Folding Forum

imzjustplayin wrote: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.

They're neither haphazard or flawed. They did it intentionally to get the most out of new hardware. Since you don't have that hardware your choices are (A) Continue to process WUs without SMP at the same PPD that you have been getting for as much as 6 years, or (B) Upgrade to the newer/faster hardware that can do more work. (i.e. Buy a nice new multiprocessor system or a PS3 or one of the supported GPU cards.)

By the way, do you know what the MP stands for in SMP?

Obviously you completely ignored my post saying your hardware wasn't fast enough to earn that PPD. I'm about to close this discussion due to your troll-like behavior.

v00d00 wrote:PEBKAC

Hopeless debate inspite of such well made points. It is about the Science, not the points. When I caught the folding bug, I went trolling for Q6600's.

I'm impressed by the kindness and patience thats been displayed by almost everyone who's weighed in to try and help this joker.

I suspect imzjustplayin's just trying to be as annoying as possible to see how much the Mods will tolerate before closing the thread or banning him/her. It isn't too hard to imagine this person purposefully created the "imzjustplayin" account with the intention that it would get banned within X hours or days. Has anyone noticed imzjustplayin's entire posting history seems to all be in this topic. The rehashing of the same weak arguments is certainly trollish behavior, same with the continued ingnor-ance of the reasonable explanations provided over & over, as are the changes to new subject to attack, etc... imzjustplayin (like the name suggests) isjust playing a game. The more people engage imzjustplayin, the longer this person goes on irritating everyone.

I'm ashamed to admit that I can see how this might be the slightest bit entertainng,,, in a mildly disturbing & pathetic sort of way.

I don't even know how to respond. He keeps bouncing from one topic to the next, and when you nail him down on one point, he tries to weasel out by trying a different route. First he complains that the benchmarks don't make sense when I explain how they are exactly equal, so then he doesn't want to use that data, and only use FAHINFO data. And when I agree to school him on that, he goes back to attacking the benchmarks again. Then he jumps on the SMP client being Intel only, and when AMD folders chime in to prove him wrong, he switches topics again. WTH? If facts don't work, let's try a history lesson.

Pande Group has a LONG history of doing everything it can to provide points parity over time and between clients, cores, etc. The original benchmark machine was a 500 MHz Celeron, as that was a very common platform at the time that system was chosen. A few years later, when that system was all but obsolete, they chose a faster computer, one that was AGAIN very common at the time of selction... A P4 2.8 GHz system. AND when they did update the benchmark system, they ran tests to see what common work units would score on both systems. Then they adjusted the points on some work units, but not others so that all work would score about equally on the new system. We had complaints back then too, but mostly from Celeron owners and not P4 owners (some things never change). If you drag your feet, you will eventually get left behind. (except that takes a very long time, as Celerons still make the deadlines (barely), and the client 2 versions back from 3 years ago is still supported, as is Win98 from 10 years ago) Pande Group tries to enable everyone to contribute to the project as much or as long as possible.

A few more years have gone by now, and Pande Group has chosen a new benchmark machine that are commonly in use now. 4 core systems for the SMP client. X1900XT for the GPU client. And, naturally, a PS3 for the PS3 client. And I know that Pande Group discusses these decision at great length, taking many many things in to consideration before setting a new benchmark level. As I said before, the primary consideration is scientific production. So naturally, a chip that processes faster would score more points because it does more work. Whether you consider it coincidence or not, the new SMP benchmark has a logical connection back to the CPU benchmark (as I demonstrated mathematically whether you choose to believe 1 + 1 = 2 or not), just as the newer P4 system linked back to the Celeron benchmark.

Yes, to some, the points benchmarks seem arbitrary because they are so diverse and wide spread from the original, but they are not arbitrary. When you consider how diverse and widespread the performance of this hardware has grown over time, it is quite logical. If that's not logical enough for you, then consider Moore's law. Chips double in power every 18 months. The original poster's 3+ GHz system is from 2 or 3 years ago. If you take the benchmark from 3 years ago (110 PPD) and then double it twice with Moore's law (440 PPD) and then take that times 4 cores in the new systems, what does that equal? 440x4 = 1760 Hey look, we get the same numbers again. I wonder why that is?

Topic locked