Early F@H project points

R

relic

[H]ard|DCer of the Month - August 2007
Joined
Mar 30, 2001
Messages
9,318
Just as an FYI for some of the newer folks who are used to 10s of thousands of PPD.
Here is a link to F@H projects with WU points.

Check out the points on some of the early WU.

It's easy to see why 6-7 year folding vets have ten's of thousands of WU and far fewer points. With 1 billion team points approaching in two weeks, these early WU and their points allocated, underscore even more dramatically the amount of effort that got us here.

Fold on
 
Yep, those were the days... Wonder where we're heading for in terms of the future?

 
Heh, adjust the older points to current standards and the number of points the team would have would probably increase by 50%.

Hell, I remember when they re-benched the Tinkers from 70 to I believe 220 points or so and what happened after that. The PPD for a lot of people running AMD machines, including myself, got a nice boost from that change. If I remember correctly, it was an uproar over the valuation of the Tinkers due to the change in benchmark machines. They had recently switched to a P4 for the benchmark machine and only did a valuation with that machine on the Gromacs core work units. Because of this, the Gromacs work units were blazing by on P4 machines with SSE but the Tinkers were still valued on the old P3 machine which was much more efficient concerning clock cycles. Once the work units were re-benched on the P4, the valuation went way up.

I guess I'll leave on that note as this thread doesn't need to turn into a Stanford bitchfest.

 
The old AMD Thunderbirds & Athlons used to crunch Tinkers at twice the PpDpGhz that the P3 & P4 hit.
Hence early pure folding farms where all AMD's.
When Stanford rebenched the Tinkers, you suddenly got a nice 2x bonus for running Tinkers on a AMD box.
Stanford didn't like it when you switched your AMD box to timeless Tinkers just to get this bonus ........... :p

The early protiens that I hated where the p799 GAH ones.
If the FAH servers ran out of protiens, you got GAH ones.
Worth 0.8 points and it took you ~10 hours to crunch it on a 2Ghz AMD box.

Luck ............ :D
 
Tigerbiten said:
The old AMD Thunderbirds & Athlons used to crunch Tinkers at twice the PpDpGhz that the P3 & P4 hit.
Hence early pure folding farms where all AMD's.
When Stanford rebenched the Tinkers, you suddenly got a nice 2x bonus for running Tinkers on a AMD box.
Stanford didn't like it when you switched your AMD box to timeless Tinkers just to get this bonus ........... :p

The early protiens that I hated where the p799 GAH ones.
If the FAH servers ran out of protiens, you got GAH ones.
Worth 0.8 points and it took you ~10 hours to crunch it on a 2Ghz AMD box.

Luck ............ :D
Click to expand...

Whoops 1,92 ppd wioth those GAH units :eek:

I knew the points value is a lot less in the past. It's surprising how much points scaled, especially since the advent of SMP folding.

 
relic said:
;) Just think of Vijay asRainman.
It get's me through the next WU.
Click to expand...

Well, it would help if I hadn't lost 800+ PPD moving back to Vista64 from XP. The GPU client is only using about 4% of a core @3.6, about 20% of a core for the console client and the SMP client using the rest (more CPU power than it got with XP) but the SMP client is about 800PPD less than it was doing before on the same work unit. I'm only getting about 1000PPD out of it instead of around 1800 on a 2665.

 
Not sure what the problem is there.
I still get about 1800 SMP in Vista 64 and 4700 GPU on a Q6600/8800GT
 
SmokeRngs said:
Well, it would help if I hadn't lost 800+ PPD moving back to Vista64 from XP. The GPU client is only using about 4% of a core @3.6, about 20% of a core for the console client and the SMP client using the rest (more CPU power than it got with XP) but the SMP client is about 800PPD less than it was doing before on the same work unit. I'm only getting about 1000PPD out of it instead of around 1800 on a 2665.
Click to expand...
So, despite the fact Vista almost entirely releases a core from the GPU client, you're still seeing a drop of 800PPD?? How do you have the clients configured? Are you using any VMs or isolating cores?
 
SmokeRngs said:
Well, it would help if I hadn't lost 800+ PPD moving back to Vista64 from XP. The GPU client is only using about 4% of a core @3.6, about 20% of a core for the console client and the SMP client using the rest (more CPU power than it got with XP) but the SMP client is about 800PPD less than it was doing before on the same work unit. I'm only getting about 1000PPD out of it instead of around 1800 on a 2665.
Click to expand...

The only answer to that is to try and run VM's without the -advmethods flag.
My VM's have crunched 3,195x p2605 @330 PpDpGhz vs 55x p2665 @250 PpDpGhz.

Also 2x VM's on a quad core only load a CPU to ~97% max.
So the 4% needed to run the GPU client should not slow the VM's by to much.

Luck ................. :D
 
APOLLO said:
So, despite the fact Vista almost entirely releases a core from the GPU client, you're still seeing a drop of 800PPD?? How do you have the clients configured? Are you using any VMs or isolating cores?
Click to expand...

Trust me, it was a surprise. I have the GPU client running higher priority with a normal install, the console client is running as a service at idle priority and the SMP client is just the normal Windows SMP client running at idle priority.

What confuses me is that the SMP client is actually using more cycles than it did before while under XP and the points are lower. The client is a direct copy from the WinXP install and I have it setup correctly with all the correct flags. SSE boost is on and everything looks fine.

I just installed SMP Seesaw Pro and I'm trying something out with it.

 
Wow, talk about a nostalgic thread. I just had to go back and see what all my old athalons did back then:

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=userpagedet&username=BillR&teamnum=33&prange=0

One to five points at a time………sheese.

Then add to that all those late nights shopping NewEgg (they binned their chips back then) for just the right stepping praying each one would over clock better then the last. Using a battery to blow up the shorting connectors on CPU’s for better over clocks. Reconnecting factory blown connections with silver lacquer at 3 in the morning with custom made toothpicks, paperclips, razorblades etc.

Know what, I miss those days in a way, yeah, I really do.;)


 
I regret not folding back in those days. I was around the site for most of it seeing as HardOCP taught me about taking a 300A up to 450 and saving a few hundred bucks. 6 years on the site and only folding for 5 months, such a waste. :(
 
IcedEmotion said:
I regret not folding back in those days. I was around the site for most of it seeing as HardOCP taught me about taking a 300A up to 450 and saving a few hundred bucks. 6 years on the site and only folding for 5 months, such a waste. :(
Click to expand...
Better late than never. :)
 
IcedEmotion said:
I regret not folding back in those days. I was around the site for most of it seeing as HardOCP taught me about taking a 300A up to 450 and saving a few hundred bucks. 6 years on the site and only folding for 5 months, such a waste. :(
Click to expand...

I wasn't doing F@H2 at the beginning either and missed out on a lot of these 0.8 point work units and the like as well. I stayed with G@H until the end of the project. By the time I was on board with F@H2 the points structure had changed a good bit. I'd probably have a few more work units to my credit but I was a small timer until about February of this year.

No one really cares when you started, just that you have. As stated, better late than never. Besides, being a bit later to the game than some just means you need to work that much harder to "prove yourself". ;)

 
Welp, I guess I'm one of those "newer folks who are used to 10s of thousands of PPD" :(

Anyhoo, I'm also one of the many people that think the point system in F@H has some major discrepancies and should be reevaluated. I'm also proud and amazed at the past scores of the many "old timer" folders (time not age, except for relic, his is probably age too :p) I'm impressed how they religiously stuck with WU's that had very little point value. (the only logical reason I can come up with is they wanted to find a cure and the "hell" with the points)

With all that said, I thank all of you old timers (time not age) for keeping the F@H project alive so that people like me can go broker than we already are, become incurable addicts, join clubs like the "mowed over" club and maybe do something for the advancement of mankind with our hobby of computers. (what's that old sayin' about "misery loves company" :D

FOLD ON!

 
Exactly right JWS.

Every WU needed was crunched...and misery loves company. (Which is why women want to marry)

Sometime people complain a bit too much about 'only' getting 2-3k PPD.
In the beginning I struggled for a month to get that many points.
My average points per WU is under 140.

Just fold....it's what we do.
We will win, because we are too [H]ard to lose.
 
Yea, as an oldtimer (in all respects), I'm averaging 134 points per WU.
No SMPs, no GPUs, 1 PS3, a handfull of CPUs, and persistence...:D

Fold on!

 
im averaging 209 p/wu but man i remember folding on tons of PII/PIII machines and the first time i broke 1000ppd it was the most awesome feeling... now one video card can do 5x that lol
 
Wow, My points per work unit overall is right at 71!

I do tend to keep older hardware running though because electricity is free in my 6000 sqft Warehouse were i keep my EBay and Amazon goods.

 
FLECOM said:
im averaging 209 p/wu but man i remember folding on tons of PII/PIII machines and the first time i broke 1000ppd it was the most awesome feeling... now one video card can do 5x that lol
Click to expand...
OOC, are your work machines configured to receive the bonus WUs for the standard client or the regular WUs?
 
APOLLO said:
OOC, are your work machines configured to receive the bonus WUs for the standard client or the regular WUs?
Click to expand...

I'm pretty sure he has it setup for the regular work units since it's a work environment and the fewer headaches the better.

 
APOLLO said:
OOC, are your work machines configured to receive the bonus WUs for the standard client or the regular WUs?
Click to expand...

standard

each machine (2x standard clients) averages like 400ppd for a Pentium D945... same machine running SMP gets a little over 1100ppd :rolleyes:
 
FLECOM said:
standard
Click to expand...
You could see a potential increase of close to 50% with the bonus WUs. Seeing that there's only a few weeks left before Stanford terminates the bonus WU run, it wouldn't be worth the hassle to reconfigure the clients now, unfortunately.
 
APOLLO said:
You could see a potential increase of close to 50% with the bonus WUs. Seeing that there's only a few weeks left before Stanford terminates the bonus WU run, it wouldn't be worth the hassle to reconfigure the clients now, unfortunately.
Click to expand...

ya whatever, im not too concerned about that

i just think its kinda crappy that a GeForce 280 can do as much work as 40(!) 3.4ghz intel cores
 
FLECOM said:
i just think its kinda crappy that a GeForce 280 can do as much work as 40(!) 3.4ghz intel cores
Click to expand...
I understand where you're coming from. Over the last year, I invested in quad-core server architecture to get the most production from limited space, and now the nVidia client comes along to render my decision moot. The error of my choices are further compounded because the motherboards I run have almost no 16X PCI-E slots, making video card upgrades for the GPU client practically impossible. It's yet another story in the continuing saga of Stanford's vagarious nature, or the nature of the beast.


 
FLECOM said:
ya whatever, im not too concerned about that

i just think its kinda crappy that a GeForce 280 can do as much work as 40(!) 3.4ghz intel cores
Click to expand...

But that GX280 has 240 stream processors. ;)
 
You must log in or register to reply here.
Back
Top