PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : General discussion : Split forum posts into pages? (Message 136258)
Posted 12 days ago by dlouweProject donor
I would greatly welcome paginated posts. While sometimes non-paginated can be nicer for being able to read a whole thread from top-to-bottom, pagination has huge benefits when attempting to find any information in the middle of a lengthy thread, or remembering your place if you need to stop reading partway through and come back later.

I've also experienced loading issues when trying to view some of the longer threads on mobile.
2) Message boards : Proth Prime Search : Top 5000 Prime not submitted correctly (Message 20902)
Posted 3632 days ago by dlouweProject donor
Checking my primes list (http://www.primegrid.com/primes/?section=primelist&userid=48166), it shows that I was primary checker for 8751*2^485041+1, which was submitted to The Prime Database (http://primes.utm.edu/primes/page.php?id=90904), but not under my name (http://primes.utm.edu/bios/page.php?id=2084).

Is there any way this can be updated?
3) Message boards : AP26 - AP27 Search : CUDA application released (Message 19652)
Posted 3693 days ago by dlouweProject donor
Will there be any chance of forcing AP26 to only get CUDA WUs?

I ask this because I have 32-bit Windows (I haven't been able to get CUDA drivers to work on my 64-bit Linux), and I would like to have CPU exclusively work on LLRing and GPU to work on the AP26.

Also I'm not sure if this is related to the slowdown you are all talking about, but, I currently have 1 CUDA AP26 WU running, as well as 2 CPU AP 26 WU running, however only one of the CPU WUs is seeing any CPU time.

CPU: GenuineIntel Intel(R) Core(TM)2 Duo CPU E7200 @ 2.53GHz
GPU: NVIDIA GeForce 9800 GT (511MB) driver: 19038
4) Message boards : Number crunching : Badges (Message 19379)
Posted 3708 days ago by dlouweProject donor
Whew, just barely squeeked in my PSA badge. It looked like it had stopped at 19,400 and I thought I was going to have to wait another month, hah.
5) Message boards : Problems and Help : No Progress Indication (Message 19305)
Posted 3712 days ago by dlouweProject donor
Well I can at least clear up one thing for ya: LLR WUs don't update progress immediately, but Sieves and AP26 do. However, this is usually 3 or 6 minutes, and I haven't the foggiest as to how a PPSE LLR could be taking 12+ hours. That I'll leave to someone more knowledgeable.
6) Message boards : Number crunching : Badges (Message 19283)
Posted 3714 days ago by dlouweProject donor
I'd guess because AP26 also benefits from 64-bit the same way that sieves do, and as such can gain credit like gangbusters in comparison to LLRing.

Edit: I say this because for what it does, AP26 definitely is not a sieve in this context.
7) Message boards : Cullen/Woodall prime search : Why did my wingman ask for so little credit on this WU? (Message 19272)
Posted 3714 days ago by dlouweProject donor
Since you were curious; the Cullen my machine did finally validated, and it recieved 1084 credits.

Doing a little bit of research and math, it looks like Primegrid assigns credit for LLRs based on the average of the claimed credit of both checkers, times some sort of correction factor. Checking a few WUs from a few different Subprojects showed that the factor seems to be set at about 1.5. (Cullen all appeared to be 1.55, PPS was about 1.4, and I checked two Woodall that were 1.4 and 1.5).
8) Message boards : AP26 - AP27 Search : Contact needed with persons that have CUDA-devices (Message 19265)
Posted 3715 days ago by dlouweProject donor
Yeah, from what I understand, in GPU applications, the CPU only needs to be employed minimally. The CPU can spend its time working on its own WUs.
9) Message boards : Cullen/Woodall prime search : Why did my wingman ask for so little credit on this WU? (Message 19227)
Posted 3716 days ago by dlouweProject donor
One thing to note when considering the discrepancy between fixed and variable rate projects, is there seems to be a premium assigned to the variable rate projects in the actual credit granted. Standard BOINC procedure is to simply award the lowest credit claimed, but at Primegrid, the granted credit is generally higher than both of the claimed.

There doesn't seem to be a consistent ratio, but it does close (or even overshoot) the gap between 32 and 64-bit projects once you also account for the 1.7x advantage.

For example, one of the Cullen LLR done by your Q9550 has validated, and received 1290 credits instead of 963, and the doublechecker only claimed 692 credits.

I obviously don't personally know the inner workings of how Primegrid awards its credit, but looking at all of these factors there does seem to be a method to the madness.
10) Message boards : Cullen/Woodall prime search : Why did my wingman ask for so little credit on this WU? (Message 19221)
Posted 3716 days ago by dlouweProject donor
Reading the following page might give you a few of the answers you are looking for regarding credit computation.

http://boinc.berkeley.edu/wiki/Computation_credit

The basic hourly claim the client submits with a Result to the project servers is the sum of Whetstone plus Dhrystone values of the last Benchmark divided by about 480 (it varies slightly with processor and OS). Thus for instance (2,000 Whetstone + 6,000 Dhrystone) / 480 = 16.66 hourly claim. This exact value will not always show on the results listing of projects. It is often substituted by their own pre-assigned credit value.


Using this formula, we can determine the rate at which these computers will gain credit (per core)

This time it's on Cullen LLR WUs. It crunches these at 127k-129k sec's per WU and is claiming 630 credits for each. My other machine is processing them at 131k-133k secs each claiming 960 cr.


Computer 128275
Measured floating point speed 2457.91 million ops/sec
Measured integer speed 6072.35 million ops/sec
(2457.91+6072.35)/480 = 17.77 hourly

128000s = 35.56h * 17.77 = 631 credits

Computer 86532
Measured floating point speed 3244.17 million ops/sec
Measured integer speed 9255.04 million ops/sec
(3244.17+9255.04)/480 = 26.04 hourly

132000s = 36.66h * 26.04 = 956 credits

So, the expected calculated credit matches up almost exactly with the actual claimed credit.

Also as an example of something even more extreme: I have a P4 that recently did a Cullen LLR, took 224k seconds, and only claimed 557 credit.


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.25, 2.59, 2.62
Generated 21 Jan 2020 | 23:16:57 UTC