teraFLOPS estimate?

Message boards : Number crunching : teraFLOPS estimate?

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile David

Send message
Joined: 2 Aug 06
Posts: 2
Credit: 277,723
RAC: 0
Message 35508 - Posted: 25 Jan 2007, 19:58:19 UTC

Why does the estimate always wabble between the high 30,000's and the low 40,000's? With all of these new users and hosts each day you'd think we'd be approaching the mid to high 40,000's. What am I missing here?
ID: 35508 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Christoph Jansen
Avatar

Send message
Joined: 6 Jun 06
Posts: 248
Credit: 267,153
RAC: 0
Message 35515 - Posted: 25 Jan 2007, 22:15:13 UTC
Last modified: 25 Jan 2007, 22:28:48 UTC

Hi David,

first of all: the active users are pretty constant during the last 2 months. In other words: Rosetta gains about as many users as it loses in the same time.

Another effect that I guess is this: many Mac users have turned away from Rosetta because of too little credits per WU. But these machines contributed to the credits dealt out by the new credit system a lot. In numbers: a certain Mac may have claimed something around 10 to 15 credits per hour but was only granted 3 to 5. In words: these machines claim many credits per decoy, but do not really give the performance. Credits were inflated by some degree by these machines. With them gone, this inflation is gone too.

On the other hand a lot of the new machines are pretty fast and their architecture brings it with it that they show in the benchmarks much poorer than they really are. This means, that e.g. a Core 2 Duo may claim 30 credits per hour but is granted 50 or 60 or 70, meaning it contributes much less to the accumulated credits dealt out than it takes out. This way the credits the project shows are lower than the really would be regarding the processing power.

In short: Macs leaving and new machines coming in which have more crunching power than they show in the benchmarks both lower the credits.

To further show what I mean:Here is the link to the thread where I did an example calculation showing how the new credit system works in this respect. Three posts higher there is another example of what happens when a Mac reports first and when a Core 2 reports first.

If you read the example carefully you will note: if all machines were Core 2s, then all machines would get the same low credits als claimed and granted would be equal. To say it another way: the more high-power machines we have the less will the credits really reflect the crunching power behind the project unless someone really comes up with a new benchmarking system that tells what a machine does on Rosetta and not on the uniform and pretty useless BOINC benchmark.

[Edit] What I am saying about "inflation" is in no way intended as a pun against pre-Intel Macs, on the contrary. The Rosetta project, the way it presently does its calculations, is not optimized for pre-Intel Macs in any way and thus underrates the true power of these machines. Leaving Rosetta was thus just natural for many Mac users who either feel disregarded or think they can contribute more in other places. I really value this project, but I am not sure if it was still my choice if I was a pre-Intel Mac user myself [/Edit]
ID: 35515 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile rochester new york
Avatar

Send message
Joined: 2 Jul 06
Posts: 2803
Credit: 1,852,420
RAC: 490
Message 35558 - Posted: 26 Jan 2007, 13:54:37 UTC - in response to Message 35515.  

Hi David,

first of all: the active users are pretty constant during the last 2 months. In other words: Rosetta gains about as many users as it loses in the same time.

Another effect that I guess is this: many Mac users have turned away from Rosetta because of too little credits per WU. But these machines contributed to the credits dealt out by the new credit system a lot. In numbers: a certain Mac may have claimed something around 10 to 15 credits per hour but was only granted 3 to 5. In words: these machines claim many credits per decoy, but do not really give the performance. Credits were inflated by some degree by these machines. With them gone, this inflation is gone too.

On the other hand a lot of the new machines are pretty fast and their architecture brings it with it that they show in the benchmarks much poorer than they really are. This means, that e.g. a Core 2 Duo may claim 30 credits per hour but is granted 50 or 60 or 70, meaning it contributes much less to the accumulated credits dealt out than it takes out. This way the credits the project shows are lower than the really would be regarding the processing power.

In short: Macs leaving and new machines coming in which have more crunching power than they show in the benchmarks both lower the credits.

To further show what I mean:Here is the link to the thread where I did an example calculation showing how the new credit system works in this respect. Three posts higher there is another example of what happens when a Mac reports first and when a Core 2 reports first.

If you read the example carefully you will note: if all machines were Core 2s, then all machines would get the same low credits als claimed and granted would be equal. To say it another way: the more high-power machines we have the less will the credits really reflect the crunching power behind the project unless someone really comes up with a new benchmarking system that tells what a machine does on Rosetta and not on the uniform and pretty useless BOINC benchmark.

[Edit] What I am saying about "inflation" is in no way intended as a pun against pre-Intel Macs, on the contrary. The Rosetta project, the way it presently does its calculations, is not optimized for pre-Intel Macs in any way and thus underrates the true power of these machines. Leaving Rosetta was thus just natural for many Mac users who either feel disregarded or think they can contribute more in other places. I really value this project, but I am not sure if it was still my choice if I was a pre-Intel Mac user myself [/Edit]

a neat web site you might like...............http://www.rense.com/general72/size.htm
ID: 35558 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1676
Credit: 88,754,000
RAC: 40,207
Message 35559 - Posted: 26 Jan 2007, 14:23:33 UTC

the scripts that credit errored tasks will probably throw the terraflops estimate out too as there'll suddenly be a jump in credits (assuming they're still running those scripts at intervals).
ID: 35559 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Tom Philippart
Avatar

Send message
Joined: 29 May 06
Posts: 183
Credit: 834,667
RAC: 0
Message 35915 - Posted: 1 Feb 2007, 15:08:57 UTC

wow it's a long time since I last saw such a low Tflops estimate:
22.138
ID: 35915 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile rochester new york
Avatar

Send message
Joined: 2 Jul 06
Posts: 2803
Credit: 1,852,420
RAC: 490
Message 53722 - Posted: 16 Jun 2008, 18:31:53 UTC

why did the tflops go down over 10 in the last 2 or 3 weeks ???????
ID: 53722 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David Emigh
Avatar

Send message
Joined: 13 Mar 06
Posts: 158
Credit: 417,178
RAC: 0
Message 53732 - Posted: 17 Jun 2008, 1:31:25 UTC

It looks to me as if the decline has taken place in just the last 6 days.

Reviewing the 60 day history, this looks like a pretty serious drop.
Rosie, Rosie, she's our gal,
If she can't do it, no one shall!
ID: 53732 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Tom Philippart
Avatar

Send message
Joined: 29 May 06
Posts: 183
Credit: 834,667
RAC: 0
Message 53738 - Posted: 17 Jun 2008, 10:37:06 UTC
Last modified: 17 Jun 2008, 10:47:43 UTC

Everybody now supports POEM@Home during the CASP race. Many crunchers moved (temporarily?) to this project as I predicted last month in the "CASP8" thread in the science section...
POEM went from 7Tflops to over 35Tflops!

I suppose a number of them will still come back for July after this race.
ID: 53738 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Greg_BE
Avatar

Send message
Joined: 30 May 06
Posts: 4875
Credit: 4,510,296
RAC: 1,069
Message 53758 - Posted: 18 Jun 2008, 11:06:10 UTC - in response to Message 53738.  

Everybody now supports POEM@Home during the CASP race. Many crunchers moved (temporarily?) to this project as I predicted last month in the "CASP8" thread in the science section...
POEM went from 7Tflops to over 35Tflops!

I suppose a number of them will still come back for July after this race.


I'm tempted to move over there due to all the errors I am getting now.
ID: 53758 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jeremy

Send message
Joined: 15 May 08
Posts: 13
Credit: 2,636
RAC: 0
Message 53760 - Posted: 18 Jun 2008, 12:26:15 UTC

Due to all the errors I did move away to Poem@home parialy, i keep checking if there are Beta WU's. The main thing keeping me at rosetta are the Graphics of the beta version. I'm sure this doesn't count for everyone but the errors are a pain.

ID: 53760 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David Emigh
Avatar

Send message
Joined: 13 Mar 06
Posts: 158
Credit: 417,178
RAC: 0
Message 53796 - Posted: 18 Jun 2008, 21:52:01 UTC

Over the past 8 days, the estimated teraflops for the project has declined by 20%

This is not a blip, but a week long trend that shows no sign of bottoming out.

What is happening?
Rosie, Rosie, she's our gal,
If she can't do it, no one shall!
ID: 53796 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David E K
Volunteer moderator
Project administrator
Project developer
Project scientist

Send message
Joined: 1 Jul 05
Posts: 1018
Credit: 4,003,213
RAC: 0
Message 53802 - Posted: 18 Jun 2008, 22:35:18 UTC

we're looking into this right now. It's a little difficult to figure out the cause.
ID: 53802 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David Emigh
Avatar

Send message
Joined: 13 Mar 06
Posts: 158
Credit: 417,178
RAC: 0
Message 53819 - Posted: 19 Jun 2008, 2:48:08 UTC
Last modified: 19 Jun 2008, 2:48:20 UTC

Thank you very much =)
Rosie, Rosie, she's our gal,
If she can't do it, no one shall!
ID: 53819 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David Emigh
Avatar

Send message
Joined: 13 Mar 06
Posts: 158
Credit: 417,178
RAC: 0
Message 53922 - Posted: 23 Jun 2008, 13:42:15 UTC

I hope I don't jinx this by taking note of it too soon but...

It appears the teraFLOP hemorrhaging has stopped.
Rosie, Rosie, she's our gal,
If she can't do it, no one shall!
ID: 53922 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile rochester new york
Avatar

Send message
Joined: 2 Jul 06
Posts: 2803
Credit: 1,852,420
RAC: 490
Message 53924 - Posted: 23 Jun 2008, 13:48:53 UTC - in response to Message 53922.  

I hope I don't jinx this by taking note of it too soon but...

It appears the teraFLOP hemorrhaging has stopped.

hope so .....it was getting very disappointing
ID: 53924 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 53958 - Posted: 24 Jun 2008, 15:26:49 UTC

while I hope I'm wrong, it's possible that folks are leaving for Folding@Home. F@H just added nVidia cards and there is considerable excitement around F@H. They are now getting over 700 teraflops from GPUs alone (not counting PS3)
ID: 53958 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile David Emigh
Avatar

Send message
Joined: 13 Mar 06
Posts: 158
Credit: 417,178
RAC: 0
Message 53959 - Posted: 24 Jun 2008, 15:33:14 UTC - in response to Message 53958.  

while I hope I'm wrong, it's possible that folks are leaving for Folding@Home. F@H just added nVidia cards and there is considerable excitement around F@H. They are now getting over 700 teraflops from GPUs alone (not counting PS3)


I think that most of the decline was due to t405 tasks stalling clients. We'll have to wait and see how much the teraFLOPS recovers over the next few days
Rosie, Rosie, she's our gal,
If she can't do it, no one shall!
ID: 53959 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BrnmccO1

Send message
Joined: 26 Jun 07
Posts: 17
Credit: 578,825
RAC: 0
Message 53977 - Posted: 25 Jun 2008, 0:13:06 UTC
Last modified: 25 Jun 2008, 0:14:14 UTC

Back up to about 57 t-flops now! :))

Drat those t405's.... keep up the good work, in a couple days it will be one year to the day I put my boxes on Rosetta. (Not going to mention the prev project I was on, that was an "Epic Fail", lol)

- Happy Camper (well, cruncher)
ID: 53977 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile JChojnacki
Avatar

Send message
Joined: 17 Sep 05
Posts: 71
Credit: 8,844,677
RAC: 0
Message 54623 - Posted: 23 Jul 2008, 2:56:37 UTC

It took a little bit, but I am glad to see the TFLOPS estimate back to 70. :-)

Crunch on.



ID: 54623 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BrnmccO1

Send message
Joined: 26 Jun 07
Posts: 17
Credit: 578,825
RAC: 0
Message 55014 - Posted: 10 Aug 2008, 5:26:41 UTC

74 now, gogogo! Lol, ;p
ID: 55014 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · Next

Message boards : Number crunching : teraFLOPS estimate?



©2021 University of Washington
https://www.bakerlab.org