p1387{3,4,6} to F@H（N卡全系A卡仅支持 species=6 Navi-based ））

金鹏

Keyco 发表于 2020-3-25 23:12
这个project从“13.82.98.119”，“fah3.eastus.cloudapp.azure.com” 过来的，微软云就是给力。下载上传比 ...

40.114.52.201，13.90.152.57，13.82.98.119都是，下载上传很给力

Lynt

所以我的2060 PPD不算高的原因可能跟CPU差有关？i3 4130拖了后腿

金鹏

Lynt 发表于 2020-3-26 09:52
所以我的2060 PPD不算高的原因可能跟CPU差有关？i3 4130拖了后腿

理论上如此。

我在32线程的机器上开算13873时，在启动计算初期CPU吃满32线程，几乎瞬间开始GPU的计算，

而在8线程双卡机器上启动计算需要过好几分钟后GPU才开始计算

Lynt

金鹏 发表于 2020-3-26 10:27
理论上如此。

我在32线程的机器上开算13873时，在启动计算初期CPU吃满32线程，几乎瞬间开始GPU的计算，

看了一下log，正常TPF大约是2分6秒，从第一个百分点开始，每5个出现一次TPF增加18秒左右，没有出现几分钟的差距，感觉还好

1. 01:53:33:WU01:FS00:Download complete
   
2. 01:53:33:WU01:FS00:Received Unit: id:01 state:DOWNLOAD error:NO_ERROR project:13873 run:0 clone:1264 gen:0 core:0x22 unit:0x000000000d5262775e7ade2bf91373d6
   
3. 01:53:33:WU01:FS00:Starting
   
4. 01:53:33:WU01:FS00:Running FahCore: /usr/bin/FAHCoreWrapper /var/lib/fahclient/cores/cores.foldingathome.org/v7/lin/64bit/beta/Core_22.fah/FahCore_22 -dir 01 -suffix 01 -version 704 -lifeline 23943 -checkpoint 3 -gpu 0 -gpu-vendor nvidia
   
5. 01:53:33:WU01:FS00:Started FahCore on PID 5240
   
6. 01:53:33:WU01:FS00:Core PID:5244
   
7. 01:53:33:WU01:FS00:FahCore 0x22 started
   
8. 01:53:33:WU01:FS00:0x22:*********************** Log Started 2020-03-26T01:53:33Z ***********************
   
9. 01:53:33:WU01:FS00:0x22:*************************** Core22 Folding[url=home.php?mod=space&uid=92741]@home[/url] Core ***************************
   
10. 01:53:33:WU01:FS00:0x22:       Type: 0x22
    
11. 01:53:33:WU01:FS00:0x22:       Core: Core22
    
12. 01:53:33:WU01:FS00:0x22:    Website: https://foldingathome.org/
    
13. 01:53:33:WU01:FS00:0x22:  Copyright: (c) 2009-2018 foldingathome.org
    
14. 01:53:33:WU01:FS00:0x22:     Author: John Chodera <[email protected]> and Rafal Wiewiora
    
15. 01:53:33:WU01:FS00:0x22:             <[email protected]>
    
16. 01:53:33:WU01:FS00:0x22:       Args: -dir 01 -suffix 01 -version 704 -lifeline 5240 -checkpoint 3 -gpu 0
    
17. 01:53:33:WU01:FS00:0x22:             -gpu-vendor nvidia
    
18. 01:53:33:WU01:FS00:0x22:     Config: <none>
    
19. 01:53:33:WU01:FS00:0x22:************************************ Build *************************************
    
20. 01:53:33:WU01:FS00:0x22:    Version: 0.0.2
    
21. 01:53:33:WU01:FS00:0x22:       Date: Dec 6 2019
    
22. 01:53:33:WU01:FS00:0x22:       Time: 21:20:17
    
23. 01:53:33:WU01:FS00:0x22: Repository: Git
    
24. 01:53:33:WU01:FS00:0x22:   Revision: f87d92b58abdf7e6bf2e173cfbc4dc3e837c7042
    
25. 01:53:33:WU01:FS00:0x22:     Branch: core22
    
26. 01:53:33:WU01:FS00:0x22:   Compiler: GNU 4.8.2 20140120 (Red Hat 4.8.2-15)
    
27. 01:53:33:WU01:FS00:0x22:    Options: -std=gnu++98 -O3 -funroll-loops
    
28. 01:53:33:WU01:FS00:0x22:   Platform: linux2 4.9.87-linuxkit-aufs
    
29. 01:53:33:WU01:FS00:0x22:       Bits: 64
    
30. 01:53:33:WU01:FS00:0x22:       Mode: Release
    
31. 01:53:33:WU01:FS00:0x22:************************************ System ************************************
    
32. 01:53:33:WU01:FS00:0x22:        CPU: Intel(R) Core(TM) i3-4130 CPU @ 3.40GHz
    
33. 01:53:33:WU01:FS00:0x22:     CPU ID: GenuineIntel Family 6 Model 60 Stepping 3
    
34. 01:53:33:WU01:FS00:0x22:       CPUs: 4
    
35. 01:53:33:WU01:FS00:0x22:     Memory: 7.74GiB
    
36. 01:53:33:WU01:FS00:0x22:Free Memory: 4.36GiB
    
37. 01:53:33:WU01:FS00:0x22:    Threads: POSIX_THREADS
    
38. 01:53:33:WU01:FS00:0x22: OS Version: 3.16
    
39. 01:53:33:WU01:FS00:0x22:Has Battery: false
    
40. 01:53:33:WU01:FS00:0x22: On Battery: false
    
41. 01:53:33:WU01:FS00:0x22: UTC Offset: 8
    
42. 01:53:33:WU01:FS00:0x22:        PID: 5244
    
43. 01:53:33:WU01:FS00:0x22:        CWD: /var/lib/fahclient/work
    
44. 01:53:33:WU01:FS00:0x22:         OS: Linux 3.16.0-23-generic x86_64
    
45. 01:53:33:WU01:FS00:0x22:    OS Arch: AMD64
    
46. 01:53:33:WU01:FS00:0x22:********************************************************************************
    
47. 01:53:33:WU01:FS00:0x22:Project: 13873 (Run 0, Clone 1264, Gen 0)
    
48. 01:53:33:WU01:FS00:0x22:Unit: 0x000000000d5262775e7ade2bf91373d6
    
49. 01:53:33:WU01:FS00:0x22:Reading tar file core.xml
    
50. 01:53:33:WU01:FS00:0x22:Reading tar file integrator.xml
    
51. 01:53:33:WU01:FS00:0x22:Reading tar file state.xml
    
52. 01:53:33:WU01:FS00:0x22:Reading tar file system.xml
    
53. 01:53:34:WU00:FS00:Upload 11.70%
    
54. 01:53:34:WU01:FS00:0x22:Digital signatures verified
    
55. 01:53:34:WU01:FS00:0x22:Folding@home GPU Core22 Folding@home Core
    
56. 01:53:34:WU01:FS00:0x22:Version 0.0.2
    
57. 01:53:40:WU00:FS00:Upload 40.69%
    
58. 01:53:46:WU00:FS00:Upload 66.56%
    
59. 01:53:52:WU00:FS00:Upload 91.39%
    
60. 01:53:55:WU00:FS00:Upload complete
    
61. 01:53:55:WU00:FS00:Server responded WORK_ACK (400)
    
62. 01:53:55:WU00:FS00:Final credit estimate, 180860.00 points
    
63. 01:53:55:WU00:FS00:Cleaning up
    
64. 01:54:35:WU01:FS00:0x22:Completed 0 out of 1000000 steps (0%)
    
65. 01:54:35:WU01:FS00:0x22:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900
    
66. 01:56:55:WU01:FS00:0x22:Completed 10000 out of 1000000 steps (1%)
    
67. 01:59:00:WU01:FS00:0x22:Completed 20000 out of 1000000 steps (2%)
    
68. 02:01:06:WU01:FS00:0x22:Completed 30000 out of 1000000 steps (3%)
    
69. 02:03:12:WU01:FS00:0x22:Completed 40000 out of 1000000 steps (4%)
    
70. 02:05:19:WU01:FS00:0x22:Completed 50000 out of 1000000 steps (5%)
    
71. 02:07:45:WU01:FS00:0x22:Completed 60000 out of 1000000 steps (6%)
    
72. 02:09:52:WU01:FS00:0x22:Completed 70000 out of 1000000 steps (7%)
    
73. 02:11:58:WU01:FS00:0x22:Completed 80000 out of 1000000 steps (8%)
    
74. 02:14:04:WU01:FS00:0x22:Completed 90000 out of 1000000 steps (9%)
    
75. 02:16:10:WU01:FS00:0x22:Completed 100000 out of 1000000 steps (10%)
    
76. 02:18:34:WU01:FS00:0x22:Completed 110000 out of 1000000 steps (11%)
    
77. 02:20:40:WU01:FS00:0x22:Completed 120000 out of 1000000 steps (12%)
    
78. 02:22:46:WU01:FS00:0x22:Completed 130000 out of 1000000 steps (13%)
    
79. 02:24:53:WU01:FS00:0x22:Completed 140000 out of 1000000 steps (14%)
    
80. 02:26:59:WU01:FS00:0x22:Completed 150000 out of 1000000 steps (15%)
    
81. 02:29:23:WU01:FS00:0x22:Completed 160000 out of 1000000 steps (16%)
    
82. 02:31:29:WU01:FS00:0x22:Completed 170000 out of 1000000 steps (17%)
    
83. 02:33:35:WU01:FS00:0x22:Completed 180000 out of 1000000 steps (18%)
    
84. 02:35:41:WU01:FS00:0x22:Completed 190000 out of 1000000 steps (19%)
    
85. 02:37:47:WU01:FS00:0x22:Completed 200000 out of 1000000 steps (20%)
    
86. 02:40:11:WU01:FS00:0x22:Completed 210000 out of 1000000 steps (21%)
    
87. 02:42:17:WU01:FS00:0x22:Completed 220000 out of 1000000 steps (22%)
    
88. 02:44:23:WU01:FS00:0x22:Completed 230000 out of 1000000 steps (23%)
    
89. 02:46:30:WU01:FS00:0x22:Completed 240000 out of 1000000 steps (24%)
    
90. 02:48:36:WU01:FS00:0x22:Completed 250000 out of 1000000 steps (25%)
    
91. 02:51:01:WU01:FS00:0x22:Completed 260000 out of 1000000 steps (26%)
    

复制代码

Keyco

那看来这个beta包在计算到某个环节上消耗不少CPU的资源，带4卡的话看来要把这个X79的U升级到2697 V2才行了。现在这个3930K捉襟见肘。
而且并非是4卡5线程就够，我看到有时候光单个FahCore_22.exe在高峰的时候要用掉我现在6核12线程CPU大概75%的资源，后面会降到37%左右。

看来以后单U平台要上X99来搭配16核的服务器U才行了。

或者dual cpu C602配4卡。

Keyco

还发现这个beta包还很吃内存，8G内存用掉100%。。。
看来以后标配是要16核CPU外加16G内存才能带动4卡跑beta了。

金鹏

Lynt 发表于 2020-3-26 10:57
看了一下log，正常TPF大约是2分6秒，从第一个百分点开始，每5个出现一次TPF增加18秒左右，没有出现几分钟 ...

1. 01:53:34:WU01:FS00:0x22:Version 0.0.2
   
2. 01:53:40:WU00:FS00:Upload 40.69%
   
3. 01:53:46:WU00:FS00:Upload 66.56%
   
4. 01:53:52:WU00:FS00:Upload 91.39%
   
5. 01:53:55:WU00:FS00:Upload complete
   
6. 01:53:55:WU00:FS00:Server responded WORK_ACK (400)
   
7. 01:53:55:WU00:FS00:Final credit estimate, 180860.00 points
   
8. 01:53:55:WU00:FS00:Cleaning up
   
9. 01:54:35:WU01:FS00:0x22:Completed 0 out of 1000000 steps (0%)

复制代码

用了1分钟，还好，后续每5%一次CPU矫正计算也是吃满所有线程

金鹏

Keyco 发表于 2020-3-26 11:56
还发现这个beta包还很吃内存，8G内存用掉100%。。。
看来以后标配是要16核CPU外加16G内存才能带动4卡跑beta ...

兄弟的3930K基本上可以带4张2080TI跑13873包，每5%矫正一次，会拖一些PPD后退。
但是内存貌似每个内核1.5G上下，你四张卡加系统占用，最好12+G内存

Keyco

金鹏 发表于 2020-3-26 12:36
兄弟的3930K基本上可以带4张2080TI跑13873包，每5%矫正一次，会拖一些PPD后退。
但是内存貌似每个内核1.5 ...

意思是CPU的话，不管多少核，都会吃满对么？
区别就在于吃满的情况下，越是核多，越快的完成矫正，PPD越不会被拖累。是这个路数么？

那么CPU的话，建议是高频还是多核呢？还是频率*核数要尽可能堆高？

牵牛星

1660Ti+8700，开头CPU需要满载30秒，每5%的时候满载大约12秒（同时挂着三个线程ARP），TPF2min45s

金鹏

Keyco 发表于 2020-3-26 12:59
意思是CPU的话，不管多少核，都会吃满对么？
区别就在于吃满的情况下，越是核多，越快的完成矫正，PPD越 ...

是的！
在我的双路E5-2687W上，一个core22内核吃满32线程，二个内核也是吃满32线程，单个时基本上顺开GPU计算


PS：感觉是多核作用更大

金鹏

牵牛星 发表于 2020-3-26 13:35
1660Ti+8700，开头CPU需要满载30秒，每5%的时候满载大约12秒（同时挂着三个线程ARP），TPF2min45s ...

看最开始的时间，你8700能30秒解决进入GPU？
下面是E3-1280V2大概1：41“后进入GPU计算

1. 06:11:32:WU00:FS01:0x22:Version 0.0.2
   
2. 06:11:32:WU00:FS01:0x22:  Found a checkpoint file
   
3. 06:11:38:Started thread 11 on PID 1896
   
4. 06:11:40:Started thread 12 on PID 1896
   
5. 06:13:09:WU00:FS01:0x22:Completed 900000 out of 1000000 steps (90%)
   
6. 06:13:09:WU00:FS01:0x22:Temperature control disabled. Requirements: single Nvidia GPU, tmax must be < 110 and twait >= 900

复制代码

Keyco

金鹏 发表于 2020-3-26 14:17
看最开始的时间，你8700能30秒解决进入GPU？
下面是E3-1280V2大概1：41“后进入GPU计算

3930K 拖4卡需要3分钟才开始计算。。。先换E5 V2的试试看。

不行只能整个平台换成X99的了。

Keyco

金鹏 发表于 2020-3-26 12:36
兄弟的3930K基本上可以带4张2080TI跑13873包，每5%矫正一次，会拖一些PPD后退。
但是内存貌似每个内核1.5 ...

前面把8G的闲置内存加上去了。确实，fahcore22在内存足够的情况下，从占用了1.3G开始慢慢变成了1.66G，加上teamviewer，msi afterbuner 以及 elsa  system graph，这些已经达到了8G+。看来是要准备至少10G以上才够，所以一步到位16G。

牵牛星

金鹏 发表于 2020-3-26 14:17
看最开始的时间，你8700能30秒解决进入GPU？
下面是E3-1280V2大概1：41“后进入GPU计算

日志倒没注意看我指的纯满载时间而已