Citrix,

Which client are you using, sr1sieve (1.0.22) or srsieve (0.6.6)?

Carlos

sr1sieve. 1.0.21

RAM=512MB on the machine.

edit: switched to 1.0.22 (Now getting 17.5M/sec)

Reserving:

25T-50T
50T-100T

Carlos

Sieve speed is almost totally irrelevant to the total time it would take to complete the project. One should always sieve until the sieve removes candidates at roughly the rate LLR tests them. If you choose a small k (where LLR is roughly twice as fast as 3^16), you would cut your required sieve depth dramatically.

Also, note that sieving is in general less than 5% of the total work; doubling LLR speed per test is MASSIVELY more efficient than picking a k that happens to sieve fast. Related to this is the fact that high-weight k's (generally those divisible by 15) sieve rather slowly, but produce more primes in a given range than a random k.

Finally, jasong's idea to sieve a fixed n with a range of k's is indeed the fastest way to sieve to a fixed bitdepth; I suggest choosing an n a few thousand above the 10M digit cutoff, as this idea for finding a 10M digit prime has been around quite a few years, and it is likely individuals have searched the first few dozen n's quite a bit. The drawback to this method is that LLR time varies more with k than it does with n; k=1000 is slower than k=5 by quite a bit, and k=1G half the speed of k=5.

Even with these hurdles, I believe a project like this does increase an individual's probability of hitting the prize compared to running GIMPS. If that is the goal, 3^16 is very much barking up the wrong tree. The easiest route may be asking SOB or Rieselsieve for 10M digit candidates from their sieve; that way you contribute to a project, while LLRing presieved candidates. Not sure if they would oblige this, though.
-curtis

What is the smallest n in the sieve? If zero, you can certainly start LLRing 0-100k or 200k now. If the sieve starts at n=1M or similar, you should start LLR on the smallest chunk of n (while still sieving the rest) when sieve speed is about half the rate of LLR on the range of n you are breaking off for LLR. One candidate an hour removed from the sieve would be a good rough place to break off 1M-1.1M and begin LLR, for instance (on P4s).

I did not notice a second page on the thread when I wrote my last post-- sorry for the sidetrack, as it seems you have decided to explore 3^16 rather than chase the prize.
-curtis

Completed 10T-15T

9T - 10T - Complete - Old dat

running my new range
P4 3 Ghz = 22M/sec

I cannot use this machine for LLR testing because of the HW implemented CPU temperature throttling.

Sieved 100G in 5 hours!

[quote=CedricVonck;104963]

Sieved 100G in 5 hours![/quote]

With the latest dat file? If so you should be doing 396G/5 hours...

Edit: Try running two threads, you will get something like 14M/sec/thread.

[QUOTE=CedricVonck;104963]9T - 10T - Complete - Old dat

running my new range
P4 3 Ghz = 22M/sec
I cannot use this machine for LLR testing because of the HW implemented CPU temperature throttling.
[/QUOTE]

If LLR causes your CPU to heat-throttle, you need more airflow in the case. There is no reason to expect a CPU to throttle in a normal installation. Check your case for dust, clogged fans, etc, and consider installing a fan in any open location meant for a fan. A P4-3.0 should NOT throttle! Also, if 400G in 5 hrs is expected, you might still be throttling while sieving. Regardless, you have a heat/airflow problem. If your case has no intake fan, take the side of the case off and see if LLR still causes throttling. Not a long-term solution, but easiest way to test if enough air is being blown in. A properly designed case has enough intake fans and airflow control that taking the side of the case off actually heats the CPU more!
-Curtis

32365 candidates left after importing 10-15T and 9-10T.

[QUOTE=VBCurtis;104970]If LLR causes your CPU to heat-throttle, you need more airflow in the case. There is no reason to expect a CPU to throttle in a normal installation. Check your case for dust, clogged fans, etc, and consider installing a fan in any open location meant for a fan. A P4-3.0 should NOT throttle! Also, if 400G in 5 hrs is expected, you might still be throttling while sieving. Regardless, you have a heat/airflow problem. If your case has no intake fan, take the side of the case off and see if LLR still causes throttling. Not a long-term solution, but easiest way to test if enough air is being blown in. A properly designed case has enough intake fans and airflow control that taking the side of the case off actually heats the CPU more!
-Curtis[/QUOTE]

I will try this.