What do i do after failing Prim95?

[mosco]

Athlon XP 2500+ overclocked at 11X200 at 1.725 Vcore on a Abit NF7-S v. 2. I also tried upping the Ram voltage to 2.7 but that did not have any effect on prime95. It fails instantly with a rounding error but works fine using XP2500 defaults.

Do i have to up the Vcore to 1.8 or 1.75? Is that safe? How much more heat is that? What about Ram, 2.8, its Corsair XMS 512MB PC3200C2 with Platinum heat-spreader.

Also, i had ram timing set to auto when i overclocked to 2200Mhz and it slowed the ram down to 2.5 3 3 8 i believe, can i override that to 2 2 2 5?

 

[Corporate Thug]

you can either up the vcore or lower the overclock

 

[MachoDonut]

Remember that the Abit board undervolts. That 1.725 volts is actually ~1.68 to 1.7 volts.
You can up your Vcore a little bit more. 1.8 is safe but you shouldn't need that much for 11x200. Set all of your other settings to default. Work on one thing at a time.

You can set your ram timings to aggressive in the BIOS or set the timings to whatever you want. But get it stable first and then mess with that.

 

[deadseasquirrel]

Here are my thoughts on Prime95. Now, lemme preface this with the fact that I'm an idiot and quite frankly never have any real world evidence to back up my asinine opinions. Prime95 is great for stressing a system, ensuring that it is stable enough to complete that one specific task. Well, in your day-to-day computing, it probably won't translate as well.

I hooked up my XP1600+ and overclocked it to 166x10.5 eight months ago. Not a huge oc by any means. Prime ran fine at default, but, for the life of me, I couldn't get it to complete more than 10 mins no matter how much vcore I threw at the chip.

Well, after a good week or two of tweaking and tweaking to no avail, I gave up and said "Screw it, I'll just have an unstable system."

Eight months later, this "unstable" system has played every single game I've thrown at it, encoded divx, looped 3dmark2001 and 2003 so many times that the system itself actually *believes* that joint press release from nvidia/futuremark, rendered some great stuff with 3dmax, burned hundreds of cds, run memtest for many hours, and done all this at default vcore (good thing cuz my temps arent the best since I keep a hot ambient room).

I'm not saying to forego testing the stability of a system. It's great for identifying future or current problems and issues. I just wouldn't stress over it if Prime's level of stability remains out of reach as it has for me. That being said, if a small bump in vcore solves the stability issue with little heat increase, perfect.

deadseasquirrel

 

[Jeff7]

Originally posted by: deadseasquirrel
Here are my thoughts on Prime95. Now, lemme preface this with the fact that I'm an idiot and quite frankly never have any real world evidence to back up my asinine opinions. Prime95 is great for stressing a system, ensuring that it is stable enough to complete that one specific task. Well, in your day-to-day computing, it probably won't translate as well.

I hooked up my XP1600+ and overclocked it to 166x10.5 eight months ago. Not a huge oc by any means. Prime ran fine at default, but, for the life of me, I couldn't get it to complete more than 10 mins no matter how much vcore I threw at the chip.

Well, after a good week or two of tweaking and tweaking to no avail, I gave up and said "Screw it, I'll just have an unstable system."

Eight months later, this "unstable" system has played every single game I've thrown at it, encoded divx, looped 3dmark2001 and 2003 so many times that the system itself actually *believes* that joint press release from nvidia/futuremark, rendered some great stuff with 3dmax, burned hundreds of cds, run memtest for many hours, and done all this at default vcore (good thing cuz my temps arent the best since I keep a hot ambient room).

Was that XP1600 a Palomino? If so, I'm amazed that you got it that high. The two Palomino's I had only OC'd 200 of those performance points that AMD uses - XP1500@XP1700, and XP1700@XP1900. I never did remember the actual GHz ratings on them.
I still wouldn't like to have a system that can't do math right, considering that computers do a LOT of it.
I'm not saying to forego testing the stability of a system. It's great for identifying future or current problems and issues. I just wouldn't stress over it if Prime's level of stability remains out of reach as it has for me. That being said, if a small bump in vcore solves the stability issue with little heat increase, perfect.

deadseasquirrel

 

[mosco]

The problem is, my games were afftected. I would get a sutter and then Windows would say an error had occured. On default settings, i don't have any problems.

What about ram and chipset voltages?

 

[MachoDonut]

Originally posted by: mosco
The problem is, my games were afftected. I would get a sutter and then Windows would say an error had occured. On default settings, i don't have any problems.

What about ram and chipset voltages?

I have always felt that if your system wouldn't Prime for 24 hours, then it wasn't completely stable. I know a lot of people don't place a lot of faith in Prime, and that's fine. But your machine won't even do one loop. Since other programs bomb on you, then you definently have a problem.

Well, you are running at 200mhz FSB, which is what your RAM is rated for, which is why I say run it at it's default settings for now untill your stability issues are worked out. The Abit board also overvolts memory a little, so if you have it set at default in the BIOS, it's probably at 1.7v already, which is fine. Also, since you board is a rev. 2.0, it is designed for 200mhz FSB Althlons, so your chipset voltage should also be fine at default. If you plan on going higher, then you may need to give it more Vdimm and Vdd. I would be more concerned with chipset voltage if your board was a rev. 1.2 or earlier.

At what point in working up your overclock did you start getting stability issues? Also, have you tried upping your Vcore a little more? I can't see it needing more than 1.75v to get it stable at that speed.

Out of curiousity, what week is your Barton?

 

[deadseasquirrel]

Originally posted by: Jeff7

Was that XP1600 a Palomino? If so, I'm amazed that you got it that high. The two Palomino's I had only OC'd 200 of those performance points that AMD uses - XP1500@XP1700, and XP1700@XP1900. I never did remember the actual GHz ratings on them.

yeah, its an XP 1600+ palomino overclocked to XP2100+ speeds. however, many others were able to push their fsb towards 200 and whatnot, unlocking them with defoggers and all. i opted for keeping it locked at default multi and just raising to 166fsb. any higher and it puts my agp outta spec and things get ugly.

deadseasquirrel

 

[HawkeyeARK]

I have an Athlon XP 2500+ and it won't do 2.2 ghz unless I up the voltage to 1.8. What kind of heatsink/fan do you have? I replaced the retail hsf and got my temps about 7deg C cooler. Either way, you should be able to hit 2.2 ghz fine. It could be your ram, so start out with the fsb at 200mhz, but run your ram at a lower speed (333). I had to take my system out of dual channel mode to get stable at 400mhz. Now (in single channel mode), I can hit 460mhz. Using a multiplier of 9.5, this will put you at 2.2 ghz. The added speed from the ram and fsb really helps. Either way, start out with your ram at a lower speed and up your voltage a bit on your processor.

 

[Jeff7]

Originally posted by: deadseasquirrel

Originally posted by: Jeff7

Was that XP1600 a Palomino? If so, I'm amazed that you got it that high. The two Palomino's I had only OC'd 200 of those performance points that AMD uses - XP1500@XP1700, and XP1700@XP1900. I never did remember the actual GHz ratings on them.

yeah, its an XP 1600+ palomino overclocked to XP2100+ speeds. however, many others were able to push their fsb towards 200 and whatnot, unlocking them with defoggers and all. i opted for keeping it locked at default multi and just raising to 166fsb. any higher and it puts my agp outta spec and things get ugly.

deadseasquirrel

Geez, you must have a good one; as far as I knew, Palomino's were not good overclockers.

Well, you are running at 200mhz FSB, which is what your RAM is rated for, which is why I say run it at it's default settings for now untill your stability issues are worked out. The Abit board also overvolts memory a little, so if you have it set at default in the BIOS, it's probably at 1.7v already, which is fine. Also, since you board is a rev. 2.0, it is designed for 200mhz FSB Althlons, so your chipset voltage should also be fine at default. If you plan on going higher, then you may need to give it more Vdimm and Vdd. I would be more concerned with chipset voltage if your board was a rev. 1.2 or earlier.

Maybe you mean 2.7 volts? DDR runs at 2.5V as far as I know. But the VDD voltage as you mentioned, that may need to be increased.

 

[mosco]

i upped the chipset voltage, ram voltage, and vcore by little and it seems to be stable now. I am going to go through them and see which ones i can lower and still have stability. I can't remember what week i have, is there any way to tell without removing the heatsink? I am using the sk-7 with a variable speed fan. I don't think its a dual channel ram problem, i am running only 1 stick of Corsair XMS.

I tried Vcore at 1.775, ram at 2.7, and everything else at default and it seems fine now. I think it might have been the ram voltage.

So now this problem is fixed, I only have to get this stupid Echo issue with the sound.

 

[MachoDonut]

Well, you are running at 200mhz FSB, which is what your RAM is rated for, which is why I say run it at it's default settings for now untill your stability issues are worked out. The Abit board also overvolts memory a little, so if you have it set at default in the BIOS, it's probably at 1.7v already, which is fine. Also, since you board is a rev. 2.0, it is designed for 200mhz FSB Althlons, so your chipset voltage should also be fine at default. If you plan on going higher, then you may need to give it more Vdimm and Vdd. I would be more concerned with chipset voltage if your board was a rev. 1.2 or earlier.

Maybe you mean 2.7 volts? DDR runs at 2.5V as far as I know. But the VDD voltage as you mentioned, that may need to be increased.[/quote]

I did mean 2.7 volts. I just fat-fingered when I was typing

 

[sambao21]

Originally posted by: deadseasquirrel
Here are my thoughts on Prime95. Now, lemme preface this with the fact that I'm an idiot and quite frankly never have any real world evidence to back up my asinine opinions. Prime95 is great for stressing a system, ensuring that it is stable enough to complete that one specific task. Well, in your day-to-day computing, it probably won't translate as well.

I hooked up my XP1600+ and overclocked it to 166x10.5 eight months ago. Not a huge oc by any means. Prime ran fine at default, but, for the life of me, I couldn't get it to complete more than 10 mins no matter how much vcore I threw at the chip.

Well, after a good week or two of tweaking and tweaking to no avail, I gave up and said "Screw it, I'll just have an unstable system."

Eight months later, this "unstable" system has played every single game I've thrown at it, encoded divx, looped 3dmark2001 and 2003 so many times that the system itself actually *believes* that joint press release from nvidia/futuremark, rendered some great stuff with 3dmax, burned hundreds of cds, run memtest for many hours, and done all this at default vcore (good thing cuz my temps arent the best since I keep a hot ambient room).

I'm not saying to forego testing the stability of a system. It's great for identifying future or current problems and issues. I just wouldn't stress over it if Prime's level of stability remains out of reach as it has for me. That being said, if a small bump in vcore solves the stability issue with little heat increase, perfect.

deadseasquirrel

My understanding is that prime95 is the one of the most stressful programs you can run on your cpu because it uses 100% of the resources at 100% of the time, and the torture test checks the values against the correct known values, so if there is a tiny mismatch, it'll fail. Where as day-to-day applications, programs will still run if there is a tiny discrepancy. For example if you're encoding a movie, you may get a weird anomaly for less than 1 second of the movie (hardly noticeable) due to your cpu being unstable, but it will still encode the entire movie. You may never notice it, but it's there. Or playing unreal2, if it renders one insignificant pixel wrong, you may never notice it, and keep on playing Those are pretty dismissable errors, but errors none the less. Now, if you are getting a lot of errors, then you better back down, or up the voltage. But in prime95, there is no difference between a lot of errors or a single error, or big error or little errors. An error is an error to prime95 and it will fail no matter how insignificant. So I say if prime95 can't run for more than 10 minutes, you definetly have a "unstable" cpu.

 

[Jeff7]

Originally posted by: MachoDonut

Well, you are running at 200mhz FSB, which is what your RAM is rated for, which is why I say run it at it's default settings for now untill your stability issues are worked out. The Abit board also overvolts memory a little, so if you have it set at default in the BIOS, it's probably at 1.7v already, which is fine. Also, since you board is a rev. 2.0, it is designed for 200mhz FSB Althlons, so your chipset voltage should also be fine at default. If you plan on going higher, then you may need to give it more Vdimm and Vdd. I would be more concerned with chipset voltage if your board was a rev. 1.2 or earlier.

Maybe you mean 2.7 volts? DDR runs at 2.5V as far as I know. But the VDD voltage as you mentioned, that may need to be increased.

I did mean 2.7 volts. I just fat-fingered when I was typing [/quote]





Voltage increases can do wonders - my XP2100@2600 (Tbred B) gave an error in Prime95 after only maybe test 14 in the sequence it does. I increased the vCore by .25v, and it just went 14 hours without an error.