Hi folks,

Credit on one of my pc's has fallen from a constant average of about 50 per 2 hour work, to under 20.

This has been for results sent in over the last 3 days. My other pc's are unaffected.

The computer in question has not been touched in any way, it's just a cruncher, and a laptop at that.

Is there any reason why credit would fall such a huge amount for no apparent reason?

Link to the results for this machine: https://boinc.bakerlab.org/rosetta/results.php?hostid=170247

---

Check your benchmarks in the Message tab....

---

Check your benchmarks in the Message tab....

Benchmark gives the same results as it always has, approximately 3200/9800.

---

Check your benchmarks in the Message tab....

Benchmark gives the same results as it always has, approximately 3200/9800.

Boinc just reran your benchmark. Your latest result got a credit of 56.

I detached from the project, uninstalled boinc, reinstalled boinc then reattached to the project. That's the first result that has been crunched since, and it's back to normal.

Why it hasn't given me full credit for the last 3 or 4 days work I don't know, but it's quite frustrating.

I'd be interested to know why, and if anyone else has had a similar experience.

---

Just having more of a think... is it possible that boinc ran a benchmark whilst the laptop was moved, and thus credit has been given based on that benchmark?

The laptop was moved to another room a few days ago and would have been running off the battery, albeit only for 10 minutes or so. The cpu would have been throttled during this brief period of not being connected to the mains.

Does that sound like a feasable explanation? I'm unsure as to how the credit system is worked out.

---

That is why I told you to look under the message tab to see if it had run a much lower benchmank......not run the benchmark now....

---

That is why I told you to look under the message tab to see if it had run a much lower benchmank......not run the benchmark now....

Why didn't you say that then? Unfortunately I don't possess a third eye.

---

"Check your benchmarks under the message tab" to me means LOOK. "Run your benchmarks....." has an entirely different word! Look and Run are not the same thing.

Look and Run are not the same thing.

Really? I'm glad you pointed that out.

The point I'm trying to make is that you come across as being very condescending when there really is no need. All I asked for was some help.

Not everyone is 100% clued up on the workings of the project, having more of a thought for why they're involved.

Please save your attitude for some kid who won't question it. Personally I'll be thinking twice about bothering to use this forum again if this is the kind of pompous behaviour to expect.

Good day to you.

---

Look and Run are not the same thing.

Really? I'm glad you pointed that out.

The point I'm trying to make is that you come across as being very condescending when there really is no need. All I asked for was some help.

Not everyone is 100% clued up on the workings of the project, having more of a thought for why they're involved.

Please save your attitude for some kid who won't question it. Personally I'll be thinking twice about bothering to use this forum again if this is the kind of pompous behaviour to expect.

Good day to you.

Ok, everyone calm down. This is a written medium that is frequently understood as to attitude. This is just a simple misunderstanding.

Your system will automatically benchmark it self from time to time. There will be a message in the listing under the BOINC manager message tab when this happens showing the new benchmark, unless BOINC is restarted and thus clears the message list. It is actually possible that going to battery could have triggered the running of a new benchmark, but in any case it probably did do a new one while it was on battery. That would definitely account for the lowered scores.

Since that system is used only for crunching, you can prevent this in the future by resetting your system for high performance operation when on battery power. Of course the risk is that this will run the battery down very fast unless you keep it plugged in.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

......Personally I'll be thinking twice about bothering to use this forum again if this is the kind of pompous behaviour to expect....

Good, responding to people who only try to help you by insulting them means we will not miss you!

Most of my machines have dropped to about 60% of their original credits/day over the last few weeks. The first issue was a large number of units that took well over 24h, and a week's worth of those ended up being past the deadline, so I aborted a whole slew of them from most of my 5 machines. Now I am back to running <2h units since about a week, and the stats look like they are levelling off at the lower level. Each machine is a different CPU and motherboard, none are laptops. I occasionally have to suspend Boinc for an hour or two when some dumb installshield apps run, as they [not!] run at the same priority, but this is very rare, like once a week. I also noticed that my queues were almost empty on two machines, as if the benchmarks were miscalculated (24H estimate each for a pair of 2h units), and as soon as a few of the 2h units were calculated, the queue filled up again, although the estimate still shows 5h.
What happens when Boinc runs when you play a game? (hey, maybe 3-4 hours a week) Does it randomly do benchmarks when CPU is busy? If I am in the middle of an Excel sheet recalc, the benchmark will take forever and get a lousy score.

---

Most of my machines have dropped to about 60% of their original credits/day over the last few weeks. The first issue was a large number of units that took well over 24h, and a week's worth of those ended up being past the deadline, so I aborted a whole slew of them from most of my 5 machines. Now I am back to running <2h units since about a week, and the stats look like they are levelling off at the lower level. Each machine is a different CPU and motherboard, none are laptops. I occasionally have to suspend Boinc for an hour or two when some dumb installshield apps run, as they [not!] run at the same priority, but this is very rare, like once a week. I also noticed that my queues were almost empty on two machines, as if the benchmarks were miscalculated (24H estimate each for a pair of 2h units), and as soon as a few of the 2h units were calculated, the queue filled up again, although the estimate still shows 5h.
What happens when Boinc runs when you play a game? (hey, maybe 3-4 hours a week) Does it randomly do benchmarks when CPU is busy? If I am in the middle of an Excel sheet recalc, the benchmark will take forever and get a lousy score.

In short yes. If BOINC is running it could benchmark at any time. As for your Queue and estimated time to completion. This will correct slowly over time. The longer the system is allowed to run the more accurate it will become. But aborted WUs will slow that process significantly, and CPU intensive programs will slow progress on the WUs pushing the system to EDF mode. This will also slow the adjustment. Slowly your time estimates will drop to 2 hours if that is your time setting. On my systems (which have run well for weeks) the estimated time for a time setting of 4 hours is 4:15, for the two hour machines it is 2:25. It is still dropping slowly with every Wu.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

[snip] ... What happens when Boinc runs when you play a game? (hey, maybe 3-4 hours a week) Does it randomly do benchmarks when CPU is busy? If I am in the middle of an Excel sheet recalc, the benchmark will take forever and get a lousy score.

In short yes. If BOINC is running it could benchmark at any time.... [snip]

In that case, how is the credit calculation affected by a loaded machine getting lower benchmarks? Do the credits drop for processing the same 2h wu?
Does that mean the souped up Boinc versions getting higher benchmarks get higher credits for the same work?
And then, why would Boinc recheck throughput on a regular basis? Any random hits of Explorer stuck looping for a slow web page could chew hours? of crunching into zero credits?

---

[snip] ... What happens when Boinc runs when you play a game? (hey, maybe 3-4 hours a week) Does it randomly do benchmarks when CPU is busy? If I am in the middle of an Excel sheet recalc, the benchmark will take forever and get a lousy score.

In short yes. If BOINC is running it could benchmark at any time.... [snip]

In that case, how is the credit calculation affected by a loaded machine getting lower benchmarks? Do the credits drop for processing the same 2h wu?
Does that mean the souped up Boinc versions getting higher benchmarks get higher credits for the same work?
And then, why would Boinc recheck throughput on a regular basis? Any random hits of Explorer stuck looping for a slow web page could chew hours? of crunching into zero credits?

Lower benchmarks equal lower credit claims. So yes, the same machine crunching the same WU once with high benchmarks and once with lower benchmarks, would claim different credits. The optimized BOINC clients are usually identical in every respect to the distributions versions, except that they are altered to claim higher benchmarks (i.e. hight credit) or compiled for a specific platform (i.e. tuned) so they run a little better. In some cases they are designed to use parts of the machine that the normal BOINC client does not take into account, so the higher benchmark may be legitimate, but the machine does not really run that fast when running the project applications because they may not be using that part of the system. A Mac system benchmarked for use of Altivec code is an example. If the BOINC client looks at the Altivec part of the system, it will bench very high, but many of the project applications do not use this part of the Mac, so they run slower than the benchmarks would otherwise indicate. But the tuned BOINC versions do not really alter credit significantly, only the benchmark altered versions have this effect.

The credit claims will never reach zero because your system will always have some benchmark value. But a benchmark run when the system is very busy can significantly effect credit claims.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact