拥有12T+ (OR CORE)牛神兄弟们可以尝试跑6903超级大包了

金鹏

貌似2.5倍于目前原子数最多的2692包的6903超级包等着大牛和大神级的兄弟们去尝鲜折磨你们的神U了

   ./fah6 -bigbeta -smp 12   或者 ./fah6 -bigbeta -smp 16    ？

http://foldingforum.org/viewtopic.php?f=66&t=18756

We are preparing to release a new A5 bigbeta/bigadv project. This one will be a little larger than the current generation of A5 projects, and it will enable us to tackle a new set of important scientific questions.
Because the work units are a bit larger, we are setting higher memory and processor limits:
2G+ RAM, 12 cores+
We may adjust both of these as necessary.
We'll also start these work units out as 64-bit Linux only.

Initial points values and deadlines are similarly subject to change but currently stand at:
Base points: 30964
Deadline: 17 days
Preferred deadline: 10.2 days
k-factor: 26.86

I'm pretty sure that a Linux system with appropriate configuration and the -bigbeta flag will be enough to get them ...

Project: 6903
Average time/frame: 00:42:17
CPU: Opteron G34 6134 @ 2.3 GHz
# of CPU sockets: 2
# of Physical cores: 16

# of FAH CPU processes: 1
# of FAH GPU Clients: 0

RAM GB installed: 32
RAM Type: DDR3
RAM Speed: 1333 (Not sure about timing)
OS / Linux kernel: Arch Linux x64
Running in VM: No

According to HFM = 131.5K PPD

Project: 6903
Average time/frame: 0:41:57 {in hh:mm:ss}
CPU: i7 970 @ 4.2 GHz
# of CPU sockets: 1
# of Physical cores: 1

# of FAH CPU processes: 12
# of FAH GPU Clients: 0

RAM GB installed: 6
RAM Type: {DDR3 /2000}
RAM Speed: 2009Mhz 9-9-9-24-110-2T
OS / Linux kernel:Ubuntu 10.10 / 2.6.35-28 Generic-ck
Running in VM: {No}

0:33:36 tpf is avg of 2 frames

Project: 6903
Average time/frame: 0:33:36 {in hh:mm:ss}
CPU: X5660 @ 2.8 GHz (turbo is on but cat/proc/cpuinfo does not show it)
# of CPU sockets: 2
# of Physical cores: 12 (24 threads)

# of FAH CPU processes:
# of FAH GPU Clients:0

RAM GB installed: 8 4x 2gb (so dual chan)
RAM Type: {DDR/DDR2/DDR3} DDR3 (should be known given cpu but w/e)
RAM Speed: 1333
OS / Linux kernel: Linux 2.6.18-194.el5

Running in VM: {Yes/No} No

Project: 6903
Average time/frame: 0:12:15 {in hh:mm:ss}
CPU: Opteron 6180SE
# of CPU sockets: 4
# of Physical cores: 48

RAM GB installed: 32
RAM Type: DDR3-1333
RAM Speed: 1386Mhz 6-7-5-20 1T

Above is running stable at 2.6ghz.
2.7ghz was unstable -- likely memory (2 modules were running at 5-6-5 1T @ 1066), OR much higher temperatures on 6903 vs. 6901. (see separate post on this.

simc

我也想试试

白银的马甲

平台压力甚大

nekoko

这儿有12核鸡么……

cym104

又要改客户端参数？我要是保持原样（-bigadv -smp）能收到么？

maxzong

刚下了个WU6903, 是Linux下的，貌似PPD 可以到10万，比原来的bigadv PPD涨1倍啊，很猛！！

log 文件如下：


--- Opening Log file [June 3 21:50:01 UTC]


# Linux SMP Console Edition ###################################################
###############################################################################

                       Folding@Home Client Version 6.34

                          http://folding.stanford.edu

###############################################################################
###############################################################################

Launch directory: /home/max/Desktop/Folding at Home
Executable: ./fah6
Arguments: -smp 12 -bigbeta -verbosity 9 -local

[21:50:01] - Ask before connecting: No
[21:50:01] - User name: maxzong (Team 3213)
[21:50:01] - User ID: ***************
[21:50:01] - Machine ID: 1
[21:50:01]
[21:50:01] Loaded queue successfully.
[21:50:01] - Preparing to get new work unit...
[21:50:01] - Autosending finished units... [June 3 21:50:01 UTC]
[21:50:01] Cleaning up work directory
[21:50:01] Trying to send all finished work units
[21:50:01] + No unsent completed units remaining.
[21:50:01] - Autosend completed
[21:50:01] + Attempting to get work packet
[21:50:01] Passkey found
[21:50:01] - Will indicate memory of 3009 MB
[21:50:01] - Connecting to assignment server
[21:50:01] Connecting to http://assign.stanford.edu:8080/
[21:50:03] Posted data.
[21:50:03] Initial: ED82; - Successful: assigned to (130.237.232.237).
[21:50:03] + News From Folding@Home: Welcome to Folding@Home
[21:50:03] Loaded queue successfully.
[21:50:03] Sent data
[21:50:03] Connecting to http://130.237.232.237:8080/
[21:50:16] Posted data.
[21:50:16] Initial: 0000; - Receiving payload (expected size: 46519909)
[21:59:15] - Downloaded at ~84 kB/s
[21:59:15] - Averaged speed for that direction ~84 kB/s
[21:59:15] + Received work.
[21:59:15] + Closed connections
[21:59:15]
[21:59:15] + Processing work unit
[21:59:15] Core required: FahCore_a5.exe
[21:59:15] Core found.
[21:59:15] Working on queue slot 01 [June 3 21:59:15 UTC]
[21:59:15] + Working ...
[21:59:15] - Calling './FahCore_a5.exe -dir work/ -nice 19 -suffix 01 -np 12 -checkpoint 15 -verbose -lifeline 1629 -version 634'

[21:59:15]
[21:59:15] *------------------------------*
[21:59:15] Folding@Home Gromacs SMP Core
[21:59:15] Version 2.27 (Thu Feb 10 09:46:40 PST 2011)
[21:59:15]
[21:59:15] Preparing to commence simulation
[21:59:15] - Looking at optimizations...
[21:59:15] - Created dyn
[21:59:15] - Files status OK
[21:59:19] - Expanded 46519397 -> 71846524 (decompressed 62.1 percent)
[21:59:19] Called DecompressByteArray: compressed_data_size=46519397 data_size=71846524, decompressed_data_size=71846524 diff=0
[21:59:19] - Digital signature verified
[21:59:19]
[21:59:19] Project: 6903 (Run 7, Clone 0, Gen 0)
[21:59:19]
[21:59:19] Assembly optimizations on if available.
[21:59:19] Entering M.D.
[21:59:27] Mapping NT from 12 to 12
[21:59:31] Completed 0 out of 250000 steps  (0%)
[22:49:14] Completed 2500 out of 250000 steps  (1%)

金鹏

回复 6# maxzong

我倒，效率提升高不少啊

将近13万的PPD

金鹏

这儿有12核鸡么……
nekoko 发表于 2011-6-4 01:28

应该是问120核机

金鹏

回复 5# cym104

肯定要加 -BIGBETA参数

大神的机器跑6903的PPD又要翻倍了

heren3

这年头我们这些跑显卡的...

jpegreader

用提高PPD的方法鼓励牛机加入……
差别待遇啊……

金鹏

回复 11# nekoko

目前大蛋白质折叠包是GPU无法完成的，而传统CPU计算量远远跟不上，阻碍了FAH研究进程，所以针对目前支持AVX和6C12T及以上CPU各种计算包的出现也是计算发展的趋势，

还是那句话，量力而行重在参与。

比如去拜佛，有些人仅烧烧香，有些人捐个几百，有些人捐上几百甚至上千万，一样道理。

nekoko

金版有意思啊，会用屏蔽了

说的必须是事实，我的本职工作就是研究这个的，所以我当然清楚
FAH做的确实有点过分，金版不要学他哟