fine. I was increasing those voltages to see if I could get 2.3ghz+ CPU_NB clock stable.
I was bringing the memory up to 1000 (250fsb x4 mem multi), and accidentally set the fsb higher than 250. It took me a minute to notice, because normally even a hair over 1000mhz memory, prime would error out. But there were no errors at 1040mhz ddr2 speed.
So I just went ahead and returned my fsb to 200, CPU_NB and HT speeds to my standard x10 multis (2ghz each), and changed the memory multi to x5. Low and behold, over an hour into Prime 95 torture test, and the memory is doing [email protected] just fine, no problems or errors in prime, and no errors or noticable problems in win7 64 either.
I wonder, does memory benefit from burn in time, or is it just that increasing either or both of the NB and CPU_NB voltages that stabilized it? Or perhaps one of the new bios versions I have upgraded too since the original F3/F4 the mobo came with (I since upgraded to F5 a few months ago, and F6 today).
Oh you know what, I just thought of this, a little while ago I swapped my old 940BE into a friends new build, and adopted the 955BE I ordered for his pc. He does not o/c, and was using ddr2, so no harm or foul. I overclocked it to 3.2ghz & 2ghz CPU_NB and HT clocks so it runs exactly as a 955be would. Perhaps the memory controller on the newer 955BE can handle the load better?
I checked, and if I lower the CPU_NB and NB voltages back down -.05v to stock, prime errors out @1066. So the extra mobo voltages are definitely helping. But back when I first built this box, bumping up those voltages still did not allow me to run 1066mhz stable.
So what do you guys think, is it possible that 8 months of burn in loosened up the memory and made it able to run at previously impossible speeds, or is it the new CPU, or new bios version? There is likely no way to find out for sure, but still, pretty interesting stuff!
So if you are in the same boat as me, running your ddr2-1066 @800 because you have all 4 slots populated (I absolutely love 8gb by the way, would never go back under that), bump those mobo voltages a bit, update your bios, and see if your system can magically handle the higher speeds!
fakedit: now 2hrs prime stable, plus watching video and editing some photos, and still no problemo
Some Images:
Everest Cache & Mem Bench, about 1000mb faster than @800 (write speed has always been a little low, maybe I should try unganged?)
Everest Sidebar/Desktop Gadget & CPUID stats
Once I stability test this a bit more, Ill bring my CPU back up to 3.7-3.8ghz and CPU_NB link back to 2.2-2.4ghz. So nice when your box is able to oc faster as time passes! Considering this is a true budget box, total cost was minuscule compared to say, an i7 box, I am in love with its performance and have been since I first built it!
Well after a new bios flash (this could have been the cause of the new functional speed too, just flashed the F6 bios to my ma790gp-ud4h in sig), and bumping both the CPU_NB and NB voltages up a minor +0.05v (this could also be it, I was unaware either of these voltages had an effect on memory overclocks), now they can do 1066 just fine. I was increasing those voltages to see if I could get 2.3ghz+ CPU_NB clock stable.
I was bringing the memory up to 1000 (250fsb x4 mem multi), and accidentally set the fsb higher than 250. It took me a minute to notice, because normally even a hair over 1000mhz memory, prime would error out. But there were no errors at 1040mhz ddr2 speed.
So I just went ahead and returned my fsb to 200, CPU_NB and HT speeds to my standard x10 multis (2ghz each), and changed the memory multi to x5. Low and behold, over an hour into Prime 95 torture test, and the memory is doing [email protected] just fine, no problems or errors in prime, and no errors or noticable problems in win7 64 either.
I wonder, does memory benefit from burn in time, or is it just that increasing either or both of the NB and CPU_NB voltages that stabilized it? Or perhaps one of the new bios versions I have upgraded too since the original F3/F4 the mobo came with (I since upgraded to F5 a few months ago, and F6 today).
Oh you know what, I just thought of this, a little while ago I swapped my old 940BE into a friends new build, and adopted the 955BE I ordered for his pc. He does not o/c, and was using ddr2, so no harm or foul. I overclocked it to 3.2ghz & 2ghz CPU_NB and HT clocks so it runs exactly as a 955be would. Perhaps the memory controller on the newer 955BE can handle the load better?
I checked, and if I lower the CPU_NB and NB voltages back down -.05v to stock, prime errors out @1066. So the extra mobo voltages are definitely helping. But back when I first built this box, bumping up those voltages still did not allow me to run 1066mhz stable.
So what do you guys think, is it possible that 8 months of burn in loosened up the memory and made it able to run at previously impossible speeds, or is it the new CPU, or new bios version? There is likely no way to find out for sure, but still, pretty interesting stuff!
So if you are in the same boat as me, running your ddr2-1066 @800 because you have all 4 slots populated (I absolutely love 8gb by the way, would never go back under that), bump those mobo voltages a bit, update your bios, and see if your system can magically handle the higher speeds!
fakedit: now 2hrs prime stable, plus watching video and editing some photos, and still no problemo
Some Images:
Everest Cache & Mem Bench, about 1000mb faster than @800 (write speed has always been a little low, maybe I should try unganged?)
Everest Sidebar/Desktop Gadget & CPUID stats
Once I stability test this a bit more, Ill bring my CPU back up to 3.7-3.8ghz and CPU_NB link back to 2.2-2.4ghz. So nice when your box is able to oc faster as time passes! Considering this is a true budget box, total cost was minuscule compared to say, an i7 box, I am in love with its performance and have been since I first built it!
Last edited:
