找回密码
 新注册用户
搜索
查看: 6453|回复: 12

[新闻] 超级毒包 project 9838 to ADVANCED/9839 for beta

[复制链接]
发表于 2017-1-27 10:19:07 | 显示全部楼层 |阅读模式
本帖最后由 金鹏 于 2017-2-7 09:37 编辑

史上最毒的包?直接PPD减半还多


捕获1.PNG


捕获2.PNG


moving p9838 to ADVANCED
by sinitskiy » Thu Feb 02, 2017 4:24 am
[size=1.3em]Dear all,

I'm moving project 9838 to advanced. The science is the same as earlier in projects 9810-9817 and 9820-9827, but now it's an insulin dimer, while those were monomers:

===

Go big by going… small! We want to study really large proteins and protein complexes, maybe as large as ribosomes or ion channels in neurons. One possible way to reach this goal is to run computations for each part of a big complex separately, and then assemble the resulting models together. We think we know how to do so, but first we need to test our approach on simpler systems.

Insulin is a small protein used by the pancreas to signal to the whole organism whether to consume more glucose or not. Disruptions in the processes of secreting and sensing insulin lead to various diseases, including diabetes mellitus, metabolic syndrome and polycystic ovary syndrome. Insulin can form dimers (complexes consisting of two insulin molecules), hexamers (complexes of six insulin molecules) and even larger aggregates resembling those formed under Alzheimer’s, but only the monomer is physiologically active. In this project, we want to run simulations of insulin monomers and dimers and see whether our methodology for studying big systems works here.

===

base credit 3068
28K atoms
Deadline 5 days, timeout 4 days, k-factor 0.75.
NVidia/ATI GPUs, only Linux

Looking forward to your feedback!
new GPU project 9839 for beta
by sinitskiy » Thu Feb 02, 2017 4:22 am
[size=1.3em]Dear beta-testers,

Here is a new project p9839. It is similar to 9838, but with longer trajectories.
Please let me know if you think I should increase the credit for it.

base credit 3068
28K atoms
Deadline 5 days, timeout 4 days, k-factor 0.75.
NVidia/ATI GPUs, only Linux

The science is the same as earlier in projects 9810-9817 and 9820-9827, but now it's an insulin dimer, while those were monomers:
===

Go big by going… small! We want to study really large proteins and protein complexes, maybe as large as ribosomes or ion channels in neurons. One possible way to reach this goal is to run computations for each part of a big complex separately, and then assemble the resulting models together. We think we know how to do so, but first we need to test our approach on simpler systems.

Insulin is a small protein used by the pancreas to signal to the whole organism whether to consume more glucose or not. Disruptions in the processes of secreting and sensing insulin lead to various diseases, including diabetes mellitus, metabolic syndrome and polycystic ovary syndrome. Insulin can form dimers (complexes consisting of two insulin molecules), hexamers (complexes of six insulin molecules) and even larger aggregates resembling those formed under Alzheimer’s, but only the monomer is physiologically active. In this project, we want to run simulations of insulin monomers and dimers and see whether our methodology for studying big systems works here.

===

Looking forward to your feedback!
Thanks a lot
new GPU project 9838 for beta
by sinitskiy » Fri Jan 27, 2017 3:50 am
[size=1.3em]Dear beta-testers,

Here is a new project for you, p9838. The science is the same as earlier in projects 9810-9817 and 9820-9827, but now it's an insulin dimer, while those were monomers:

===

Go big by going… small! We want to study really large proteins and protein complexes, maybe as large as ribosomes or ion channels in neurons. One possible way to reach this goal is to run computations for each part of a big complex separately, and then assemble the resulting models together. We think we know how to do so, but first we need to test our approach on simpler systems.

Insulin is a small protein used by the pancreas to signal to the whole organism whether to consume more glucose or not. Disruptions in the processes of secreting and sensing insulin lead to various diseases, including diabetes mellitus, metabolic syndrome and polycystic ovary syndrome. Insulin can form dimers (complexes consisting of two insulin molecules), hexamers (complexes of six insulin molecules) and even larger aggregates resembling those formed under Alzheimer’s, but only the monomer is physiologically active. In this project, we want to run simulations of insulin monomers and dimers and see whether our methodology for studying big systems works here.

===

base credit 3068
28K atoms
Deadline 5 days, timeout 4 days, k-factor 0.75.
NVidia/ATI GPUs, Linux/Win

PS: I'm aware that it doesn't show on psummary, and I'm working on it.

Looking forward to your feedback!

回复

使用道具 举报

发表于 2017-1-29 12:09:28 | 显示全部楼层
如何中止正在计算的包?
回复

使用道具 举报

发表于 2017-1-29 13:07:37 | 显示全部楼层
还有就是怎么样才能提前下载WU?每次都是算完一个再下载,浪费时间了。PPD上不去。
回复

使用道具 举报

 楼主| 发表于 2017-1-29 14:52:53 | 显示全部楼层
barenboim 发表于 2017-1-29 13:07
还有就是怎么样才能提前下载WU?每次都是算完一个再下载,浪费时间了。PPD上不去。 ...

客户端界面>configure>expert>左侧add>分别输入 next-unit-percentage  和   98% > ok > 右下侧save


捕获3.PNG
回复

使用道具 举报

发表于 2017-2-4 17:56:16 | 显示全部楼层
金鹏 发表于 2017-1-29 14:52
客户端界面>configure>expert>左侧add>分别输入 next-unit-percentage  和   98% > ok > 右下侧save

谢谢金版!还有个问题就是下载WU的时候,经常断流。我必须得重启一次客户端重新下载一个WU。这个问题怎么破?

回复

使用道具 举报

发表于 2017-2-4 18:24:55 | 显示全部楼层
barenboim 发表于 2017-2-4 17:56
谢谢金版!还有个问题就是下载WU的时候,经常断流。我必须得重启一次客户端重新下载一个WU。这个问题怎么 ...

本地或者路由挂个SS,百万PPD每天消耗流量200M上下。

评分

参与人数 1基本分 +20 收起 理由
金鹏 + 20 很给力!

查看全部评分

回复

使用道具 举报

 楼主| 发表于 2017-2-4 20:25:14 | 显示全部楼层
barenboim 发表于 2017-2-4 17:56
谢谢金版!还有个问题就是下载WU的时候,经常断流。我必须得重启一次客户端重新下载一个WU。这个问题怎么 ...

除了楼上的挂SS办法,最简单的可以用@Lynt 兄弟的FAH守护精灵    FAH_Daemon_Beta_20150826.rar
回复

使用道具 举报

发表于 2017-2-6 07:32:02 | 显示全部楼层
捕获.PNG

我也遇到低分包
回复

使用道具 举报

 楼主| 发表于 2017-2-6 10:27:19 | 显示全部楼层

这个还是没有9838/9839毒
回复

使用道具 举报

发表于 2017-2-6 17:05:04 | 显示全部楼层
Linux下970表现还可以,查log计算出以下结果:

双卡机上只跑过9838,1470MHz PCIE4X的970能跑到34万PPD,1480MHz的980Ti也有接近67万PPD

单卡970@1470MHz@PCIE16X跑9838和9839都能到38万,又是吃频率的包
回复

使用道具 举报

 楼主| 发表于 2017-2-6 18:54:56 | 显示全部楼层
Lynt 发表于 2017-2-6 17:05
Linux下970表现还可以,查log计算出以下结果:

双卡机上只跑过9838,1470MHz PCIE4X的970能跑到34万PPD,1 ...

不完全是吃频率,否则1080的2062和980的1506都很高频率了,但PPD可怜,

赶脚这货完全没有优化,
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 新注册用户

本版积分规则

论坛官方淘宝店开业啦~
欢迎大家多多支持基金会~

Archiver|手机版|小黑屋|中国分布式计算总站 ( 沪ICP备05042587号 )

GMT+8, 2024-4-20 10:27

Powered by Discuz! X3.5

© 2001-2024 Discuz! Team.

快速回复 返回顶部 返回列表