I go to "Computers on this account" and select to view the stats for my computer: 229890.... and there is nothing.

If I go to top_hosts.php and select computer 237876, I see stats. So where can I find mine?

My last completed task was on 27-Mar-2024 06:35:09. I see entries for 26-Mar-2024 on computer 237876. I thought it might be related to date but not so.

Bobby,

I assume that all your results have been purged from the database.

I doubt that there is anything wrong with the system, because I can see your list of computers and their average and total credit on the website.

Looking at my own list of tasks, I currently have only 3 left showing, all reported on 27 March, 1 at 10:12:28 UTC, 1 at 15:35:11 UTC and 1 at 21:13:22 UTC.

The order in which tasks are purged will not necessarily follow the date order in which you return them. This is particularly because your task will need to be validated against a wingman, and the wingman's task may be returned much later than when your task was returned.

In the case of each of the 3 tasks mentioned above, my initial wingman did not respond and a further task was sent out to a second wingman before validation.

Cheers,
Mark

WOW that is quick. So I guess, unless you are on the top, you are SOL

I'm trying to figure out how Denis counts points vs WCG vs Rosetta as per my 5950x thread in WCG. I recognise your name and moniker.

I'm trying to figure out how Denis counts points vs WCG vs Rosetta as per my 5950x thread in WCG.

I'd be interested to hear if you find differences between the performance of the CPUs on the different projects. I have similar AMD CPUs to you. I've mainly been running Rosetta recently.

Cheers,
Mark

I have no data for this project but I did find that Rosetta seems to give points based on time and floating point MIPS (Whetstone) per CPU which it finds in in client_state.xml from <p_fpops> and <p_iops>. When you do a Boinc benchmark it stashes it there.

WCG seems to assign credits per job regardless of time.

It's written up in the hardware forum over there.

WOW that is quick. So I guess, unless you are on the top, you are SOL

I'm trying to figure out how Denis counts points vs WCG vs Rosetta as per my 5950x thread in WCG. I recognise your name and moniker.

Only WCG is "internally" counting different, though it is the same for external stats, as with all BOINC projects. That's due to WCG have started out before BOINC came into existence and they stick to their way of calculating points (7x that of BOINC).

How much each project is "paying" in points varies wildly across BOINC, with some actually inflating the points values, so others (incl, DENIS) look like a rather meager "payout". But then not everyone is one of those pointw****s...

Ralf

I have no data for this project but I did find that Rosetta seems to give points based on time and floating point MIPS (Whetstone) per CPU which it finds in in client_state.xml from <p_fpops> and <p_iops>. When you do a Boinc benchmark it stashes it there.

WCG seems to assign credits per job regardless of time.

It's written up in the hardware forum over there.

That is definitely NOT the case. The only project that I am aware of (and at times participate in as backup project) is Einstein@Home.

In general, most projects award points depending on the amount of computations (cobblestones, as you can see for example mentioned on the certificate you can get at some projects, including DENIS). That's why a lot of pointw*****s favoring GPU projects, as there they get the most points in the shortest amount of time due to the high number of floating point operations per second compared to regular CPU/FPU computation...


Ralf

I have no data for this project but I did find that Rosetta seems to give points based on time and floating point MIPS (Whetstone) per CPU which it finds in in client_state.xml from <p_fpops> and <p_iops>. When you do a Boinc benchmark it stashes it there.

WCG seems to assign credits per job regardless of time.

It's written up in the hardware forum over there.

That is definitely NOT the case. The only project that I am aware of (and at times participate in as backup project) is Einstein@Home.

In general, most projects award points depending on the amount of computations (cobblestones, as you can see for example mentioned on the certificate you can get at some projects, including DENIS). That's why a lot of pointw*****s favoring GPU projects, as there they get the most points in the shortest amount of time due to the high number of floating point operations per second compared to regular CPU/FPU computation...


Ralf

I noticed that as well. My GPU projects go way higher than my CPU. Figured it was something like you mentioned.

I know about the 7x Boinc credits for WCG.

I was not trying to figure out the biggest bang for the buck between Boinc projects but just trying to figure out how each does it, matching it to what I see in my logs and their results and seeing if it all jives.

I think you are both saying the same thing as I did about WCG credits and Rosetta points and about your GPU vs CPU/FPU credits comparison: the more operations/second the more credits over a time period. This tells me each unit gets n credits and if you can chew up 10 units in a time period you get 10n credits but if the slower CPU/FPU (or low-end 4-core machine) only does 3 units in the same time period then it only gets 3n credits. Fair.

I've analysed all three now.
Both Denis and WCG report CPU/RUN times in the same manner where the 5950x is twice as long as the 3900x but the resulting credits are similar in points. This supports my theory that the 5950x is throttling itself. So I get less for less.

Funny that here, in Denis, the sent time matches my log start time but the time reported is one hour later than my log finished time. My suspicion is that the former is using my UTC offset which is -4 and the later is using UTC.

Now Rosetta must be using a fixed time WU. On my 4-core i5-2520M the run time is about 3 hours; same as my 5950x and 3900x. So I'm guessing it must know how much work gets done in that time and assigns credits accordingly. My 5950x gets less than the 3900x. Now this also supports my theory that the 5950x is throttling itself. Or does it use my Boinc benchmark which it stores in my profile for that computer. I'm waiting for more units to be available to test this.

Bobby - Europe is UTC+1. Since this project is in Spain his local time on the system is +1 to the UTC reported time.

I know about the 7x Boinc credits for WCG.

I was not trying to figure out the biggest bang for the buck between Boinc projects but just trying to figure out how each does it, matching it to what I see in my logs and their results and seeing if it all jives.

I think you are both saying the same thing as I did about WCG credits and Rosetta points and about your GPU vs CPU/FPU credits comparison: the more operations/second the more credits over a time period. This tells me each unit gets n credits and if you can chew up 10 units in a time period you get 10n credits but if the slower CPU/FPU (or low-end 4-core machine) only does 3 units in the same time period then it only gets 3n credits. Fair.

I've analysed all three now.
Both Denis and WCG report CPU/RUN times in the same manner where the 5950x is twice as long as the 3900x but the resulting credits are similar in points. This supports my theory that the 5950x is throttling itself. So I get less for less.

Funny that here, in Denis, the sent time matches my log start time but the time reported is one hour later than my log finished time. My suspicion is that the former is using my UTC offset which is -4 and the later is using UTC.

Now Rosetta must be using a fixed time WU. On my 4-core i5-2520M the run time is about 3 hours; same as my 5950x and 3900x. So I'm guessing it must know how much work gets done in that time and assigns credits accordingly. My 5950x gets less than the 3900x. Now this also supports my theory that the 5950x is throttling itself. Or does it use my Boinc benchmark which it stores in my profile for that computer. I'm waiting for more units to be available to test this.

Europe is UTC+1. Since this project is in Spain his local time on the system is +1 to the UTC reported time

Yes I know. I've been there. I was wondering why both dates don't reflect this. There is a 3 hour difference between the sent time and receive time. My logs show a 2 hour difference between my received (download) time and my sent (upload) time respectively.

It was only a FYI comment. It makes no difference to anything.

Like this:

Name 	HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1
Workunit 	19686803
Created 	9 Apr 2024, 12:39:59 UTC
Sent 	9 Apr 2024, 13:25:06 UTC
Report deadline 	12 Apr 2024, 13:25:06 UTC
Received 	9 Apr 2024, 15:36:13 UTC

MY TIME: UTC-4
2024-04-09 9:25:15 | DENIS@home | Starting task HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1
2024-04-09 10:35:19 | DENIS@home | Computation for task HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1 finished

Europe is UTC+1. Since this project is in Spain his local time on the system is +1 to the UTC reported time

Yes I know. I've been there. I was wondering why both dates don't reflect this. There is a 3 hour difference between the sent time and receive time. My logs show a 2 hour difference between my received (download) time and my sent (upload) time respectively.

It was only a FYI comment. It makes no difference to anything.

Like this:

Name 	HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1
Workunit 	19686803
Created 	9 Apr 2024, 12:39:59 UTC
Sent 	9 Apr 2024, 13:25:06 UTC
Report deadline 	12 Apr 2024, 13:25:06 UTC
Received 	9 Apr 2024, 15:36:13 UTC

MY TIME: UTC-4
2024-04-09 9:25:15 | DENIS@home | Starting task HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1
2024-04-09 10:35:19 | DENIS@home | Computation for task HuVeMOp_20240409142215232055_ORdInc_k_15-Pieske_1000-conf_131_1 finished

I had a theory, but it doesn't jive with what I found for 'sent time', etc.
I'll have to dig on friday when I have more time. I never paid attention to those lines.

Is it possible that the machine which sends is different from the one which receives and one of them has not done the adjustment to summer time? I checked, Madrid is now on CEST (+2)

How much each project is "paying" in points varies wildly across BOINC, with some actually inflating the points values, so others (incl, DENIS) look like a rather meager "payout". But then not everyone is one of those pointw****s...

I used to run MilkyWay and Universe. I no longer do. One has work only for GPUs, and the operator of the other died. I do not remember which is which. Don't care either. They both gave out so many points for each work unit, even though those work units went very fast, that it was embarassing to me how many points I got. The five I try to run are all different, and all of them have long dry spells with no work, but at least the credits are not ridiculously large for the amount of work done.

---

The point of my initial post was not really about credits as such. I was trying to figure out the method used. I see different points for my 3900x than my 5950x on the same project whether it is Denis here or Rosetta or WCG and my concludion is that the 5950x is throttling itself.

My post Last modified: 9 Apr 2024, 18:02:26 UTC explains it.
https://denis.usj.es/denisathome/forum_thread.php?id=290&postid=2359

In Rosetta the task run time is configured on Rosetta side and you can tweak it in preferences (i think default is 6h?). During that time, your computer calculates lowest energy structure and reports whatever lowest structure it could find in given time. In all these 3 projects I see different credits for each task, so probably computation time is related to task data and how that data fits to given CPU cache and RAM

---

--
I crunch for Ukraine