• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

PPD: Was and is now.

K

Kendrak

[H]ard|DCer of the Year 2009
Joined
Aug 29, 2001
Messages
21,141
What was your ppd before The Drop?

I had mine up to about 50k ppd with my 2nd 9800GX2

Now I'm at 40k or so.

 
The stats have been so funky as of late, I couldn't tell you for sure. I was ~8kppd I think, now its ~6kppd ish.
 
Too hard for me to tell, maybe it's lowered by a few thousands. It's hard since at the same time the new units appeared, I also added 2 extra GPU.

 
Yeah the stats have been so funky and I think there are missing W/U's out there... because I'm consistently showing 110k still and I've had a 98 day in there which is well below my norm. Regardless I'm just happy I have not really had to touch a box in weeks... I had one reboot the other day that has been it.



 
I ramped up while this was being knocked down, so I'm still up prior to last week.

Looking at others, 10% penatly for the heavy GPU guys seems the norm.

BYW: While GPU's were "leveled" did they up the CPU PPDs? I've been seeing some higher point values the past few days.
 
oraldlight said:
I ramped up while this was being knocked down, so I'm still up prior to last week.

Looking at others, 10% penatly for the heavy GPU guys seems the norm.

BYW: While GPU's were "leveled" did they up the CPU PPDs? I've been seeing some higher point values the past few days.
Click to expand...

If your running a linux version, yes the new A2 cores are great :D

 
Yeah, my GPU losses is mitigated by the increase of A2 units. Also, I agree with Sunin that I think there are some missing points (maybe 10,000 or 20,000, hard to follow with lots and lots of clients).

 
I've never gotten around to getting my linux boxen to report what they're workign on, so this is just Windows CPU observations.

Project: 2620 (R85,C42,G37)= 292points.
Project: 2170 (Run 6, Clone 316, Gen 6) I think this was about 171 points
Had another one that was in the 240 range that started me noticing the higher values. Prior to that I've never seen greater than about 119 points for Windows CPU/consoles.
 
Real hard to tell, I was around 25k but the SMP crashed so much that I stopped running it, now it's only 4 GPU clients...maybe 14k now.
 
What was the average drop for an 8800GS? Xil, you stated a drop to around 2.7k ppd... But you also quoted a FAH developer saying he was looking into optimizations... Optimizations that might lead to a PPD increase equal to half of the drop.

Any insight on what we *could* expect in some later work units? 3500 ppd or so?

Right now I've only got 4 CPU clients and an ATI 4870 folding, so I didn't see a drop at all... Well, I mean, I'm getting these shitty amber WU's worth around 84 points :mad:



 
I was seeing high 4k ppd on my 8800 GTS, now seeing middle to high 2k ppd. :(

Why don't they give those WU's higher point values to help even everything out?
 
Arcygenical said:
What was the average drop for an 8800GS? Xil, you stated a drop to around 2.7k ppd... But you also quoted a FAH developer saying he was looking into optimizations... Optimizations that might lead to a PPD increase equal to half of the drop.

Any insight on what we *could* expect in some later work units? 3500 ppd or so?

Right now I've only got 4 CPU clients and an ATI 4870 folding, so I didn't see a drop at all... Well, I mean, I'm getting these shitty amber WU's worth around 84 points :mad:



Click to expand...

A 8800GS get about 4300 ppd with non-Lambda units. with the new Lambda units, the PPD is about 2700 with the same clocks, about 1600 PPD. If Scott is right, it should be around 3500 PPD after optimisations.

 
I was around 13K avg and now I'm down to 8K, big drop in my opinion.
 
Xilikon said:
A 8800GS get about 4300 ppd with non-Lambda units. with the new Lambda units, the PPD is about 2700 with the same clocks, about 1600 PPD. If Scott is right, it should be around 3500 PPD after optimisations.
Click to expand...

Hey, 3500ppd would be absolutely fine for me! 4300ppd would be wicked for those odd non-lambda units too.
 
I've actually gotten hit in both the GPU & CPU area's and I'm not exactly sure why. At one point I was averaging around 80-90k PPD, but now, with all the same hardware, and then some more, I can't seem to break 70k PPD on a good day, but I average in the low 60's. So it's been quite a hit for me. Oh well, just as long as the science get's done, and winter comes soon. My A/C isn't able to pump out all the hot air in my condo, so I'm looking for a home, if only my job would be more stable so I could plan better.

 
Old WUs took about 2h45m on my 9600GSO, so (in theory) 8/day X 480pernt/WU = 3840ppd w/ shaders @ 1596.

New WUs take almost 4h each, so 6/day X 430pernt/WU = 2580 ppd...

So I'd call that about a third less ppd, and about 75% of WU - of course, my one SMP reporting back a 1920 pernt WU ever other day makes my ppd chart bounce up and down constantly, so it's hard to really tell! :rolleyes:

 
PPD are ever changing in the 7 mos I've been here I've seen it go all over the place, and I know there still are some enhancements for the windows clients to come. I remember when you could get 4200 PPD on a Quad Core/windows machine, now your lucky if you crack 3200. We already know F@H is re-evaluating the point structure. They know right now that the CPUs are so unattractive(besides linux) that noone in there right mind would spend for a quad core. The cost per point per day for a cheapy cpu, and a single gpu is far better then going quad and gpu. So they are working on that. Again it is and always will be ever changing. Just flow with it don't worry about the points and have fun with it.

 
I've seen FahMon give a reading of 107,500 PpD and also 93,400 PpD.
And thats with a mix of slow and fast protiens.
I could probably drop another 10,000 PpD with all slow protiens.

A Quad with 2 or 3 cards is still a good mix.
The trouble with the GPU work units is that you crunch them in a couple of hours.
If Stanford goes down, your soon out of work.
Add a couple of Linux clients inside VM, they'll crunch for 24 hours a work unit.
So there's more leeway with Stanfords servers.

My Q6600 with 2x 9800GTX+'s only pulls 340 Watts and does between 11,000 and 16,000 PpD.

Luck ............ :D
 
Tigerbiten said:
I've seen FahMon give a reading of 107,500 PpD and also 93,400 PpD.
And thats with a mix of slow and fast protiens.
I could probably drop another 10,000 PpD with all slow protiens.

A Quad with 2 or 3 cards is still a good mix.
The trouble with the GPU work units is that you crunch them in a couple of hours.
If Stanford goes down, your soon out of work.
Add a couple of Linux clients inside VM, they'll crunch for 24 hours a work unit.
So there's more leeway with Stanfords servers.

My Q6600 with 2x 9800GTX+'s only pulls 340 Watts and does between 11,000 and 16,000 PpD.

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

I agree, it's wise to have a good mix to avoid being hit too hard by changes.


 
I'm currently running on GPU & 3 SMP. My PpD has dropped a little. The biggest drop is obviously the GPU. Currently running an 8800GTX. Fast Proteins around 5300 PpD. With the newer Proteins, currently around 3800. I'm still getting a mix of fast and slow proteins so I'm averaging 4550 PpD on that client. I'm moving to all notfred VMs on my SMPs to make up the difference

 
Well I'm hoping still that once they have the issues worked out that they will move to WU sizes that take 24hrs to crunch, thus an outage at stanford has zero impact on users. Maybe they could set this up for the adv methods flag and if you don't use it you get the 2-4hr variety. If your stable those 24hr WUs would be perfect!

 
Sunin said:
Well I'm hoping still that once they have the issues worked out that they will move to WU sizes that take 24hrs to crunch, thus an outage at stanford has zero impact on users. Maybe they could set this up for the adv methods flag and if you don't use it you get the 2-4hr variety. If your stable those 24hr WUs would be perfect!

Click to expand...

AMEN Brother!!!!
 
I actually didn't see a big time increase for the new WU. Maybe an extra 15-20 minutes. Still went from ~24k PPD to 18k PPD.

 
We just had this same conversation a week or so ago. For as long as I have been folding the points have varied, sometimes after several months but it has been as short as a few weeks.

Either way everyone involved on every team is going through the same thing so if you are in this for the competition nothing is lost.

The major source of aggravation to me is simple; harder longer work units should be worth more not less. For some reason the math folks at Stanford are having a painful time with this concept. The points system used to based on tests run on a given machine but unfortunately nobody could ever duplicate Stanford’s results, even Stanford. The current system(?) is far better than the 5 to 15 point days of old but…………well I have ranted on this before to no avail.

The point remains, everyone on every team is facing the same thing.

Above all remember in the end what we do is for the cure, the point are just to make it fun.

Of course the “fun” would be more “fun” if we didn’t have to deal with VJ’s funny math…………;):rolleyes:


 
You must log in or register to reply here.
Back
Top