Duron - stock cooling question + prime 95 question

[Cat]

I originally bought a KT7 w/ a retail Duron 800, and was pretty happy with the performance I got.
Eventually I decided that I'd like 200 extra MHz, so I took off the heatsink, closed the L1 bridges with a pencil, and went straight to a gig (10x100).
Things ran perfectly, with the notable exception of prime95; I'd get rounding errors (49.99999 > 40.0 percent), so I dropped back to 800. I'd upped the core and IO voltage, btw. (Using wink2pro)

So, this leads me to 2 questions:
1.) Is it normal for prime95 to report these errors? I have a friend who bought a TBird from www.ocz.com, and he also had the same rounding errors I did, until he dropped to the normal multiplier.

2.) When I removed the STOCK HS/F, most of the gummy thermal material came off, and I didn't have any paste to use. So, I'm running at 36C idle, which is higher than before, but still not so bad. Any thoughts about this?

Thanks!

 

[Cat]

bump.

 

[sad]

<< Is it normal for prime95 to report these errors? >>

When you get rounding errors, or any errors of any kind, it means your cpu is not rock-solid stable at all. Prime95 is very demanding, and I consider it the ultimate cpu stability test. You can help that by upping your core voltagege beyond stock. It's pretty safe to have it max at 1.85v if you have a good quality heatsink and thermal compound such as Artic Silver. What you have to watch for is heat, and running without any thermal compound isn't a good idea at all. Hope that helped.

 

[AdamK47]

I agree. Prime95 is my defacto CPU stability tester. When I overclock a CPU the first program I go for is Prime95. There hasn't been a better program for me to use that would detect the slightest form of instability with FSB/Memory/CPU overclocking.

 

[gtd2000]

Prime 95 is a test of mathmatical accuracey - which is not necessarily anything to do with stability

 

[yodayoda]

that &quot;gummy thermal material&quot; was the thermal pad from the stock taisols that come with the retail-boxed amd processors, right? i would just scrape that sh*t off and use some good arctic silver. there is a guy on fleabay who sells a squirt of it for a buck or two, instead of having to buy a whole tub for $10--very smart, hehehe.

 

[AdamK47]

<< Prime 95 is a test of mathmatical accuracey - which is not necessarily anything to do with stability >>

Sure it does when it's the computer that has to accuratly run the mathimatical tests.

 

[gtd2000]

What I mean is it might &quot;fail&quot; Prime 95 but be stable at speeds higher than the failure of Prime 95 at overclocked speeds.
Because it fails Prime 95 does not necessarily mean you will have an &quot;unstable&quot; system.
There are lots of people on here posting that they can play Q3 &quot;stable&quot; etc at a frequency that Prime 95 fails at...and wondering why it fails Prime 95 but runs stable during normal/regular tasks.
So as I said it's not not necessarily anything to do with stability.
It's probably more a guide as to the maximum &quot;error free&quot; setting

 

[AdamK47]

Prime95 has a way of catching the slightest form of instability. Windows will ignore small CPU glitches caused by overclocking too far. Prime95 will error on these glitches and halt. I know I'm not alone when I say that I feel much more comfortable knowing my CPU can perform a task without worrying if the program will fail. Instability is instability and denial is not just a river in Egypt.

 

[xcourse]

On my system (listed below), Prime 95 would quit during the &quot;torture test&quot; and tell me that there were rounding errors if cpu speed got higher than 1313. In spite of this, everything would start OK up to a processor speed of around 1330. Nevertheless, I keep it at 1313Mhz.

 

[gtd2000]

hehe...denial.

I agree with the logic - but if for whatever reason you can play Q3 at a higher speed than Prime 95 allows then that is perfectly alright - if you choose to do so! (even if windows &quot;ignores&quot; the errors).

There are varying levels of &quot;stability&quot; I'm sure that even by using the Prime 95 test logic you cannot be guaranteed a 100% stable &quot;system&quot; during all situations assuming it passed the Prime 95 test?

I run my system at the maximum speed that will not produce windows errors (this may well coincide with Prime 95 results) because it's just not worth the hassle (worry/concern) of &quot;potential&quot; instability.

My system (Duron 700/Abit KT7A) can run at 1000Mhz but the odd gdi.exe error occurs now and again, I normally run at 900Mhz because for me it's the best trade off between performance and temperature levels

 

[esung]

It's the peace of mind that's important.. I don't want to have a extra variable in my system that I know under certain condition my CPU will be unstable.. or doing the wrong calculation..

It's the same thign with those software CPU coolers.. almost pointless IMHO. If you can not run your CPU at 100% speed, what's the point OC it?