Super Pi Acting funny

[Pbirmingham1988]

Ok well I have reduced my timings from 4-4-4-12 to 4-4-4-10 then on to 4-4-4-8, 4-4-4-6, 4-4-4-5, and finally 4-4-3-5. Well the thing is as soon as I boot my pc up and run super pi, I run the test at 1 megabyte and it goes very fast at lets say 14-15 secs. If I wait another minute or 2 I see a decrease in performance if I retry the test. Now its at 22-23 seconds. 14-15 made me so happy when I saw it but then I was greatly disappointed from the results when I retried.

What could possibly be wrong with my memory and super pi??? My fsb:ram ratio is 1:1 and it only seems to be 1:1 if I have my memory set half of the fsb frequencies.

For example,
FSB : 1600 RAM : 800 = 1:1 for me in CPU-Z
FSB : 1500 RAM : 750 = 1:1 for me in CPU-Z

Nothing else gives me an equal 1:1 ratio.

So what could be wrong with super pi and my memory. I ran out of conclusions for the decrease in performance after a few minutes after boot. My OC, I believe, is stable for the most part but I'm having thoughts that my pc is just plain screwed up.

[stinger608]

Well for one, your not reducing the timings!!! The lower the number the tighter the timings are

so to reduce the timings on memory, you would need to go something like this

4-4-4-12 to say 4-4-4-15!!!

Lowering the number in the timings puts more of a stress on the memory

[Capper]

Acutally Stinger you have it backwards, he is tightening the timings....i think a little too much....4-4-4-12 is tighter than 4-4-4-15, as those numbers represent clock cycles for memory executons.


Run those tests at 32m to make sure you are stable first....unstable memory tends to act up.

A good rule of thumb is to repeatedly run Memtest86+ v1.70 at least 20 cycles to test the memory for stability at a given setting, then boot into Windows and run benchmarks like Super Pi to further test for stability.

Also, to get a good, honest benchmark, reboot your machine between tests, and turn off all unnecessary applications....hidden/or background apps will influence test results as they use system resources

[Pbirmingham1988]

I ran these tests under memtest once and it passed. Twenty to me is about 6 hours with about 20 minute tests. Do you think my oc could be the result of my memory acting weird??

I'm just wondering why super pi seems exceptionally well at first and then a few minutes go by and 1mb timings are increased by 7-8 seconds!! It's getting me so confused as to why. No matter how tight i made the timings it would always average around 22-23 seconds...

If I increase the memory freq to lets say 1066 @ 5-5-5-15 (default) doesn't seem to be that good because in super pi 1mb takes up to 24-25 seconds. In the bios when I put optimal timings it says weird numbers like 6-6-6-17 instead of what the stock timings are @ 1066... My bios doesnt read my memory @ 1066 unless I make it unlinked and set it there. I'm thinking my board doesnt like the Corsair XMS2 8500

[Capper]

Originally Posted by Pbirmingham1988

I ran these tests under memtest once and it passed. Twenty to me is about 6 hours with about 20 minute tests. Do you think my oc could be the result of my memory acting weird??

I'm just wondering why super pi seems exceptionally well at first and then a few minutes go by and 1mb timings are increased by 7-8 seconds!! It's getting me so confused as to why. No matter how tight i made the timings it would always average around 22-23 seconds...

If I increase the memory freq to lets say 1066 @ 5-5-5-15 (default) doesn't seem to be that good because in super pi 1mb takes up to 24-25 seconds. In the bios when I put optimal timings it says weird numbers like 6-6-6-17 instead of what the stock timings are @ 1066... My bios doesnt read my memory @ 1066 unless I make it unlinked and set it there. I'm thinking my board doesnt like the Corsair XMS2 8500

I wish sometimes you would slow down, take your time,and do things the right way. You seem to be in too much of a hurry, and are pushing your system pretty hard.....combined with your "stubborness" and I think you are headed for disaster. I think you get way too caught up in benchmarking and pushing your system to the very edge, and aren't really listening to us when we are trying to help you. I know you are having fun, and are probably learning a lot.

First, Benchmark results are always going to fluctuate, even on a fresh Windows install with nothing running in the background, results can fluctuate quite a bit....which is why we run tests over and over and over....and do them under different conditions. Run Super Pi until you get consistent results.

One pass of memtest does not mean your system is 100% stable, and there is a reason testing is time consuming.....because it needs to be thorough, and accurate. I know you want to use your PC......so set it up to run just as you go to bed, and let it run from the time you go to sleep, to the time you wake up........unless you are like me and don't sleep at all. Have you tried running Stress Prime to make sure your overall system setup is stable?

As to timings....the tighter latencies are always going to give better results....Meaning if you compare 4-4-4-12 and 5-5-5-15 at the same FSB speed, 4-4-4-12 will run much better. However, when you look at system performance as a whole, the most important thing to remember is the CPU speed, which trumps everything....the best possible scenario is to run your processor and memory as close to 1:1 as possible, while remaining 100% stable........however, if you have slower memory (You don't) running a divider with the tightest timings possible is best.

As to the memory and the motherboard, most motherboard don't recognize memory correctly during the initial setup, or with the BIOS set to AUTO.......which is a good thing as enthusiast memory settings typically are out of JEDEC specs, and with motherboards set to recognize default speeds (Who would want their motherboard to automatically void their warranty by OCing their components without their approval?) many systems wouldn't boot.....so to be on the safe side, many boards automatically boot at very conservative settings, and you then have the option to tinker with it yourself.....so in short, there is nothing wrong with your memory.

[stinger608]

Originally Posted by Capper

Acutally Stinger you have it backwards, he is tightening the timings....i think a little too much....4-4-4-12 is tighter than 4-4-4-15, as those numbers represent clock cycles for memory executons.

That is exactly what I was trying to say, that 4-4-4-12 is a tighter timing than say 4-4-4-15, thus putting more of a strain on the system memory.

[Capper]

Originally Posted by stinger608

That is exactly what I was trying to say, that 4-4-4-12 is a tighter timing than say 4-4-4-15, thus putting more of a strain on the system memory.

http://forums.hardwarelogic.com/54001-post15.html

Its alright Stinger, we understand if YOUR latencies are off a little.....

[stinger608]

Originally Posted by Capper

http://forums.hardwarelogic.com/54001-post15.html

Its alright Stinger, we understand if YOUR latencies are off a little.....

your just so damn funny

[Pbirmingham1988]

Originally Posted by Capper

I wish sometimes you would slow down, take your time,and do things the right way. You seem to be in too much of a hurry, and are pushing your system pretty hard.....combined with your "stubborness" and I think you are headed for disaster. I think you get way too caught up in benchmarking and pushing your system to the very edge, and aren't really listening to us when we are trying to help you. I know you are having fun, and are probably learning a lot.

First, Benchmark results are always going to fluctuate, even on a fresh Windows install with nothing running in the background, results can fluctuate quite a bit....which is why we run tests over and over and over....and do them under different conditions. Run Super Pi until you get consistent results.

One pass of memtest does not mean your system is 100% stable, and there is a reason testing is time consuming.....because it needs to be thorough, and accurate. I know you want to use your PC......so set it up to run just as you go to bed, and let it run from the time you go to sleep, to the time you wake up........unless you are like me and don't sleep at all. Have you tried running Stress Prime to make sure your overall system setup is stable?

As to timings....the tighter latencies are always going to give better results....Meaning if you compare 4-4-4-12 and 5-5-5-15 at the same FSB speed, 4-4-4-12 will run much better. However, when you look at system performance as a whole, the most important thing to remember is the CPU speed, which trumps everything....the best possible scenario is to run your processor and memory as close to 1:1 as possible, while remaining 100% stable........however, if you have slower memory (You don't) running a divider with the tightest timings possible is best.

As to the memory and the motherboard, most motherboard don't recognize memory correctly during the initial setup, or with the BIOS set to AUTO.......which is a good thing as enthusiast memory settings typically are out of JEDEC specs, and with motherboards set to recognize default speeds (Who would want their motherboard to automatically void their warranty by OCing their components without their approval?) many systems wouldn't boot.....so to be on the safe side, many boards automatically boot at very conservative settings, and you then have the option to tinker with it yourself.....so in short, there is nothing wrong with your memory.

Well put Casper. I agree that I am stubborn, but I'll tell you why I wanted my fsb @ 1600 to make it 3.6ghz. It was the only time I ever saw my fsb:ram @ 1:1. I was thinking, what if I made my multiplier lets say x8 but increase frequencies like 1700 or hell leave it at 1600 for 3.2ghz.

Basically, the only reason why I left my fsb frequencies so high is because I wanted the ram to be equal, but I was scared to use the multiplier... For reasons I thought would make the computer extremely slower, and I was scared to increase fsb speeds due to the fact I might damage my hardware(cpu).

So anyways when I get back the pc I will retest and tone it down a notch. What yall think?