XP2100 Prime95 sum out error

[Mattlock]

just built the following rig:

XP2100 rev B, AIUHB, 0248
Epox 8rda+
Slk-800, Smart Fan II
Antec True430 (power supply)
Zalman Northbridge HS (passive)
1x512 Corsair XMS3200 (low latency) w/ platinum Heatsinks

My first attempt at overclocking was to up my FSB to 166= 13x166= 2,157mhz
Upped CPU voltage to 1.7 (Vdimm, still default), RAM timings (default)


1) I ran CPU Hot Tester for 3 hours, no problem (will run for 6 tonight)
2) Ran MemTest for 3 hours, no errors
3) However, Prime95 ran for 1 hour with a "Sumout error".


My CPU temperature never got above 38C under stress. In fact with CPU Hot tester temp was 35C with fan wide open.

Questions
1) Is the Prime95 error a hardward or possible software(read that somewhere?) error?
2) If CPU Hot Tester runs 6 hours with no problems....am I stable?

3) I want to overclock further...what settings shoudl I change next?
4) What's probability of getting to 2.3 or 2.4ghz based on my results so far?

thanks
Mattlock

 

[shan]

I got the same ****ty error. However, after I lower my fps from 185 to 181, everything seems to be fine now.

 

[linlibj]

It's your hardware error. That means your overclocked system failed to pass prime95 test. I am wondering how many of those who oc their 2100+ to 2.3G or 2.4G have passed prime95 test for a period longer than 24 hr without error.

 

[jonspd]

I have ran prime for 24 hours @ 12 x 190 vcore 1.875. I think you need more vcore just try upping it alittle and see if the sumout goes away.

 

[[xeno]neophyte]

0248 is probably the best date i've seen. 1.7 vcore is fairly low and if your temps are 38°C load you could bump it up. My xp2100 0302 can run stable at 2300mhz at 1.8v and from what i've seen from the 0248 2300mhz is probably attainable at 1.75, 2400mhz or more if you maxed out your vcore to 1.85. But keep an eye on your temps!

 

[c0nfusion]

most common cause for prime95 errors is to not have enough vcore...try bumping it to 1.725 and that might do the trick

 

[Mattlock]

change multiplier or FSB

To get from 2157mhz to 2400mhz should I?

1) Up my FSB to 185 (13x185) or

2) Drop my multi to 12x and up fsb to 200? I think my RAM can handle it.

Never quite understood which is best....

thanks,

Mattlock

 

[jonspd]

Higher fsb will always increase overall preformance.

 

[shan]

I just failed prime95 after a period of time. now I increased my voltage to 1.80v, still burning now.

 

[jonspd]

There Is really no need to see how high you can get your cpu to when burning.

1 You need to set the vcore to the max.

2 Set the cpu speed as low as possible.

3 Then burn your heart out for @least 24 hours.

 

[SpaceRangerJoe]

your ram shouldnt have any problem running at 200fsb with enough voltage. something else you might want to try, is upping the voltage a lot, 1.85 or perhaps more, then burn in for a few days. after that, try to run the same speed at a lower voltage than before. burning my cpu in at higher voltages has seemed to help me. next time i restart, im going to go to 2.0vcore, while its still cold outside (my room is under 16C right now). so hopfully it will run over 2gig at 1.9volts or lower.

 

[jonspd]

(my room is under 16C right now).

Where do you live alaska?

 

[[xeno]neophyte]

In the end what you want is the highest fsb possible combined with the highest mhz you can get out of your cpu.
So what you need to do is find the max for both of them.
Setting the fsb really low and using a high multiplier factors out fsb problems. So by incrementing the fsb by 1's you can get your cpu high without losing stability from your fsb.
Once you reach an unstable situation, you can fine tune the cpu speed by lowering the multiplier and increasing the fsb more (keeping the fsb low enough its not causing the problems). Using this method you should be able to find the cpu max fairly quick.

After you know the cpu max speed, you do the same thing but maximizing the fsb by using a low multiplier. This keeps the cpu speed low and factors out the cpu as causing the problems. Now you increase the fsb until you see problems, and you get your max fsb.

At this point you know your cpu will run at X and the fsb will run at Y
And what you want is Y*multiplier = X if its possible. Chances are you are going to have to concede a bit of cpu speed or fsb since the multipliers are in 0.5 increments and at that point its better to have the higher fsb since it increases overall system performance.

This is the method I used to OC my system, any suggestions?

 

[SpaceRangerJoe]

Where do you live alaska?

tulsa oklahoma. its been cold here recently. i leave my heater vent shut all the time, and it gets about 17c or so, then i open my window for a little while. i like cold weather .

 

[Yodums]

Burn in after a week and that should change. Also more voltage would definitely help since you are getting it to get into windows and stress testing but it isn't able to maintain that stability.

 

[Gnufsh]

If prime95 comes back with an error message (like the sumout error) it is a hardware problem.

 

[BomberBear]

I've found the best way to found max speed is just the opposite of what neophyte suggested. I lower the mutiplier way way down to keep the cpu out of the way. Then I find the max point that the fsb will reach. Then raise the multi. until you reach the cpu max.

 

[mata2974]

Prime is useless and overrated.............dont base everything on it.

 

[[xeno]neophyte]

bomber: well its really the same thing.. your just finding the fsb part first


mata: I tend to agree with you about p95. I use superPI and i play sof2 and those work fine for hours.

 

[Gnufsh]

mata2974: why do you say that?