Optimal CPU - GPU comparison table [moderate to larger WUs]

[X1900AIW]

Q6600 (9*333) @3,0GHz/1,17V (1x GPU2 @core0+1+2+3, 1x BOINC @core0, 1x BOINC @core1, 1x BOINC @core2, 1x BOINC @core3)
Sapphire HD3850/512MB (669 / 829 =stock) @PCIe x4 99% GPU activity (no GPU2 @PCIe x16)
Project: 4736 (Run 2, Clone 10, Gen 40)
Client Version 6.20, GPU Core - Beta Version 1.04 (Fri Jun 6 15:46:09 PDT 2008)
1925 PPD

Projekt : 4736
Core : GPUv2 Gromacs
Frames : 100
Punkte : 234


-- ID3_GPU2_systray_ATI_gpu1 --

Minimale Frame-Dauer : 1mn 45s - 1925,49 ppd
Gemittelte Frame-Dauer : 1mn 45s - 1925,49 ppd
Aktuelle Frame-Dauer : 1mn 47s - 1889,50 ppd
L3F-Frame-Dauer : 1mn 45s - 1925,49 ppd
Effektive Frame-Dauer : 1mn 45s - 1925,49 ppd

OC: Sapphire HD3850/512MB (799 / 829)
2150 PPD

-- ID3_GPU2_systray_ATI_gpu1 --

Minimale Frame-Dauer : 1mn 34s - 2150,81 ppd
Gemittelte Frame-Dauer : 1mn 35s - 2128,17 ppd
Aktuelle Frame-Dauer : 1mn 35s - 2150,17 ppd
L3F-Frame-Dauer : 1mn 34s - 2128,17 ppd
Effektive Frame-Dauer : 1mn 43s - 1962,87 ppd

[Karamiekos]

Project 4739 (R2, C16, G30)
Core Version 1.4

X2 3800+ @ 2.5GHz
4850 @ 670/1050
XP SP3

2128 PPD

[Karamiekos]

Same Computer as above but:

Project 4741 (R2, C12, G38)
Core Version 1.4
2128 PPD (again)

FAHMON set to last three frames.

[X1900AIW]

Q6600 (9*333) @3,0GHz/1,17V (2x GPU2 @core0+1+2+3, 1x BOINC @core0, 1x BOINC @core1, 1x BOINC @core2, 1x BOINC @core3)
Client Version 6.20, GPU Core - Beta Version 1.04 (Fri Jun 6 15:46:09 PDT 2008)
Edit: update with 4736 + 4729 Edit2: update with 4739 + 4736 Edit3: Core 1.10 with 4731 + 4727 / Edit4: update 4745 + 4727 (first BIG WU !)

gpu 0 @stock
Sapphire HD3850/512MB (669 / 829 =stock) @PCIe x16 / ~65-83% GPU activity / same fluctuation / / upward trend / ~80% / 99% GOOD !
Project: 4727 (Run 0, Clone 501, Gen 3) / Project: 4736 (Run 3, Clone 29, Gen 37) / Project: 4739 (Run 3, Clone 4, Gen 51) / Project: 4731 (Run 2, Clone 789, Gen 1) / Project: 4745 (Run 1, Clone 30, Gen 3)
1863 PPD / 1925 PPD / 2150 PPD / 1916 PPD / 1956 PPD

gpu 1 @796 overclocked
Sapphire HD3850/512MB (796 / 829 =stock) @PCIe x4 / ~73-91% GPU activity / same fluctuation / upward trend /~85% /37-40% Problem !
Project: 4736 (Run 3, Clone 15, Gen 39) / Project: 4729 (Run 1, Clone 19, Gen 38) / Project: 4736 (Run 1, Clone 28, Gen 22) / Project: 4727 (Run 1, Clone 670, Gen 5) / Project: 4727 (Run 1, Clone 837, Gen 5)
2150 PPD / 1775 PPD / 2197 PPD / 1764 PPD / 1764 PPD

together 4014 PPD / 3701 PPD / 4348 PPD /3681 PPD / 3721 PPD
> ups and downs, with higher GPU activity the performance is really satisfying (based on HD3850s)

/ / / /

____________________________________

Edit4: Trying to increase PPD with overclocking:

gpu 0 @796 / 999 overclocked
Sapphire HD3850/512MB (796 / 999 OC) @PCIe x16 / 99% - same activity, looks like CPU-clock is not sufficient; Edit6: Decrease to ~90% in the end.
Project: Project: 4745 (Run 1, Clone 30, Gen 3)
2006 PPD - slightly positive effect, with every FahMon-refresh a bit more ... (will edit last PPD when finished; Edit6: 2171 PPD = +200 PPD)

gpu 1 @796 overclocked (remember: did this as before to "regulate" differences between PCIe x4 and x16)
Sapphire HD3850/512MB (796 / 829 =stock) @PCIe x4 / 45-50% - "better"
Project: Project: 4727 (Run 1, Clone 837, Gen 5)
1764 PPD - same

Cconclusion:
- no more PPD with HD3850, in spite of overclocking - New WUs are made just for HD4000-series !?
- but heating the GPU up to 102 degree Celsius - with or without OC (see FAN speed in screenshots)
> old FAN speed ~53-55% is not enough to cool the GPU down to "normal" temperatures, with 60% a niveau at 92-93 degree can be reached (meantime 88-90 degree)

/ / /

Edit6: FahMon about the new 4745:

-- ID2_GPU2_systray_ATI_gpu0 --

Minimale Frame-Dauer : 3mn 24s - 2320.94 ppd
Gemittelte Frame-Dauer : 3mn 38s - 2171.89 ppd
Aktuelle Frame-Dauer : 3mn 51s - 2049.66 ppd
L3F-Frame-Dauer : 3mn 53s - 2032.07 ppd
Effektive Frame-Dauer : 3mn 50s - 2058.57 ppd

[Karamiekos]

With My same Setup I would confirm:

Project 4715
2151 PPD

[Karamiekos]

Same Setup

Project 4742 (R0, C56, G0)
3381ppd

[legoman666]

HD4850 @ stock
Q6600 @ 3.1
P4743
3100PPD
Core 1.10

[jrweiss]

HD3850
Proj 4742
2132 PPD
Core 1.10

[legoman666]

HD4850 @ stock
Q6600 @ 3.1
P4743
3100PPD
Core 1.10

same PPD with p4742

My hd2400 doesn't like the new WU's... 200.79PPD down from 315.

[Toxic Crow]

HD4870 @ 790/1050
Q9450 @ 3.2GHz
P4745
4046PPD
Core 1.10

I´ve got two of those nice P4745 now

[klefreak]

Hd 2900 Pro @ 2d clock --> 510mhz/510ram
c2d E6850 @ 3,6ghz
P4737 (r2c61g3)
PPD1742,9
Core 1.10
--> got up from 1500 to 1700ppd !!

[Karamiekos]

And update on 47xx projects.
I increased my 4850 clocks to 690/1125 and my ppd went up to 3507. GPU is still saturated at 99%. Before with older work units, the OC didn'ts seem to matter at all, cept for it dropped GPU usage from 94%to roughly 88%. Just and interesting observation, FYI.

[legoman666]

Heh, I tried increasing my 4850 to 690. It lasted for about 45 minutes before the machine locked up. Then when it rebooted, it was still at 690, so it locked up again. It did this a few times until I booted into safe mode and disabled the autostart of the GPU client. I then rebooted into normal mode. I had to try a few times before I could open CCC and reduce the clock speed before it locked up. PITA. It even locked up at 640. Suffice to say, I have the worst overclocking 4850 ever.

[Karamiekos]

Its always the luck of the draw. Mine has been stable for almost 24 hrs. but I am a skeptic, so I have been keeping a keen eye on it. What brand card do you have if you don't mind sharing.

[legoman666]

Asus apparently. Bought it 6/25, shortly after they were released.