p8102是个超级甜包

cuda

回复 12# 金鹏

我这里下载速度倒是很正常. 不过p8102在重新启动之后似乎很容易爆包, 今天已经爆掉2个了, 现在正在跑第3个.
出错代码是CoreStatus = FF (255). 我觉得应该不是硬件问题, 软件bug可能性比较大.

1. [04:49:45] Project: 8102 (Run 0, Clone 45, Gen 0)
   
2. [04:49:45]
   
3. [04:49:45] Assembly optimizations on if available.
   
4. [04:49:45] Entering M.D.
   
5. ......
   
6. [04:50:22] Resuming from checkpoint
   
7. [04:50:25] Verified work/wudata_01.log
   
8. [04:50:25] Verified work/wudata_01.trr
   
9. [04:50:25] Verified work/wudata_01.xtc
   
10. [04:50:25] Verified work/wudata_01.edr
    
11. [04:50:27] Completed 2850 out of 250000 steps  (1%)
    
12. 
    
13. NOTE: Turning on dynamic load balancing
    
14. 
    
15. [05:13:11] Completed 5000 out of 250000 steps  (2%)
    
16. [05:39:03] Completed 7500 out of 250000 steps  (3%)
    
17. [06:04:49] Completed 10000 out of 250000 steps  (4%)
    
18. [06:30:34] Completed 12500 out of 250000 steps  (5%)
    
19. [06:56:18] Completed 15000 out of 250000 steps  (6%)
    
20. [07:01:40] CoreStatus = FF (255)
    
21. [07:01:40] Sending work to server
    
22. [07:01:40] Project: 8102 (Run 0, Clone 45, Gen 0)
    
23. [07:01:40] - Error: Could not get length of results file work/wuresults_01.dat
    
24. [07:01:40] - Error: Could not read unit 01 file. Removing from queue.
    
25. ......
    
26. [12:59:45] Project: 8102 (Run 0, Clone 45, Gen 0)
    
27. [12:59:45]
    
28. [12:59:46] Assembly optimizations on if available.
    
29. [12:59:46] Entering M.D.
    
30. [13:01:11] Resuming from checkpoint
    
31. [13:01:15] Verified work/wudata_02.log
    
32. [13:01:16] Verified work/wudata_02.trr
    
33. [13:01:16] Verified work/wudata_02.xtc
    
34. [13:01:16] Verified work/wudata_02.edr
    
35. [13:01:20] Completed 32885 out of 250000 steps  (13%)
    
36. 
    
37. NOTE: Turning on dynamic load balancing
    
38. 
    
39. [13:16:36] CoreStatus = FF (255)
    
40. [13:16:36] Sending work to server
    
41. [13:16:36] Project: 8102 (Run 0, Clone 45, Gen 0)
    
42. [13:16:36] - Error: Could not get length of results file work/wuresults_02.dat
    
43. [13:16:36] - Error: Could not read unit 02 file. Removing from queue.

复制代码

Keyco

回复  xiaoren6412


    V6 linux    -bigbeta下载速度真坑啊
金鹏 发表于 2012-7-29 15:28

    最后金版下了多久下下来了啊？我这里2个小时没有下载成功过。。。LINUX下的。

Keyco

回复  金鹏

我这里下载速度倒是很正常. 不过p8102在重新启动之后似乎很容易爆包, 今天已经爆掉2个了, 现 ...
cuda 发表于 2012-7-30 00:59

    我之前的3960X跑BA12，在VB下，也是和你一样，一直爆掉。。。

金鹏

回复 17# Keyco

5Kb下载速速，最终是个8101，坑啊

cuda

回复 19# 金鹏

似乎直接用-bigadv参数接到p8102的几率更大些。
此外有人成功接到p6905/p6906吗？这两个还不知道是甜包还是毒包。
还有，发现跑p8102的时候功率计显示不稳定，有20W左右的跳动，有没有人也遇到这个现象？

maxzong

我用VPN下载，速度倒是没问题。只是连下3个8101， 接不到8102啊。

Keyco

表示暂时看不到有8102包，还是8101.

金鹏

回复 22# Keyco

使用-bigbeta表示完全接不到8102，今日换回WEB下的 -bigadv -verbosity 9 -smp 32命令了，

期待能像O版那样只接接到8102

金鹏

回复 21# maxzong

兄弟使用的是什么参数？

O版直接使用 -smp64 -bigadv -verbosity 9 接到了8102

金鹏

回复 20# cuda


    6905/6906木有见到,目前12T机器全线6901

另外斯坦福论坛上7IM说6905/6906属于BA16的范畴,不知道兄弟是否有更详细的资料?

ocw

我現在一台機接 8101, 另一台接 8102.

cuda

回复 25# 金鹏

6905/6906应该是BA16, BA12今天正式退出舞台. 金版的提高PPD技巧教程又要再次修订了.

金鹏

回复 27# cuda

BA12退出舞台后,那些技巧真的无从说起了,适用面很狭小了,感谢兄弟提醒,已经全面修改了

BA16转正后3213的PPD要大跌了

金鹏

回复 1# cuda

我現在兩台 6272 都接到了8102, 11分11秒 1%.
ocw 发表于 2012-7-29 11:20

看到了8101和8102对比

by Macaholic » Mon Jul 30, 2012 10:16 pm

4P 6128 stock;

1. p8101 (DLB did not engage)
   
2. 
   
3. [22:15:03] Completed 237500 out of 250000 steps  (95%)
   
4. [22:37:26] Completed 240000 out of 250000 steps  (96%)
   
5. [22:59:46] Completed 242500 out of 250000 steps  (97%)
   
6. [23:22:08] Completed 245000 out of 250000 steps  (98%)
   
7. [23:44:27] Completed 247500 out of 250000 steps  (99%)
   
8. [00:06:49] Completed 250000 out of 250000 steps  (100%)

复制代码

1. p8102
   
2. 
   
3. [03:53:46] Completed 237500 out of 250000 steps  (95%)
   
4. [04:09:54] Completed 240000 out of 250000 steps  (96%)
   
5. [04:26:00] Completed 242500 out of 250000 steps  (97%)
   
6. [04:42:04] Completed 245000 out of 250000 steps  (98%)
   
7. [04:58:08] Completed 247500 out of 250000 steps  (99%)
   
8. [05:14:13] Completed 250000 out of 250000 steps  (100%)
   

复制代码

可以看出8101的TPF=22：3   8102 TPF=16：03    TPF提升27%   PPD提升61%，达到27万


推测一下俺的双路3.1G  E5  8101 TPF=16：21   ，假设8102 TPF提升25%达到 12:16      8102   PPD则可达到40.4万

基本上与CUDA兄弟预测的差不多


坏消息是8102将被调低分值点

by kasson » Mon Jul 30, 2012 11:19 pm
Thanks, everyone, for testing 8102. The good news and the bad news is that we will adjust points to bring them in line with 6901, 6902, 6905, 6906, 8101. So by all means enjoy a points bump in our initial stages of release, but please do be aware that it may not last.