RC5 OGR 25 cracking speed problems

gmbrker

Member
Oct 19, 1999
84
0
0
My first 24 OGR 25 (not sure if it was 25) packets completed in about 18 hours on my Athlon Classic 700 mhz Took about 38-48 minutes per individual packet.

Now with my second batch of 24 packets, it is taking around 5 hours per packet. This can't be right? Anyone with input?

Update: I'm getting about 4.06 Mnodes/s
 

NT4Mike

Senior member
Oct 9, 1999
536
0
0
gmbrker

don't be alarmed if your stubs take different times to complete. The size of the stubs can vary greatly although it seems that there has been a trend lately of smaller stubs. When the contest first started, some of the stubs were huge requiring days to complete. Dnet finally got things in line to make the stubs more manageable and that's where we are today.

Just keep an eye on your summary to see what your overall rate is and you should be good. Off the top of my head I would say that you should be somewhere in the 5.x Mnode/sec range with the 700, but you would have to check one of the speed pages for a more exact mark. My PIII700@784 does 5.6 Mnode/sec.

-Mike
 

JonB

Platinum Member
Oct 10, 1999
2,126
13
81
www.granburychristmaslights.com
To roughly calculate how long an OGR-25 stub will take to complete, look at the the numbers listed. If they are all small numbers (say, less than 15), then it will take a while because they have more nodes. If they are all less than 12, then it will take even longer. I can't really explain why, but it works.

[Oct 01 14:27:00 UTC] Completed OGR stub 25/2-8-5-45-6-3 (25,145,703,786 nodes)
[Oct 01 15:35:56 UTC] Completed OGR stub 25/2-8-5-47-1-3 (16,765,483,751 nodes)
[Oct 02 19:58:22 UTC] Completed OGR stub 25/6-11-9-5-8-15 (243,186,462,880 nodes)
[Oct 03 20:51:20 UTC] Completed OGR stub 25/6-9-10-13-11-7 (219,546,329,738 nodes)
[Oct 04 00:24:05 UTC] Completed OGR stub 25/6-9-10-13-11-5 (233,832,146,299 nodes)
[Oct 04 03:06:31 UTC] Completed OGR stub 25/1-3-5-17-2-39 (71,532,042,482 nodes)
[Oct 04 08:34:21 UTC] Completed OGR stub 25/1-3-5-17-2-33 (93,708,787,212 nodes)
[Oct 04 15:57:30 UTC] Completed OGR stub 25/6-9-10-13-3-2 (145,553,029,892 nodes)
[Oct 04 21:32:10 UTC] Completed OGR stub 25/6-9-11-17-18-1 (150,104,308,730 nodes)
[Oct 04 21:50:13 UTC] Completed OGR stub 25/1-7-28-19-13-2 (68,763,038,599 nodes)
[Oct 05 15:42:51 UTC] Completed OGR stub 25/6-9-11-18-3-1 (108,426,649,749 nodes)
[Oct 05 21:56:15 UTC] Completed OGR stub 25/6-11-9-5-10-29 (185,569,842,918 nodes)
[Oct 06 11:49:34 UTC] Completed OGR stub 25/1-10-12-3-35-4 (134,906,565,524 nodes)
[Oct 06 20:16:43 UTC] Completed OGR stub 25/6-9-11-18-1-24 (156,578,647,283 nodes)
[Oct 07 01:25:38 UTC] Completed OGR stub 25/6-11-9-5-23-12 (147,942,763,296 nodes)
[Oct 07 07:43:26 UTC] Completed OGR stub 25/6-9-13-20-7-3 (126,106,526,077 nodes)
[Oct 07 17:25:05 UTC] Completed OGR stub 25/4-15-3-20-14-12 (180,565,050,156 nodes)
[Oct 07 19:38:16 UTC] Completed OGR stub 25/4-37-1-13-12-3 (25,326,718,031 nodes)
[Oct 08 01:43:23 UTC] Completed OGR stub 25/4-14-11-9-17-7 (207,527,401,980 nodes)
[Oct 08 20:23:20 UTC] Completed OGR stub 25/6-11-9-5-23-10 (182,064,620,525 nodes)
[Oct 09 02:58:40 UTC] Completed OGR stub 25/6-11-9-5-10-28 (190,577,544,350 nodes)
[Oct 10 15:05:30 UTC] Completed OGR stub 25/2-10-5-9-21-6 (309,353,144,766 nodes)
[Oct 10 16:46:45 UTC] Completed OGR stub 25/7-1-40-4-5-10 (19,217,291,614 nodes)
 

Joe O

Senior member
Oct 11, 1999
961
0
0
What OS are you running? If it has a good built in task manager then look to see what is getting the CPU cycles. If not, then you should get one and see what is getting the cycles. Look at this thread and this thread for more ideas.

Something is getting 20% of your CPU!
 

gmbrker

Member
Oct 19, 1999
84
0
0
Thanks Joe O for the links but doesn't seem to help me here. I'm running Win 2k professional and DNet is getting 99% of CPU time. I do use the cpu occasionally but this value was taken from it running all night when I was not using it. I should be getting 5.06 MNodes/s, but nothing is taking my CPU cycles. Maybe my Athlon is screwed up. It's a 550 classic that has been overclocked to 700. My score is inline with an Athlon 550, but I have benched my computer and it is running at 700 mhz.

In RC5 I am getting 2.23 Mkeys/s which is close to the 2.3 Mkeys that a athlon 700mhz should get/

Just strange how my OGR rate is off by so much
 

Joe O

Senior member
Oct 11, 1999
961
0
0
gmbrker, Just as a diagnostic measure, why don't you try running RC5 for a day or two. If it runs at full speed, then it's likely your cache that's slowing down OGR. If it also runs at 80% then we need to find out what's running at night. Or it could be power management.
 

imported_Thunder

Senior member
Oct 14, 1999
509
0
0
gmbrker,

I'm not sure if you considered this yet... There was a newer Athlon-specific core released in the latest client available. Specifically, it's the 2.8010.463b release. All of the prior versions that were OGR capable (2.8009 and all prior 2.8010) had a MUCH less efficient core that produced just about the output speed that you are describing.

Is there any chance that this is the problem?

-Brian
 

RC

Golden Member
Jun 23, 2000
1,358
4
91
Thunder - "MUCH less efficient" - On Linux about 30% less efficient :)
I assumed that since gmbrker is using WIN 2K and he seems to be new to OGR that he probably has the 463b client. However, it's possible that he somehow downloaded an old version.

Thunder - You probably don't know anymore about this than I do, but it doesn't hurt to ask - Do you know when dnet intends to implement this new core into their Linux client? Perhaps a better question might be whether dnet has any intentions of doing this. :)

I recall an article on amdzone.com that was focused on the person that developed the new Athlon core. Perhaps I could contact him. Nothing ventured - nothing gained.
 

gmbrker

Member
Oct 19, 1999
84
0
0
Yup, I got the most recent version. Well last night I left my machine on cracking RC5 and was getting 2.23 Mkeys/s . It then switched to OGR25 and this morning I finished one packedt and got around 5.63 MNodes/s. I guess its finally cracking at the right speed. Just strange how it was going so slow before.