BOINC benchmark figures

[The Borg]

Hi all.

Some clarity please. This is what I did:

Gigabyte mobo with 939 3500+ processor
BOINC benchmarks (OC @ 2.64 up from 2.2 - 20%):
- 4200 Floating
- 13900 Integer

Bought a Foxconn mobo and a 939 X2 4800+
Put the X2 in the Gigabyte mobo (damn good mobo and in main machine) and the 3500+ in the Foxconn (to fit in the crack-rack)

X2 benchmarks (OC @ 2.63 up from 2.4 - 9.5%):
- 4400 Floating
- 14400 Integer
This is per CPU and exactly what I would have expected. I was always confused as to what I would get becasue of the 4800+ rating. This is actually closer to 7000+ (2 x 3500+) becasue both cores are used fully.

3500+ benchmarks (OC @ 2.57 up from 2.2 - 17%):
- 3150 Floating (!!!!)
- 13600 Integer.

So here is the query. Why have the Floating values dropped so much on the 3500+? There are only 2 main differences. One is the mobo, and the second is the fact I used 512 meg of the 2 x 512 in the Foxconn mobo. Need to upgrade the memory in the X2 now (got the other 512 meg), but that is going to be expensive.

 

[Philippart]

it's the mainboard.

many forget the effect of mainboards, they do a big part in performance!

I already had a few similar cases...

 

[petrusbroder]

I agree with Philippart. I have done some testing using the BOINC 5.2.9-version: the same mobo, CPUs with differnt frequencies: scales nicely... Then the same CPU with a different mobo: some 15% difference. Too low RAM slows the crunching too, but the we are talking about 256 Mbytes RAM vs 512 MBytes RAM (the latter is some 2 - 4% faster). Above 512 MBytes no difference.

 

[The Borg]

Hi guys,

Thanks for the comments, but I still don't understand why only the Floating score drops so much. The Integer score stays pretty much the same. Surely that is the result of the CPU? so how does the mobo affect that?

 

[petrusbroder]

Hmmm that was much harder - missed that part of the data.
Is the cooler the same?
Is the PSU the same?
Is the mobo's powersystem different?
(and should that affect the FPU)?
Food for thoughts. More research is needed. :Q

 

[The Borg]

Cooler the same. Actually this board is now sitting on a table with no case, so everything is exposed and should be cooler.

Different PSU.
Mobo's powersystem - ?

 

[petrusbroder]

I was thinking:
Has the Foxconn mobo the same powersystem onboard as the Gigabyte? Probably not. The power system has quite some impact on the stability of the mobo and especially of the CPU ... I do not know more - and my thinking may be wrong, but in many tests a mobo's capability to provide smooth and steady power to the different components.
Yes, the power comes from the PSU, buit a lot is done on the mobo, that where the "caps" come in ... I think.

 

[The Borg]

Well, I only have a 20 pin PSU, not the newer 24 pin, but the plug fitted and the mobo is working. The manual did say it was recommended (not a must) to have a 24 pin connection. I only think this would be necessary with high loads - lots of expansion cards and periferals connected. I don't have this. CPU, memory, a network connection, HDD and a fan.