Need help OverVolting

[BBigJ]

I'm getting pretty jealous seeing all these people with 2.0 volts or more for their vcore, because I can't seem to keep my system stable with anything over 1.85 volts. There must be something I am doing wrong and maybe somebody can help me.

Let me start with my standard overclocking proceedure. First I determine my stable FSB range. I lower the multi, and push the fsb until Prime95 torture test gives me errors (right now I appear to be capped at about 146, but that's another tale). Then I drop the fsb back to where it is stable and start pushing the multi. At every step I run p95 for about 5 mins to check stability. When I reach a multiplier where p95 gives errors, I try the same speed again with both lower and higher voltages. If I don't get errors, I start pushing the speed more (either by multi or fsb.) I consider myself finished when I can run p95 for an hour or more without errors (at which point I usually get bored an turn folding back on) and I can't hit a higher speed with either a higher or lower voltage. With my tbird 1.2 this happened at about 1.48, 1.85v. Now I have a week 13 AGOIA Y and I hit the ceiling at 1796mhz 1.81v. I'm a little disappointed, cuz I know this chip can do more.

At first I thought that the problem was my cooling. I figured that my heatsink couldn't keep the chip cool enough at higher voltages. When I went to watercooling, sure enough, my tbird was able to handle another .05v. But with all the people reporting stable overclocks at 2 volts or more with air cooling or simple water, I'm starting to think that temps might not be the entire story. As I write this (on a rather warm night) mbm5 is reporting a core temp of 39 (at full load) with 1.79 vcore and 1796mhz (don't know why the vcore is so low, I have it set to +0.075).

Please let me know if there is some fundamental flaw to my approach. I can get another 0.05 to 0.1v out of this mobo without doing a volt mod and I'd love to get the most out of my chip. My specs are in the sig.

Thanks is advance.

[Gnufsh]

You may be starting to hit the edge of your proc's performance. Therefore, it will take more and more volts to make the same difference. Also, before folding you should run prime95 for well more then an hour. More then 24 hours would be good, but at a bare minimum 6-8 hours. You do not want to return bad data due to proc errors. Your procedure is sound, other then not stability testing enough. If you get board, try throwing 1.9-.0 volts at it and see if that makes a diff.

[juliendogg]

although the bios in your ak31 says it is adjustable so much, the adjustments only work to + or - .1v, any more and it just drops back to default. to get more than that you'll need a vmod. your temps look good, i'd think you'd be able to squeeze more out of that chip. have you tried setting it as high as you can stable and run either p95 or sandra burn in for a good 24 hours? usually that will help you get a little more. other than that, i can't see anything you'd be doing wrong, you're method of finding highest stable speeds looks good, what is your fastest stable fsb?

[BBigJ]

You may be starting to hit the edge of your proc's performance.

I was afraid of that.

When I think I'm done tweaking I'll run p95 for several hours to make sure. I'm not terribly concerned about sending bad data back to folding. There is a certain amount of randomness built into the program. One bad fpu operation every hour will simply raise the effective temperature of the simulation slightly. No big deal. One bad fpu operation every few minutes is another story.

My fsb looks like it is capped at 146. I get major errors at 148. For some reason I can no longer throw 2.7v at my memory. I used to be able to run at 148 at 2.7 with no problem. Now it flips out at 2.7v at 133mhz. Go figure. Next time I'm tinkering around I'll swap the dimms and see if that makes a difference.

This chip has been running folding about 6 hrs/day for the last week and a half. I hope there is still some burning in to be done.

[BBigJ]

I'd like to hear from some people that are using voltages over 1.9 or so. What kind of cooling are you using and what temps are you getting? Are your comps really stable or are they just "stable" (nudge, nudge, ;), ;) ).

Juliendogg, I see you are at 1.95v watercooled with the same mobo and chip I have. What are your temps? What are the components in your wc system? Are you able to run the Torture Test without errors?

[Milkman]

I am running my Agoia 1600+ @ 1.92 volts, 1846 10.5 x176 and am getting 40 at idle and 48 max load watercooled.

Mamer

[BBigJ]

Thanks, are you able to run torture test without errors?

[Skiing Squirrel]

BBigJ, I have the same prob as you... I got an agioa 2013 and Im maxxed out at about 1780mhz. I guess we got crappy chips. :(

[Gnufsh]

Yes, crappy chips. They only run 380MHz faster then they are supposed to...

[AZZKICKER]

sounds like a power problem to me

what power supply you running
and what are your 12V 5V and 3.3V under a load?

[Silver]

Originally posted by BBigJ
I'd like to hear from some people that are using voltages over 1.9 or so. What kind of cooling are you using and what temps are you getting? Are your comps really stable or are they just "stable" (nudge, nudge, ;), ;) ).

Juliendogg, I see you are at 1.95v watercooled with the same mobo and chip I have. What are your temps? What are the components in your wc system? Are you able to run the Torture Test without errors?

Listen and listen well as I do have experience on this board and have run the 1.4 T-bird at 2.35v and could have done more if I thought the processor would do more. The fsb is about right however (as mentioned earlier in the thread) the vcore on this board was faulty. Overall the board was a good one however there was a major flaw with the vcore adjustment. It would always give a touch less than desired at default and would not allow any greater adjustment to the core than .1 (which usually brought it back to default). Regardless of what the bios is telling you, you are not getting that voltage. The vmod is one of the easiest to do and requires only a couple of pieces of wire and a Pomona grabber as well as a pot. With this simple clip on (read no soldering) the vcore could be adjusted using mbm. I highly recommend the mod (though as always the disclaimer on damage at your own risk) as this will open up some very high voltages should you desire them. Look for an article on the front by 'Doc' as well as look for a vmod thread in the shuttle mobo under AMD written by Doc. This will definately do the trick on the vcore and you will no longer wonder as you will be able to get higher voltages than most here, inclusive. As always on the vmod, watch out for temps under higher voltages. Nice thing is you can adjust with mbm running to give you read outs on temp/voltage. Take Care. :)

[BBigJ]

Silver, I am familiar with Doc's voltage mod. I will consider that when the time comes. My problem is that my optimal voltage is at 1.80 which is still below the 1.85 that I can get with this board without a mod (actually, I was able to get up to 1.87 with my tbird, but that might have been a fluke.) It seems to me that there are more Mhz to be gained by going to 1.85, but I lose stability by doing so.

[rogerdugans]

I am not familaiar with that board, but I am familiar with vcore over 2v. Increasing voltage has only caused me problems in two ways: increased cpu temps(causing instability) and lower 5v line (also causing instability.)

I am lucky, in that my agoia will hit 1850 on aircooling at 1.85v.
To get to 2 ghz I had to go all the way to 2.1v vcore- and I was only able to run that speed for about 2 hours.
I am now running at 2.03v in my water cooled system @1927mhz, and yes, it is completely stable at this speed. If ambients were lower I would be stable up to 1960.

I also have two XP1700s volt-modded: one is running at 1.95v and the other is 1.85- both can go higher but are limited by the 5v rails.

Increasing voltage to the core, by itself, should not cause instability, but it may be impacting something else that is: 39c is great, so it isn't temp; how is the 5v line?; you mentioned that you are also having memory error problems: fix that and you may see a big difference;)

[BBigJ]

As I write this, my voltages (according to MBM) are 1.79, 3.20, 4.89, and 12.16. I am currently running folding. I have a 350W, AMD approved Enhance psu. I don't think that is the problem.

I am a little frustrated with my fsb, but right now I am only at 143. Although I suppose it wouldn't hurt to investigate that some more.

[BBigJ]

As I mentioned earlier, I was unable to run my RAM at 2.7v at any speed without major errors. Last night I swapped my dimms (moved one stick from slot 1 to slot 3 and the other from slot 3 to slot 1). Now all of a sudden I can use 2.7v again, and when I started pushing my fsb I didn't get any errors until I hit 158 (I only tested for 5 min at each speed, though. I don't think I'll actually be able to run it above 155.) I think that is pretty darn good for two sticks of pc2100 at the most aggressive settings. The bad news is that the proc is still capped at about 1.8 gigs. Rather dissappointing, but now that my fsb is back up I can live with it.

[bobt17]

y not try lowering your ram timeings some if u want to take it a little higher? i had some crucial pc 2100 running at 162fsb for the longest time with 2.8v tho...... still useing less agressive timeings u should be able to squeeze a bit more from ur cpu if thats whats limiting u right now

[BBigJ]

Actually I have a 12.5 multiplier on this board, so the board isn't holding me back. It just seems like this cpu doesn't want to go higher than 1.8 whether it is 151*12 or 144*12.5. Since the fsb isn't holding me back, I think I'll leave the timings where they are. I don't have a 1/5 divider on this thing, so I really shouldn't take it any higher.