CoreStatus = 7A (122)--急求高手帮着找出病因

金鹏

Z9PE D8 WS刷了3206新BIOS后,按照以前的设置上了103外频,结果跑了一天不到出现这个CoreStatus = 7A (122),貌似两个2687W ES 负载不均衡造成?
自动核心电压,进入BIOS查看CPU1 1.12V  CPU2  1.06V  内存电压1.51V  1600频率 时序9-9-9-24-128-1 ,开启内存交错和NUMA   ,功耗限制185瓦,高性能服务级别设置, 2个 CPU温度50-52度
求各路大神帮着分析原因

1. (40%)
   
2. [04:13:22] Completed 102500 out of 250000 steps  (41%)
   
3. [04:29:15] Completed 105000 out of 250000 steps  (42%)
   
4. [04:45:09] Completed 107500 out of 250000 steps  (43%)
   
5. 
   
6. -------------------------------------------------------
   
7. Program Gromacs, VERSION 4.5.3
   
8. Source code file: /vspm58/VM/fah-converted/mnt/fah_windows_build/LinuxBuilds/gromacs-4.5.3/src/mdlib/pme.c, line: 534
   
9. 
   
10. Fatal error:
    
11. 6 particles communicated to PME node 16 are more than 2/3 times the cut-off out of the domain decomposition cell of their charge group in dimension x.
    
12. This usually means that your system is not well equilibrated.
    
13. For more information and tips for troubleshooting, please check the GROMACS
    
14. website at http://www.gromacs.org/Documentation/Errors
    
15. -------------------------------------------------------
    
16. 
    
17. Thanx for Using GROMACS - Have a Nice Day
    
18. 
    
19. [04:54:22] mdrun returned 255
    
20. [04:54:22] Going to send back what have done -- stepsTotalG=250000
    
21. [04:54:22] Work fraction=162.9188 steps=250000.
    
22. [04:54:26] logfile size=89648 infoLength=89648 edr=25 trr=1
    
23. [04:54:26] logfile size: 89648 info=89648 bed=25 hdr=1
    
24. [04:54:26] - Writing 90186 bytes of core data to disk...
    
25. [04:54:26] Done: 89674 -> 11605 (compressed to 12.9 percent)
    
26. [04:54:26]   ... Done.
    
27. [04:59:27]
    
28. [04:59:27] Folding@home Core Shutdown: UNSTABLE_MACHINE
    
29. [04:59:27] CoreStatus = 7A (122)
    
30. [04:59:27] Sending work to server
    
31. [04:59:27] Project: 8101 (Run 21, Clone 0, Gen 71)
    
32. 
    
33. [04:59:27] + Attempting to send results [November 9 04:59:27 UTC]
    
34. [04:59:27] - Reading file work/wuresults_01.dat from core
    
35. [04:59:27]   (Read 12117 bytes from disk)
    
36. [04:59:27] Connecting to http://128.143.231.201:8080/
    
37. [04:59:35] Posted data.
    
38. [04:59:35] Initial: 0000; - Uploaded at ~1 kB/s
    
39. [04:59:35] - Averaged speed for that direction ~147 kB/s
    
40. [04:59:35] + Results successfully sent
    
41. [04:59:35] Thank you for your contribution to Folding@Home.
    
42. thekraken: The Kraken 0.7-pre15 (compiled Sun Jul  1 20:38:22 MST 2012 by root@Test)
    
43. thekraken: Processor affinity wrapper for Folding@Home
    
44. thekraken: The Kraken comes with ABSOLUTELY NO WARRANTY; licensed under GPLv2
    
45. thekraken: PID: 1963
    
46. thekraken: Logging to thekraken.log
    
47. [04:59:35] Trying to send all finished work units
    
48. [04:59:35] + No unsent completed units remaining.
    
49. [04:59:35] - Preparing to get new work unit...
    
50. [04:59:35] Cleaning up work directory
    
51. [05:01:50] + Attempting to get work packet
    
52. [05:01:50] Passkey found
    
53. [05:01:50] - Will indicate memory of 32132 MB
    
54. [05:01:50] - Connecting to assignment server
    
55. [05:01:50] Connecting to http://assign.stanford.edu:8080/
    
56. [05:01:53] Posted data.
    
57. [05:01:53] Initial: 8F80; - Successful: assigned to (128.143.231.201).
    
58. [05:01:53] + News From Folding@Home: Welcome to Folding@Home
    
59. [05:01:53] Loaded queue successfully.
    
60. [05:01:53] Sent data
    
61. [05:01:53] Connecting to http://128.143.231.201:8080/
    
62. [05:02:01] Posted data.
    
63. [05:02:01] Initial: 0000; - Receiving payload (expected size: 30306108)
    
64. [05:13:04] - Downloaded at ~44 kB/s
    
65. [05:13:04] - Averaged speed for that direction ~211 kB/s
    
66. [05:13:04] + Received work.
    
67. [05:13:04] Trying to send all finished work units
    
68. [05:13:04] + No unsent completed units remaining.
    
69. [05:13:04] + Closed connections
    
70. [05:13:09]
    
71. [05:13:09] + Processing work unit
    
72. [05:13:09] Core required: FahCore_a5.exe
    
73. [05:13:09] Core found.
    
74. [05:13:09] Working on queue slot 02 [November 9 05:13:09 UTC]
    
75. [05:13:09] + Working ...
    
76. [05:13:09] - Calling './FahCore_a5.exe -dir work/ -nice 19 -suffix 02 -np 32 -checkpoint 15 -verbose -lifeline 1844 -version 634'
    
77. 
    
78. thekraken: The Kraken 0.7-pre15 (compiled Sun Jul  1 20:38:22 MST 2012 by root@Test)
    
79. thekraken: Processor affinity wrapper for Folding@Home
    
80. thekraken: The Kraken comes with ABSOLUTELY NO WARRANTY; licensed under GPLv2
    
81. thekraken: PID: 1967
    
82. thekraken: Logging to thekraken.log
    
83. [05:13:09]
    
84. [05:13:09] *------------------------------*
    
85. [05:13:09] Folding@Home Gromacs SMP Core
    
86. [05:13:09] Version 2.27 (Thu Feb 10 09:46:40 PST 2011)
    
87. [05:13:09]
    
88. [05:13:09] Preparing to commence simulation
    
89. [05:13:09] - Looking at optimizations...
    
90. [05:13:09] - Created dyn
    
91. [05:13:09] - Files status OK
    
92. [05:13:11] - Expanded 30305596 -> 33158020 (decompressed 109.4 percent)
    
93. [05:13:11] Called DecompressByteArray: compressed_data_size=30305596 data_size=33158020, decompressed_data_size=33158020 diff=0
    
94. [05:13:11] - Digital signature verified
    
95. [05:13:11]
    
96. [05:13:11] Project: 8101 (Run 25, Clone 6, Gen 56)
    
97. [05:13:11]
    
98. [05:13:11] Assembly optimizations on if available.
    
99. [05:13:11] Entering M.D.
    
100.                          :-)  G  R  O  M  A  C  S  (-:
     
101. 
     
102.                    Groningen Machine for Chemical Simulation
     
103. 
     
104.                             :-)  VERSION 4.5.3  (-:
     
105. 
     
106.         Written by Emile Apol, Rossen Apostolov, Herman J.C. Berendsen,
     
107.       Aldert van Buuren, Pär Bjelkmar, Rudi van Drunen, Anton Feenstra,
     
108.         Gerrit Groenhof, Peter Kasson, Per Larsson, Pieter Meulenhoff,
     
109.            Teemu Murtola, Szilard Pall, Sander Pronk, Roland Schulz,
     
110.                 Michael Shirts, Alfons Sijbers, Peter Tieleman,
     
111. 
     
112.                Berk Hess, David van der Spoel, and Erik Lindahl.
     
113. 
     
114.        Copyright (c) 1991-2000, University of Groningen, The Netherlands.
     
115.             Copyright (c) 2001-2010, The GROMACS development team at
     
116.         Uppsala University & The Royal Institute of Technology, Sweden.
     
117.             check out http://www.gromacs.org for more information.
     
118. 
     
119. 
     
120.                                :-)  Gromacs  (-:
     
121. 
     
122. Reading file work/wudata_02.tpr, VERSION 4.5.5-dev-20120903-d64b9e3 (single precision)
     
123. [05:13:18] Mapping NT from 32 to 32
     
124. Starting 32 threads
     
125. Making 2D domain decomposition 8 x 4 x 1
     
126. starting mdrun 'FP_membrane in water'
     
127. 14250000 steps,  57000.0 ps (continuing from step 14000000,  56000.0 ps).
     
128. [05:13:24] Completed 0 out of 250000 steps  (0%)
     
129. 
     
130. NOTE: Turning on dynamic load balancing
     
131. 
     
132. [05:29:25] Completed 2500 out of 250000 steps  (1%)
     
133. [05:45:01] Completed 5000 out of 250000 steps  (2%)
     

复制代码

cuda

回复 1# 金鹏

在过度超频的机器上常见这个错误，不过按理讲103外频并不高。建议再观察观察。

wpf999

UNSTABLE_MACHINE，根据提示：不稳定的机器。估计是超频引起的

zflowers

100-103意义不大吧

金鹏

回复 3# wpf999
回复 2# cuda

1. Program Gromacs, VERSION 4.5.3
   
2. Source code file: /vspm58/VM/fah-converted/mnt/fah_windows_build/LinuxBuilds/gromacs-4.5.3/src/mdlib/pme.c, line: 534
   
3. 
   
4. Fatal error:
   
5. 6 particles communicated to PME node 16 are more than 2/3 times the cut-off out of the domain decomposition cell of their charge group in dimension x.
   
6. This usually means that your system is not well equilibrated.

复制代码

排除核心电压问题，我估计是出在内存电压问题上，8跟1600的4G内存默压是1.65V，我设置在1.51V，现在设置成1.55V跑几天看看咋样

金鹏

100-103意义不大吧
zflowers 发表于 2012-11-9 17:46

103下跑8101包比100下每帧快了30秒上下，一个包正好节省出1-2个存盘上传下载的时间，相当于增加了30K的PPD

cuda

回复 5# 金鹏

内存降压恐怕没有什么好处，也省不了多少电，建议改回默认电压。金士顿这款内存跑1.51V感觉有点悬，以前也是跑这个电压吗？

金鹏

回复 7# cuda

以前的板子上也是这个电压，木有一点问题

换新的这个板子不论是CPU核心还是内存电压都不是很准确，而且有小幅掉压