Folding Forum

Does the PPD of a computer (no change in configuration) decrease over the years?

The inflation I'm talking about is all home and office computers (that participate in Folding@Home) getting faster with time and and generating greater points.

There's been a huge inflation in the points. It started with PS3 WU's, got 'worse' with the GPU WU's and got (imo) out of control with bigadv.

Not much need to introduce any artificial curve, Moore's law pretty much does that for you. Looking at a blast from the past when I was folding for team Anonymous with old rig(s):



IIRC, it took me at least a couple of years to reach 470 000 points in the old times. My lowly Atom330 setup has yearned about 3x that in a year. Some i7 2600K setup can outdo that in a couple of weeks. And some BigAdv monster can do so in a day... are you saying that old points should be devalued even more?

There has also been a huge increase in computational power.
If you google some research on FLOPS/watt and you will see that the bar is always rising.

Perhaps improved efficiency in PPD/watt or PPD/socket or PPD/client(due to computational "machines" having 2048 processing units and 3.79 TFLOPS Single Precision compute power; 947 GFLOPS Double Precision compute power) would be better terms to describe the exponential growth in processing power.

Inflation really doesn't accurately describe what is going on.

Points are based on scientific value. Any increased value in points is not inflation. It is simply an increase in processor power, as others have noted already.

Sorry 7im, butI have to take issue with the assertion that "It is simply an increase in processor power...". Points inflation has outpaced Moore's law and actual performance increases since the introduction SMP work and more especially the introduction of the QRB. I hang on to computers and data for a long time. My Q6600s have gone from making 3600 ppd to as much as 12,000 ppd, without any increase in processor power. What can you call that other than 'inflation".

A typical WU, p7133 on a Q6600 @ 3.2 GHz: You can probably explain away some, but not all of the production increase.

Edit:
While we're on this topic, I'd love to see my points adjusted for inflation. 100,000,000 would be close to right.

A points program adjustment to more accurately reflect the scientific value of the work being completed is NOT points inflation.

The Quick Return Bonus gives additional points for turning in the work units more quickly, which is scientifically more valuable. More value, more points. This is not time related nor an inflationary adjustment as a function there of.

Have the points continued to rise since the QRB was implimented? No. We've ACTUALLY seen them reduced (corrected), right?

An "improvement" is what I'd call it.


To answer the question in the OP, not the title question, the PPD of existing hardware does not descrease over time. As you noted, newer hardware will speed up but that doesn't change anything.

All work units are benchmarked against a standard hardware configuration. Your PPD in relation to that unchanging benchmark will remain relatively equal for several years. Do note, however, that the benchmark hardware is updated every few years. But not to worry, the new points benchmarks are scaled to match the old hardware in a linear manner. So again, your PPD will remain unchanged (assuming no more points program adjustments, bonuses, etc.)

One thing is for sure, PPD is almost never reduced. (the one exception was in a beta trial program, so it doesn't really count)

It's all a matter of perspective.

Baseline points have increased with Moore's law as processing power has increased but one day of P4 processing @ 2.8 is still worth 110 points, so there is no explicit inflation.

QRB has been added on, so PPD has increased for both reasons, not just one. Does FahCore_a4 give higher PPDs than FahCore_78? Most certainly. That same P4@2.8 now earns more points, so there has been inflation.

Suppose your first job starts with a salary of X $/Day. Suppose some time later the boss says we've noted how good a job you have been doing and in additions to your base salary of X $/Day we're going to put you on commission and pay you Y $/unit that you sell.

Did you get a raise? Of course. Would it be classified as inflation? Probably not.

If, instead, the boss said they were going to pay you X1 $/day where X1 = 100%*X + 5%*X did you get a raise? Certainly. Would it be classified as inflation? Probably -- unless they fired somebody else and you are now going to be doing 5% of the work he was doing plus your own.

Actually the current ppd on the "average" regular SMP WU is significantly higher than it was when the QRB was introduced. IIRC, the WUs values were set so that a Q6000 @ 3.2 GHz is supposed to make 6172 ppd. They did for a while, but the ppd has crept up dramatically. I found another example on a Windows SMP machine with 2 GPUs folding:
I consider the QRB to be highly inflationary. I folded every SMP WU the same way before as after it was introduced, 24/7/365. Now I get up to 12,000 ppd for the same work I used to get 3600 ppd for.

Enough though. You're trying to help the OP, while I'm just using his post as an opportunity to bitch.

I can't address whether pratice hasn't followed policy. As a matter of policy, points do not creep.