PDA

View Full Version : Bad performance with 64 bit Linux?


the-mk
07-15-2007, 06:27 AM
Hi!

Has anybody experience with d.net client on a 64 bit linux?

Today I've installed Mandriva Free 2007 Spring edition (http://www.mandriva.com/en/download/free) on a AMD 4800+ X2 (AM2 platform) with 2 GB of RAM.

I've downloaded the 32 bit and the 64 bit d.net client and "tested" them. After some time running (more than one hour), the logs showed overall performance at about 30 MNodes/s (2 cores) and one "block" finishes with about from 15 MNodes/s to 20 MNodes/s

64-bit-client-log:


dnetc v2.9011-496-CFR-05060815 for Linux (Linux 2.6.17-14mdv).
Using email address (distributed.net ID) '*******@***.**'

[Jul 15 05:39:20 UTC] Automatic processor detection found 2 processors.
[Jul 15 05:39:20 UTC] Loading crunchers with work...
[Jul 15 05:39:21 UTC] The perproxy says: "Free-DC DNet Proxy!! - Where DC is
open and Fun!"
[Jul 15 05:39:21 UTC] Retrieved project state data from server. (cached)
[Jul 15 05:39:28 UTC] OGR-P2: Retrieved 48 packets from server.
[Jul 15 05:39:28 UTC] OGR-P2 #a: Loaded 25/7-9-28-12-17-18
[Jul 15 05:39:28 UTC] OGR-P2 #b: Loaded 25/7-9-28-12-17-19
[Jul 15 05:39:28 UTC] OGR-P2: 46 packets remain in buff-in.ogf
[Jul 15 05:39:28 UTC] OGR-P2: 0 packets are in buff-out.ogf
[Jul 15 05:39:28 UTC] 2 crunchers ('a' and 'b') have been started.
[Jul 15 06:11:52 UTC] OGR-P2 #a: Completed 25/7-9-28-12-17-18 (28.29 stats units)
0.00:32:23.95 - [14,553,245 nodes/s]
[Jul 15 06:11:52 UTC] OGR-P2 #a: 25/7-9-28-12-17-18 [28,290,781,139 nodes]
[Jul 15 06:11:52 UTC] OGR-P2 #a: Loaded 25/7-9-28-12-18-17
[Jul 15 06:11:52 UTC] OGR-P2: Summary: 1 packet (28.29 stats units)
0.00:32:23.95 - [14.55 Mnodes/s]
[Jul 15 06:11:52 UTC] OGR-P2: 45 packets remain in buff-in.ogf
[Jul 15 06:11:52 UTC] OGR-P2: 1 packet (28.29 stats units) is in
buff-out.ogf
[Jul 15 06:13:16 UTC] OGR-P2 #b: Completed 25/7-9-28-12-17-19 (29.70 stats units)
0.00:33:47.90 - [14,647,268 nodes/s]
[Jul 15 06:13:16 UTC] OGR-P2 #b: 25/7-9-28-12-17-19 [29,703,195,674 nodes]
[Jul 15 06:13:16 UTC] OGR-P2 #b: Loaded 25/7-9-28-12-19-17
[Jul 15 06:13:16 UTC] OGR-P2: Summary: 2 packets (57.99 stats units)
0.00:33:55.98 - [28.48 Mnodes/s]
[Jul 15 06:13:16 UTC] OGR-P2: 44 packets remain in buff-in.ogf
[Jul 15 06:13:16 UTC] OGR-P2: 2 packets (57.99 stats units) are in
buff-out.ogf
[Jul 15 06:41:34 UTC] OGR-P2 #a: Completed 25/7-9-28-12-18-17 (26.89 stats units)
0.00:29:41.45 - [15,094,255 nodes/s]
[Jul 15 06:41:34 UTC] OGR-P2 #a: 25/7-9-28-12-18-17 [26,889,661,524 nodes]
[Jul 15 06:41:34 UTC] OGR-P2 #a: Loaded 25/7-9-28-12-29-6
[Jul 15 06:41:34 UTC] OGR-P2: Summary: 3 packets (84.88 stats units)
0.01:02:13.86 - [22.73 Mnodes/s]
[Jul 15 06:41:34 UTC] OGR-P2: 43 packets remain in buff-in.ogf
[Jul 15 06:41:34 UTC] OGR-P2: 3 packets (84.88 stats units) are in
buff-out.ogf
[Jul 15 06:45:22 UTC] OGR-P2 #b: Completed 25/7-9-28-12-19-17 (28.06 stats units)
0.00:32:06.03 - [14,566,822 nodes/s]
[Jul 15 06:45:22 UTC] OGR-P2 #b: 25/7-9-28-12-19-17 [28,056,137,230 nodes]
[Jul 15 06:45:22 UTC] OGR-P2 #b: Loaded 25/7-9-28-12-31-4
[Jul 15 06:45:22 UTC] OGR-P2: Summary: 4 packets (112.94 stats units)
0.01:06:02.02 - [28.51 Mnodes/s]
[Jul 15 06:45:22 UTC] OGR-P2: 42 packets remain in buff-in.ogf
[Jul 15 06:45:22 UTC] OGR-P2: 4 packets (112.94 stats units) are in
buff-out.ogf
[Jul 15 06:56:45 UTC] OGR-P2 #a: Completed 25/7-9-28-12-29-6 (16.35 stats units)
0.00:15:10.63 - [17,953,203 nodes/s]
[Jul 15 06:56:45 UTC] OGR-P2 #a: 25/7-9-28-12-29-6 [16,348,725,956 nodes]
[Jul 15 06:56:45 UTC] OGR-P2 #a: Loaded 25/7-9-28-12-33-2
[Jul 15 06:56:45 UTC] OGR-P2: Summary: 5 packets (129.29 stats units)
0.01:17:25.09 - [27.83 Mnodes/s]
[Jul 15 06:56:45 UTC] OGR-P2: 41 packets remain in buff-in.ogf
[Jul 15 06:56:45 UTC] OGR-P2: 5 packets (129.29 stats units) are in
buff-out.ogf
[Jul 15 06:57:36 UTC] OGR-P2 #b: Completed 25/7-9-28-12-31-4 (13.48 stats units)
0.00:12:12.16 - [18,412,170 nodes/s]
[Jul 15 06:57:36 UTC] OGR-P2 #b: 25/7-9-28-12-31-4 [13,480,654,989 nodes]
[Jul 15 06:57:36 UTC] OGR-P2 #b: Loaded 25/7-9-28-12-34-1
[Jul 15 06:57:36 UTC] OGR-P2: Summary: 6 packets (142.77 stats units)
0.01:18:15.21 - [30.41 Mnodes/s]
[Jul 15 06:57:36 UTC] OGR-P2: 40 packets remain in buff-in.ogf
[Jul 15 06:57:36 UTC] OGR-P2: 6 packets (142.77 stats units) are in
buff-out.ogf
[Jul 15 07:07:03 UTC] OGR-P2 #b: Completed 25/7-9-28-12-34-1 (11.52 stats units)
0.00:09:26.69 - [20,326,300 nodes/s]
[Jul 15 07:07:03 UTC] OGR-P2 #b: 25/7-9-28-12-34-1 [11,518,711,152 nodes]
[Jul 15 07:07:03 UTC] OGR-P2 #b: Loaded 25/7-9-28-12-35-1
[Jul 15 07:07:03 UTC] OGR-P2: Summary: 7 packets (154.29 stats units)
0.01:27:42.17 - [29.32 Mnodes/s]
[Jul 15 07:07:03 UTC] OGR-P2: 39 packets remain in buff-in.ogf
[Jul 15 07:07:03 UTC] OGR-P2: 7 packets (154.29 stats units) are in
buff-out.ogf
[Jul 15 07:10:58 UTC] OGR-P2 #a: Completed 25/7-9-28-12-33-2 (13.26 stats units)
0.00:14:12.12 - [15,565,864 nodes/s]
[Jul 15 07:10:58 UTC] OGR-P2 #a: 25/7-9-28-12-33-2 [13,263,984,674 nodes]
[Jul 15 07:10:58 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-1-33
[Jul 15 07:10:58 UTC] OGR-P2: Summary: 8 packets (167.55 stats units)
0.01:31:37.21 - [30.48 Mnodes/s]
[Jul 15 07:10:58 UTC] OGR-P2: 38 packets remain in buff-in.ogf
[Jul 15 07:10:58 UTC] OGR-P2: 8 packets (167.55 stats units) are in
buff-out.ogf
[Jul 15 07:18:48 UTC] OGR-P2 #b: Completed 25/7-9-28-12-35-1 (12.45 stats units)
0.00:11:49.41 - [17,550,522 nodes/s]
[Jul 15 07:18:48 UTC] OGR-P2 #b: 25/7-9-28-12-35-1 [12,450,515,929 nodes]
[Jul 15 07:18:48 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-2-32
[Jul 15 07:18:48 UTC] OGR-P2: Summary: 9 packets (180.00 stats units)
0.01:39:33.05 - [30.14 Mnodes/s]
[Jul 15 07:18:48 UTC] OGR-P2: 37 packets remain in buff-in.ogf
[Jul 15 07:18:48 UTC] OGR-P2: 9 packets (180.00 stats units) are in
buff-out.ogf
[Jul 15 07:31:14 UTC] *Break* Shutting down...
[Jul 15 07:31:15 UTC] OGR-P2 #b: Saved 25/7-9-28-13-2-32 (12.54 Gnodes done)
0.00:12:26.40 - [16,800,981 nodes/s]
[Jul 15 07:31:15 UTC] OGR-P2 #a: Saved 25/7-9-28-13-1-33 (22.92 Gnodes done)
0.00:20:22.24 - [18,756,201 nodes/s]
[Jul 15 07:31:15 UTC] OGR-P2: Summary: 9 packets (180.00 stats units)
0.01:51:59.46 - [26.79 Mnodes/s]
[Jul 15 07:31:15 UTC] OGR-P2: 39 packets are in buff-in.ogf
[Jul 15 07:31:15 UTC] OGR-P2: 9 packets (180.00 stats units) are in
buff-out.ogf
[Jul 15 07:31:15 UTC] Shutdown complete.


My AMD 3800+ X2 (939 Socket) on Windows XP 32 bit does d.net-work much faster!! (~25 MNodes/s per core!!) while that new 4800+ X2 on linux is so slow!! (~16 MNodes per core)
the-mk
07-15-2007, 06:28 AM
32-bit-client-log from same linux-machine (4800+ X2) for comparison:


dnetc v2.9013-498-CFR-06082120 for Linux (Linux 2.6.17-14mdv).
Using email address (distributed.net ID) 'the_mk@gmx.at'

[Jul 15 07:32:11 UTC] Automatic processor detection found 2 processors.
[Jul 15 07:32:11 UTC] Loading crunchers with work...
[Jul 15 07:32:11 UTC] Automatic processor type detection found
an AMD K8-11 processor.
[Jul 15 07:32:11 UTC] OGR-P2: using core #4 (GARSP 6.0-asm-rt1-mmx-amd).
[Jul 15 07:32:11 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-1-33
Packet was from a different user/core/client cpu/os/build.
[Jul 15 07:32:11 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-2-32
Packet was from a different user/core/client cpu/os/build.
[Jul 15 07:32:11 UTC] OGR-P2: 37 packets remain in buff-in.ogf
[Jul 15 07:32:11 UTC] OGR-P2: 0 packets are in buff-out.ogf
[Jul 15 07:32:11 UTC] 2 crunchers ('a' and 'b') have been started.
[Jul 15 08:01:17 UTC] OGR-P2 #b: Completed 25/7-9-28-13-2-32 (27.29 stats units)
0.00:29:04.60 - [15,643,952 nodes/s]
[Jul 15 08:01:17 UTC] OGR-P2 #b: 25/7-9-28-13-2-32 [27,292,440,165 nodes]
[Jul 15 08:01:17 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-4-31
[Jul 15 08:01:17 UTC] OGR-P2: Summary: 1 packet (27.29 stats units)
0.00:29:04.60 - [15.64 Mnodes/s]
[Jul 15 08:01:17 UTC] OGR-P2: 36 packets remain in buff-in.ogf
[Jul 15 08:01:17 UTC] OGR-P2: 1 packet (27.29 stats units) is in
buff-out.ogf
[Jul 15 08:05:50 UTC] OGR-P2 #a: Completed 25/7-9-28-13-1-33 (31.04 stats units)
0.00:33:37.27 - [15,386,331 nodes/s]
[Jul 15 08:05:50 UTC] OGR-P2 #a: 25/7-9-28-13-1-33 [31,038,385,584 nodes]
[Jul 15 08:05:50 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-10-25
[Jul 15 08:05:50 UTC] OGR-P2: Summary: 2 packets (58.33 stats units)
0.00:33:38.82 - [28.89 Mnodes/s]
[Jul 15 08:05:50 UTC] OGR-P2: 35 packets remain in buff-in.ogf
[Jul 15 08:05:50 UTC] OGR-P2: 2 packets (58.33 stats units) are in
buff-out.ogf
[Jul 15 08:29:22 UTC] OGR-P2 #b: Completed 25/7-9-28-13-4-31 (25.81 stats units)
0.00:28:04.96 - [15,317,316 nodes/s]
[Jul 15 08:29:22 UTC] OGR-P2 #b: 25/7-9-28-13-4-31 [25,809,065,923 nodes]
[Jul 15 08:29:22 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-11-23
[Jul 15 08:29:22 UTC] OGR-P2: Summary: 3 packets (84.14 stats units)
0.00:57:10.91 - [24.52 Mnodes/s]
[Jul 15 08:29:22 UTC] OGR-P2: 34 packets remain in buff-in.ogf
[Jul 15 08:29:22 UTC] OGR-P2: 3 packets (84.14 stats units) are in
buff-out.ogf
[Jul 15 08:38:49 UTC] OGR-P2 #a: Completed 25/7-9-28-13-10-25 (25.75 stats units)
0.00:32:58.74 - [13,013,717 nodes/s]
[Jul 15 08:38:49 UTC] OGR-P2 #a: 25/7-9-28-13-10-25 [25,750,763,310 nodes]
[Jul 15 08:38:49 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-12-22
[Jul 15 08:38:49 UTC] OGR-P2: Summary: 4 packets (109.89 stats units)
0.01:06:37.92 - [27.49 Mnodes/s]
[Jul 15 08:38:49 UTC] OGR-P2: 33 packets remain in buff-in.ogf
[Jul 15 08:38:49 UTC] OGR-P2: 4 packets (109.89 stats units) are in
buff-out.ogf
[Jul 15 09:15:55 UTC] OGR-P2 #a: Completed 25/7-9-28-13-12-22 (30.08 stats units)
0.00:37:05.82 - [13,515,689 nodes/s]
[Jul 15 09:15:55 UTC] OGR-P2 #a: 25/7-9-28-13-12-22 [30,083,491,276 nodes]
[Jul 15 09:15:55 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-14-20
[Jul 15 09:15:55 UTC] OGR-P2: Summary: 5 packets (139.97 stats units)
0.01:43:44.00 - [22.49 Mnodes/s]
[Jul 15 09:15:55 UTC] OGR-P2: 32 packets remain in buff-in.ogf
[Jul 15 09:15:55 UTC] OGR-P2: 5 packets (139.97 stats units) are in
buff-out.ogf
[Jul 15 09:16:30 UTC] OGR-P2 #b: Completed 25/7-9-28-13-11-23 (36.25 stats units)
0.00:47:07.97 - [12,817,277 nodes/s]
[Jul 15 09:16:30 UTC] OGR-P2 #b: 25/7-9-28-13-11-23 [36,246,877,561 nodes]
[Jul 15 09:16:30 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-17-18
[Jul 15 09:16:30 UTC] OGR-P2: Summary: 6 packets (176.22 stats units)
0.01:44:19.11 - [28.15 Mnodes/s]
[Jul 15 09:16:30 UTC] OGR-P2: 31 packets remain in buff-in.ogf
[Jul 15 09:16:30 UTC] OGR-P2: 6 packets (176.22 stats units) are in
buff-out.ogf
[Jul 15 09:45:23 UTC] OGR-P2 #b: Completed 25/7-9-28-13-17-18 (33.40 stats units)
0.00:28:51.74 - [19,289,681 nodes/s]
[Jul 15 09:45:23 UTC] OGR-P2 #b: 25/7-9-28-13-17-18 [33,404,712,905 nodes]
[Jul 15 09:45:23 UTC] OGR-P2 #b: Loaded 25/7-9-28-13-19-17
[Jul 15 09:45:23 UTC] OGR-P2: Summary: 7 packets (209.62 stats units)
0.02:13:11.18 - [26.23 Mnodes/s]
[Jul 15 09:45:23 UTC] OGR-P2: 30 packets remain in buff-in.ogf
[Jul 15 09:45:23 UTC] OGR-P2: 7 packets (209.62 stats units) are in
buff-out.ogf
[Jul 15 09:51:54 UTC] OGR-P2 #a: Completed 25/7-9-28-13-14-20 (36.48 stats units)
0.00:35:57.49 - [16,907,992 nodes/s]
[Jul 15 09:51:54 UTC] OGR-P2 #a: 25/7-9-28-13-14-20 [36,478,825,087 nodes]
[Jul 15 09:51:54 UTC] OGR-P2 #a: Loaded 25/7-9-28-13-22-12
[Jul 15 09:51:54 UTC] OGR-P2: Summary: 8 packets (246.10 stats units)
0.02:19:42.20 - [29.36 Mnodes/s]
[Jul 15 09:51:54 UTC] OGR-P2: 29 packets remain in buff-in.ogf
[Jul 15 09:51:54 UTC] OGR-P2: 8 packets (246.10 stats units) are in
buff-out.ogf
[Jul 15 09:59:30 UTC] *Break* Shutting down...
[Jul 15 09:59:30 UTC] OGR-P2 #b: Saved 25/7-9-28-13-19-17 (13.29 Gnodes done)
0.00:14:07.46 - [15,677,094 nodes/s]
[Jul 15 09:59:30 UTC] OGR-P2 #a: Saved 25/7-9-28-13-22-12 (6.60 Gnodes done)
0.00:07:36.44 - [14,450,020 nodes/s]
[Jul 15 09:59:30 UTC] OGR-P2: Summary: 8 packets (246.10 stats units)
0.02:27:18.64 - [27.84 Mnodes/s]
[Jul 15 09:59:30 UTC] OGR-P2: 31 packets are in buff-in.ogf
[Jul 15 09:59:30 UTC] OGR-P2: 8 packets (246.10 stats units) are in
buff-out.ogf
[Jul 15 09:59:30 UTC] Shutdown complete.


Any idea why a 64 bit Linux decreases the MNodes/s so much?
the-mk
07-15-2007, 06:34 AM
dnetc -bench OGR-P2 on 32 bit client:


distributed.net client for Linux Copyright 1997-2006, distributed.net
Please visit http://www.distributed.net/ for up-to-date contest information.


dnetc v2.9013-498-CFR-06082120 for Linux (Linux 2.6.17-14mdv).
Please provide the *entire* version descriptor when submitting bug reports.
The distributed.net bug report pages are at http://www.distributed.net/bugs/
Using email address (distributed.net ID) 'the_mk@gmx.at'

[Jul 15 10:30:01 UTC] Automatic processor type detection found
an AMD K8-11 processor.
[Jul 15 10:30:01 UTC] OGR-P2: using core #0 (GARSP 6.0-A).
[Jul 15 10:30:21 UTC] OGR-P2: Benchmark for core #0 (GARSP 6.0-A)
0.00:00:17.70 [20,318,163 nodes/sec]
[Jul 15 10:30:21 UTC] OGR-P2: using core #1 (GARSP 6.0-B).
[Jul 15 10:30:40 UTC] OGR-P2: Benchmark for core #1 (GARSP 6.0-B)
0.00:00:16.93 [14,190,934 nodes/sec]
[Jul 15 10:30:40 UTC] OGR-P2: using core #2 (GARSP 6.0-asm-rt1-gen).
[Jul 15 10:31:00 UTC] OGR-P2: Benchmark for core #2 (GARSP 6.0-asm-rt1-gen)
0.00:00:17.18 [23,528,537 nodes/sec]
[Jul 15 10:31:00 UTC] OGR-P2: using core #3 (GARSP 6.0-asm-rt1-mmx).
[Jul 15 10:31:19 UTC] OGR-P2: Benchmark for core #3 (GARSP 6.0-asm-rt1-mmx)
0.00:00:16.76 [30,172,332 nodes/sec]
[Jul 15 10:31:19 UTC] OGR-P2: using core #4 (GARSP 6.0-asm-rt1-mmx-amd).
[Jul 15 10:31:39 UTC] OGR-P2: Benchmark for core #4 (GARSP 6.0-asm-rt1-mmx-amd)
0.00:00:17.02 [32,081,815 nodes/sec]


dnetc -bench OGR-P2 on 32 bit client:


distributed.net client for Linux Copyright 1997-2005, distributed.net
Please visit http://www.distributed.net/ for up-to-date contest information.


dnetc v2.9011-496-CFR-05060815 for Linux (Linux 2.6.17-14mdv).
Please provide the *entire* version descriptor when submitting bug reports.
The distributed.net bug report pages are at http://www.distributed.net/bugs/
Using email address (distributed.net ID) 'the_mk@gmx.at'

[Jul 15 10:32:43 UTC] OGR-P2: Benchmark for core #0 (GARSP 6.0-64)
0.00:00:17.40 [34,386,874 nodes/sec]


does the -bench show the overall performance of all cores together? or does it just show the result of one core?

I'm confused now! :confused:

BTW: I used three posts because of the 10,000 character limit per post...
alpha
07-15-2007, 11:27 AM
nodes/sec after each benchmark line is the performance of that particular core. So in your case on 32-bit platform core #4 is fastest and on 64-bit core #0 is fastest.

When you start up the 64-bit client, does it say that it is using core #0?
the-mk
07-15-2007, 12:18 PM
Since there is only one core at the 64-bit client, it doesn't say, which one it uses...

I wanted to test the 32-bit OS with the 32-bit client, but I sold my car (http://www.free-dc.org/forum/showthread.php?t=5888) this afternoon, so I had no time to investigate :( I'll do further testing next weekend, if I'll find some time...

If you are a good logfile-reader, you should see that the numbers while running are not so high as the numbers of the benchmark (only about a half of it). That worries me!
the-mk
07-20-2007, 02:03 PM
Now I installed 32 bit version of Mandriva Linux free, seems to be the same things: poor performance! A 4800+ X2 must perform for higher numbers!


dnetc v2.9013-498-CFR-06082120 for Linux (Linux 2.6.17-14mdv).
Using email address (distributed.net ID) 'the_mk@gmx.at'

[Jul 20 17:41:23 UTC] Automatic processor detection found 2 processors.
[Jul 20 17:41:23 UTC] Loading crunchers with work...
[Jul 20 17:41:23 UTC] Automatic processor type detection found
an AMD K8-11 processor.
[Jul 20 17:41:23 UTC] OGR-P2: using core #4 (GARSP 6.0-asm-rt1-mmx-amd).
[Jul 20 17:41:23 UTC] OGR-P2 #a: Loaded 25/1-6-4-8-73-2
[Jul 20 17:41:23 UTC] OGR-P2 #b: Loaded 25/1-6-4-8-73-3
[Jul 20 17:41:23 UTC] OGR-P2: 44 packets remain in buff-in.ogf
[Jul 20 17:41:23 UTC] OGR-P2: 0 packets are in buff-out.ogf
[Jul 20 17:41:23 UTC] 2 crunchers ('a' and 'b') have been started.
[Jul 20 17:43:37 UTC] OGR-P2 #a: Completed 25/1-6-4-8-73-2 (1.64 stats units)
0.00:02:13.87 - [12,272,749 nodes/s]
[Jul 20 17:43:37 UTC] OGR-P2 #a: 25/1-6-4-8-73-2 [1,642,953,024 nodes]
[Jul 20 17:43:37 UTC] OGR-P2 #a: Loaded 25/1-6-4-8-70-5 (633.42 Mnodes done)
[Jul 20 17:43:37 UTC] OGR-P2: Summary: 1 packet (1.64 stats units)
0.00:02:13.87 - [12.27 Mnodes/s]
[Jul 20 17:43:37 UTC] OGR-P2: 45 packets remain in buff-in.ogf
[Jul 20 17:43:37 UTC] OGR-P2: 1 packet (1.64 stats units) is in buff-out.ogf
[Jul 20 17:43:56 UTC] OGR-P2 #b: Completed 25/1-6-4-8-73-3 (1.87 stats units)
0.00:02:32.71 - [12,253,347 nodes/s]
[Jul 20 17:43:56 UTC] OGR-P2 #b: 25/1-6-4-8-73-3 [1,871,208,622 nodes]
[Jul 20 17:43:56 UTC] OGR-P2 #b: Loaded 25/1-6-4-8-72-3 (625.10 Mnodes done)
[Jul 20 17:43:56 UTC] OGR-P2: Summary: 2 packets (3.51 stats units)
0.00:02:33.13 - [22.95 Mnodes/s]
[Jul 20 17:43:56 UTC] OGR-P2: 44 packets remain in buff-in.ogf
[Jul 20 17:43:56 UTC] OGR-P2: 2 packets (3.51 stats units) are in
buff-out.ogf
[Jul 20 17:45:53 UTC] OGR-P2 #b: Completed 25/1-6-4-8-72-3 (2.04 stats units)
0.00:01:55.31 - [12,282,285 nodes/s]
[Jul 20 17:45:53 UTC] OGR-P2 #b: 25/1-6-4-8-72-3 [2,041,366,036 nodes]
[Jul 20 17:45:53 UTC] OGR-P2 #b: Loaded 25/1-6-4-9-3-72
[Jul 20 17:45:53 UTC] OGR-P2: Summary: 3 packets (5.55 stats units)
0.00:04:29.25 - [18.31 Mnodes/s]
[Jul 20 17:45:53 UTC] OGR-P2: 43 packets remain in buff-in.ogf
[Jul 20 17:45:53 UTC] OGR-P2: 3 packets (5.55 stats units) are in
buff-out.ogf
[Jul 20 17:46:16 UTC] OGR-P2 #a: Completed 25/1-6-4-8-70-5 (2.61 stats units)
0.00:02:38.67 - [12,427,346 nodes/s]
[Jul 20 17:46:16 UTC] OGR-P2 #a: 25/1-6-4-8-70-5 [2,605,268,632 nodes]
[Jul 20 17:46:16 UTC] OGR-P2 #a: Loaded 25/1-6-4-9-5-69
[Jul 20 17:46:16 UTC] OGR-P2: Summary: 4 packets (8.16 stats units)
0.00:04:53.07 - [23.55 Mnodes/s]
[Jul 20 17:46:16 UTC] OGR-P2: 42 packets remain in buff-in.ogf
[Jul 20 17:46:16 UTC] OGR-P2: 4 packets (8.16 stats units) are in
buff-out.ogf
[Jul 20 17:49:44 UTC] OGR-P2 #b: Completed 25/1-6-4-9-3-72 (2.83 stats units)
0.00:03:50.26 - [12,275,304 nodes/s]
[Jul 20 17:49:44 UTC] OGR-P2 #b: 25/1-6-4-9-3-72 [2,826,511,547 nodes]
[Jul 20 17:49:44 UTC] OGR-P2 #b: Loaded 25/1-6-4-9-18-56
[Jul 20 17:49:44 UTC] OGR-P2: Summary: 5 packets (10.99 stats units)
0.00:08:20.27 - [19.45 Mnodes/s]
[Jul 20 17:49:44 UTC] OGR-P2: 41 packets remain in buff-in.ogf
[Jul 20 17:49:44 UTC] OGR-P2: 5 packets (10.99 stats units) are in
buff-out.ogf
[Jul 20 17:53:12 UTC] OGR-P2 #a: Completed 25/1-6-4-9-5-69 (5.27 stats units)
0.00:06:56.07 - [12,667,966 nodes/s]
[Jul 20 17:53:12 UTC] OGR-P2 #a: 25/1-6-4-9-5-69 [5,270,761,009 nodes]
[Jul 20 17:53:12 UTC] OGR-P2 #a: Loaded 25/1-6-4-9-18-57
[Jul 20 17:53:12 UTC] OGR-P2: Summary: 6 packets (16.26 stats units)
0.00:11:49.14 - [21.15 Mnodes/s]
[Jul 20 17:53:12 UTC] OGR-P2: 40 packets remain in buff-in.ogf
[Jul 20 17:53:12 UTC] OGR-P2: 6 packets (16.26 stats units) are in
buff-out.ogf


while ./dnetc -bench OGR-P2 on that system says:


dnetc v2.9013-498-CFR-06082120 for Linux (Linux 2.6.17-14mdv).
Please provide the *entire* version descriptor when submitting bug reports.
The distributed.net bug report pages are at http://www.distributed.net/bugs/
Using email address (distributed.net ID) 'the_mk@gmx.at'

[Jul 20 17:56:59 UTC] Automatic processor type detection found
an AMD K8-11 processor.
[Jul 20 17:56:59 UTC] OGR-P2: using core #0 (GARSP 6.0-A).
[Jul 20 17:57:18 UTC] OGR-P2: Benchmark for core #0 (GARSP 6.0-A)
0.00:00:17.00 [20,128,228 nodes/sec]
[Jul 20 17:57:18 UTC] OGR-P2: using core #1 (GARSP 6.0-B).
[Jul 20 17:57:37 UTC] OGR-P2: Benchmark for core #1 (GARSP 6.0-B)
0.00:00:16.41 [14,250,542 nodes/sec]
[Jul 20 17:57:37 UTC] OGR-P2: using core #2 (GARSP 6.0-asm-rt1-gen).
[Jul 20 17:57:56 UTC] OGR-P2: Benchmark for core #2 (GARSP 6.0-asm-rt1-gen)
0.00:00:16.50 [23,606,015 nodes/sec]
[Jul 20 17:57:56 UTC] OGR-P2: using core #3 (GARSP 6.0-asm-rt1-mmx).
[Jul 20 17:58:14 UTC] OGR-P2: Benchmark for core #3 (GARSP 6.0-asm-rt1-mmx)
0.00:00:16.15 [30,262,322 nodes/sec]
[Jul 20 17:58:14 UTC] OGR-P2: using core #4 (GARSP 6.0-asm-rt1-mmx-amd).
[Jul 20 17:58:33 UTC] OGR-P2: Benchmark for core #4 (GARSP 6.0-asm-rt1-mmx-amd)
0.00:00:16.34 [32,107,240 nodes/sec]

--> no other high performance app is running on this machine, while dnetc crunches...
--> core #4 is the fastest core for one processor, what dnetc -bench states
--> core #4 is used while crunching
--> since I have a dual core, the overall MNodes/s should be around 64 MNodes/s and not around 20 MNodes/s :bonk: :confused:

I will do some further testing on Windows, but this is not the OS of choice for this machine :swear:

Anybody any idea what could be going wrong?
the-mk
07-20-2007, 02:16 PM
Maybe I've found the issue: AMD cool'n'quiet...

took a look at /proc/cpuinfo where it states that the CPU runs at 1000 mhz (even during dnetc)

disabled that feature in bios, now /proc/cpuinfo shows about 2500 mhz :D

hopefully the MNodes/s increase :cheers:
the-mk
07-20-2007, 02:23 PM
Now it behaves like a 4800+ X2 should do:



[Jul 20 18:13:44 UTC] Automatic processor detection found 2 processors.
[Jul 20 18:13:44 UTC] Loading crunchers with work...
[Jul 20 18:13:44 UTC] Automatic processor type detection found
an AMD K8-11 processor.
[Jul 20 18:13:44 UTC] OGR-P2: using core #4 (GARSP 6.0-asm-rt1-mmx-amd).
[Jul 20 18:13:44 UTC] OGR-P2 #a: Loaded 25/1-6-4-9-18-57 (5.65 Gnodes done)
[Jul 20 18:13:44 UTC] OGR-P2 #b: Loaded 25/1-6-4-9-18-56 (8.28 Gnodes done)
[Jul 20 18:13:44 UTC] OGR-P2: 40 packets remain in buff-in.ogf
[Jul 20 18:13:44 UTC] OGR-P2: 0 packets are in buff-out.ogf
[Jul 20 18:13:44 UTC] 2 crunchers ('a' and 'b') have been started.
[Jul 20 18:19:38 UTC] OGR-P2 #b: Completed 25/1-6-4-9-18-56 (19.47 stats units)
0.00:05:53.75 - [31,616,409 nodes/s]
[Jul 20 18:19:38 UTC] OGR-P2 #b: 25/1-6-4-9-18-56 [19,468,188,016 nodes]
[Jul 20 18:19:38 UTC] OGR-P2 #b: Loaded 25/1-6-4-9-23-51
[Jul 20 18:19:38 UTC] OGR-P2: Summary: 1 packet (19.47 stats units)
0.00:05:53.75 - [31.62 Mnodes/s]
[Jul 20 18:19:38 UTC] OGR-P2: 39 packets remain in buff-in.ogf
[Jul 20 18:19:38 UTC] OGR-P2: 1 packet (19.47 stats units) is in
buff-out.ogf
[Jul 20 18:19:58 UTC] OGR-P2 #a: Completed 25/1-6-4-9-18-57 (17.55 stats units)
0.00:06:14.04 - [31,815,687 nodes/s]
[Jul 20 18:19:58 UTC] OGR-P2 #a: 25/1-6-4-9-18-57 [17,546,561,193 nodes]
[Jul 20 18:19:58 UTC] OGR-P2 #a: Loaded 25/1-6-4-9-51-23
[Jul 20 18:19:58 UTC] OGR-P2: Summary: 2 packets (37.02 stats units)
0.00:06:14.19 - [61.69 Mnodes/s]
[Jul 20 18:19:58 UTC] OGR-P2: 38 packets remain in buff-in.ogf
[Jul 20 18:19:58 UTC] OGR-P2: 2 packets (37.02 stats units) are in
buff-out.ogf


:bonk: cool'n'quiet :bonk:

but since the x64-client has a little bit more MNodes/s I'll set up the x64 environment once again...
IronBits
07-20-2007, 06:49 PM
Now it behaves like a 4800+ X2 should do:

:bonk: cool'n'quiet :bonk:

but since the x64-client has a little bit more MNodes/s I'll set up the x64 environment once again... Yup! Nothing new there, keep it disabled at all times. ;)
alpha
07-21-2007, 12:11 PM
That doesn't explain why the 64-bit client was benchmarking faster than the 32-bit client, but ended up being slower during actual crunching. Or does cool n' quiet only kick in after x seconds of high load?
the-mk
07-21-2007, 05:50 PM
IMHO cool'n'quiet leaves the cpu at 1000 MHz, while there is no process using 100% when having a high priority. dnetc has only low priority, so it stays at 1000 MHz... So I think it'll increase MHz if you have a process with higher priority and higher load, but not so with dnetc, since it has low priority.
IronBits
07-21-2007, 09:28 PM
Cool and Quiet will try to reduce the processor speed every chance it gets.
When you are crunching, you never want the processor to slow down to save energy ($$$) or to run cooler (Get a bigger HSF).
http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_9485_9487%255e10272,00.html?redir=dtqc04

I have it disabled via the BIOS on any computer that I want full performance at all times.
It's a good thing to use when you have a laptop and want to conserve battery life.