Granting of credits

Message boards : Number crunching : Granting of credits
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 725
Credit: 16,445,605
RAC: 5
Germany
Message 701 - Posted: 10 Sep 2014, 15:31:08 UTC

Which RAC do the 4 systems have?
ID: 701 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 702 - Posted: 10 Sep 2014, 18:03:39 UTC - in response to Message 701.  

ID: 702 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 703 - Posted: 11 Sep 2014, 17:32:06 UTC

Todays numbers:

2 core: 1 point ~ every 143 seconds RAC: 491.60
4 core: 1 point ~ every 65 seconds RAC: 1,527.82
8 core: 1 point ~ every 38 seconds RAC: 2,127.60
16 core: 1 point ~ every 151 seconds RAC: 1,942.50
ID: 703 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 704 - Posted: 12 Sep 2014, 13:01:57 UTC - in response to Message 703.  

Todays numbers:

VM's were down all night. No update.
ID: 704 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 705 - Posted: 13 Sep 2014, 15:21:02 UTC

Todays numbers:

2 core: 1 point ~ every 115-120 seconds RAC: 525.30
4 core: 1 point ~ every 43-44 seconds RAC: 1,561.98
8 core: 1 point ~ every 23-24 seconds RAC: 2,143.19
16 core: 1 point ~ every 143-146 seconds RAC: 1,723.84
ID: 705 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 706 - Posted: 15 Sep 2014, 15:29:20 UTC

Todays numbers:

2 core: 1 point ~ every 95-99 seconds RAC: 579.23
4 core: 1 point ~ every 40 seconds RAC: 1,612.81
8 core: 1 point ~ every 24 seconds RAC: 2,433.22
16 core: 1 point ~ every 38-39 seconds RAC: 1,370.22
ID: 706 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 707 - Posted: 16 Sep 2014, 13:20:28 UTC

Todays numbers:

2 core: 1 point ~ every 91-95 seconds RAC: 623.78
4 core: 1 point ~ every 39-40 seconds RAC: 1,687.23
8 core: 1 point ~ every 22 seconds RAC: 2,534.79
16 core: 1 point ~ every 38 seconds RAC: 1,462.75
ID: 707 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 725
Credit: 16,445,605
RAC: 5
Germany
Message 708 - Posted: 16 Sep 2014, 18:52:10 UTC

For the 2,4,8 core systems it looks fine for me.
ID: 708 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 709 - Posted: 17 Sep 2014, 13:18:11 UTC
Last modified: 17 Sep 2014, 13:19:40 UTC

2 core: 1 point ~ every 89-91 seconds RAC: 645.93
4 core: 1 point ~ every 38-39 seconds RAC: 1,724.71
8 core: 1 point ~ every 23 seconds RAC: 2,666.55
16 core: 1 point ~ every 36-37 seconds RAC: 1,531.51

Quad core looking to be the sweet spot at this point. I wont be running this test much longer as we have a team challenge elsewhere. I just thought these numbers would be helpful to the efforts.
ID: 709 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 710 - Posted: 18 Sep 2014, 13:19:21 UTC - in response to Message 709.  

2 core: 1 point ~ every 96-99 seconds RAC: 676.60
4 core: 1 point ~ every 45-46 seconds RAC: 1,734.23
8 core: 1 point ~ every 33-35 seconds RAC: 2,733.29
16 core: 1 point ~ every 41-42 seconds RAC: 1,603.03
ID: 710 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 711 - Posted: 20 Sep 2014, 14:01:01 UTC - in response to Message 710.  

2 core: 1 point ~ every 91-93 seconds RAC: 720.49
4 core: 1 point ~ every 38-40 seconds RAC: 1,805.00
8 core: 1 point ~ every 29 seconds RAC: 2,753.78
16 core: 1 point ~ every 71-72 seconds RAC: 1,520.98

Well...I'm going to wind down my testing. 18 days of crunching and the numbers still have not leveled even close to consistent. I would not recommend anyone to run anything past a quad core at this time on these apps. This is just my opinion, but Credit New certainly does not do MT apps justice and should be reconsidered here.
ID: 711 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Werinbert

Send message
Joined: 9 May 13
Posts: 10
Credit: 100,312
RAC: 0
United States
Message 712 - Posted: 21 Sep 2014, 21:15:56 UTC

After leaving my computer to run 24/7 this week the 3-core results are returning only 13-14 credits per hour per core. This well below what I would call acceptable on this computer. Last month the 3-core results were a consistent 21-21 credits per hour per core. CreditNew maybe, but I think it is something else.
ID: 712 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 713 - Posted: 22 Sep 2014, 2:06:46 UTC

Last test units are in.

2 core: 1 point ~ every 93 seconds RAC: 712.19
4 core: 1 point ~ every 39 seconds RAC: 1,774.69
8 core: 1 point ~ every 28 seconds RAC: 2,713.55
16 core: 1 point ~ every 68-71 seconds RAC: 1,477.31
ID: 713 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 725
Credit: 16,445,605
RAC: 5
Germany
Message 714 - Posted: 22 Sep 2014, 7:00:01 UTC

Looks good for me. Seems that creditnew scales the credits also with number of cores. Only 16 cores do not get enough credits. But this might be caused by to small workunits.

yoyo
ID: 714 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 715 - Posted: 22 Sep 2014, 12:53:31 UTC

Wait, you think that going from 4 cores to 8 cores scales good?

2 x 4 cores = 2 points in 39 seconds
1 x 8 cores = 1 point in 28 seconds

What I would expect is 2 x 4 cores to = 1 x 8 cores.
However what we are seeing is that 8 cores is scoring less points. Then the 16 core gets really stupid. The same logic could be applied to the 2 core.
2 x 2 cores should = 4 cores but it does not. And we aren't talking a minor difference. This adds up. Look at it this way.

At 39 seconds per point, the quad core earns 2215.38 points per day. At 38 seconds per point, the octacore only makes 3085.71. Now, if you were to take that octacore and chop it into 2 virtual machines, then your 3085.71 would magically transform into 4430.76 points. That is almost a 44% increase in points for the same cores on the same rig.

So, I ask again, how is this scaling " looking good". I will concede that other factors may be involved but as the numbers have shifted over the last few weeks, I think the facts point to a disparity.
ID: 715 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 716 - Posted: 22 Sep 2014, 17:05:03 UTC - in response to Message 715.  
Last modified: 22 Sep 2014, 17:05:23 UTC

Wait, you think that going from 4 cores to 8 cores scales good?

2 x 4 cores = 2 points in 39 seconds
1 x 8 cores = 1 point in 28 seconds

What I would expect is 2 x 4 cores to = 1 x 8 cores.
However what we are seeing is that 8 cores is scoring less points. Then the 16 core gets really stupid. The same logic could be applied to the 2 core.
2 x 2 cores should = 4 cores but it does not. And we aren't talking a minor difference. This adds up. Look at it this way.

At 39 seconds per point, the quad core earns 2215.38 points per day. At 38 seconds per point, the octacore only makes 3085.71. Now, if you were to take that octacore and chop it into 2 virtual machines, then your 3085.71 would magically transform into 4430.76 points. That is almost a 44% increase in points for the same cores on the same rig.

So, I ask again, how is this scaling " looking good". I will concede that other factors may be involved but as the numbers have shifted over the last few weeks, I think the facts point to a disparity.


Too late to edit, but the octacore at 28 seconds per point not 38 seconds per point was intended.

ID: 716 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 725
Credit: 16,445,605
RAC: 5
Germany
Message 717 - Posted: 24 Sep 2014, 10:03:18 UTC

The rac is still converging over time. As you can see the rac is still going down over different days. Only the 2-core rac increased.
I don't know how long it will take until the rac is nearly stable.

yoyo
ID: 717 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 718 - Posted: 24 Sep 2014, 13:21:54 UTC - in response to Message 717.  
Last modified: 24 Sep 2014, 13:39:27 UTC

The rac is still converging over time. As you can see the rac is still going down over different days. Only the 2-core rac increased.
I don't know how long it will take until the rac is nearly stable.

yoyo



Well, looking at the RAC as I have broken it down, it looks like the 16 core also went back up. Don't look at the RAC number I posted because that is from this site. You need to look at how many seconds it was taking to earn each point. The 4 and 8 were the only ones really lowering over time while the 2 and 16 jumped around. I don't see this being "stable" at all after over 2 weeks of processing work. Maybe you are more relaxed on the issue, but I prefer something a lot more consistent.

2 core: 54, 60, 60, 107, 160, 137, 143, 115, 95, 91, 89, 96, 91, 93
4 core: 10, 30, 32, 30, 29, 52, 58, 65, 43, 40, 39, 38, 45, 38, 39
8 core: 2, 17, 17, 16, 17, 31, 39, 38, 23, 24, 22, 23, 33, 29, 28
16 core: 3, 13, 16, 18, 117, 137, 151, 143, 38, 38, 36, 41, 71, 68

You would also think that the scoring would be closer. The 8 core should logically be closer to half the time needed to get the same points as a 4 core. But it isn't. It is still off quite a bit. As noted above, I could get a 40%+ increase in points by running two 4 core VM's instead of an 8 core.
ID: 718 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile yoyo_rkn
Volunteer moderator
Project administrator
Project developer
Project tester
Volunteer developer
Volunteer tester
Project scientist

Send message
Joined: 22 Aug 11
Posts: 725
Credit: 16,445,605
RAC: 5
Germany
Message 719 - Posted: 24 Sep 2014, 14:23:38 UTC

I don't know how creditnew is really working, but it might be a problem for credit new, that there are phases in a yafu run where only 1 core is used.
ID: 719 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Coleslaw
Avatar

Send message
Joined: 6 Sep 11
Posts: 31
Credit: 10,963,803
RAC: 3,985
United States
Message 720 - Posted: 24 Sep 2014, 14:26:11 UTC - in response to Message 719.  

I don't know how creditnew is really working, but it might be a problem for credit new, that there are phases in a yafu run where only 1 core is used.


Yes, I am assuming it is CreditNew that is the problem. It is just good for people to know this issue before they get discouraged by comparing numbers later and seeing heavy hitting machines not get rewarded accordingly.
ID: 720 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : Granting of credits




Datenschutz / Privacy Copyright © 2011-2024 Rechenkraft.net e.V. & yoyo