dual processor F@H help!

H

hacktor2

Limp Gawd
Joined
Oct 18, 2004
Messages
263
Just looking to see if i've setup everything correctly.

Each node will have dual 2.0 ghz xeon with 8 gigs. I am creating an image for a 100 node cluster that will eventually all be running F@H. Yes, again, I have permission. In writing.

Here is my current setup config for each instance. I have each instance in its own directory labeled C:\FAH\fah6 & C:\FAH2\fah6.2 respectively.
Code: 
User Name [Anonymous]? 
Team Number [0]?
Passkey []?
Ask before fetching/sending work (no/yes) [no]?
Use proxy (yes/no) [no]?

Acceptable size of work assignment and work result packets (bigger units
may have large memory demands) -- 'small' is <5MB, 'normal' is <10MB, and
'big' is >10MB (small/normal/big) [normal]? Big

Change advanced options (yes/no) [no]? yes
If you want to change the advanced options, enter yes but it's not mandatory.
Launch automatically, install as a service in this directory (yes/no) [no]? yes

Core Priority (idle/low) [idle]?
CPU usage requested (5-100) [100]?
Disable highly optimized assembly code (no/yes) [no]?
Pause if battery power is being used (useful for laptops) (no/yes) [no]?
Interval, in minutes, between checkpoints (3-30) [15]?
Memory, in MB, to indicate (8192 available) [4000]?

Set -advmethods flag always, requesting new advanced scientific cores and/or work units if available (no/yes) [no]?

Ignore any deadline information (mainly useful if system clock frequently has errors) (no/yes) [no]?

Machine ID (1-16) [1]? [1] and [2] (each processor gets a different machine ID)
Disable CPU affinity lock (no/yes) [no]? yes
Additional client parameters []?
IP address to bind core to (for viewer) []?
Does this look like it should work? When I config'd each instance, it downloaded everything and seemed like it was working.

EDIT: I know there is no team or username. We are still deciding if we want to fold for the [H]
 
Haktor2 - That looks like it should work. I don't have any direct experience with dual CPU rigs but someone here will chime in that has. I don't know that you have to reserve so much memory for each processor. I believe it will just take what it needs. I lurked here for a long time before jumping in and beginning to post. I don't believe there is a better group of people or folders around. We would be pleased to have you join the [H].
 
So you're using the standard client then? I recommend that you go with Linux VMs if possible, but if not, that would be the most maintenance-free way to go. Everything looks fine, although I would recommend that you enable the advmethods flag. Aside from that, you're good to go.
 
Each node is running windows 2008 HPC. In the near future, we are planning to rebuild one of our 96 processor linux clusters at which time I'll be able to install some VM's for f@h.

Standard client meaning console? Then yes. Once all the racks are deployed, this will be a 400 Ghz, 800 gig folding farm. Add the other cluster and we'll have a 592 Ghz folding farm!!! :)
 
Sorry about that... Yes, as of right now uniprocessor.

I've been reading up on SMP clients. Would 1 SMP client work better then 2 instances of the standard?

If I were to run SMP, would I turn on hyperthreading to create 4 logical processors?

Any recommendations? If you had the ability to set up a 100 node, dual xeon cluster would you run SMP on each instead of 2 standard?
 
hacktor2 said:
Each node is running windows 2008 HPC. In the near future, we are planning to rebuild one of our 96 processor linux clusters at which time I'll be able to install some VM's for f@h.

Standard client meaning console? Then yes. Once all the racks are deployed, this will be a 400 Ghz, 800 gig folding farm. Add the other cluster and we'll have a 592 Ghz folding farm!!! :)
Click to expand...
That is incredible! Anyone calculate what PPD this will amount to ballpark figure? :eek:
 
hacktor2 said:
Sorry about that... Yes, as of right now uniprocessor.

I've been reading up on SMP clients. Would 1 SMP client work better then 2 instances of the standard?
Click to expand...
You mean in stability or production? The SMP client is less stable but vastly more productive in both science and points production.

If I were to run SMP, would I turn on hyperthreading to create 4 logical processors?
Click to expand...
What kind of processors are you running? You can turn on HT but I cannot tell you how much more if at all it will benefit you. Others who have HT processors can tell you more.

Any recommendations? If you had the ability to set up a 100 node, dual xeon cluster would you run SMP on each instead of 2 standard?
Click to expand...
SMP no question.
 
Last edited:
hacktor2 said:
I've been reading up on SMP clients. Would 1 SMP client work better then 2 instances of the standard?
Click to expand...
Like APOLLO said. Less reliability but much more points output. Using Linux VMs (which you can run under a Windows host OS), you get the best of both worlds for the most part. Even better performance than the Windows SMP client, and very good stability.
hacktor2 said:
If I were to run SMP, would I turn on hyperthreading to create 4 logical processors?
Click to expand...
With the Windows SMP client, you wouldn't benefit much since the client can only really take advantage of two CPUs. However, if you use Linux VMs you'll see a big improvement.
hacktor2 said:
Any recommendations? If you had the ability to set up a 100 node, dual xeon cluster would you run SMP on each instead of 2 standard?
Click to expand...
I'd run the Windows SMP client out of those two options, but I still recommend Linux VMs as you'll get much better points output with them.
 
Well if you have 2008 HPC, you can try enabling the Hyper-V functionality although it may not work depending on your CPU.
It must have Intel-V or whatever its called.

But in General LinuxSMP>WinSMP>Uniprocessor.

BTW that is an awesome setup.
Dont forget in the config we are Team 33;)
 
O also im not sure how much HT will help...depends on your CPU, but if you do that then you would need 4 Uniprocessor clients per box or 2 WinSMP clients per box.
 
nitrobass24 said:
Well if you have 2008 HPC, you can try enabling the Hyper-V functionality although it may not work depending on your CPU.
It must have Intel-V or whatever its called.
Click to expand...
I believe that all current Xeons have VT-x support.
 
Yea all the dual cores and up had VT-x, except for the Paxville Xeons, but it didnt have HT.

So your CPUs should be compatible.
 
nitrobass24 said:
BTW that is an awesome setup.
Dont forget in the config we are Team 33;)
Click to expand...
If he applies most of the potential hardware mentioned with the SMP client, that should go a long way to keeping EVGA at bay... :cool:
 
Thanks for all the help!

As of right now i'll be switching over to the windows SMP with big and advmethods flags.

It will be awhile before we get all the racks up because we still are in the build process (including getting more power to the room).
 
hacktor2 said:
Thanks for all the help!

As of right now i'll be switching over to the windows SMP with big and advmethods flags.

It will be awhile before we get all the racks up because we still are in the build process (including getting more power to the room).
Click to expand...
Excellent!. Looking forward to see your debut on the stats pages! :cool:
 
just wanna jump in here, HOLY CRAP, 33 welcomes you with arms wide open! While not all of us can answer your questions, don't be afraid to ask and I'm positive someone will help you. Again, welcome to the folding forums and best of luck to you!!
 
You must log in or register to reply here.
Back
Top