Author |
Message |
|
What sort of computer takes less than 30 seconds for the 1418488458 839143435 10 Dec 2022 | 4:03:43 UTC 26 Dec 2022 | 14:06:20 UTC Completed and validated (1st) [Proof task] 21.19 0.00 955,403.42 Do You Feel Lucky?
Anonymous platform (CPU)
Just curious.
Thanks |
|
|
Crun-chi Volunteer tester
 Send message
Joined: 25 Nov 09 Posts: 3247 ID: 50683 Credit: 152,646,050 RAC: 18,212
                         
|
There is answer on discord: that computer is just gateway from computer that process WU.
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|
|
There is answer on discord: that computer is just gateway from computer that process WU.
Do you have a link to that discord post? I'd like to read it. Or what channel and when was it posted? |
|
|
|
Essentially it was a user who has lots of hardware and manages it through that linked computer. The send/recieve duration was 9 days i believe (the discord example was). No unfortunately this does not mean someone has perfected quantum computing or anything else exotic. |
|
|
|
1418488458 839143435 10 Dec 2022 | 4:03:43 UTC 26 Dec 2022 | 14:06:20 UTC Completed and validated (1st) [Proof task] 21.19 0.00 955,403.42 Do You Feel Lucky?
Anonymous platform (CPU)
For those who like to click and see:
Task: 1418488458
Workunit: 839143435
All valid DYFL tasks for that formal host
(These links will expire quite soon when the workunit is purged from the database of the server.)
/JeppeSN |
|
|
|
So the run time and cpu time reported are not real and the actual run time was probably 11-12 days? I didn't know a WU result could be constructed with fake stats and pass validation but I guess the project admins are aware of how it works and it's all legit. |
|
|
|
If the credit remains many others will do the same and the point system will have to be revised or abandoned? |
|
|
|
If the credit remains many others will do the same and the point system will have to be revised or abandoned?
The run time and cpu time stats are faked but the result is apparently legit. I'm sure the project admins are aware of this custom app and have blessed it
On the Discord channel, Magic Mike commented on the computer and is aware of how it works. I assume Magic Mike is either Michael Gutierrez or Michael Goetz, both are project admins.
There were no details about the computer or app that actually completed any of the WUs but it wasn't actually the 4 thread computer where the tasks are listed. That 4 thread computer acts as an interface that feeds the actual computer(s) WUs because there is no Boinc application for that computer(s) hardware-OS combo. DYFL actually doesn't have CPU app. Dunno. The actual app could be something like the current GPU app for DYFL but written for the Nvidia Jetson GPUs that be used on raspberry pi's and other SBCs. That would make a lot of crunchers happy if an app were released that ran on those |
|
|
|
Time will tell if any damage to the stats or it makes a new type of parameters to be used for the stats.
It would be nice if the numbers/speed was real and could be used toward medical research. |
|
|
|
Credit is not based on the time used by users to complete a task. If you Liquid NO2 voltage shunt a 4090 or put a 4090 next to a space heater and have its fans off so it thermal throttles to nothing makes no difference to credit per task. It may mess with statistics but not credit. |
|
|
|
If that is how fast the 4090's are we have entered a new age. I understand the project is scored by the wu's value and not the speed. However it does help if one can do 1000 a day versus 1/10th. |
|
|
|
If that is how fast the 4090's are we have entered a new age. I understand the project is scored by the wu's value and not the speed. However it does help if one can do 1000 a day versus 1/10th.
Miklos,
Again, those times are not real. If they were using a GPU, it's likely that system could have run the boinc client from the repository. I imagine they were running a custom CPU app. BTW, a RTX 3090 Ti can turn around a DYFL WU in about 15 hours. I can't imagine a RTX 4090 is significantly faster than the RTX 3090 Ti. |
|
|
|
Thank you for the explanation. |
|
|
Michael Goetz Volunteer moderator Project administrator
 Send message
Joined: 21 Jan 10 Posts: 14037 ID: 53948 Credit: 477,161,398 RAC: 289,514
                               
|
The timing isn't used to calculate the credits, so faked times don't matter at all. In fact, sometimes we fake those numbers ourselves, for various reasons.
This is hardly the first person to have used a gateway to enable non-BOINC computers to run PrimeGrid tasks.
This particular person happens to be one of the developers here. Everything is Kosher. :)
____________
My lucky number is 75898524288+1 |
|
|
|
This type of feeder system has been in operation for years at different sites and has been brought up before. You can find older messages about it. Often times in is in an environment that may have specialized firewalls in place and can do the work inside these firewalls, but the feeder is just a feeder and it does not crunch, so it has access outside the firewall. I heard of some companies/schools that only allow this type of processing. |
|
|
tng Send message
Joined: 29 Aug 10 Posts: 499 ID: 66603 Credit: 50,812,070,359 RAC: 31,720,205
                                                    
|
If that is how fast the 4090's are we have entered a new age. I understand the project is scored by the wu's value and not the speed. However it does help if one can do 1000 a day versus 1/10th.
Miklos,
Again, those times are not real. If they were using a GPU, it's likely that system could have run the boinc client from the repository. I imagine they were running a custom CPU app. BTW, a RTX 3090 Ti can turn around a DYFL WU in about 15 hours. I can't imagine a RTX 4090 is significantly faster than the RTX 3090 Ti.
My 4080 does a DYFL in 9 hours: https://www.primegrid.com/result.php?resultid=1426017579
____________
|
|
|
|
Wow. My laptop 3080 150W GPU took 102000 seconds(1d4h) only to end up in compute error.
40XX series is a whole new era in GPU computing, also cost-wise. |
|
|
|
I assume Magic Mike is either Michael Gutierrez or Michael Goetz, both are project admins.
Magic Mike is Michael Goetz. /JeppeSN
|
|
|
Crun-chi Volunteer tester
 Send message
Joined: 25 Nov 09 Posts: 3247 ID: 50683 Credit: 152,646,050 RAC: 18,212
                         
|
I assume Magic Mike is either Michael Gutierrez or Michael Goetz, both are project admins.
Magic Mike is Michael Goetz. /JeppeSN
Oh no, you reveal true identity, he now must kill you :P
____________
92*10^1585996-1 NEAR-REPDIGIT PRIME :) :) :)
4 * 650^498101-1 CRUS PRIME
2022202116^131072+1 GENERALIZED FERMAT
Proud member of team Aggie The Pew. Go Aggie! |
|
|