Welcome to the team, John. If you run into anything that baffles you on folding, don't hesitate to post up the question. I know that you are pretty familiar with it already though, so
Fold On!
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
edited September 2003
Well, it helps with Linux, as the Beta (large RAM) client and the kernel think I have an SMP processor. Client happily loads 2 Core sessions, one of which sits there, as the CPU is not TRUELY an SMP processor-- rather it is one of the first P4's made and is an older core on a 478 package adapter fro a modenr 478 socket.
In Windows, with more RAM I end up regualrly with bigger WUs, with less, smaller onces and slower output as it can swap if RAM is really tight. Over 512 MB RAM in windows gives nothing, but I like ! GB in Linux. So long as WU work is never swapped and it can grab all the RAM it needs to uncompress WUs in RMA and then calc space to work on them, it is set. BUT, a box with 64-128 MB could well be swapping if it tries to run a big WU and things could bog. Client wqorkspace might get swapped, or WU uncompressed file, or the client or core itself or all the above. XP wants 256 MB to run BEST, before major apps. I get mostly 50-70 point WUs, which is why I am over 680 points with 11 WUs done since I restarted (I had 4 from before on Team-SM, about 170 points IIRC).
With 512 RAM in Linux, parts of Folding DID get swapped out to HD and things slowed down. If enough RAM for your normal apps+O\S+Folding, then fine and not much improvement with more RAM, though FASTER SPEED RAM might change things some. I went the route of figuring it this way:
O\S RAM
MAX use APP RAM
Folding for largest WU sent (a 50 day 70 pointer, done in about 1.8 days actual as it was a tinker). The WU alone uncompressed to aboutr 9 MB, and lessee what client has grabbed and core right now-- The FahCore_78 running on a WU now is using about 8 MB, the client itself about 2.2 MB. That, to my mind, is 20 MB used in Linux just for FAH, so your boxes, to be FULLY effective, need 20 MB RAM free roughly to handle big WUs.
Littler ones, more like 15-16 MB total with file area for uncompressed WU. But the average is getting bigger and bigger, so RAM will become more and more an issues, and DDR266 is slower than DDR333 and that is slower than DDR400-- RAM response time is a key and if you find your folding swapping then you have file reload time(HD->RAM) every time it is unswapped for active use.
CPU MHz is major factor, RAM below a certain level given apps you use a lot allloaded at once can cause folding to bog while things like video or motion graphics or digital photo processing are active. Adn I DO do video editing on the folding Barton, and accounting, and scheduling-- active box. And bogs add up to slower trhoughput overall. So, after CPU, gettign fastest RAM you can afford in minimum of twice to three times recommended base for O\S in use will improve your overall throughput of work including Folding as major RAM-hungry apps will not take priority over a compliant client which will get pended and swapped to HD if needed.
RAM of same type ALONE will not, faster RAM or RAM AND CPU will especially if RAM is faster and CPU has fewer wait states in embedded BIOS Setup due to values in RAM access parms.
John.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited September 2003
Welcome to the team, John. If you run into anything that baffles you on folding....
You're kidding, right?
0
Straight_ManGeeky, in my own wayNaples, FLIcrontian
edited October 2003
Well, not perfect, but in fact about 50 more points will show today, so... Last week's stats in production:
The other WU that should hit today is a p638. So, in the 800-825 point range for this 7 day period. Not QUITE what I'd hoped, but not gawdawful either for two single instance machines running FAH. My stats page is predicting 9 K by or on Dec. 15, 2003 and 10K by about Jan. 3-4 of 2004. Will see what can do about that.
Comments
Fold On!
In Windows, with more RAM I end up regualrly with bigger WUs, with less, smaller onces and slower output as it can swap if RAM is really tight. Over 512 MB RAM in windows gives nothing, but I like ! GB in Linux. So long as WU work is never swapped and it can grab all the RAM it needs to uncompress WUs in RMA and then calc space to work on them, it is set. BUT, a box with 64-128 MB could well be swapping if it tries to run a big WU and things could bog. Client wqorkspace might get swapped, or WU uncompressed file, or the client or core itself or all the above. XP wants 256 MB to run BEST, before major apps. I get mostly 50-70 point WUs, which is why I am over 680 points with 11 WUs done since I restarted (I had 4 from before on Team-SM, about 170 points IIRC).
With 512 RAM in Linux, parts of Folding DID get swapped out to HD and things slowed down. If enough RAM for your normal apps+O\S+Folding, then fine and not much improvement with more RAM, though FASTER SPEED RAM might change things some. I went the route of figuring it this way:
O\S RAM
MAX use APP RAM
Folding for largest WU sent (a 50 day 70 pointer, done in about 1.8 days actual as it was a tinker). The WU alone uncompressed to aboutr 9 MB, and lessee what client has grabbed and core right now-- The FahCore_78 running on a WU now is using about 8 MB, the client itself about 2.2 MB. That, to my mind, is 20 MB used in Linux just for FAH, so your boxes, to be FULLY effective, need 20 MB RAM free roughly to handle big WUs.
Littler ones, more like 15-16 MB total with file area for uncompressed WU. But the average is getting bigger and bigger, so RAM will become more and more an issues, and DDR266 is slower than DDR333 and that is slower than DDR400-- RAM response time is a key and if you find your folding swapping then you have file reload time(HD->RAM) every time it is unswapped for active use.
CPU MHz is major factor, RAM below a certain level given apps you use a lot allloaded at once can cause folding to bog while things like video or motion graphics or digital photo processing are active. Adn I DO do video editing on the folding Barton, and accounting, and scheduling-- active box. And bogs add up to slower trhoughput overall. So, after CPU, gettign fastest RAM you can afford in minimum of twice to three times recommended base for O\S in use will improve your overall throughput of work including Folding as major RAM-hungry apps will not take priority over a compliant client which will get pended and swapped to HD if needed.
RAM of same type ALONE will not, faster RAM or RAM AND CPU will especially if RAM is faster and CPU has fewer wait states in embedded BIOS Setup due to values in RAM access parms.
John.
You're kidding, right?
Last 7 Days Production:
Time
Day Total
Points
WUs
10/06/03 51.40 1
10/05/03 122.30 2
10/04/03 51.40 1
10/03/03 122.30 2
10/02/03 173.70 3
10/01/03 51.40 1
09/30/03 122.30 2
The other WU that should hit today is a p638. So, in the 800-825 point range for this 7 day period. Not QUITE what I'd hoped, but not gawdawful either for two single instance machines running FAH. My stats page is predicting 9 K by or on Dec. 15, 2003 and 10K by about Jan. 3-4 of 2004. Will see what can do about that.
FOLD ON, EVERYONE!
John-- who just folds.