I.M.O.G. dit :

11 year and 4 months ago –

I was looking to see how my score stacked up against the Pro's, despite my CPU being more average than awesome... Looking at efficiency by using the performance product. SP32M could be seen to separate the knowledgeable benchers from the less so, as its simple but elegant to compare against other scores. You can calculate an efficiency rating, and easily extrapolate scores with a good degree of accuracy.

 

If you don't know, you can estimate times in SuperPi by converting minutes to seconds, calculating performance product, then dividing by the MHz you want to estimate time for.

 

So with my tweaks and RAM settings (tightened secondary and tertiary timings), doing the calculations:

 

6.4GHz Ivy - My time was 5 minutes, 8 seconds

6.9GHz Ivy - This time would be 4 minutes, 46 seconds.

6.8GHz Ivy - This time would be 4 minutes, 50 seconds.

6.6GHz Ivy - This time would be 4 minutes, 59 seconds.

 

 

So this tells me a few things. My efficiency is in the right ballpark. But all of the Pro cup SP32M scores so far, but one, have beaten me twice. Beaten once by CPU frequency, and beaten as well by RAM settings. Phil is an exception, as his RAM might be running stock for his current submission. If he and I were running the same CPU clocks I'd be 2 or 3 seconds ahead.

 

Here's a comparison of current submissiosn ranked by Performance Product (PP), lower numbers mean better efficiency:

 

1. PP = 1959884 - Splave 4 minuts 47 seconds

2. PP = 1963601 - SF3D 4 minutes, 41.641 seconds

3. PP = 1966001 - Smoke 4 minutes 43 seconds

4. PP = 1978741 - I.M.O.G. 5 minutes 8 seconds

5. PP = 1994286 - Phil 4 minutes 53 seconds

6. PP = 2024610 - K404 7 minutes 57 seconds

7. PP = 2156912 - Oj0 6 minutes 54 seconds

 

So Splave is currently 6 seconds out of first place, but if he had that 6.972GHz Ivy SF3D has, he'd be in first place by half a second.

 

Anyways, my curiosity is satisfied. This isn't scientifically accurate, but it ballparks things pretty well. If you want to get better at SuperPi32M... Start paying attention to your Performance product, and keep making changes that make it lower. Do that, and once you do have god's own chip some day, you'll have what it takes to run it the way god intended.