• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Standard vs Timeless

J

justakid

Weaksauce
Joined
Oct 3, 2005
Messages
91
Is it better to run standard wu's or timeless? I understand that in slower machines running timeless is a given because of how fast they finish, but I noticed timeless gives me a much more linear and stable ppd then standard. Another thing I noticed is standard wu's gives me a considerably lower ppd.

Example - On my main machine (sempon 3000@2.6ghz) I get about 225ppd with timeless and with standard I've seen anywhere from 130-195ppd, never matching timeless.

Also, does turning on large work units help at all? I've gotten a couple of 400 pointers or so, but they ran soooooo slow. I dunno... do the standard wu's run that much slower because of my small L2 cache? I have a sempron 2800@2.1Ghz folding too, I think I'm going to switch it over to standard wu's and see how it does. These are things YURT is supposed to help answer in the future no?
 
Lately timeless units have given better ppd. Especially AMDs since they are very efficient with Tinker cores (Pentium-Ms absolutely suck at them to my recollection). It used to be that standards would give a better PPD like the 1479 protien, but they have been few and far between lately so quite a few people have switched over to timeless. I've only do so on one core. I have seen about a 50PPD increase, averaged, for that core.
 
The number of points assigned to a WU is based on how long that WU takes to complete on a Pentium 4 2.4 GHz (IIRC). AMD's, with a different architecture, perform much better on Tinkers and somewhat worse on Gromacs. So if you're running Intel procs, timeless/normal won't make much of a ppd difference, but if you're running AMDs, there's a big difference.

The only problem is that Stanford's timeless WU server is getting toasted by all the people looking to boost their ppd, and those WUs aren't as valuable to Stanford as the normal ones. Most of my machines are AXPs, but I run normal units on them, just because they're better for Stanford.
 
Mohonri covered exactly what I was going to suggest.

If your system is fast enough to turn in standard wus, has SSE optimizations, then please run standard wus. They are the most important projects to Stanford(They give them a deadline), so we should be working to get the standard wus done, and leave the timeless to the slower machines.

 
Agreed. I have switched some of my boxen from running the timeless to running standard WU. I wanna make sure Marty's crapper farm keeps chugging along.
 
Also if you have enough ram, turning on big units and advmethods will give you a points boost over the normal units. Getting GBgro's on a sempy is better than standard units, but the new double gro's are really nice. If you've got 512 for a dedicated folder I'd turn both on, but if you're using it alot then have a gig before turning on those flags. That's just me tho.

 
Depends on how Stan4rd wants to tighten the screws.
The entire point debacle is fu%^ed up.
They may release WU and give them a 50% bonus, or not.
They release WU that are rated to take X cpu hours on their reference machine but take X+10 on everybody elses.
Just let your machines crunch.


 
all my home machines run the large wu and i don't run timeless on anything because the servers don't go down as oftern.
 
MN Scout said:
Mohonri covered exactly what I was going to suggest.

If your system is fast enough to turn in standard wus, has SSE optimizations, then please run standard wus. They are the most important projects to Stanford(They give them a deadline), so we should be working to get the standard wus done, and leave the timeless to the slower machines.

Click to expand...

Forgot to add the Genuine Intel part....specifically, INTEL will not let some of their optimisation code work on AMD processors (purely a liscensing issue).... Yes Intel sucks.
 
Hrm... I have a gig of ram so I guess I should turn on large wu's. The genuine Intel thing sucks, cause all I have right now are Amd's. Another thing is, standard wu's always mess up for me, I know I'm overclocked but I can prime for over a day without errors, game, do anything except fold those damn wu's that aren't timeless.
 
gnewbury said:
Depends on how Stan4rd wants to tighten the screws.
The entire point debacle is fu%^ed up.
They may release WU and give them a 50% bonus, or not.
They release WU that are rated to take X cpu hours on their reference machine but take X+10 on everybody elses.
Just let your machines crunch.


Click to expand...

I have to agree with you on this for sure. There simply is no parity in the scoring system at all. I “got into it” with V.J. over at the community forums about that situation and the fact that nobody at Stanford seems to be able to write code for the assignment server to send the appropriate units to the proper cpus.

The issue with the timeless sever is very real, it simply can’t handle the load.

For my own part if a machine is 2.0 gigs or above and supports SSE I set up for big units and add –forceasm and –advmethods to the command line. I noticed a jump in completed work units right away.

justakid said:
Hrm... I have a gig of ram so I guess I should turn on large wu's. The genuine Intel thing sucks, cause all I have right now are Amd's. Another thing is, standard wu's always mess up for me, I know I'm overclocked but I can prime for over a day without errors, game, do anything except fold those damn wu's that aren't timeless.
Click to expand...

After a very long time at this I have found sometimes the simple things are the easiest. The most vulnerable part of the folding program is the “core”. The core seems very sensitive to things such as defragging or bad shutdown etc.

I find that if I go through and delete the cores once and a while the problems you mentioned simply go away. Give it a shot, like the man says, it don’t cost nothing and it just might help you out.

Luck
 
BillR said:
For my own part if a machine is 2.0 gigs or above and supports SSE I set up for big units and add –forceasm and –advmethods to the command line. I noticed a jump in completed work units right away.
Click to expand...

I totally agree here with you bill. The low end P4's <2.0ghz aren't really that useful for crunching some of the newer gromacs that are worth 400ish points but don't get a bonus. Hence they take FOREVER and a day to fold them. Also any older athlons that are pre XP variety should definitely be set to run timeless. My old 1.4 athlon put out 110ppd on timeless, while my 2.4ghz p4 lappy only put out 83 on double gromacs.

 
You must log in or register to reply here.
Back
Top