Amount of blocks in 1 test

[Quatar]

@project leaders.

Isn't there a way to reduce the amount of blocks in a test?

When I started SOB there were only 1,3K blocks in one test, now we are reaching almost 4K blocks in a test. For the users with slowers PC's it almost imposslible to reach the end of a test in one month.

With smaller tests, I think, there would be more users active on the project!
I'm problably noy the only one, who could spare some idle hours if one test wouldn't take at least 500 hours............ Not al PC's produce 2M a second.

the stat are counted in P90 years, I'm producing 2 years a day, I could add some more................

New test/
[Sun Aug 14 03:57:23 2005] n.high = 5348 . 3614 blocks left in test

[vjs]

There isn't a way to reduce the number of blocks or break the test up into pieces. You can check some of the other threads its been explained in great detail.

Most tests finish in three months even on the slowest computer so stick it out these numbers are starting to get to be acomplishments in themselves.

You make also wish to check into the sieve a factor forums for alternatives or run secondpass. I.E. login as your usernameQQQsecondpass. These blocks are quite small and finish within a wekk on slow machines.

[Quatar]

I've read several topics, but it dazzles me.

I want to stay to stay at sob for a few more months (I'm a small DCer, who wants to be big at a lot of projects.......)

I'm a statslover, but to contribute more to the project, should I use all my boxes (about 6M) on secondpass testing, or just on the normal tests?

[wirthi]

I really understand you concerns, but: you get stats points everytime you upload a block. What does it matter how many blocks actually need to be done for a complete test? Unless you reinstall your machine's operating systems once a month it should be no problem.

I have used gimps before and they have really huge blocks when using on a slow machines. They do not upload intermediate blocks so you really have to wait for a month or two to see a result in the stats. That's not true for SOB; you can see yourself improving in stats even while you have not finished the complete test.

[Nuri]

and, in my opinion, if you're for the stats, first time (normal) PRP is the place to be

[djgordon20794]

Is it possible to adjust the size of the blocks up or down?

Derek

[maddog1]

Short answer: no, it's not.
Details can be found around this forum.

If you want to do tests that last a reasonable amount of time, do secondpass tests. Easy to configure the client to do that (just change your username to usernameQQQsecondpass ,of course replacing username with your actual username. Ie, for me it's maddogQQQsecondpass.
See also the separate thread regarding this:
http://www.free-dc.org/forum/showthread.php?s=&threadid=8282
Tests there are currently at n=2.55M range and last less than a day on any decent PC (sized about 68 blocks) . My 3000+ can do 2 in a day quite easily, even when used for other things too.
Also note that the calculated cEMs are artificially higher with higher n, so don't complain about reduced stats-secondpass is not a good area for stats junkies!
But it helps the project a lot and some of us find it far more enjoyable than waiting 2-3 weeks for a normal test to finish. :)

[djgordon20794]

I know its not possible now, but is it theoretically possible further down the line? I was thinking if you could increase the size of work done in a block you could increase throughput. Right now it takes about 1- 2 seconds to switch between blocks, this might not sound like much, but when you have 3800 + blocks it adds up. Soon we will be over 4000 blocks in a test and the number of blocks is not going down.

Derek

[Frodo42]

Well with the next client I think it could be worthwhile to consider the size of a block ... as djgordon says it's a waste of CPU to report in as often as the clients do right now and it is also imagine it is hard on the server.

As I understand it blocks doesn't really have anything to do with the maths involved, they are just measured as a certain amount of work that is done before you report, which makes sense. The problem is the way we measure work (cems/s) is with flaws and when n increases so does the work done on the same CPU in the same time.

I personally turn of intermediate reporting and then, if I have a slow computer I let the computer report in once in a while, by allowing intermediate reporting for one or a few blocks.

[kelman66]

Originally posted by djgordon20794
I know its not possible now, but is it theoretically possible further down the line? I was thinking if you could increase the size of work done in a block you could increase throughput. Right now it takes about 1- 2 seconds to switch between blocks, this might not sound like much, but when you have 3800 + blocks it adds up. Soon we will be over 4000 blocks in a test and the number of blocks is not going down.

Derek

The only delay that I see there is when a block is finished and it wants to report.
I save those seconds and dont transmit intermediate blocks (for the most part)
as it moves on to the next block without delay. Even more time is saved when the server isnt responding, which happens from time to time.