Let's Play CreditNew (Credit & RAC support thread)

Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread)
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 13 · Next

AuthorMessage
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933877 - Posted: 5 May 2018, 22:51:02 UTC

A thread for CreditNew troubleshooting in regard to seti@home. Boinc-wide credit is a bit beyond the scope of this thread.
All questions welcome. I'll be asking quite a few myself. I'll also be snipping quotes from older threads to get the ball rolling.

Problem numero uno:
Personally I can't catch CreditNew being anything other than the flopcounter it was meant to be. The math isn't too hard and I'm certainly not the sharpest tool in THIS shed so this should be easy for a lot of people here.
I've taken "the credit system is broken" as dogma for years and to be honest, I suddenly realized I haven't seen any real evidence. Could it be 5-10% wrong? Yeah sure. But 30% or 50%? Compared to -let's say- 2008? I'm not seeing it. Not on a monthly scale anyway.

We've all had our share of making fun of the quirks, how 'bout we put our heads together and figure out what (if anything) needs fixing? The math is pretty elementary and if DA ever bothered to explain the thing in a video, said video would likely be 90 seconds long to take us through the path of a stock SETI task (sans cheating). Maybe 5 min. total if he wanted to bring up all the "if found cheating then" detours.

Let's crack this.

(On a side note, any Swedish humor about us all being a bunch of masochists is warmly welcome)
ID: 1933877 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933880 - Posted: 5 May 2018, 23:03:03 UTC

Let's take it from the top and see if we don't end up with Cobblestones instead of Bank of Zimbabwe dollars :)

Let's say I have a CPU task whose flopcounter says 4500 Gflops. Next step for this number is to get run by the measured floating point speed of my CPU (in my case 4.01 billion ops/sec), correct?
ID: 1933880 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1933885 - Posted: 5 May 2018, 23:34:37 UTC - in response to Message 1933877.  
Last modified: 5 May 2018, 23:36:00 UTC

But 30% or 50%? Compared to -let's say- 2008? I'm not seeing it. Not on a monthly scale anyway.

Because the amount of Credit now paid out is considerably less than what it was- have you got historical records of WUs from back then, with their AR & credit granted to compare?
If you're lucky, while you're searching you'll come across a graph someone posted that had a cruncher that had no hardware changes & was a Seti only cruncher, and you could see the drop over time from when Credit New was introduced. No change on work mix or applications or hardware, just the change in Credit system.
Then there was the period some time ago now where we had no Arecibo work at all, just GBT. And the Credit per WU dropped off significantly. When Arecibo work started flowing again, Credit per WU rose again.

In the other thread, Tbar posted some WUs for comparison- as you pointed out the GPU pairs were getting more Credit than they should. But the fact is GPUs paired with CPUs tend to get less Credit (this is by design due to the "efficiency" section of Credit New). And all processors are subject to differing amounts of Credit depending on who their wingman is, even for WUs of the same AR.

I've forgotten their name, but i'm pretty sure in one of the other threads on Credit someone posted a short list of AR with Credit granted, and the numbers were all over the place.
Of course now we've got the added mix of Arecibo VLARs on Nvidia GPUs to go through as well.


If we're lucky, we'll get another extended period of GBT only tasks. That should allow the Credit for a given AR to take a dive, and it will make it easier to find multiple WUs of a given AR to compare Credit granted to each.
Grant
Darwin NT
ID: 1933885 · Report as offensive
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34744
Credit: 261,360,520
RAC: 489
Australia
Message 1933889 - Posted: 6 May 2018, 0:15:21 UTC

But 30% or 50%? Compared to -let's say- 2008? I'm not seeing it. Not on a monthly scale anyway.
We used to get paid around 140-160% more for doing similar tasks back in those days compared to now and with the old hardware that I had back then makes a mockery of what I have now.

Cheers.
ID: 1933889 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933893 - Posted: 6 May 2018, 0:25:44 UTC - in response to Message 1933885.  

All I'm saying is we've fried our brains enough trying to reverse engineer CreditNew. In audio you follow the signal path to figure out where the problem is. Let's follow a task and see what it does. How hard can it be?

If you're lucky, while you're searching you'll come across a graph someone posted that had a cruncher that had no hardware changes & was a Seti only cruncher, and you could see the drop over time from when Credit New was introduced.


I'm 90% sure you're talking about the v7 rollout.

If there was an uprising about CN I either need to refresh my memory or I wasn't around. But for some reason I had the feeling that when CN was introduced, nobody noticed. Do we have a date? I'm also pretty sure that during The Golden Age of Seti Credit back when Jason's 560Ti rig was doing a million credz a month easy... not a soul was complaining about credits even though CreditNew was firmly in place. (I'm talking about monthly throughput, not the odd task getting whacky credit).

Richard would know.
ID: 1933893 · Report as offensive
Sirius B Project Donor
Volunteer tester
Avatar

Send message
Joined: 26 Dec 00
Posts: 24876
Credit: 3,081,182
RAC: 7
Ireland
Message 1933897 - Posted: 6 May 2018, 0:45:55 UTC

Well this thought was going to be put in your other thread but...

Over the years on checking tasks I used to get the occasional inconclusive, however this year I've had quite a few.

This one I'm safely assuming I'll get credit for.

This one I'm safely assuming I won't
ID: 1933897 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933900 - Posted: 6 May 2018, 1:06:46 UTC - in response to Message 1933885.  

I'm really good at understanding the things I want to understand but not good at conveying that understanding to others. IOW I'm a crap teacher :) Reason I'm saying this is because what I'm about to say is definitely correct and I'll do my best to explain why, but you may again need someone like Richard to explain it in a way that will help you understand :)

(Raistmer has explained this countless times but I'm guessing he's not getting through to you guys either)

Then there was the period some time ago now where we had no Arecibo work at all, just GBT. And the Credit per WU dropped off significantly. When Arecibo work started flowing again, Credit per WU rose again.

Unfortunately I'm going to have to borrow DA's "and that's OK" here. And give you a dreaded car analogy to explain why.

My car does 75mph on the highway. So in one hour my car adds 75 miles to my meter.

(Here's what a lot of you are saying and so was I, a while back)

I've been on this mountain road for exactly one hour so why the f* has my meter only added 32 miles? It should be 75 ffs!

:)
ID: 1933900 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1933903 - Posted: 6 May 2018, 1:28:41 UTC - in response to Message 1933900.  

Unfortunately I'm going to have to borrow DA's "and that's OK" here.

Why is it OK for WUs with the same AR to receive less Credit? Or as you pointed out with the GPUs paring, get more?

My car does 75mph on the highway. So in one hour my car adds 75 miles to my meter.
(Here's what a lot of you are saying and so was I, a while back)
I've been on this mountain road for exactly one hour so why the f* has my meter only added 32 miles? It should be 75 ffs!

If you're going to use an analogy, it needs to be analogous.

What we're complaining about is that a car doing 120km/h on the highway will travel 120km in an hour. And if you're still doing 120km/h on a mountain you should still travel 120km in that hour.
And just because someone decides it's not possible to do 120km/h on the mountain, doesn't mean you won't travel 120km in 1 hour if you are able to do 120km/h there.

Credit payment should be based on AR and you shouldn't loose Credit based on some perception of efficiency or perceived cheating or other quirks resulting from Estimates of work done & estimates of hardware performance.
The amount of Credit paid shouldn't vary just because you are processing a different type of WU- If it takes 1,250 GFLOPS to process a Arecibo WU, and it takes 1250 GFLOPS to process a GBT WU, they should both get the same Credit.
Grant
Darwin NT
ID: 1933903 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933916 - Posted: 6 May 2018, 3:57:15 UTC - in response to Message 1933903.  

Whoa wait!

I've only had this laptop about a month and still playing around with it, with little time for Seti...
Are Arecibo and Green Bank sharing any CPU apps?
ID: 1933916 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1933917 - Posted: 6 May 2018, 4:07:10 UTC - in response to Message 1933916.  

The cpu apps are the same for Green Bank and Arecibo.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1933917 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933921 - Posted: 6 May 2018, 4:23:30 UTC - in response to Message 1933917.  

That's asking for all kinds of trouble.
ID: 1933921 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1933937 - Posted: 6 May 2018, 7:13:59 UTC - in response to Message 1933893.  

Do we have a date? Richard would know.
CreditNew was introduced in the same server update that deployed the v6.10 CUDA application to support Fermi GPUs. That would have been in May 2010, but no, I don't have an exact date.

Some of us had concerns about CN, but no - there wasn't an uprising. People were too busy thrashing their Fermis to the max.
ID: 1933937 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22160
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1933965 - Posted: 6 May 2018, 12:19:13 UTC - in response to Message 1933921.  

That's asking for all kinds of trouble.

Why??
The data format that we see is the same for the two sources, and the application was designed around that format.
There are two differences in the actual files, first, and most obvious, is the naming conventions used, and second the Angular Range, which is generally much smaller for the data coming from the GBT than from Arecibo, and this is mostly due the GBT doing target data collection whereas Arecibo has done a lot of sky-sweeps.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1933965 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1933967 - Posted: 6 May 2018, 12:34:23 UTC

Uhhh,,,,,'troubleshoot' CreditNew?
That's a good one....LOL!
We all know it's a random number generator.
Meowmeowmeow!!!
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1933967 · Report as offensive
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 22160
Credit: 416,307,556
RAC: 380
United Kingdom
Message 1933969 - Posted: 6 May 2018, 13:05:41 UTC

Something to chew on - There is nothing that we, as users, can do to influence the way credit is awarded, but there are many ways we can disturb the so-called balance of the system.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 1933969 · Report as offensive
mmonnin
Volunteer tester

Send message
Joined: 8 Jun 17
Posts: 58
Credit: 10,176,849
RAC: 0
United States
Message 1933971 - Posted: 6 May 2018, 13:24:56 UTC
Last modified: 6 May 2018, 13:25:04 UTC

There might be some results in the Internet achieves of the forum. A random date in 2009.
https://web.archive.org/web/20090204071307/http://setiathome.berkeley.edu:80/forum_forum.php?id=10
ID: 1933971 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14649
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1933977 - Posted: 6 May 2018, 14:19:55 UTC - in response to Message 1933971.  

Don't need the Wayback Machine - we've got our own here. Here are a couple of samples from about 7,550 posts back in my history.

message 1001547
message 1002275
ID: 1933977 · Report as offensive
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 173
Credit: 509,430
RAC: 0
Greece
Message 1933981 - Posted: 6 May 2018, 14:40:00 UTC

My apologies Grant. I think it's obvious I thought you were talking about each app's efficiency. I've been pretty much away from Seti these past couple of years with a new daughter and all :) So I missed GB rollout.

But yeah, I understand what you are saying now. If that's true then yes, technically you can blame this behaviour on CN. I'm guessing if the GFLOPS are the same between telescope tasks but the credits are not, then that means the runtimes are (likely) not either. Which of course will throw CN into a hissy fit. Decoupling the 2 telescopes and giving them their own app (even just a dummy duplicate for GB) would fix that.

If it takes 1,250 GFLOPS to process a Arecibo WU, and it takes 1250 GFLOPS to process a GBT WU, they should both get the same Credit.


IOW CreditNew will have no problem giving these two the same amount of credit as long as it's told the apps are different.

Or, you know, we could just sit back and introduce Parkes tasks under the same umbrella for s* & giggles and watch CN go crazy... Petri's gonna have enough horsepower in a couple of years to crunch the whole sky in real-time anyway... we may as well enjoy ourselves in the meantime :D
ID: 1933981 · Report as offensive
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1668
Credit: 623,086,772
RAC: 156
Finland
Message 1934013 - Posted: 6 May 2018, 16:56:43 UTC

Hi,

*Laughing : Real time crunching* :)

If I remember correctly a task is a 108 second recording from the sky.

Most of the tasks can be done in under 108 seconds. Shorties take 16 seconds, 0.44 Arecibo tasks take 37 seconds, GBT vlars take 38 seconds and Arecibo vlars about 100 seconds on TITAN V. Those 1080s and a Ti take about twice the time to process them.

So you'd need as many TITAN Vs as there are simultaneous channels and polarisations recorded.

It was a couple of years ago when jason gee predicted that the real time analysis would some day be possible.

Petri
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1934013 · Report as offensive
kittyman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 9 Jul 00
Posts: 51468
Credit: 1,018,363,574
RAC: 1,004
United States
Message 1934017 - Posted: 6 May 2018, 17:34:30 UTC - in response to Message 1934013.  

That's one helluva GPU, for shore.
Meeeeeeeeeeeeeeeeow, what the kitties could do with a couple of those!
"Freedom is just Chaos, with better lighting." Alan Dean Foster

ID: 1934017 · Report as offensive
1 · 2 · 3 · 4 . . . 13 · Next

Message boards : Number crunching : Let's Play CreditNew (Credit & RAC support thread)


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.