PrimeGrid report deadlines incorrect?

[Neil]

Hi all,

Recently I have noticed that PrimeGrid keeps downloading work units for my PC that it reckons are going to take over 1,000 hours to complete despite having a report deadline that is way less than this.

Have you seen this as well?

Cheers,

Neil.

[ChertseyAl]

over 1,000 hours to complete despite having a report deadline that is way less than this.

Which sub-project are they for? I reckon the SoB LLRs would take around 700 hours on my slow machines.

Maybe your DCF has got screwed up? Mine look pretty reasonable, but were completely thrown by PPS Sieve units at first.

Al.

[Neil]

Hi Al,

Thanks for the reply.

It's always the "Woodall Prime Search (LLR) 5.11" work units that show the issue.

Sorry to be naive, but what does DCF stand for?

Cheers,

Neil.

[ChertseyAl]

It's always the "Woodall Prime Search (LLR) 5.11" work units that show the issue.

Sorry to be naive, but what does DCF stand for?

Hmmm. WPS LLRs supposedly around 70 hours.

DCF is the Duration Correction Factor. Take a look in client_state.xml for <duration_correction_factor>

How quickly do they run? Do they meet the deadline? What happens if you try Sophie Germain or Proth Prime LLRs, both of which are really short? What about other projects?

Sorry, not an expert on Primegrid, even though it's my joint-favourite project along with SIMAP

Al.

[Neil]

Hi Al,

Thanks for your help.

The correction factor is: 1.021687

Apart from the Woodall and Spierpinski Primegrid units, the rest complete fine within their deadlines. It just seems that the BOINC client is downloading work units that my PC has no chance of completing in time. It's very odd. I will try updating the BOINC client as I haven't done that for a while.

Cheers,

Neil.

[Neil]

D'oh! Turns out I am using the latest version after all

[ChertseyAl]

D'oh! Turns out I am using the latest version after all

Which is very strange, as I find the later versions are very mean when taking new work, sometimes running almost dry before downloading anything else

Probably worth asking over in the PG forums in case it's a duff batch of WUs.

Meanwhile just get a batch of work, go NNW, and keep aborting them until the client comes out of panic/hipri mode. I'm currently aborting ABC WUs because I've got half my machines clogged up with too much work

Oh, the DCF looks sensible.

Al.

[ChertseyAl]

Update: I'm now seeing a similar problem

I run a 1 day cache, but am getting 4 days work with PPS Sieve, i.e. enough to just complete in time, but putting the client into panic mode

Al.

[Neil]

Hi Al,

I've decided to detach from PrimeGrid for now. I did log into my PrimeGrid account and checked that I had told it not to download Woodall etc. work units, but no, it still does.

Cheers,

Neil.