• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Help me explain something to a daft friend?

T

TN_Chelle

Junior Member
Hi all,

I'm an old cruncher from Team-Ninja (DC-Vault). On another forum I frequent (non-DC related), someone wandered onto an old DC page I had on my old website, and saw where I listed my total crunching power as GHz. This is someone who considers himself tech savvy, and apparently everyone in the crunching world is 'stupid' because you can't calculate crunching power this way.

I tried to explain to him how DC works, with work units and clock speed, etc.. but he's really dense and doesn't get it. His argument is:

----------------------

"Hertz is a unit of measure of frequency. How do you correlate that to computing power?
More importantly, I don't think you can just add CPU frequencies together... "

and

"Seriously... does it make any sense to you to add Hz together?
It's like taking a marathon team and adding up their shoe sizes to determine how fast they will run as a team.
I'm seriously curious as to who came up with that kind of measure."

and

"The question at hand is your (defense of the) use of FREQUENCY as a measurement of CPU power or capacity.
Your logic above only works if all the processors in your DC system use exactly the same CPU. If you were right, then according to your logic, if I upgraded my relic 1GHz Pentium 3 processor to a 1GHz multi-core i7 processor, I've just wasted my money because I'm still contributing 1GHz to my team.
jpshakehead.gif


These "geeks" that you refer to are more like... wannabe nerds. I don't think you belong to nor want to be associated with that group, so stop defending them."

------------------------

I know I shouldn't bother, but it's like watching a train wreck, lol. Anyone care to take a stab at explaining it?
 
Last edited:
I suggested long ago, that perhaps the units of credit that DC projects give should be expressed as a "Mhz-hour". Kind of like a Kilowatt-hour. But different CPUs give different speeds at the same Mhz, so that would give more credit to those with older machines.
 
Uh, the measure of a CPU is by Gigaflops. There is no frequency involved. Hell a 5ghz Pentium 4 is probably slower than a 1.3ghz Core i7.
 
Generally, MHz used to work pretty well. My C2Q9400@3GHz would be 3GHz*4 cores = 12GHz of crunching power. Sure there were differences between CPUs instructions/cycle, but not usually an order of magnitude.

However, now that we (some of us anyway) are crunching with GPUs, that's pretty much out the window. Gigaflops does seem to be what BOINC is using, but (1) I do mostly integer math, and (2) GFlops numbers for GPUs seem somewhat inflated, especially when I'm trying to do 64-bit math on them.

I guess there's never going to be a perfect benchmark. :\
 
Thanks.. yeah.. the page he found was going back a few years ago when I was crunching G@H and F@H. Everything was single core back then. Now that we have multi-core processors the GHz means nothing. Just hard to explain to someone who has never crunched a work unit, lol.
 
yes, we used to have several users with things like "49 ghz crunching power" or whatnot in their sig. that was a decent measurement back in the day as long as you a good mix of old/new cpus, saw a 2x3 ghz q6600 = 24 ghz, 1xe6600 3gz = 6ghz, 4 pentium d's , a couple of athlon 64's, etc to get your number. things started to get skewed with ht numbers, then again with nehalem and of course gpu crunching.

the biggest thing now for me is the different credit that you get from different projects. for example, I can get ~ 50k ppd on my i7 920 @3.6 on aqua@home. it doesn't appear to touch my gpu, but obviously makes good use of all 8 threads on the cpu. I'm only getting 10-15k ppd on seti even with my gtx 260 going full bore. is seti still the standard and aqua is literally 3x + more efficient at doing work? is aqua giving bonus credit just to get extra crunchers who care only about points? what about milky way? if so, why doesn't a truly worthy project like folding@home quadruple their points reward? ah... I miss the good old days.
 
"I upgraded my relic 1GHz Pentium 3 processor to a 1GHz multi-core i7 processor, I've just wasted my money because I'm still contributing 1GHz to my team. "

Basic math, 1Ghz.x 1 =1 Ghz.
1 Ghz.x 4=4 Ghz. = More work done,
 
lane42 said:
"I upgraded my relic 1GHz Pentium 3 processor to a 1GHz multi-core i7 processor, I've just wasted my money because I'm still contributing 1GHz to my team. "

Basic math, 1Ghz.x 1 =1 Ghz.
1 Ghz.x 4=4 Ghz. = More work done,
Click to expand...
Exactly! LOL
 
TN_Chelle said:
What's the main project you're running here? I'm so out of the loop these days!
Click to expand...
GLeeM, like myself, does Folding@Home and we would welcome you to join us. If that's not your cup of tea then there are other projects here at AnandTech. The members involved in those would be equally appreciative but we are the ones you want to be seen with. 😉
 
Except - of course - the granddad of all projects - Seti@Home and then all those biology projects such as SIMAP, WorldCommunityGrid, etc. This are also good projects "to be seen with"!
 
"I upgraded my relic 1GHz Pentium 3 processor to a 1GHz multi-core i7 processor, I've just wasted my money because I'm still contributing 1GHz to my team. "

Basic math, 1Ghz.x 1 =1 Ghz.
1 Ghz.x 4=4 Ghz. = More work done,
Click to expand...
An atom at 1.6GHz is nearly as fast as a Pentium 4 at 3GHz. Look up the megahertz myth.
 
Mr. Pedantic said:
An atom at 1.6GHz is nearly as fast as a Pentium 4 at 3GHz. Look up the megahertz myth.
Click to expand...

We all know these facts. But it is still a grand way to compare bragging rights - and not crunching power. Simple, easy to calculate (compared to e.g. GFLOPS, which in turn depend on the method used, compilers used, and so on).

And what is this comparing good for? Not comparing crunching power but to see who can brag more. And for that any method is good as long as all who want to brag use the same method and as long as it is simple. And to quite cynical: it is the competitive side which is important for most: "I have done more", "I have more crunching power", "I am among the best", "Look at me ...", your 15 minutes in the lime light.
This is what at least in part drives quite a few crunchers. Of course, this is not the only reason and probably not the main one either. Most of us do not understand most of the science in all differnt projects anyway - at least not enough to see more than the lay of the land. We trust those who understand and say: this is good.

Thus: if one cruncher has 4GHz and an other one 40GHz then the latter has more crunching power. That is about how accurate these numbers are.
 
Last edited:
We all know these facts. But it is still a grand way to compare bragging rights - and not crunching power. Simple, easy to calculate (compared to e.g. GFLOPS, which in turn depend on the method used, compilers used, and so on).

And what is this comparing good for? Not comparing crunching power but to see who can brag more. And for that any method is good as long as all who want to brag use the same method and as long as it is simple. And to quite cynical: it is the competitive side which is important for most: "I have done more", "I have more crunching power", "I am among the best", "Look at me ...", your 15 minutes in the lime light.
This is what at least in part drives quite a few crunchers. Of course, this is not the only reason and probably not the main one either. Most of us do not understand most of the science in all differnt projects anyway - at least not enough to see more than the lay of the land. We trust those who understand and say: this is good.

Thus: if one cruncher has 4GHz and an other one 40GHz then the latter has more crunching power. That is about how accurate these numbers are.
Click to expand...
This got me thinking; does that mean that a 4870, say, has 600GHz (0.75*800) of compute power?
 
Sometimes you just have to walk away. You shouldn't try to teach a pig to sing, it just frustrates you and annoys the pig.....

😀

Back in the old days, 😉 , using total Ghz of your cruncher farm made sense, before the GPU, multi-core and many flavors of CPUs. At least for the projects I ran. X number of WU for so much CPU Mhz (yes, Mhz, I started running RC5 back when OCing a Celeron 300mhz to 450mhz and 500mhz was the latest and greatest thing :O ) and Ghz add it up and you had a pretty good idea of what your farm could produce. 🙂

Now days, it's pretty confusing as I haven't kept up with it. So all I got to say is:

You punks get the H*## off my lawn!

😀
 
LoL, Where is your cane, RaySun2Be? Or was it the fish?

Well, I really agree with you. And I still think that for the past 3 -4 years the GHz is only for bragging rights ... Calculating in that way I could write that I have xyz.ab GHz (where x is > 1, y > 2, z > 3 and a < 9 and b < 8).
What does that matter?

I could alsop write that my Whetstones > 250K, my Dhrystones > 550K but that does not take into account all the GPUs I have crunching.

What does this matter? For me only the results matter - not the number of credits, but the number of WUs crunched. It is the WUs that help the scientists to do their work and to do science. And that value is hard to calculate; many man years have been devoted to understand the efficiency of scientific research - because the effects of a certain finding may not become apparent until some - somtimes many - years after the publication.

EG: one of my own papers started again to get quoted in other scientists' papers some 3 years ago. The work was done in 1984 - 1985, published in 1986, and until 2006 quoted 90 times. Between Jan 2007 and Dec 2009 it was quoted 340 times ...

QED
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top