Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Well the 8P is still 2 sticks for each 6 cores, just DDR2 instead of 3..
Yes, they are talking about memory bandwidth. DDR2 has less bandwidth than DDR3.
DDR2 667: 12.0ns Cas 4
DDR2 800: 12.5ns Cas 5
DDR3 1333: 9.0ns Cas 6
So you can see the difference there.
Since they came out all I get is 8101 (3 so far).
At this rate I might as well use the 4P to render home made porn instead...
Are other people just getting 8101's as well?
I have 50% of the components on order now for my third 4P rig.
Ref the points,found this very informative post by ChasR over on OCF
http://www.overclockers.com/forums/showpost.php?p=7172418&postcount=12
Ref the points,found this very informative post by ChasR over on OCF
That would answer a lot of questions, but not this one - Why does it take a user to provide definitive information? Better than before, yes, but it's like pulling teeth getting Pande Group communicate effectively with their users. Stereotypical case of (unintentionally) aloof professors? Perhaps the 'definitive' information has always been out there, but it takes a concerted effort to find it?The points on p8101 are lower, because they are closer to right than the points on p6903 and p6904. PG actually used a 4P machine to determine what they should be, then added to that so that they would be closer to the p690x series to avoid mutiny from the 4P donors.
THe BA points are so huge because the WUs were benchmarked on an i5. THe WUs filled the cache completely, slowing them way down, and made about 20% more than regular smp WUs. However run on a 4P machine, without any cache constraint, the BA WUs make much more than regular smp work. Even though reduced from p690x, p8101 still makes 70% more ppd than regular smp according to the data I've seen. Let me know if that data isn't correct.
I hope this is one of the rare teams that thanks PG for the nice bonus provided by p690x, rather than whines incessantly about the loss of ppd from p8101. So far so good.
Nathan linked: That would answer a lot of questions, but not this one - Why does it take a user to provide definitive information? Better than before, yes, but it's like pulling teeth getting Pande Group communicate effectively with their users. Stereotypical case of (unintentionally) aloof professors? Perhaps the 'definitive' information has always been out there, but it takes a concerted effort to find it?
Don't be a hata 4!
I know it's getting [H]ard to keep up though.
That said I am impressed by VJ's comments in this thread : http://foldingforum.org/viewtopic.php?f=16&t=21367
4. GPU folding in Linux. People who run Linux know what they're doing, and I think there's enough computing potential there to justify the pursuit.
Point was made in fun. I guess I should ahve added the smiley face.
Point was made in fun. I guess I should ahve added the smiley face.
Nathan linked: That would answer a lot of questions, but not this one - Why does it take a user to provide definitive information? Better than before, yes, but it's like pulling teeth getting Pande Group communicate effectively with their users. Stereotypical case of (unintentionally) aloof professors? Perhaps the 'definitive' information has always been out there, but it takes a concerted effort to find it?
You actually have to visit the folding forum and folding web site once in a while to get information from PG. They don't deliver to other forums.
For example, here is where kasson from PG talked about 6903s/6904s thrashing the cache and getting over-benchmarked vs. the p8101 work units...http://foldingforum.org/viewtopic.php?p=207086#p207086 A thread several of your fellow [H]er's were active in, but neglected to relay that info.
The i5 benchmark is described in the Points FAQ on the fah web site.
etc... Not trying to pick a fight, just saying it's not about pulling teeth, or aloof professors. The info is usually out there.
One guy ventures a guess, blind followers take it as revealed truth.Grandpa_01 said:(...) very seriously doubt that Kasson or anybody else even tested that theory
Linden you do realise this is I believe another one of those thread manipulations that the three amigos are so fond of, I suggested in that thread that 7im pointed to that they bench with a 4P machine and somehow it got turned into they did benchmark with a 4P. Some people change things to fit there wants and needs and the three amigos do it on a regular basis. I would like to see the thread where it says they were benchmarked on a 4P because if it exists I have never seen it. I try not to miss things at the FF, I could have but I doubt it. I could always PM Kasson and ask him, if it comes to that.That would answer a lot of questions, but not this one - Why does it take a user to provide definitive information? Better than before, yes, but it's like pulling teeth getting Pande Group communicate effectively with their users. Stereotypical case of (unintentionally) aloof professors? Perhaps the 'definitive' information has always been out there, but it takes a concerted effort to find it?Nathan linked:
Quote:
The points on p8101 are lower, because they are closer to right than the points on p6903 and p6904. PG actually used a 4P machine to determine what they should be, then added to that so that they would be closer to the p690x series to avoid mutiny from the 4P donors.
THe BA points are so huge because the WUs were benchmarked on an i5. THe WUs filled the cache completely, slowing them way down, and made about 20% more than regular smp WUs. However run on a 4P machine, without any cache constraint, the BA WUs make much more than regular smp work. Even though reduced from p690x, p8101 still makes 70% more ppd than regular smp according to the data I've seen. Let me know if that data isn't correct.
I hope this is one of the rare teams that thanks PG for the nice bonus provided by p690x, rather than whines incessantly about the loss of ppd from p8101. So far so good.
I have 50% of the components on order now for my third 4P rig.
I hate you.
Well he is getting a little closer now he just needs to edit the rest of the fud out of it and he might get it right. What was actually said is that they we're also running these on a few of their own systems. and that preliminarily, the scaling data on these looked fairly good. And Kasson said I think the issue may be that 6903/6904 thrashed the cache on their benchmark machine and took much longer (hence the big jump in points assigned). But that's speculation. Kasson also said as they get timing data from a number of systems, they will re-evaluate the points assignment.Nathan linked:
Quote:
The points on p8101 are lower, because they are closer to right than the points on p6903 and p6904. PG actually used a 4P machine to determine what they should be, then added to that so that they would be closer to the p690x series to avoid mutiny from the 4P donors.
THe BA points are so huge because the WUs were benchmarked on an i5. THe WUs filled the cache completely, slowing them way down, and made about 20% more than regular smp WUs. However run on a 4P machine, without any cache constraint, the BA WUs make much more than regular smp work. Even though reduced from p690x, p8101 still makes 70% more ppd than regular smp according to the data I've seen. Let me know if that data isn't correct.
I hope this is one of the rare teams that thanks PG for the nice bonus provided by p690x, rather than whines incessantly about the loss of ppd from p8101. So far so good.
The points on p8101 are lower, because they are closer to right than the points on p6903 and p6904. PG actually used data from a variety of machines to help determine what they should be.
THe BA points are so huge because the WUs were benchmarked on an i5. THe WUs filled the cache completely slowing them way down, and made about 20% more than regular smp WUs. However run on a 4P machine, without any cache constraint, the BA WUs make much more than regular smp work. Even though reduced from p690x, p8101 still makes 70% more ppd than regular smp according to the data I've seen. Let me know if that data isn't correct.
I hope this is one of the rare teams that thanks PG for the nice bonus provided by p690x, rather than whines incessantly about the loss of ppd from p8101. So far so good.
I triple hate you, i don't have any parts for my 1st 4p yet
Quoted statement 2 days ago
Well he is getting a little closer now he just needs to edit the rest of the fud out of it and he might get it right. What was actually said is that they we're also running these on a few of their own systems. and that preliminarily, the scaling data on these looked fairly good. And Kasson said I think the issue may be that 6903/6904 thrashed the cache on their benchmark machine and took much longer (hence the big jump in points assigned). But that's speculation. Kasson also said as they get timing data from a number of systems, they will re-evaluate the points assignment.
Quote statement from today
Now you all know what was actually said and do not go and spread this fud any further.
http://www.overclockers.com/forums/showpost.php?p=7172418&postcount=12
Come on then! What are you waiting for?
Just fortunate to be able to fund it, and got the itch bad when I read about these....
One guy ventures a guess, blind followers take it as revealed truth.
Kasson said what he said about 6903s/6904s. Kendrak, am I guessing blindly or not?
On another note, I did not see where PG said they benched on a 4P system as ChasR said, but then I don't have access to private conversations ChasR may or may not have had with kasson. The rest is accurate. Your own PPD tracking charts show the 8108s inline with most projects, while 6903s/6904s are the out-liers in the group. I'm open to any explanation, even one that contradicts what kasson has said. But until someone has a better idea, the statements fit the facts so far. Sorry Gramps.
You will be asking the wrong question, and won't get an answer. Others have tried.
He did not "change" what he originally posted. He added to it, reinforcing what he already said. 6903/4s were benched too high (and your charts prove it), but it would cause a bigger stink over [H]ere to fix the PPD (lower them on 6903/4s) than to simply let those two projects finish out quietly and go away without so muc[H] stink.
I don't need any additional verbal confirmation from kasson. He'll provide plenty of confirmation by releasing more and more new bigadv projects at the points level of the 8101s and 6901s.