MAD-ness
01-31-2003, 04:05 AM
This isn't really a question for Louie, as I doubt he has time to look into individual accounts, etc.
Hopefully someone here might be able to give me some ideas on what to check out next. Here is a problem that a teammates is having:
He has 2 computers running SoB.
1 - AthlonXP 1700+ (overclocked)
2 - Athlon MP 1800+ X 2 (default clock)
His stats yesterday were around 100k cEMs. Today it is up around 150k. I know that the cEM unit isn't totally reliable and I also know that the stats can be misleading or slow to update or whatever, but the magnitude of the difference between his stats and his expected stats is large enough that I can't figure out a logical explanation.
Things that have been checked:
- The clients are still running.
- the cEMs rate in the client (~170k for each of the MPs and ~180k for the overclocked 1700+ - these numbers jive with the last two tests I have run on an AXP 1800+) shows that just one of his clients should be producing what his stats show (or more).
- the name in the clients is the same as the name of his account. This has been verified.
- The CPU utilization has been verified (that SoB is actually getting the bulk of the clock cycles, though the in client cEM counter should drop if the client is sharing cycles). This isn't a result of a virus or another DC client stealing cycles.
Here is the link to his stats:
http://www.seventeenorbust.com/stats/users/user.mhtml?userID=640
Note that it shows approx. 3.73ghz, which is not totally accurate, but this number is only ever a rough estimate anyways, but ~3.73 ghz should produce more than 150k, unless the n value is very small (wasn't it short n values that slowed the cEMs but made complete tests faster?). Also note that he summary shows he finished two proth tests in the last twenty four hours but still only had 150k as his rate.
So...
I am going to try to get log files from him. The only thing I can really think of at this point is that he has "transmit" intermediate blocks turned off so the stats don't update until he returns a whole test, but then you should see big spikes on a somewhat regular basis.
Oh, one final thing, I feel sort of dumb for asking, when running a Dual CPU system with SoB, you run two copies of the client and each client/CPU works on a different proth test, correct? So, the question is, what should the CPU utilization of each CPU show? 100% on each? He says he gets 50%, but I am not 100% clear how he means this.
Anyways, any suggestions would be welcome, trouble-shooting these things is always difficult, without being able to have direct access to the machine.
Hopefully someone here might be able to give me some ideas on what to check out next. Here is a problem that a teammates is having:
He has 2 computers running SoB.
1 - AthlonXP 1700+ (overclocked)
2 - Athlon MP 1800+ X 2 (default clock)
His stats yesterday were around 100k cEMs. Today it is up around 150k. I know that the cEM unit isn't totally reliable and I also know that the stats can be misleading or slow to update or whatever, but the magnitude of the difference between his stats and his expected stats is large enough that I can't figure out a logical explanation.
Things that have been checked:
- The clients are still running.
- the cEMs rate in the client (~170k for each of the MPs and ~180k for the overclocked 1700+ - these numbers jive with the last two tests I have run on an AXP 1800+) shows that just one of his clients should be producing what his stats show (or more).
- the name in the clients is the same as the name of his account. This has been verified.
- The CPU utilization has been verified (that SoB is actually getting the bulk of the clock cycles, though the in client cEM counter should drop if the client is sharing cycles). This isn't a result of a virus or another DC client stealing cycles.
Here is the link to his stats:
http://www.seventeenorbust.com/stats/users/user.mhtml?userID=640
Note that it shows approx. 3.73ghz, which is not totally accurate, but this number is only ever a rough estimate anyways, but ~3.73 ghz should produce more than 150k, unless the n value is very small (wasn't it short n values that slowed the cEMs but made complete tests faster?). Also note that he summary shows he finished two proth tests in the last twenty four hours but still only had 150k as his rate.
So...
I am going to try to get log files from him. The only thing I can really think of at this point is that he has "transmit" intermediate blocks turned off so the stats don't update until he returns a whole test, but then you should see big spikes on a somewhat regular basis.
Oh, one final thing, I feel sort of dumb for asking, when running a Dual CPU system with SoB, you run two copies of the client and each client/CPU works on a different proth test, correct? So, the question is, what should the CPU utilization of each CPU show? 100% on each? He says he gets 50%, but I am not 100% clear how he means this.
Anyways, any suggestions would be welcome, trouble-shooting these things is always difficult, without being able to have direct access to the machine.