KWSN Orbiting Fortress

Putting_things_on_top

I just noticed that we have an opportunity to re-take our place in the top 100 F@H teams...in less than a year!

We have been languishing around position #122 for the last few weeks...
...but the teams that are ahead of us (in total points) are not rising as quickly as we are.
See http://folding.extremeoverclocking.com/team_overtake.php?s=&t=117
There are more conquests ahead of us, and fewer threats behind us.

I think it would be to our collective advantage to seize upon this opportunity and accelerate our ascent.
The point-rigging system that F@H has recently put in place (which I have openly disagreed with) has actually worked to our team's advantage.

Without launching into a long, drawn-out appeal about the 'worthiness' or 'fairness' of F@H, the salient point I'd like to make is this:
If we get back into the top 100 teams, the formidable power of the loonie magnets will improve...therefore: more Kaniggits and more Lunacy!
Many F@H n00bs will be searching for a home...why not the KWSN Orbiting Fortress?
If we're on the first page of any "list of teams" that they see...well...that's how I came to be a resident loonie!
And I'm sure many others found our little home in this way as well.

So...what I'm asking for is a little extra help over at F@H to propel ourselves back into the top 100 as fast as we can.

Despite the stupid point-rigging (which gives extra points for faster hdwe #evil

), it comes down to the total team PPD!
Anyone that can spare some shrubberies for the next few weeks [or months?] would be greatly appreciated.
Sure, GTX-600 series GPUs (running in 'beta' mode), and higher-end SMP CPUs will be given [unfair] point advantages.
But - the goal here is to up-the-ante with the overall team PPD...so any points will help!
Once we've crossed over into our rightful Rolling Eyes

place, I'm sure that the perennial folders (like me) can keep us from falling into the abyss again.

And please make sure that you are absolutely confident about your favorite color!
_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

sir spuddly buddly · Posted: Sun Nov 18, 2012 2:17 am Post subject:

Plats # 658 on KWSN Team list stats, sir_spuddly_buddly
Date of last work unit None (so long ago that the Stats 'puter can n't remember it! Sometime in 2005 is the correct answer)
Total score 64
Donor Rank 1462221 of 1662193
WU 2

I'm in! Still using the same machines though! Laughing

Edit - Jeezus! 99% GPU load, phew! (stands up & applauds) I remember one of the guys who worked at AMD or NVIDIA moved to Stanford, and it shows. Bowman I think his name was - or am I thinking of 2001? TURN UP THE FANS, QUICK!! Razz

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

Putting_things_on_top · Posted: Sun Nov 18, 2012 4:45 pm Post subject:

Yeah, F@H runs everything (by default) at full-speed.
There ARE some switches/settings that can tweak this back, but I've forgotten what they are.

If you're using an nvidia card, there is a very useful tool called
nvidiaInspector that will allow you to tweak various options on the card...like fan speed.

One of my shrubbers is an ASUS GTX-560ti, and it runs a little hotter than I like.
It also doesn't hold the fan setting for too long, so...
I rigged-up a Windows scheduled-task [to run every 5 minutes] that runs nvidiaInspector to set fan-speed at 85%.
This keeps it from exceeding 70c (which is my personal upper thermal tolerance for GPUs).

At some point, though, I will need to take that GPU apart and give it a thorough cleaning & re-thermaling.
It's been running 24x7 at near 100% for almost 6 months.
It's due for some TLC.
Besides, the way most manufacturers apply thermal paste is atrocious!
It seems like mortar that was troweled on by a blind stone-mason!

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

sir spuddly buddly · Posted: Mon Nov 19, 2012 1:32 am Post subject:

I've got EVGA Precision X running to help me pedal the cycles as fast as possible.
#ni-1

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

Putting_things_on_top · Posted: Sat Dec 01, 2012 1:35 am Post subject:

Well, the stats are in for Nov 2012...

Another record month for the KWSN - 11,732,545 points!

Small brag: 4,425,463 of those are mine! (37.7%) Wink

However, we're still languishing in position #122. Sad

Hopefully, December 2012 PotM will give us some vertical acceleration

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

Antigens · Posted: Wed Dec 05, 2012 11:43 am Post subject:

Putting_things_on_top How does one get the Beta points for using faster video cards? I have currently running 2 GTX570's and a 560ti. I'd like to optimize the farm to help the cause. Smile

_________________
Check out our website
[img]http://www.katrinashome.com/antigens_counter.php[/img]

Putting_things_on_top · Posted: Wed Dec 05, 2012 10:37 pm Post subject:

As far as I know, setting "client-type" to "beta" is the way it works.
EDIT: Also, set "max-packet-size" to "big".

If you already have this set, then you're in the same boat as I am (as well as others).
The "beta" option doesn't guarantee the big-point WUs - it just means that you're willing to accept projects that are not fully QA-tested.

The WU assignment server(s) select an appropriate WU from various work queues based on:
1) OS (Windows seems to dominate heavily)
2) Client-type (F = normal, A = advmethods, B = beta, and I = internal)
3) Equipment type (GPU type, etc)
4) Project priority (internally set by Stanford)
5) Misc things (such as max-packet-size, etc)

Although somewhat cryptic, the F@H server status page is here: http://fah-web.stanford.edu/pybeta/serverstat.html
You might want to select "GPU" from the drop-down list above the "client" column heading.
The next-to-last column shows some info about the types of WUs (like OS and Client-type) - float your cursor over the the blue-circled "i" for details.

I have had streaks with beta WUs that won't actually "start" (sit in a running state at 0% for more than 5 mins).
In that case, I "pause" the WU (slot) and then delete the slot (to 'dump' the WU back to Stanford), and then re-add the slot.
In the past few days, the WUs have been really crappy (running longer and/or hotter with lower-than-expected points).

So...long story short..."beta" is a really craps shoot.
Sometimes you get a streak of very meaty WUs, followed by prolonged period of slim-pickings.

Additionally, if you run "beta" and are NOT an official beta-team member at Stanford, you'll get absolutely no help from them!
I am not a beta-team member - just a rogue!

Hope this helps - for what little it's worth!

Oh BTW, I have had more success with getting beta WUs on my GTX-690 than on my 560s.
But I do recall some of my 560s got the occasional QRB WU.

As an aside - I just ordered another shrubbery (another GTX-690) on Monday, and I should have that incorporated into my "farm" by end-of-week.
I may even have a little path running down the middle!
That's a total of 6144 GPU cores in my biggest rig!
_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

Putting_things_on_top · Posted: Fri Dec 07, 2012 9:47 pm Post subject:

There's a whole bunch of WUs now available that have sizeable points.
Project 762x GPU (14093 pts/WU) - more than 28,000 WUs queued.
These run quite long, however - so it skews the PPD lower.

My rigs are showing:
GTX-560ti: ~7.5 mins TPF [~12.5 hours/WU]
GTX-690: ~5.5 mins TPF [~9.2 hours/WU]
#ni-1

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

Antigens · Posted: Sat Dec 08, 2012 10:39 am Post subject:

Thanks for the info. I didn't have any flags set was just running the ver 7 client as it installed. What is the Big packet for? Does it help if running without beta? I have installed the beta client on one workstation to see if it does any better.
_________________
Check out our website
[img]http://www.katrinashome.com/antigens_counter.php[/img]

Putting_things_on_top · Posted: Sat Dec 08, 2012 7:44 pm Post subject:

Setting "max-packet-size" to "big" will enable the GPU to accept WUs that are larger than normal (>500Kb?) - but not necessarily more points. Crying or Very sad

This option is for those who have a good amount of spare (temporary) disk space -AND- are not concerned about IP traffic quotas (from your ISP).

Here's a handy reference for V7 client switches (and V6 equivalents):
https://fah-web.stanford.edu/projects/FAHClient/wiki/ClientDifferencesV6ToV7

OH - I just installed my 2nd GTX-690 on my big-rig! Very Happy

Now I have 4 GPU slots + 1 SMP slot. Woohoo!!!!!

The UPS is reporting an average 780 watts draw. Surprised

V7 client estimates the daily PPD > 150k points. Cool

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

Putting_things_on_top · Posted: Sun Dec 09, 2012 6:20 am Post subject:

Well - I have just found proof positive that SMP can interfere with GPU shrubbing.

Long story, but here's the gist of it:

I just got done folding some large WUs (14093 points each) on each of my 4 GPUs (two 690s).
I was also folding an SMP WU at the same time.
No problems - I was watching real-time utilization graphs.
Everything running nicely at/near 100%.

Took a nap, came back after the 4 big GPU WUs completed.
Smaller (albeit quicker) WUs had loaded-up for GPU.
*BUT* they were running an average of 65% GPU utilization.
Apparently, these GPU WUs are not "self-sufficient" and rely on CPU services - quite a bit. (specifically, project 8054).
So, when I "paused" the SMP WU, the GPUs went immediately to 100%.
The CPU (8-core) was chomping an overall average of 27% - WITHOUT THE SMP WU - just to service GPU requests!

I am going to suspend SMP folding on my BIG rig for about a week or two - just to see if it makes a difference.
-OR-
I might pare down the number of SMP cores (from 8 to 6) and see if that provides adequate servicing for GPUs.

Point is: some GPU WUs are almost completely independent of CPU, while others are definitely not. #evil

More on this later (after a few experiments)...

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

sir spuddly buddly · Posted: Sun Dec 09, 2012 8:07 am Post subject:

Ni!
Project: 7623
points 14093.00 Shocked

What! That's more then half of my work done so far Laughing

See if I can get it done in time!

#ni-1

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

Putting_things_on_top · Posted: Sun Dec 09, 2012 3:28 pm Post subject:

Putting_things_on_top · Posted: Sun Dec 09, 2012 3:30 pm Post subject:

Oh, BTW...

I trimmed back my FX-8350 to use 6 cores.
GPU slots are all running at ~100%

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

sir spuddly buddly · Posted: Sun Dec 09, 2012 4:24 pm Post subject:

After needing to download the latest drivers for the GTS 450, it's TPF at 30 mins.
#ni-1

Edit/update - with a bit of careful OC-ing down to circa 25 min now = 25 * 100 = a lonnnnnng time Laughing

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

Putting_things_on_top · Posted: Sun Dec 09, 2012 6:27 pm Post subject:

sir spuddly buddly · Posted: Sun Dec 09, 2012 6:43 pm Post subject:

Zotac GeForce GTS 450 ECO Edition but it's in an ol' Dell with PCI Express 1.0A according to the manual. It's running at 99% load and only 64C with a little OC.
#ni-1

_________________
a monster can be excused for his behaviour . . . The problem is not how a monster could do it, but how a human being did it.
Scaring 11 year olds since 2005
"There is only one path to Ni!-dom - through the shubbery"
Click every day at http://naturarvet.se/en/ to save a forest!

Putting_things_on_top · Posted: Sun Dec 09, 2012 9:37 pm Post subject:

That might be the weakest link in the chain - PCI Express 1.0A

A 450 should be a respectable shrubber.
I found your card specs:

http://videocardz.com/nvidia/geforce-400/geforce-450gts

#ni-1

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC

sir spuddly buddly · Posted: Mon Dec 10, 2012 2:19 am Post subject:

The GPU only-client is not talking to the 15 core obviously. Will not pause when asked - even though it says so in the log

Putting_things_on_top · Posted: Mon Dec 10, 2012 7:17 am Post subject:

WOW, I have not seen that before!

Did you set "client-type" to "beta"?
If so, then you're up-the-creek (unless you're an actual F@H Beta team member).
If not (and the WU runs amok), then you may want to sign-up on the Stanford Folding forum and submit the log (and configs) for review.

Looks like some errant code got into a loop and is doing a whole bunch of nothing very quickly!
#ni-1

_________________
Click here for...KWSN F@H team summary at EOC

Or here for...KWSN F@H team overtake at EOC