Folding@home points system IS UNFAIR! [NOT]

[imzjustplayin]

First off that is wrong, the SMP benchmark is NOT 1780PPD. One of the SMP projects IS 1760 points. If you run the SMP project on two processors and it finishes in one day, that comes out to 880PPD. The PPD are calculated PER CORE, if you're not going to read what I have to say and the evidence I have provided, then need not reply.

Actually 7im is correct.

All SMP projects are benchmarked such that they will achieve 1760PPD on the SMP benchmark machine.

All regular projects are benchmarked such that they will achieve 110PPD on the regular benchmark machine.

Saying that PPD is calculated per core is total nonsense. PPD is just how many points per day you get for any given project.

PPD is only given per core on fahinfo.org to simplify the data analysis.

That is wrong. For the sake of simplicity, the screen capture of the webpage I posted above uses its calculation of PPD PER CORE. Go to that website before stating that 7im is right about how PPD are calculated.

They calculate PPD per core because there are single and multiple processor systems out there. If you were to say I get 1100PPD! I'd be like, wow that's great. Then you say "well I do have a Core2quad" and I'd be like, oh, I see.. Saying the PPD is meaningless unless you know how many processors are involved and then that just adds unnecessary confusion and calculation and therefore it's much simplier and smarter to use PPD PER CORE and not per project as a whole. Don't forget, we're comparing not only single and multi processor systems, but regular and SMP enabled projects.

[wilding2004]

First off that is wrong, the SMP benchmark is NOT 1780PPD. One of the SMP projects IS 1760 points. If you run the SMP project on two processors and it finishes in one day, that comes out to 880PPD. The PPD are calculated PER CORE, if you're not going to read what I have to say and the evidence I have provided, then need not reply.

Actually 7im is correct.

All SMP projects are benchmarked such that they will achieve 1760PPD on the SMP benchmark machine.

All regular projects are benchmarked such that they will achieve 110PPD on the regular benchmark machine.

Saying that PPD is calculated per core is total nonsense. PPD is just how many points per day you get for any given project.

PPD is only given per core on fahinfo.org to simplify the data analysis.

That is wrong. For the sake of simplicity, the screen capture of the webpage I posted above uses its calculation of PPD PER CORE. Go to that website before stating that 7im is right about how PPD are calculated.

They calculate PPD per core because there are single and multiple processor systems out there. If you were to say I get 1100PPD! I'd be like, wow that's great. Then you say "well I do have a Core2quad" and I'd be like, oh, I see.. Saying the PPD is meaningless unless you know how many processors are involved and then that just adds unnecessary confusion and calculation and therefore it's much simplier and smarter to use PPD PER CORE and not per project as a whole. Don't forget, we're comparing not only single and multi processor systems, but regular and SMP enabled projects.

Oh dear, you are so very, very wrong. Why would Uncle_Fungus need to check your screen grab? After all it is his own site and I imagine he is well aware of how he programmed it.

Why don't you try to understand the FAQ's. Read them again, maybe some more will sink in.

[imzjustplayin]

Oh dear, you are so very, very wrong. Why would Uncle_Fungus need to check your screen grab? After all it is his own site and I imagine he is well aware of how he programmed it.

Why don't you try to understand the FAQ's. Read them again, maybe some more will sink in.

Just because you say or wrote something, doesn't mean you have any recollection of doing so. And if he did write the site, then he shouldn't be arguing with me in the first place as he'd understand what I'm talking about, or at least have the gist of it.

[wilding2004]

Oh dear, you are so very, very wrong. Why would Uncle_Fungus need to check your screen grab? After all it is his own site and I imagine he is well aware of how he programmed it.

Why don't you try to understand the FAQ's. Read them again, maybe some more will sink in.

Just because you say or wrote something, doesn't mean you have any recollection of doing so. And if he did write the site, then he shouldn't be arguing with me in the first place as he'd understand what I'm talking about, or at least have the gist of it.

The point is YOU ARE WRONG. I take it you still haven't read the FAQ.

P.S The benchmark PPD for SMP is 1760. Just because that is also the value of a WU doesn't mean it can't be the benchmark value, which is what 7im was trying to explain to you.

[imzjustplayin]

Oh dear, you are so very, very wrong. Why would Uncle_Fungus need to check your screen grab? After all it is his own site and I imagine he is well aware of how he programmed it.

Why don't you try to understand the FAQ's. Read them again, maybe some more will sink in.

Just because you say or wrote something, doesn't mean you have any recollection of doing so. And if he did write the site, then he shouldn't be arguing with me in the first place as he'd understand what I'm talking about, or at least have the gist of it.

The point is YOU ARE WRONG. I take it you still haven't read the FAQ.

P.S The benchmark PPD for SMP is 1760. Just because that is also the value of a WU doesn't mean it can't be the benchmark value, which is what 7im was trying to explain to you.

What does the benchmark PPD have to do with the fact that the PPD for this given project on a Core2Duo is 880? The PPD are calculated per processor, not for the project as whole. You seem to be confused as to what point of reference I am using. For SIMPLICITY SAKE, I'm sticking with the PPD being PER PROCESSOR. If you don't do it this way, comparing the Mac pro benchmark system's PPD and the P4 2.8 benchmark's PPD would be completely invalid with out any translating. Do you really think that using the PPD of all the processors combined in the SMP project is directly comparable to using the PPD of one processor in the regular projects? NO

In fact, if I were to be stupid enough to compare it that way, my argument about a points discrepancy would be much much much more justified and I would have people on my side immediately, instead of what is happening now where people are convinced the difference is due to the bonus points and so all the points being handed out are completely justified which they AREN'T.

[uncle_fungus]

PPD isn't "given" by anyone.

Each WU is assigned a point value based on how long it takes on the benchmark machine.

For SMP WUs, that is how long it takes for the 4 cores on that machine to process the WU. If it takes 0.5 days to process a WU the point value will be 1760PPD*0.5D = 880points.

If your machine takes 1 day to process the WU, regardless of how many cores you have, your PPD is 880, since 880/1 is 880.
If your machine take 0.5days to process the WU, regardless of how many cores your have, your PPD is 1760, since 880/0.5 is 1760.

PPD is totally unrelated to how many points you score for any given WU, it is just a measure of how fast your computer is processing that WU.

[bruce]

What does the benchmark PPD have to do with the fact that the PPD for this given project on a Core2Duo is 880?

Nothing whatsoever.

The benchmarks are done by Stanford. Each project is run on a "standard" machine and a time is measured. From that information they ASSIGN a number of points to WUs in that project.

When you run a project on a Core2Duo (or anything else except on hardware identical to Stanford's standard machine), you will necessarily get a different time and a different PPD.

The PPDs that you find on fahinfo.org are what individuals have reported and Stanford takes no responsibility for them.

[John Naylor]

Can someone from the Pande Group please put this guy right, as he evidently does not respect the collective correctness of users, moderators and admins of this forum!

The PPD are calculated per processor, not for the project as whole.

As you have been told before, by people who have been with this project since very soon after it launched, 7 and a half years ago, this statement is incorrect. The PPD are calculated for the project as a whole, that is how the Pande group do it, and any of them would (and probably will) come and post on here to prove myself and the various other contributors to this thread correct.

[uncle fuzzy]

Either you're not doing something right, or you just proved your cpu is obsolete. You say you've never gotten more than 137PPD from a cpu client. My Athlon XP 3000+ (2.17GHz) is currently making 151PPD, each core of my X2 6000+ (3.15GHz) makes 200-350PPD when it isn't doing the SMP, each core of the q6600 (3.3GHz) does an easy 525PPD on the cpu client. If my slow, old 3000+ (soon to be replaced by another q6600) gets better points than you, your hardware is obsolete.

They keep writing the newer projects to take advantage of the capabilities of the newest processors. If you don't keep up on the hardware, you get the leftovers.

[7im]

The FAH project has been running 6 years. imzjustplayin is not the first to make claims like this, and he won't be the last to be proven less than fully knowledgable about how things work either.

Fine, if he won't accept the explanation about how the CPU and SMP benchmarks are aligned using the project FAQs (that Pande Group wrote), then we'll do it his way. We can use any stats in the FAHINFO.ORG website that Uncle_Fungus has written and administers, in addition to being a well versed Admin of this forum. But like Bruce said above, any hardware referenced to discuss will NOT score the same as the benchmark systems, so your mileage may vary.

But imzjustplayin will have to wait to be schooled on the topic again, unless someone else wants to take up the hat. I have to pick up my car from the repair shop. I'll be back later tonight to see how this thread progresses.

[Nonymoussurfer]

imzjustplayin:
Dude. Stop it. You gotta be trolling, right? No one could be so thick.

In case this is not a joke... Please heed the advice of the various sages who have offered you explanations. Read through the various links provided & fight the urge to continue with your replies. You've left the realm of the inquisitiive & are deep within foolish territory.
Of course the troll is really only concerned with keeping the conversation going, so... my bet is this silly discussion will continue.

edit: the name say it all (hezjustplayin)

[John Naylor]

Either you're not doing something right, or you just proved your cpu is obsolete. You say you've never gotten more than 137PPD from a cpu client. My Athlon XP 3000+ (2.17GHz) is currently making 151PPD, each core of my X2 6000+ (3.15GHz) makes 200-350PPD when it isn't doing the SMP, each core of the q6600 (3.3GHz) does an easy 525PPD on the cpu client. If my slow, old 3000+ (soon to be replaced by another q6600) gets better points than you, your hardware is obsolete.

They keep writing the newer projects to take advantage of the capabilities of the newest processors. If you don't keep up on the hardware, you get the leftovers.

Hear Hear. My E6750 @ stock is doing 380PPD on each core using the standard client.

[imzjustplayin]

The FAH project has been running 6 years. imzjustplayin is not the first to make claims like this, and he won't be the last to be proven less than fully knowledgable about how things work either.

Fine, if he won't accept the explanation about how the CPU and SMP benchmarks are aligned using the project FAQs (that Pande Group wrote), then we'll do it his way. We can use any stats in the FAHINFO.ORG website that Uncle_Fungus has written and administers, in addition to being a well versed Admin of this forum. But like Bruce said above, any hardware referenced to discuss will NOT score the same as the benchmark systems, so your mileage may vary.

Ok. But does anyone else see that Pande Group's method of benchmarking is flawed? By having the SMP projects benched on a different system than the regular projects, they've haphazardly created the differential that I'm speaking of.

How can Pande Group expect the PPD for the SMP projects to be an apples to apples comparison with the non SMP projects if they're using a different benchmark machine? SMP benchmark machine is different from the non SMP benchmark machine and that is a real problem.
You guys have sort of answered your own question, why am I not using the pande group's benchmark machine? Because their methodology for benchmarking is faulty, plain and simple.

In short, not only has the use of two separate benchmark systems for their SMP and non SMP clients created an unfair points differential, a gap if you will, but has made apples to apples comparison much more difficult since they (and I assume you guys too) insist on doing PPD for the given project and not per the core/processor. However the very act of doing it this way makes the comparisons invalid, but if you do it by core, the comparions can then become valid, to some extent and that extent is what I have been outlining. The only way to get around this issue is to benchmark both the SMP client and non SMP client on one system, which is what I have done and will post about.

[uncle fuzzy]

And they should benchmark a multi-core client on a single core how...?

[uncle_fungus]

How can Pande Group expect the PPD for the SMP projects to be an apples to apples comparison with the non SMP projects if they're using a different benchmark machine?

No-one has ever stated that it is an apples-apples comparison. The SMP benchmark machine, and the point values were set arbitrarily to provide a reasonable PPD due to the client being A) In beta testing and not stable, B) Running science faster than the regular clients and C) Having tight deadlines to enforce quick return of WUs.
The same factors affect the GPU client as well. They're called HPCs for a reason.