Project 65xx units are too slow?

[AgrFan]

I've been getting a steady streem of 65xx units lately on a dedicated single core AMD Sempron box. My PPD has been reduced by 25% from 100PPD or higher to 75PPD with these units. The machine specs are: AMD Sempron 2500 (Skt-A, T-bird/B, 1.75Ghz, SSE), 1 GB RAM, Windows XP.

The reduced production seems to be not worth the electricity and the bother.

Any advice on when to retire/upgrade a machine? I know electricity costs are a big factor. This machine uses 150 watts under load and puts out alot of heat. Does CPU architecture play a role in the decision (Intel vs AMD, 32-bit vs 64-bit, instruction sets, cache size, etc)? It appears dual/quad cores released in the last couple of years provide the most bang for the buck. I have three dual cores running A3 units and they put out decent numbers.

Maybe the points for the 65xx units need to be adjusted?

[P5-133XL]

If the machine is useful to you on its own, then just let it keep folding. That being said, it is hard to justify folding on it, just for foldings sake. Modern machines are much more power effective. Upgrading it, may pay for itself over the long haul. How about an i7 920 getting 25,000-30,000 PPD while only using 250W...

[toTOW]

Strange, the PPD of the 6501 on my P4-m 1.4 GHz is normal for no bonus WU :

Code: Select all

Projet : 6501
 Core    : Gromacs
 Frames  : 100
 Credit  : 123


 -- P4-m --

 Min. Temps / Frame : 32mn 05s  - 55.21 ppd
 Avg. Temps / Frame : 32mn 11s  - 55.03 ppd

[jrweiss]

My Q9450 is crunching a 6502 and 3 6503s right now at 4-5 min/frame and 210 PPD each. My T9400 laptop has one of each at 196 and 199 PPD. I suppose that's about right, since the Q9650 running an A3 WU on all 4 cores shows 1054 PPD without bonus points...

[AgrFan]

FAHMon is showing <75PPD. This machine's days are numbered for folding I'm afraid. I can get more output using less power with a Intel Atom setup.

Project : 6503
Core : Gromacs
Frames : 100
Credit : 75

-- s2500 --
Min. Time / Frame : 13mn 35s - 79.51 ppd
Avg. Time / Frame : 14mn 32s - 74.31 ppd
Cur. Time / Frame : 14mn 38s - 73.80 ppd
R3F. Time / Frame : 14mn 38s - 73.80 ppd
Eff. Time / Frame : 14mn 42s - 73.47 ppd
Eff. Time / Frame : 14mn 30s - 74.48 ppd

[Wrish]

I generally discourage people from running any dedicated crunching hardware over 4 years old. 150 watts at 15 cents a killowatt-hour is almost $200 per year. You could buy a second-hand dual core system for that amount... and put out over ten times the points per day (~1400 for a DC Athlon64 3 GHz). Better yet, $200 would net you an even higher-yielding GPU that you could probably just add to one of your dedicated dual-core systems.

Currently, the Intel CPUs (starting with Core 2) put out considerably more points than the AMD chips but are priced higher as well. I expect this trend to continue for the next few years because of the past record with compilers and seeing Intel/AMD FPU roadmaps. Pretty much any x86 CPU in the last four years is 64-bit capable. The usual folding performance hierarchy is full i7's at the top; regular quads, 4-thread i5's, and Nvidia GPUs in the middle, and dual-cores and Atoms at the bottom.

[wan.ardiles]

Code: Select all

 Project : 6503
 Core    : Gromacs
 Frames  : 100
 Credit  : 75

 -- 1 --

 Min. Time / Frame : 10mn 27s  - 103.35 ppd
 Avg. Time / Frame : 11mn 44s  - 92.05 ppd
 Cur. Time / Frame : 10mn 31s  - 102.69 ppd
 R3F. Time / Frame : 10mn 46s  - 100.31 ppd
 Eff. Time / Frame : 11mn 02s  - 97.89 ppd

Currently running 6503 on P4 2.8GHz 1GB RAM w/ XP SP3. Looks normal to me.

[toTOW]

It's a bit less than what this CPU should get ... you should see 110 PPD if it's a dedicated CPU ...

[AgrFan]

My P4 2.53ghz box gets 100PPD on the 6503 units. My AMD box gets 75PPD on the same units. Could this be a AMD vs Intel compiler difference? That may explain why my AMD chip is getting 25% less than similar Intel chips.

Project : 6503
Core : Gromacs
Frames : 100
Credit : 75


-- Dell 4550 --

Min. Time / Frame : 10mn 29s - 103.02 ppd
Avg. Time / Frame : 10mn 31s - 102.69 ppd
No Cur. Time / Frame
No R3F. Time / Frame
No Eff. Time / Frame


-- s2500 --

Min. Time / Frame : 13mn 35s - 79.51 ppd
Avg. Time / Frame : 14mn 44s - 73.30 ppd
No Cur. Time / Frame
No R3F. Time / Frame
No Eff. Time / Frame

[7im]

No, sorry, Gromacs has been hand coded to avoid any compiler issues between chip makers. Intel and AMD have been almost identical in speeds on the CPU client for many years.

I supposed something could have changed in Gromacs, but it's unlikely.

[Flathead74]

P4-M @ 2.2GHz

6501
Min. Time / Frame : 00:21:16 - 83 PPD
Avg. Time / Frame : 00:22:35 - 78 PPD

p6503
Min. Time / Frame : 00:12:30 - 86 PPD
Avg. Time / Frame : 00:13:24 - 81 PPD

[Wrish]

That Sempron is a reduced-cache Athlon at 1750 MHz, so I don't see any surprises with the low and somewhat unsteady PPD.

With three dual-cores folding, why run the legacy systems at all? Save some electricity, save up for a more advanced system.