Race 4 - Finished!

[NeoGen]

Boinc is just so easy to use and so versatile compared to most non-boinc projects...

[gamer007]

That's true. Easier for the not-so-tech-savvy people to join projects.

I joined before BOINC started and it's existance was just announced on the classic SETI@Home site. But I haven't been active until BOINC actually started. Guess I've always been a fan of BOINC, but I will run other projects if I want to.

[vaughan]

I agree and disagree

eOn is non-BOINC and is extremely easy to d/l setup and run.
SciLink is BOINC and is a disaster even if its excuse is ALPHA. Sortof reminds me of Rectilinear Crossing Number in the beginning where it had the problem of too many small tasks bogging down the computer.

TSC & D2OL leave a bad taste in my mouth after the incompetent stats debacle. Only run it to keep the team alive and participate in this Race.

I guess the reason AMD Users is mainly a bunch of BOINC'ers is the fact that there are so many BOINC projects out there. Did I see that PoorBoy has connected to >50 of the critters? My Projects Tab has 48 so I'm just shy of my half century

[gamer007]

I have 42 projects. I know I'm missing some like Seasonal Attribution & BBC Climate Prediction [takes too long for my liking].

[PcManiac]

What is the cut off for pharmers? 700?

I'm finally back home guys!

In the end I opted to collect the data but not alter the processing scheme I had implemented, and neither to set my uncle's machine to my own preferences because it's not my machine to mess with... so I was left with daily data collections waiting to be processed.

So, I'm now back home and on my own AMD machine and the stats have been processed and uploaded like a charm... AMD Style!

Stats links here:
(Tanpaku) http://neogen.amdusers.com/contest2007/race4b.htm
(TSC) http://neogen.amdusers.com/contest2007/race4nb.htm


p.s. - Mitchell... you are having the first real head-to-head race since the beggining of the contest. Jason is one strong pharmer too! The next few days will be interesting to watch...

[NeoGen]

As with all boinc projects, 700/day.
For TSC however is 100/day.

[mitchellds]

Ok Nflight,

I just water cooled one of the quads and bumped it up a couple hundred mhz. "I'm given it all shes got captain"

[moderna]

It seems it is easy to exceed the pharmers mark with one dual processor in Tanpaku. Machines are getting fast enough that maybe the mark should be raised in future races. I will loose first place in non-pharmers unless I sandbag the last day or so which makes it a game rather than a race.

Oh well. Crunching on.

[Lagu]

I have 16 project as use Boinc and 8 project as use their own platform. I can say so much as project as use their own agent/node is sometimes stable and more nice than Boinc as for me is a sterile platform as when I watch the agent it is always the same picture. Non-Boinc platforms are nicer than Boinc. More colour and you can watch the process in different ways. These clients is not the other alike.

Boinc has the advantage one can load a bunch of WU´s but sometimes one only got 12 WU´s and yet one is allowed 500/day. TSC, D20L and PI Segment allows one download more than 50. Up to 2000 WU´s is allowed if one run TSC/D20L.

But we can only run 1 or 2 WU´s regardless we have 500 tasks on the queue. In the reality I only receive 1 or 2 tasks, crunch them and then upload them and get 1 or 2 new tasks. We can’t speed up and run 3-4 WU´s when we not have more than 1 or 2 cores.

How many times have we not tried get more work using Boinc and got a message such: “Not uploading a new work, deferring 1 second something”. Running a non-Boinc client these word never come up. You crunch the WU, upload it, download a new task and forgot you have it running. Boinc however claim a certain surveillance every day. What has happen when we was sleeping. Had there been any server failure as stopped us to get new work and how long was the stop?

A perfect example is MoneyBee. It has been running at least 1 year on my Intel PII 350 MHz without interruption as far I’m aware and am is not forced baby-sit. It runs for months and sometimes I’m not aware it is ongoing. It is very seldom I checked it. eOn is as Vaughan stated easy download, run and also forgot. However when this client has the bad habit ZZZZZZZZ a lot I use to check it. Why? Yes, because I’m curious see how many points I got but also because I don’t rely on the client just because it sleep to much. Sometimes eOn use too much recourses as can slow down other project as I run why I’m forced check it sometimes.

[NeoGen]

It seems it is easy to exceed the pharmers mark with one dual processor in Tanpaku. Machines are getting fast enough that maybe the mark should be raised in future races. I will loose first place in non-pharmers unless I sandbag the last day or so which makes it a game rather than a race.

Oh well. Crunching on.

The 700 credits mark was just a figure I came up with, I didn't do actual measurements with machines and projects. Last year I set it at 600 and it went pretty well, so I upped it by 100 points this year. I just thought about setting it in a way that no single cpu machine can reach the pharmers league, even if its dual core. I guess the set value pretty much does the job for almost all situations.
There's just too many variables to be able to calculate an optimal value, for example different projects have different credits/hour rates, optimized clients work on some projects and not on others, applications optimized for specific cpu's get more credits for people with that cpu, people with overclocked cpus get more than those at stock speeds, etc...