How long to run short and long Genfer on C2D P8400 and Geforce 9300M?
I am already running 2 short for Intel so what is the time every one else took?

How long to run short and long Genfer on C2D P8400 and Geforce 9300M?
I am already running 2 short for Intel so what is the time every one else took?

Ok I was wondering at the estimate it gave me. 3000 hrs :s ahhhhh. Or how about an i3 or i5 or i7. How does it compare with those?

Ok I was wondering at the estimate it gave me. 3000 hrs :s ahhhhh. Or how about an i3 or i5 or i7. How does it compare with those?

So I got 57 hrs is that accurate?

So I got 57 hrs is that accurate?

• "While Computer is in Use" is not checked
  
• "While computer is on batteries" is not checked
  
• "While processor usage is less than $$ percent" is greater than 0
  
• "Every day between the hours of..." has different hours
  
• "Use at most ###% CPU time" is less than 100
  

Also note that if you're on a laptop and you're running on batteries the CPU will slow down, so Genefer will take longer to run. Of course, crunching on batteries is never a great idea anyway since it will run down the battery exceptionally quickly.

Essentially the estimate is accurate as long as Genefer is allowed to run at full speed.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

I have an Nvidia GT640 and started a WR on the 13th, running 24x7. It is now the 18th and it's less than 20%. At this rate it will take 25+ days to complete. The units only have a 21 day deadline. I know that if it finished it will be OK and get credit, but why a newer video card is taking longer than the deadline? Has anyone else run on a GT640 and what are the results?

Thanks.

... but I'll see if I can look through the database and try to find some results from another 640.

4386^4194304+1 is complete. (15275944 digits) (err = 0.0000) (time = 858:39:04) 18:14:32

Clock=901 MHz
Estimated total run time for 3396^4194304+1 is 664:18:15

Clock=901 MHz
Estimated total run time for 4824^4194304+1 is 612:23:40

Clock=901 MHz
Estimated total run time for 6396^4194304+1 is 730:07:38

Clock=901 MHz
Estimated total run time for 6214^4194304+1 is 706:28:40

According to WIKI a 640 runs at 1/5 the speed of a 680. A 580 takes 100 hours to run one of the current units. A 680 is around 20% slower than that so 120 hours. That means the 640 will take around 600 hours.

I have two 640s running the GFN 262144 units on PRPnet. They run those units in about 4hrs 3mins (factory OC'ed one) to 4hrs 20mins (stock clocked card). Perhaps Mike can extrapolate from those numbers?

Where the 640s really shine are on the Wall-Sun-Sun prime search on PRPnet. On that project, they were as good or better than a GTS450, GTX550 Ti, and Quadro 4000.


EDIT (Remembered that I had replied to a PM about the 640s with some relevant info that I will copy to here):

Well, I have tried it out on the GFN prime searches, CW sieve, PPS sieve, and GFN sieve with mixed results compared to the GTX 550 Ti. On the GFN prime, it is much slower (on the 262144 work, the 550 is a bit more than twice as fact as the 640). On the GFN sieve, the 640 does better, but is still much slower than the 550 (20P/day vs 32P/day). Things are similar on the CW sieve (about 700 seconds on the 550 vs. 1100 seconds on the 640). Things are much better, however, on the PPS sieve. The 640 is just a bit faster than the 550 (sorry, I have those times at work in a spreadsheet and cannot remember them at the moment).

So why the mixed results? My guess is that the programs have not yet caught up to Kepler. Fermi was different than the previous architecture, but it built on the strengths of the previous GPUs. Kepler is fundamentally different (for example, it can do 3+3+3 faster than it can multiply 3x3...MUCH faster!). When the code catches up to this, the 640 will be an excellent card. Overall at PG, at least for now, the 550 is a bit better.

That said, the 550 does use about twice as much power, so the production per Watt is similar even on GFN prime searching, and much better for the 640 at PPS sieve.

On the GT6xx series, particularly with portables, I would advise the short tasks. The credit per hour is about the same.

I am over 60% and still some time to go. No one has finished the unit I am working on, so even though I will be probably over a week late, I may still be the first to finish it. Wouldn't it ironic if it was a GFN and I am the first to finish it?

too bad I cannot make the deadlines. Sure a hefty payout.

Well I can see the new short workunits have to compute on the cpu for 227 hours on a 2nd gen Core i5 processor. What the hell is going on with that? The Short workunits are getting bigger and bigger? Or is the client getting slower and slower? Before it was 57 hours on a C2D P8400 and now it takes 243 hours? Oh and even the GPU 525M needs 196 hours? What the Hell?
____________

Then N=524288 search range was exhausted as of November and the 'short' tasks now have N=1048596. These take about 3.5 times as long to process as the 524288 tasks.

____________
My lucky number is 75898⁵²⁴²⁸⁸+1

Genefer80 (CPU only) for n=15 on PRPNet only takes about 40 minutes for an average CPU. Maybe this could be ported over to BOINC and be the "Short", and n=20 renamed Medium?
____________

Genefer80 (CPU only) for n=15 on PRPNet only takes about 40 minutes for an average CPU. Maybe this could be ported over to BOINC and be the "Short", and n=20 renamed Medium?

There's limitations as to how many projects we can run on BOINC, which is why we're not running all of the ranges simultaneously on the BOINC server. Since there's a lot more computing power running on the BOINC side, that's where we want the largest tasks to be run.

There's faster CPU apps in the pipeline, especially for AVX CPUs. I expect CPU computation times to drop by about 25% for non-AVX CPUs and a whopping 60% for AVX. We're not going to think about moving projects around until after those new apps are in place.