Best way to benchmark performance?

[bovinda]

I got John's post on how to test for stability using Prime 95 and Memtest86+. But how do you benchmark performance? I just finished building a computer and am looking for ways to benchmark it before I start trying to OC.

(I know synthetic benchmarks may or may not indicate realworld performance. But I'm looking to sidestep that debate. I just want to get some objective indication of changes, before and after tweaking and OC'ing.)

Some programs I am considering using:

3DMark05 and/or later
Cinebench
OpenSourceMark

That's all I got. Any others? What methods or programs do you recommend, or not?

(If it matters, this will be on a Vista 64 system. But any and all ideas are appreciated!)

 

[myocardia]

For CPU's, I like SuperPi, NuclearMC, and either CPUMark 2.1 or CPUMark 99.

 

[JustaGeek]

3DMark06

PCMark Vantage (Vista 32 and 64)

Everest by Lavalys

SiSoft SANDRA (excellent Bandwidth tests)

 

[bovinda]

Thanks guys, when I get a chance this weekend, I will utilize those programs you suggested!

 

[Duvie]

I know you wanted to sidestep the argument of real world versus synthetics....but....other then stroking e-penis most synthetic benchmarks are about worthless...

I say test it with things you plan on running....

I use CAD apps, video encoding apps, etc...

cinebench is a pretty good indicator but is far more multithreaded in rendering then most real world apps....so it can be deceiving

 

[tigersty1e]

Super Pi.

1M and 2M benchies are really short. This will give you a good indicator of whether or not your OC is really in fact making your system faster.

 

[JustaGeek]

Originally posted by: Duvie

I say test it with things you plan on running....

I use CAD apps, video encoding apps, etc...

cinebench is a pretty good indicator but is far more multithreaded in rendering then most real world apps....so it can be deceiving

Absolutely!

But only the "real benchmark" programs allow you to compare the performance of your computer to other configurations.

However "synthetic" they might be...

 

[LOUISSSSS]

best way to benchmark is to actually do the things you normally do with your machine. if u can't actually feel a difference in speed, then you'll realize that overclocking gives you minimal real-world benefits...

 

[demiurge3141]

folding

 

[LOUISSSSS]

Originally posted by: demiurge3141
folding

how does folding benchmark a computer?

 

[tigersty1e]

Originally posted by: LOUISSSSS
best way to benchmark is to actually do the things you normally do with your machine. if u can't actually feel a difference in speed, then you'll realize that overclocking gives you minimal real-world benefits...

What kind of tasks are you doing that you'll notice a difference in speed?

BTW, this is a bad and inefficient way of measuring your overclock. So many factors affect real-world applications... especially the hard drive. When your OCing the CPU, you only want to test the CPU.

 

[LOUISSSSS]

Originally posted by: tigersty1e

Originally posted by: LOUISSSSS
best way to benchmark is to actually do the things you normally do with your machine. if u can't actually feel a difference in speed, then you'll realize that overclocking gives you minimal real-world benefits...

What kind of tasks are you doing that you'll notice a difference in speed?

BTW, this is a bad and inefficient way of measuring your overclock. So many factors affect real-world applications... especially the hard drive. When your OCing the CPU, you only want to test the CPU.

thats what i was kind of implying. that u barely notice any real world performance increases with an overclock

 

[tigersty1e]

Originally posted by: LOUISSSSS

Originally posted by: tigersty1e

Originally posted by: LOUISSSSS
best way to benchmark is to actually do the things you normally do with your machine. if u can't actually feel a difference in speed, then you'll realize that overclocking gives you minimal real-world benefits...

What kind of tasks are you doing that you'll notice a difference in speed?

BTW, this is a bad and inefficient way of measuring your overclock. So many factors affect real-world applications... especially the hard drive. When your OCing the CPU, you only want to test the CPU.

thats what i was kind of implying. that u barely notice any real world performance increases with an overclock

I see.

Would you classify games as real-world performance?

I've actually noticed a very nice performance boost from overclocking my cpu from 1.86 to 3.15. Considering it's a 70% overclock, I should.

 

[Conjugal Visit]

Originally posted by: LOUISSSSS

Originally posted by: tigersty1e

Originally posted by: LOUISSSSS
best way to benchmark is to actually do the things you normally do with your machine. if u can't actually feel a difference in speed, then you'll realize that overclocking gives you minimal real-world benefits...

What kind of tasks are you doing that you'll notice a difference in speed?

BTW, this is a bad and inefficient way of measuring your overclock. So many factors affect real-world applications... especially the hard drive. When your OCing the CPU, you only want to test the CPU.

thats what i was kind of implying. that u barely notice any real world performance increases with an overclock

Seriously, Louisss - as someone above me just said, real world apps definitely benefit from overclocks. You know that, yes? A whole lot of overclockers are gamers on the side, and I can tell you that overclocking the big three - mem, CPU, and vid card - have real, tangible benefits. We ain't talkin' web surf and word processing here....

Regards,

 

[j0j081]

I used to go for the more complex route but I think I've figured out a reliable way to measure real world performance. You MUST be running Windows Vista to do this. Go to your start menu, click on the all programs button, find and open the games folder, then launch solitaire. Carefully examine how smoothly the cards are placed out for you. Take detailed notes of your findings in your favorite text program (I use wordpad), then make the changes you want to your system/oc and repeat. Hint* The smoother the better.

 

[ajemm]

PerformanceTest tests all the subsytems and does it in only a few minutes.

 

[tigersty1e]

Originally posted by: j0j081
I used to go for the more complex route but I think I've figured out a reliable way to measure real world performance. You MUST be running Windows Vista to do this. Go to your start menu, click on the all programs button, find and open the games folder, then launch solitaire. Carefully examine how smoothly the cards are placed out for you. Take detailed notes of your findings in your favorite text program (I use wordpad), then make the changes you want to your system/oc and repeat. Hint* The smoother the better.

That's really inefficeint.

A better way would be to set your mouse speed to the lowest and then intuitively "measure" how long it takes for you to move your mouse pointer from one end of the screen to the other.

 

[BoboKatt]

Best way I always test is to use Winrar and see how long it takes me to compress or uncompress a 4.8GB ISO, or how long it takes for a program like quickpar which rebuilds lost or missing files from archives (again for a 4.8GB movie file). I can always tell the diffeernce between builds with these.

As for 3dmark variants, I can't stand them. I was so into them trying to get a few extra points here and there I never really noted any difference in performance in daily use either way.