Dual core CPU's should run 2 Instances of Super pi

[[GF]Duke]

I'm seeing a ton of people running Pi with only one Instance of it. You must make 2 separte Super-Pi folders and name the exe'z pi-1 and pi-2 respectively. Then just copy/paste them to desktop. Open each folder and run them at the same time.

[Somebodynew]

Originally Posted by [GF]Duke

I'm seeing a ton of people running Pi with only one Instance of it. You must make 2 separte Super-Pi folders and name the exe'z pi-1 and pi-2 respectively. Then just copy/paste them to desktop. Open each folder and run them at the same time.

Why? Im just curious?

[[GF]Duke]

Originally Posted by Somebodynew

Why? Im just curious?

Because they are only using 50% of their cpu power. Check my post in the top pi section for details.
Setting the affinity dosnt change a thing either so dont worry about that. I read somewhere someone suggested setting the affinities to cpu0 and cpu1. I tried it and saw no difference in cpu usage or performance.
http://forums.extremeoverclocking.co...115816&page=15

[ARSkemp]

other than showing that both cores are stable enough to run superpi does it matter?

[technodanvan]

I think most people are wanting to get their fastest SuperPi times, so they don't bother with two instances. When I'm trying to find my max CPU speed I'll use two and run them both at 8M, then continue on to the next level if it passes...but for most of the guys here its not a question of stability, just speed (not that it isn't stable, just saying stability can be guarateed other ways). And (at least on my Conroe rig) running one instance instead of two actually shaves off a good 3 seconds or so.

*shrug*

[ARSkemp]

i usually run 2 instances of prime95 for stability, and i can understand using 2 instances of superpi for stability too, but should i be running 2 instances of superpi for speed too? He linked to POST YOUR SUPERPI's topic which im assuming means i should be using it for speed as well since that topic is all about speed....right?

[slash_2_2000]

It's for stability testing on dual cores. Sometimes one core will fail before the other, when it comes to overclocking. Odd, but true.

Thus, when you only SuperPI one core, it's not a real stability test of your dual core. Only one of them. What if the second untested core is unstable at that clock?

[[GF]Duke]

If the 2nd core is unstable then the test will crash and you will have to adjust your tweaks. The cpus will always test faster with 1 instance because the 2nd core will kick in a little to help out.

[deadpass]

We should have to do this with HT procs as well then right? since there is a physical core and a virtual one. Personally I've never seen anyone post their superpi as a measure of stability, it's always about speed, what I have seen is people actually mention that their proc will run superpi but it's not prime95 stable. Superpi is to measure speed, prime is for stability, I'm willing to bet that the world record holder for a 1M run didn't use two instances, so then we shouldn't either since they set the standard.

[AruisDante]

Originally Posted by [GF]Duke

If the 2nd core is unstable then the test will crash and you will have to adjust your tweaks. The cpus will always test faster with 1 instance because the 2nd core will kick in a little to help out.

no, it won't. A single threaded app is a single threaded app no matter what. Only one CPU will ever be working on it at a time. You will still only get whatever your CPU speed worth of calulations done on it per second. Whether they come from core 0 or core 1 is irellevent. That's why a single threaded app will never use more then 50% of your CPU power on a dual core, because it will never exceed the max calculations of one core. Windows will sometimes bounce the app back and forth between the two cores, which is why in the Performance tab it migh not all be processed on one core, but rest assured, it is still only getting the same amount of CPU power as a single core would, at least froma threaded standpoint. The onyl advantage would be that any backgound tasks would be handeled by the other core, allowing a full CPU's worth of power to be devoted to pi. In fact, in theroy you would get the best performance by setting the afinity to one core, because when it can be bounched back and forth between the two cores there is a slight delay and a few wasted cycles as the data for the app is transfured from one core to the other (this is greatly reduced by Conroe, which has shared cache, but is so small that it's negligable anyway)

[[GF]Duke]

I stand corrected on the 2nd core kicking in. Thanks. But being a single threaded app you do need to run 2 instances of it to get an accurate score.
Every forum i posted a Pi score in got on my butt for posting with 1 instance with pi when i first got my FX60. Including this one i believe. And it was explained to me this way. Except the 2nd core kicking in part.

[AruisDante]

no. you don't. You're still getting an acurate score for said core. I suppose if you want to say you're dual core stable..... but A lot of benches are done when it's not necissarily dual core stable, and those scores are still equally valad. The only point I see in runing a dual pie is for stability purposes, not for bench score purposes.

[SOLDNER]

Ohhh what it would be like to have to wonder which one of my cores where working the most

Poor single-core me

[[GF]Duke]

So super-pi testing with an overclock has nothing to do with stability of the chip? hmm Dosnt make sense. You're only testing one core with one instance of pi. how can you say thats a stable overclock? I've had many a overclock where one core crashed and the other kept going to the finish.

[K31TH3R]

SuperPi is more of an extreme benchers stability test, because running 32m SPI only takes about 20-30 minutes, where as running Prime95 can take all night.

32M SPI is not as good for checking stability as Prime95 is.

i've run dual instances of 1m SPI and both of my cores score the same, 31 seconds @ 2.8GHz/310MHz/255MHz 3-4-4-8 1T

in the end it all comes down to personal preference when stability testing.

but yes, i agree, if you are running SPI to stability test, you need 2 instances.

but for a 1M score, only 1 instance needs to be run

[AMDcracker]

but what if i run super pie1, wait and then run super pie 2 after one is done? My score woould be better then running them at the same time.

[AruisDante]

In theory, it shoudn't be, assuming you set the affinity of both to their own core.

My point was that to recive a valad super pi score for comparison purposes, then you shouldn't have to run two instances. If you're testing for stability, of course you should run both instancies, but you should be testing stability with prime or SP2004 (orthos), or folding, all of which are far more stressful the Super Pi.

[[GF]Duke]

Usually one core scores better than the other. Not by alot but they do score differently. Whether ran together or alone they score differently. 2 Instances proves that the score is stable. If the cpu is stable there shouldn't be a problem running 2 instances. What good is a overclocked score if it isnt dual core stable on a dual core cpu. Thats my question. I think most people would want to know that their investment is working.