Super PI Contest

[Duvie]

Originally posted by: michaelpatrick33

Originally posted by: Duvie

Originally posted by: coomar
it wouldn't be twice as fast, the best you could probably hope for is 90% efficiency, 2x boost is the ideal, most of the time in multi-threaded apps it will be 50-90% increase

I see we are speaking from experience knowledge...oops wait a minute...nope!!!

Tested at 2.574ghz

My speed with 1 instance = 5:59

My speed with 2 instances no affinity set = 6:16-6:18
My speed with 2 instances w/ affnity set = 6:19-6:21

so it looks like 2 instances was 94.6 to 93.1%...not 2x but above your ideal 90 and more then your 50-90 range!!!

screenies will be linked momentarily...


NOtice how I had one core better then the other....

I went from a 5:43 single instance score to a 5:57 dual instance score (5:57 for both superpi runs). Not too shabby.

I am going to test higher speeds.....I may tweak the memory a bit or too but I cant imagine I will get much better then a 5:40ish for my single speed...

 

[chinkgai]

ill post my time tnite or something.

school started...been pwning me hard

 

[imported_michaelpatrick33]

Originally posted by: Duvie

Originally posted by: michaelpatrick33

Originally posted by: Duvie

Originally posted by: coomar
it wouldn't be twice as fast, the best you could probably hope for is 90% efficiency, 2x boost is the ideal, most of the time in multi-threaded apps it will be 50-90% increase

I see we are speaking from experience knowledge...oops wait a minute...nope!!!

Tested at 2.574ghz

My speed with 1 instance = 5:59

My speed with 2 instances no affinity set = 6:16-6:18
My speed with 2 instances w/ affnity set = 6:19-6:21

so it looks like 2 instances was 94.6 to 93.1%...not 2x but above your ideal 90 and more then your 50-90 range!!!

screenies will be linked momentarily...


NOtice how I had one core better then the other....

I went from a 5:43 single instance score to a 5:57 dual instance score (5:57 for both superpi runs). Not too shabby.

I am going to test higher speeds.....I may tweak the memory a bit or too but I cant imagine I will get much better then a 5:40ish for my single speed...

I can get into Windows at 2850 or so on my X2 but the second core isn't stable. Too bad, LOL. My memory is nasty when it comes to dividers but maybe I will tinker a bit, LOL.

 

[Duvie]

Originally posted by: Duvie

Originally posted by: coomar
it wouldn't be twice as fast, the best you could probably hope for is 90% efficiency, 2x boost is the ideal, most of the time in multi-threaded apps it will be 50-90% increase

I see we are speaking from experience knowledge...oops wait a minute...nope!!!

Tested at 2.574ghz

My speed with 1 instance = 5:59

My speed with 2 instances no affinity set =6:16-6:18
My speed with 2 instances w/ affnity set = 6:19-6:21
so it looks like 2 instances was 94.6 to 93.1%...not 2x but above your ideal 90 and more then your 50-90 range!!!

screenies will be linked momentarily...


NOtice how I had one core better then the other....

proof???

SSE2 gets me 6:00

Yeah thanks for SSE3.... :roll:

 

[McGeyser]

Shenkoa,

I noticed in your first post or the results table, that michaelpatrick33's score is listed wrong. He actually acheived 5m 43s single core, 5m 57s dual core simultaneous.

I had 5m 52s single run and 6m 04s for a dual simultaneous calculation.

 

[coomar]

duvie i was talking about a multithreaded program for one calculation, not each core running an instance of the program

for that it would be mostly shared bandwith loss but superpi doesn't use that much ram

 

[Perpetual]

Shenkoa -

Hey, I was wondering if you could change my DDR time it's down as 2-2-5 and should be down as 2-2-2 or 2-2-2-5

Thanks man!

 

[Shenkoa]

I am sorry, I had someone else update the scores because I was very busy today and as you can see, he did not bother to look twice nor did that someone list RAM frequency.

I will have a word with that mofo.

 

[dnavarro]

Well here are my results on my X23800+ @3Gigs...

5mins 10 secs (single) and 5mins 22secs (dual)



SSE3 8M SuperPi Dual run
SSE3 8M SuperPi



X2's rock

D

 

[Shenkoa]

Originally posted by: dnavarro
Well here are my results on my X23800+ @3Gigs...

5mins 10 secs (single) and 5mins 22secs (dual)



SSE3 8M SuperPi Dual run
SSE3 8M SuperPi



X2's rock

D

Dnavarro, can you list your system specs please and also is there any way we can get a higher resolution picture

 

[chinkgai]

rofl, my cell phone has higher res

 

[Duvie]

SUPERPI SSE3 8MB SINGLE CORE 2640MHZ 233 DIVIDER CAS 3-4-4-10 = 5:48

SUPERPI SSE3 8MB DUAL 2640MHZ 233 DIVIDER CAS 3-4-4-10 = 6:00-5:59

 

[chinkgai]

here are my results:

single instance - 5:05

dual intances - 5:12

4400 @ 2894
ocz vx 4000

 

[Shenkoa]

And taking number 1 spot for AMD Dual Core. Thats one hell of a system you got there!

 

[chinkgai]

Originally posted by: Shenkoa
And taking number 1 spot for AMD Dual Core. Thats one hell of a system you got there!

thanks
and single too, but who's counting

 

[dnavarro]

Guys, I would post the pics directly in this thread, but they are at 1920x1200. (I have been warnerd not to post large pics before). You should be able to download them from the link however, and view them up close.

System specs:

DFI NF4 Ultra D modded to SLI
Gskill 4400LE @ 300MHz 2-4-3-8 2.9V
X2 3800+ @ 3.0Ghz 1.6V Vapo LS cooled
Dual Raptors RAID 0
PCP&C 510 SLI Power Supply

Also, I am still tweaking this setup and I should have 3.1Gig stable at SuperPi 32m. I'll post my 8m results tonight hopefully

D

SSE3 8M SuperPi Dual run
SSE3 8M SuperPi

 

[chinkgai]

please make it prime stable...as many of us can post times at higher speeds that arent close to prime stable.

nice chip btw

 

[Shenkoa]

The point is to try to get the best times by any means.

 

[Duvie]

mine are stable....otherwise I can suicide come much tighter timings and a 250 divider.....all in fun though....

 

[chinkgai]

if its by any means, we should all get ram drive....and run it from there...HUGE boost

i dont have ram drive tho, but i guess i could look...

 

[dnavarro]

Mine are prime stable also, for over 2 hours anyways (haven't tried more). But that is not the point of this thread. Also mine is dual SuperPi 32m stable which I always find to be a better judge of stability than Prime.

see here: (download the pic for a larger look)Dual SuperPi 32M Stable

D

 

[Duvie]

Originally posted by: dnavarro
Mine are prime stable also, for over 2 hours anyways (haven't tried more). But that is not the point of this thread. Also mine is dual SuperPi 32m stable which I always find to be a better judge of stability than Prime.

see here: (download the pic for a larger look)Dual SuperPi 32M Stable

D

quite the diff between the two...did you set affinity for each instance or not mess with it???

I am actually suprised it is not lower with the 600ddr ram speed you have as well....

 

[dnavarro]

Duvie, when you consider I was running 3Ghz DUAL 32m runs (yes affinity was set for each core) at a little over 25 minutes and my FX55@3150 on the same setup does 24 mins I think that is pretty good. see here:
FX55@3150MHz Super Pi 32m

Dual runs stresses the system much more and each core only has access to half the system memory (over 600MB). This brings the performance down a tad.

But a single 32m run with this chip brings it into the mid 24 minute range, very close to my (higher clocked) FX. And note the FX has a full meg of cache onboard while the 3800+ has only 512K per core (and SuperPi loves cache....see Dothan).

Also Super Pi loves LOW LATENCY memory. Any BH5 or VX at 2-2-2-5 timings in the 250-260Mhz range would be better score wise than my GSKILL TCCD @300Mhz (or 315Mhz for that matter). The TCCD runs at 2.5-3-3-7 timings at that speed.

Anyways I have done enough overclocking over the years to know these times are whats expected. This is not my only Super Pi contest

D

P.S. in MY personal Experience, I have had builds that can do over 12 hours prime stable and yet fail a Super Pi 32m run. Thats why I always end my testing with a 32m run as the final indicator of stability.

 

[Duvie]

Originally posted by: dnavarro
Duvie, when you consider I was running 3Ghz DUAL 32m runs (yes affinity was set for each core) at a little over 25 minutes and my FX55@3150 on the same setup does 24 mins I think that is pretty good. see here:
FX55@3150MHz Super Pi 32m

Dual runs stresses the system much more and each core only has access to half the system memory (over 600MB). This brings the performance down a tad.

But a single 32m run with this chip brings it into the mid 24 minute range, very close to my (higher clocked) FX. And note the FX has a full meg of cache onboard while the 3800+ has only 512K per core (and SuperPi loves cache....see Dothan).

Also Super Pi loves LOW LATENCY memory. Any BH5 or VX at 2-2-2-5 timings in the 250-260Mhz range would be better score wise than my GSKILL TCCD @300Mhz (or 315Mhz for that matter). The TCCD runs at 2.5-3-3-7 timings at that speed.

Anyways I have done enough overclocking over the years to know these times are whats expected. This is not my only Super Pi contest

D

P.S. in MY personal Experience, I have had builds that can do over 12 hours prime stable and yet fail a Super Pi 32m run. Thats why I always end my testing with a 32m run as the final indicator of stability.

That is not what I am talking about intially...I am wondering why one core got sooo much better then the other...up unti,l now ihave only see a few seconds but not as many as yours....

The second part I am conparing yours to Chinkgai who is running less....

 

[dnavarro]

Well Chinkgai has a 4400+ at only roughly 1 hundred MHZ less than mine. Yet he has 1Meg of L2 cache per core (Which AMD usually gives a +200 MHZ rating for that extra cache). And since Super Pi loves L2 cache you cn see why he got a better score. Also, he may be running lower latency memory, I can't see from his A64 Tweaker screenshot (it is not showing them correctly on there) but that is another factor that could be helping his score.

I should be there soon. I am priming 3150 Dual on this thing right now I just have a cache deficit to make up...

D