1.0 running slower than .99? [Archive]

View Full Version : 1.0 running slower than .99?

CyberWire

11-21-2002, 01:04 AM

I noticed a drop again after i upgraded to 1.0. This is not due to the new K value as i was getting those with the .99 client also. Anyone else experiencing the same?

DIstComp

11-21-2002, 01:29 AM

I've dropped from a peak of 130-140kcEM/sec to about 60. I'm running an Athlon XP 1700+ on winXP.

Alien88

11-21-2002, 02:12 AM

its not the client, it's the new k value... louie will explain it eventually.. i think he briefly talked about it in one of the other threads

shifted

11-21-2002, 02:42 AM

Originally posted by DIstComp
I've dropped from a peak of 130-140kcEM/sec to about 60. I'm running an Athlon XP 1700+ on winXP.

I've dropped from 133 to 73 myself... that's on the Athlon 1400 (Thunderbird) on Windows eXactProblem

CyberWire

11-21-2002, 02:55 AM

as i said before this is a second drop after installing 1.0 this is after i noticed the first drop from the new K values...

shifted

11-21-2002, 03:36 AM

I've experienced changes even though i've had the same amount of cpu power running for the last day. I can't explain the 20 cEMs/s difference between the two levels :/

CyberWire:
http://www.seventeenorbust.com/cgi-bin/mkgraph.cgi?CyberWire,24hourreal

shifted:
http://www.seventeenorbust.com/cgi-bin/mkgraph.cgi?shifted,24hourreal

Not sure what it is...

CyberWire

11-21-2002, 04:10 AM

mind checking the logs to see is im mistaken about my K values switching louie? maybe it just took an oddly long time for some of my machines to finish some of the old work units? I hate to see my rate drop so dramatically, esp. when others rates dont seem to be doing so badly.

shifted

11-21-2002, 04:32 AM

My rates did drop by about 40%, but that was over a day ago now. It's still visible on the overall graph:

http://www.seventeenorbust.com/cgi-bin/mkgraph.cgi?shifted,overallreal

The small disturbance around 11/19 08h was when i upgraded the cooling on one athlon so i could run it full speed (was at 1.1, now at 1.2). The dip at about 11/18 20h was when i updated the client... there was no immediate drop as it was still processing the smaller k's.

CyberWire

11-21-2002, 08:37 AM

speed seems to be picking up.. guess as we progress thru the N values this will happen. interesting.

smh

11-21-2002, 10:25 AM

Yes, also for the old K value, the higher the N the higher the cEMs.

Remember, the 'c' in cEMs stands for Corrected. Because different FFT's are used if the N value gets larger it takes longer to do a Modulo. Thats why it needs to be corrected, to have about the same value throughout the whole range.

I guess (wild guess) that Louie made the correction with the 'old' (0.90) client which used GMP (and was 30 times slower). With the new client the correction factor should be a bit different.

Even more so, i think that different cpu's need different correction if you want everything equal