 |
 |
Folding @ Home Tweaker's Guide (for the over 30 crowd only)
|
 |
|
 |
|
Senior User
Join Date: Oct 2002
Location: God's Country, The South
Status:
Offline
|
|
Remember these instructions are for the "Over 30" crowd to use, there will be a seperate posting for the other teams later. I will post instructions for the other teams on or around February 1, I promise
Maximum PC has an excellent article on how to maximize your performance of the folding client under Windows:
http://www.maximumpc.com/features/fe...003-06-10.html
His instructions are excellent, even though they are slightly out of date. Hence my addendums:
There is a new version of the client, version 4.0, since he wrote that article. Use it! If you have machines that have been folding for a while, stop the old version of the client and replace it with the new version. it is faster on everything that I have tested it on.
Read all the way down his article, it has some advanced tips on using the CLI client. The CLI client is definitely the way to go, it is much faster and has less system overhead if setup correctly.
Instead of using the flags that he specifies in step 5, use these:
if you are using an AMD processor:
-forcesse -advmethods -local
if you are using an Intel processor:
-forceasm -advmethods -local
If you can, I would recommend using a program called Firedaemon to start the folding client as a background service. That will be the topic of a future post. Scott has already done an excellent tutorial on using Firedaemon, if I can find that posting from a year or so ago I will repost it or a link to it.
|
|
|
| |
|
|
|
 |
|
 |
|
Administrator 
Join Date: May 2000
Location: California
Status:
Offline
|
|
As of the version 4 client on AMD, use both the -forceasm and -forcesse flags. The v4 client returns them to their intended uses.
-forceasm causes optimized assembly loops to be used.
-forcesse causes AMD processors that support SSE to use it instead of 3dNow (SSE is faster). This flag is ignored if optimized assembly loops are turned off, which is why you still want -forceasm. Note, keep an eye on AMD boxes using the -forcesse flag, especially if the boxes are overclocked. This flag has been known to make some AMD boxes unstable.
Also, keep using the -advmethods flag as OneMacGuy says. It is the way to get more Gromacs units, which give better points per day on modern CPUs. Tinker isn't too bad on x86, but Gromacs are still worth more per day. The project keeps threatening to reserve -advmethods for beta work units only, but in practice it still insures mostly Gromacs.
AMD
-forceasm -forcesse -advmethods -local
Intel
-forceasm -advmethods -local
PPC (uh, some of us have Macs you know)
-forceasm -advmethods -local
Other useful control flags can be found here.
-------------------------
Instructions for the Over 30 team only: The latest hot tip is that the screen saver is now the way to go. Best performance per MHz.
(Last edited by reader50; Jan 19, 2004 at 03:20 PM.
)
|
|
|
| |
|
|
|
 |
|
 |
|
Professional Poster
Join Date: May 2001
Location: Switzerland
Status:
Offline
|
|
Thanks for the advice OneMacGuy
(Reader50: Mods implemented Capt.)
|
|
|
| |
|
|
|
 |
|
 |
|
Mac Elite
Join Date: Nov 2003
Location: Minnesota
Status:
Offline
|
|
*peeks*

|
|
|
| |
|
|
|
 |
|
 |
|
Professional Poster
Join Date: May 2001
Location: Switzerland
Status:
Offline
|
|
Wow. Things have certainly speeded up on my Athlon XP box!
|
|
|
| |
|
|
|
 |
|
 |
|
Forum Regular
Join Date: Nov 2002
Location: Santa Barbara CA
Status:
Offline
|
|
The -forcesse flag has made a huge difference with my Athlons.
|
|
|
| |
|
|
|
 |
|
 |
|
Senior User
Join Date: Oct 2002
Location: God's Country, The South
Status:
Offline
|
|
Originally posted by reader50:
As of the version 4 client on AMD, use both the -forceasm and -forcesse flags. The v4 client returns them to their intended uses.
-forceasm causes optimized assembly loops to be used.
-forcesse causes AMD processors that support SSE to use it instead of 3dNow (SSE is faster). This flag is ignored if optimized assembly loops are turned off, which is why you still want -forceasm. Note, keep an eye on AMD boxes using the -forcesse flag, especially if the boxes are overclocked. This flag has been known to make some AMD boxes unstable.
Also, keep using the -advmethods flag as OneMacGuy says. It is the way to get more Gromacs units, which give better points per day on modern CPUs. Tinker isn't too bad on x86, but Gromacs are still worth more per day. The project keeps threatening to reserve -advmethods for beta work units only, but in practice it still insures mostly Gromacs.
AMD
-forceasm -forcesse -advmethods -local
Intel
-forceasm -advmethods -local
PPC (uh, some of us have Macs you know)
-forceasm -advmethods -local
Other useful control flags can be found here.
According to some research I was doing in the Folding@Home forums, the -forcesse switch is to be used for both Intel and AMD processors. Their understanding is that both Intel and AMD both support the SSE instruction set. They recommended to use both the -forceasm and -forcesse on BOTH processor types. The moderator, I can't remember his name, said that the -forceasm switch was really only useful if you had frequent crashes and restarts or power outages.
Here is the link to the postings:
http://forum.folding-community.org/viewforum.php?f=10
As usual, this information is for the "over 30" crowd only, until after February 1, 2004!!! 
|
|
|
| |
|
|
|
 |
|
 |
|
Forum Regular
Join Date: Nov 2002
Location: Santa Barbara CA
Status:
Offline
|
|
Looking at my FAHlog.txt I could see that the AMDs were definitely using 3DNow before:
[14:20:08] Assembly optimizations on if available.
[14:20:08] Entering M.D.
[14:20:27] Protein: p522_BBA5_pf
[14:20:27]
[14:20:27] Writing local files
[14:20:27] Extra 3DNow boost OK.
[14:20:31] Writing local files
[14:20:33] Completed 0 out of 500000 steps (0%)
Where now they are using SSE:
[05:54:35] Assembly optimizations on if available.
[05:54:35] Entering M.D.
[05:54:41] Protein: p849_p53dimer849
[05:54:41]
[05:54:41] Writing local files
[05:54:42] Extra SSE boost OK.
[05:54:44] Writing local files
[05:54:46] Completed 0 out of 500000 steps (0%)
But looking at the log for the one P4 that I run FAH on it already used SSE without the -forcesse switch. I added it anyway, but I don't think it matters.
|
|
|
| |
|
|
|
 |
|
 |
|
Mac Enthusiast
Join Date: Dec 2002
Location: Newport Beach, CA
Status:
Offline
|
|
Alright, so what about boosting performance on a Linux box?

|
Be a traveler, not a tourist
|
| |
|
|
|
 |
|
 |
|
Forum Regular
Join Date: Nov 2002
Location: Santa Barbara CA
Status:
Offline
|
|
All of those flags apply to linux boxen as well as windows. All of my AMDs are linux boxen, actually.
The hot ticket for linux used to be to run the windows version in Wine, I would think that still holds. I never got around to messing with that myself.
|
|
|
| |
|
|
|
 |
|
 |
|
Mac Enthusiast
Join Date: Dec 2002
Location: Newport Beach, CA
Status:
Offline
|
|
Anyone know of any good addons for F@H in Linux?
I'll try the -forceasm, see how it goes....
|
Be a traveler, not a tourist
|
| |
|
|
|
 |
|
 |
|
Senior User
Join Date: Oct 2002
Location: God's Country, The South
Status:
Offline
|
|
Originally posted by EnVoy:
Anyone know of any good addons for F@H in Linux?
I'll try the -forceasm, see how it goes....
The -forcesse and -forceasm flag are processor switches, not anything to do with the operating system. They should provide a performance increase with any OS.
|
|
|
| |
|
|
|
 |
|
 |
|
Mac Enthusiast
Join Date: Dec 2002
Location: Newport Beach, CA
Status:
Offline
|
|
The -forcesse and -forceasm flag are processor switches, not anything to do with the operating system. They should provide a performance increase with any OS.
I realize that. As a seperate issue, I was looking for some good addons for F@H on Linux, like the addons they have for Seti. Anyone know of any?
|
Be a traveler, not a tourist
|
| |
|
|
|
 |
 |
|
 |
|
|
|
|
|

|
|
 |
Forum Rules
|
 |
 |
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is Off
|
|
|
|
|
|
 |
 |
 |
 |
|
 |
|