Large WU for non smp clients?

[conway]

Is it worth the time to use large work units or advanced methods(either/or)? I have some spare p4/axp/754 a64 machines I fold with, and want to get the best ppd I can with them. Are those flags beneficial, or should I just leave them as is? They seem to only get the smaller 186 point Gromacs units. Lately it seems as if they are ONLY getting project 3040.

 

[dfonda]

I know when I ran that client I always chose yes to large work units and used the Advanced methods flag in case something came up for it. Not so sure there is anything using the advanced flag right now. I never saw where it would anything to use it tho.

I know I am not using it on the SMP client until someone tells me otherwise.

 

[ChasR]

Bonus WUs are all large WUs so answer yes in config to accept WUs > 5MB. You'll get at least 60% better production than you will answering no, but only when bonus work is available. -advmethods at times gives you bonus WUs earlier than you'd get them without it, but you may also run out of them earlier when they leave -advmethods. -advmethods is a flag you add or subtract depending on the WU assignments of the time.
Large WUs are either large downloads or use large amounts of memory (>100MB) or both.

 

[conway]

Thanks for the help.

 

[conway]

So after converting my nonsmp folders to large workunits, fahmon reports that my ppd is down. Should I stick with it, or turn it off?

 

[ChasR]

THe only chance you have of getting bonus work is with large WUs. On average, you'll do much better with with large WUs on than off and you get the same WUs with it on that you get with it off when there aren't any large WUs being assigned.

 

[conway]

THe only chance you have of getting bonus work is with large WUs. On average, you'll do much better with with large WUs on than off and you get the same WUs with it on that you get with it off when there aren't any large WUs being assigned.

I guess I'll stick with it then, just hate to see a drop in production when I was aiming for a gain. We're not all lucky enough to have several quads laying around.

 

[ChasR]

The way bigpackets (large WUs) works is as an acceptance flag. You're saying I'll accept Large WUs if they are available. If there aren't any available, you get the same WUs you'd have gotten without the setting. The only time I've ever seen it be a disadvantage points wise to accept large WUs is when there were large quantities of Double Gromacs WUs available and thats not the case now nor has it been for over two years. (Double Gromacs aren't bonus WUs but if your cpu has SSE2, it will make more points per day than most bonus work).

 

[ChasR]

Production examples for large WUs and regular WUs from FAHmon:

A ribosome series Large WU:

Project : 1169
Core    : Gromacs
Frames  : 100
Credit  : 574


 -- WS-Kristy XP-M @ 2.4 --

Min. Time / Frame : 31mn 31s - 262.26 ppd
Avg. Time / Frame : 31mn 33s - 261.98 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Marie XP-M @ 2.4 --

Min. Time / Frame : 31mn 24s - 263.24 ppd
Avg. Time / Frame : 31mn 35s - 261.71 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Clealand_1 P4c 3.2 HT on --

Min. Time / Frame : 46mn 11s - 178.97 ppd
Avg. Time / Frame : 49mn 28s - 167.09 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Clealand_2 P4c 3.2 HT on --

Min. Time / Frame : 47mn 56s - 172.44 ppd
Avg. Time / Frame : 48mn 15s - 171.31 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Ben XP-M @ 2.3 --

Min. Time / Frame : 32mn 00s - 258.30 ppd
Avg. Time / Frame : 32mn 31s - 254.20 ppd
No Cur. Time / Frame
No R3F. Time / Frame



 -- Comp-Room P3 800 --

Min. Time / Frame : 1h 42mn 06s - 80.96 ppd
Avg. Time / Frame : 1h 42mn 27s - 80.68 ppd
No Cur. Time / Frame
No R3F. Time / Frame

A membrane series Large WU:

Project : 2620
Core    : Gromacs
Frames  : 100
Credit  : 292


 -- WS-Kristy XP-M @ 2.4 --

Min. Time / Frame : 16mn 32s - 254.32 ppd
Avg. Time / Frame : 16mn 33s - 254.07 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Marie XP-M @ 2.4 --

Min. Time / Frame : 17mn 53s - 235.12 ppd
Avg. Time / Frame : 17mn 53s - 235.12 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Clealand_2 P4c 3.2 HT on --

Min. Time / Frame : 22mn 56s - 183.35 ppd
Avg. Time / Frame : 22mn 57s - 183.22 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Ben XP-M @ 2.3 --

Min. Time / Frame : 18mn 12s - 231.03 ppd
Avg. Time / Frame : 18mn 15s - 230.40 ppd
Cur. Time / Frame : 18mn 16s - 230.19 ppd
R3F. Time / Frame : 18mn 15s - 230.40 ppd

A normal Gromac on the slow side of normal production:

Project : 3402
Core    : Gromacs
Frames  : 100
Credit  : 206


 -- WS-Kristy XP-M @ 2.4 --

Min. Time / Frame : 27mn 34s - 107.61 ppd
Avg. Time / Frame : 27mn 41s - 107.15 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Marie XP-M @ 2.4 --

Min. Time / Frame : 27mn 35s - 107.54 ppd
Avg. Time / Frame : 27mn 43s - 107.03 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Clealand_1 P4c 3.2 HT on --

Min. Time / Frame : 36mn 53s - 80.43 ppd
Avg. Time / Frame : 39mn 14s - 75.61 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Clealand_2 P4c 3.2 HT on --

Min. Time / Frame : 40mn 50s - 72.65 ppd
Avg. Time / Frame : 1h 02mn 48s - 47.24 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-Ben XP-M @ 2.3 --

Min. Time / Frame : 28mn 39s - 103.54 ppd
Avg. Time / Frame : 28mn 45s - 103.18 ppd
Cur. Time / Frame : 29mn 12s - 101.59 ppd
R3F. Time / Frame : 29mn 16s - 101.36 ppd


 -- Comp-Room P3 800 --

Min. Time / Frame : 1h 41mn 02s - 29.36 ppd
Avg. Time / Frame : 1h 43mn 48s - 28.58 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- WS-24 XP-M @ 2.4 --

Min. Time / Frame : 27mn 38s - 107.35 ppd
Avg. Time / Frame : 27mn 38s - 107.35 ppd
No Cur. Time / Frame
No R3F. Time / Frame


 -- CR-Linux5 SMP Q6600 @ 3.312 --

Min. Time / Frame : 12mn 15s - 242.16 ppd
Avg. Time / Frame : 12mn 15s - 242.16 ppd
No Cur. Time / Frame
No R3F. Time / Frame

Production on regular WUs varies from WU to WU and if you have a machine different from the benchmark machine (P4c @ 2.8) your production will vary. Your production going down is due to this variability. For instance p3402 is particularly slow on XP-Ms. but makes 40 ppd/GHz on my P4c (single instance), same as the benchmark machine. All my machines aren't Quads, in fact I only have 4 quads folding for ChasR..... ATM.

 

[NedClocker]

I don't have any quads folding.

 

[pscout]

I don't have any quads folding.

Well then fix it

 

[conway]

What a variable smorgasbord of information this thread has become. Thanks for the valuable insights. I'll leave it on.

 

[NedClocker]

What a variable smorgasbord of information this thread has become. Thanks for the valuable insights. I'll leave it on.

Flattery will get you everywhere!

 

[conway]

If it ever got me anywhere it would be a first for me.