My oc settings for my new Abit AN78GS

[ATrebeau]

I had been looking for oc help for my new mobo, related to FSB and RAM but believe I figured it out and just wanted to share what I came up with. I've always been use to only having a FSB adjustment available but with my latest build, I have a multitude of oc'ing features at my fingertips.

My goal? No major temp increase with a noticeable increase in system perforance, which I believe I achieved. And having my FSB:RAM ratio at 1:1. A member here, sorry I do not recall who, highly rec' that I run my FSB:RAM ratio at 1:1 on my Barton system years ago and after doing so, my system ran smoother and better than ever.

My settings now:

Core Speed: 3.2 GHz (from 2.6 GHz)
Multiplier: 16
Bus Speed: 200 MHz
HT Link: 1000 MHz
HT Clock: 1000 MHz
HT Width: AUTO
HT Speed: 16-bit
RAM: 400 MHz 4-4-4-12 tRC 23 2T

Vcore: 1.4 V (from 1.3 V)
RAM Voltage: 1.9 V

CPU Temp: 43.5C
Mainboard Temp: 28C

Core #1 Temp: 4.8C
Core #2 Temp: 20C

Related - EVGA 8800GS

Stock Speed Settings: 550|1375|800
Core Temp: 44C
Ambient: 38C
Fan Speed: 50%

Monitoring closely, took CPU to 3.4 with no stability issues but since I didn't want to maintain that, there was no reason in going any higher for experimental purposes. I can go 3.0 GHz with no voltage increase and may do that, haven't decided yet. The timings for the RAM I listed are the tightest I have been able to get. OCZ warrants this RAM for 2.1-2.2 V and when I was attempting tighter settings, I did set the V up to 2.1 just so I knew voltage wouldn't be hampering my trys.

Even though I'm still unclear whether 42.5 is my idle temp for my cpu (5000+ Black Editions are noted for falsely higher than actual temp readings) or not, I only noted a 1C increase in temp between 42.5C with 3.0 GHz and 43.5C with 3.2 GHz. Individual core temps only rose by 0.5C each.

You'll notice I did not increase any voltages on the mobo, my HT is running perfectly on'spec and there was no reason to. Why heat things up more with no reason, when NB's do their own job fine of running warm, lol.

Well, that's where I'm at. I notice the difference between 3.0 & 3.2 GHz but not necessarity enough to warrant an increase in voltage or the minor 1C increase in temp. 3.2 GHz does appear to be the sweet spot for this CPU.

If you have any suggestions or notice any changes you think I should make, please let me know, they would be appreciated. Otherwise, I help my post helps other Abit AN78GS owners out.

Cheers!

[ajrox]

have you benched it stock and then OC'ed?
you might find interesting numbers there. will also help you find the sweet spot.
just my 2 cents....
AJ

[ATrebeau]

The only thing I have to use from benchmarking is PCWizard2008 and it doesn't seem to be truly made for such and I realize it isn't one of the standards, it's just all I've had to use thus far.

Great suggestion and I'll do that. I know I need to find a suitable program to do so. Would 3dMark05 (free basic version) suffice or does it include enough parameters in it's free form to do such? I'm running XP Pro so really have no need for Direct 10 support. Or would you rec' I use the free edition of 3dMark06?

Thanks for the reply, app'it.

[ATrebeau]

Went ahead and dl'd both 3dMark05 and 3dMark06 free editions and Prime95.

EVGA 8800GS oc'd 700|1750|1000 no voltage change or bios update and fan at 60%.

3.0 voltages = Vcore 1.35|RAM V 1.9
3.2 voltages = Vcore 1.425|RAM V 2.1

RAM set to 4-4-4-12 23 2T.


3dMark05

CPU|RAM|FSB|HT
3.0 375 200 1000
15321 3DMarks
7504 CPUMarks

CPU|RAM|FSB|HT
3.2 400 200 1000
16084 3DMarks
8565 CPUMarks
_____

3dMark06

CPU|RAM|FSB|HT
3.0 3.75 200 1000
3DMark Score 10312
SM 2.0 Score 4732
SM 3.0 Score 4877
CPU Score 2289

CPU|RAM|FSB|HT
3.2 400 200 1000
3DMark Score 10773
SM 2.0 Score 4971
SM 3.0 Score 4993
CPU Score 2441


I ran Prime95 Torture Test x2 Large FTT's for 15 minutes on each, both stable. Also ran Blend for 5+ minutes on each, both stable.

[MrMozart]

Running Prime95 for 15min. is OK just to test your OC. Once you are happy with your OC, you should let it run for at least 4 hours, many overclockers run it for 8 hours or more. I usually let it run overnight. Letting Prime run for this length of time, really stresses your system. If you are Prime stable after 4 hours or so, great. Prime stresses your system much more than anything you would most likely do.

[ATrebeau]

Thanks, MrMozart, I'll do that, run it overnight.

I actually would like to run at 3.0 GHz (so I can remain at 1.35 Vcore) or thereabouts but doing so won't allow my achieving FSB:RAM 1:1 ratio, which I really want to do. I've tried so many different FSB and multiplier combos to this point, I have nightmares with numbers chasing me in my sleep, lol.

Good of you to reply and help

[ATrebeau]

Prime95 talks about a "divisor error" that may occur r/t software incompatibility issues on some systems. I'm one of those systems. At the exact same point every single time I run it, about 3 minutes in, Core #1 stops testing. Doesn't matter if I run stock at 2.6 or no voltage change at 3.0 or oc at 3.2 GHz, exact same place everytime. So I won't be able to run it coninuously overnight.

I dl'd wPrime and it runs fine with no compatibility errors. I've run it several times in a row at 3.0 and 3.2 GHz and completely stable; ie, close to an hour on each CPU clock setting.

I am going to dl and run MemTest to check out my RAM and to get experience working with it, as well. Will report back my results.