Howdy folks,

This was not the original aim for my work on re-compiling and distributing the optimized apps, but it's confusing me nonetheless. When I started checking individual computer stats here Top Computers my box was around #21 on the list (which I thought wasn't bad for an AMD Quad Core.) When I first started with the new optimized app, I was around 1775 Recent Average Credit on my box, and it has been going up ever since. With the 20% increase in optimization, I figured I would be somewhere around 2100-2200 RAC after I've run things long enough. Initially, this was working well as I shot up to #16 on the list, and I thought I would end up around #11 as things progressed. Not so, apparently.

As of this morning, I'm right back at #21. Puzzled, I started looking at the other stats on the other computers. For instance, ID #1363 who is running an Intel Q6600 at 2.4GHz (or maybe overclocked, not certain) has completed this WU hceyz72_0_5509443_r0 in 1,963.41 seconds. His RAC is 2,120.15

I, on the other hand, completed a similar WU in less time. ID #9385 hceyz72_0_6197779_r0 1,112.37 seconds. My RAC is 2,108.66. I understand there is a "lag" between my new WU's and the old average values being eliminated. However, if my productivity is going UP how can my position be falling father behind?

This then begs the question....

If one Quad Core computer can complete a task(s) in less time than another Quad Core computer, and push more WU's up in any day because of it, how can it be losing position statistically?

Enquiring minds want to know....and can't figure it out right now....;-)

PS I understand that an 8 or 16 core computer, even w/o the optimized app will beat me. This is about completion time and # of WU's completed in a day for quad-cores. There's no way a computer taking longer on each individual WU can beat another computer with the same number of processing cores if it's taking longer to complete each task. Even if this Q6600 is overclocked, the completion time must be LESS than my completion time in order to pull ahead, yes?

PPS Unless my optimized app is screwing up completion times, but this is unlikely....

Mike Doerner

---

If one Quad Core computer can complete a task(s) in less time than another Quad Core computer, and push more WU's up in any day because of it, how can it be losing position statistically?

Well, that's simple: the other machine can be used to upload tasks from another PCs (without Internet connection). Since you don't need great computing power to upload and report results, theoretically any machine can hit top10 easily.

---

M4 Project homepage
M4 Project wiki

You can do that?!?!?!?! (I didn't think you could transfer tasks from the download machine to another machine in BOINC)

That begs the question - Why? I mean you have several users with multiple computers. Why skew the results for one box? (And give me a mental complex...Why isn't AMD competitive? Why WHY? Why?) ;-) I don't think so, here's my reasoning...

Looking at the stats yet again, it looks like the TOTAL WU's completed comes into play. For example, danzigrules has a Q6600 near the top of the heap (#7 currently), and his completion times are almost as good as mine, yet he has an extra 200,000 total credits ahead of me.

The only thing I can think of is I gotta let this app run for a MONTH at least to flush out my old completion times. Maybe two months, we'll see.....at least my stats aren't going down again.

Mike Doerner

---

You can do that?!?!?!?! (I didn't think you could transfer tasks from the download machine to another machine in BOINC)

That begs the question - Why? I mean you have several users with multiple computers. Why skew the results for one box? (And give me a mental complex...Why isn't AMD competitive? Why WHY? Why?) ;-) I don't think so, here's my reasoning...

Yes it's possible, I believe it requires some .xml editing, copy&paste and probably both machines have to run under the same cpid. I never did that myself because all my computers are connected to the network, but I know that some people are using tricks like that to download work for offline machines.

---

M4 Project homepage
M4 Project wiki

Sounds like a real PITA to me, not sure it would be worth the results returned for the labor involved.

---

Click Here to see My Detailed BOINC Stats

Actually, it's just a benefit of running the new 64-bit optimized app on Linux, so no PITA, really. I was initially confused as I knew I was pushing out many more WU's than before, but other users with less credits and higher completion times were passing me by statistically. I know there are guys out there who go "CPU Racing" and such, but that's not why I did it. Open64 could have been a dog and I'd be in the middle of the pack with everyone else.....but I'm not complaining either :-)

Currently, the stats look good. I'm already at 2350 RAC and I still seem to be on the way upwards. I guess I'll just let it run it's course and see where I end up.

Mike Doerner

---