Using a 4 way dual core server for SETI-BOINC

留言板 : Number crunching : Using a 4 way dual core server for SETI-BOINC
留言板合理

To post messages, you must log in.

前 · 1 · 2

作者消息
Profile Landroval

发送消息
已加入:7 Oct 01
贴子:188
积分:2,098,881
近期平均积分:1
United States
消息 178367 - 发表于:14 Oct 2005, 22:19:13 UTC - 回复消息 178360.  



I assume I can find copies of each to D/L where I found the original BOINC programme?


No, but a full description of the process and links to optimized clients (and utilities to help ensure you're using the right optimized client) can be found here in the Wiki.

Happy crunching....

If you think education is expensive, try ignorance.
ID: 178367 · 举报违规帖子
Ned Slider

发送消息
已加入:12 Oct 01
贴子:668
积分:4,375,315
近期平均积分:0
United Kingdom
消息 178365 - 发表于:14 Oct 2005, 22:16:57 UTC - 回复消息 178362.  
最近的修改日期:14 Oct 2005, 22:17:32 UTC


Hi

I agree with you, despite all my machines being Intel based. The AMD CPU is a far better number crunching design.

In time I will need to explore the price of a dual core dual CPU AMD Optron system against an equivalent (if they exost) dual core dual CPU Xeon system.

John


...and the next logical step would be to ditch the overly expensive Windows Server platform you would require for a 4-way system in favour of a true enterprise OS such as Linux, which also happens to be free, and spend the saved cash on more hardware.

It's still beyond me why Microsoft insists you have to run a Server OS on a 4-way workstation!

Ned

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***
ID: 178365 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178362 - 发表于:14 Oct 2005, 22:06:11 UTC - 回复消息 178359.  

[quote][quote]
I agree with you for 50%. If you go with Intel 4 way, they indeed share the FSB and the available memory bandwidth. The result will be less impressive.

But if you would go for a WINAMD combination, that is quad-opteron dual core, each CPU will have it's own dual channel memory controller. Now that'll be impressive!

Now don't start about stability or heat production. Stability will be at least the same, and the Opteron will run cooler.


Hi

I agree with you, despite all my machines being Intel based. The AMD CPU is a far better number crunching design.

In time I will need to explore the price of a dual core dual CPU AMD Optron system against an equivalent (if they exost) dual core dual CPU Xeon system.

John
It's good to be back amongst friends and colleagues



ID: 178362 · 举报违规帖子
Hans Dorn
志愿者开发人员
志愿者测试人员
Avatar

发送消息
已加入:3 Apr 99
贴子:2262
积分:26,448,570
近期平均积分:0
Germany
消息 178361 - 发表于:14 Oct 2005, 21:58:29 UTC - 回复消息 178359.  


I agree with you for 50%. If you go with Intel 4 way, they indeed share the FSB and the available memory bandwidth. The result will be less impressive.

But if you would go for a WINAMD combination, that is quad-opteron dual core, each CPU will have it's own dual channel memory controller. Now that'll be impressive!

Now don't start about stability or heat production. Stability will be at least the same, and the Opteron will run cooler.


True. But have you seen the price tag on the 8-Way dual core opteron :o)

I was fancying a n-way opteron cruncher for a while, but there's no way I could
afford a setup like this.

Regards Hans


ID: 178361 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178360 - 发表于:14 Oct 2005, 21:55:31 UTC - 回复消息 178353.  

[quote][quote]
There are optimized versions of both the science application (which processes the work units) and the BOINC manager (which does scheduling, benchmarking, and calculates claimed credit).

Hi Brian

I assume I can find copies of each to D/L where I found the original BOINC programme?

John
It's good to be back amongst friends and colleagues



ID: 178360 · 举报违规帖子
[BOINCstats] Willy
志愿者测试人员

发送消息
已加入:4 Mar 01
贴子:201
积分:152,243
近期平均积分:0
Netherlands
消息 178359 - 发表于:14 Oct 2005, 21:54:05 UTC - 回复消息 178348.  


Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John


Hi John!

I'd advice against a 4-way system. The 4 CPUs will share a single front side bus.
Since seti is memory hungry, the CPUs would be competing for memory bandwith.
You'd get more WUs per invested pound if you set up several dual core crunchers
instead.
(Provided, you have the space, that is.)


Regards Hans



I agree with you for 50%. If you go with Intel 4 way, they indeed share the FSB and the available memory bandwidth. The result will be less impressive.

But if you would go for a WINAMD combination, that is quad-opteron dual core, each CPU will have it's own dual channel memory controller. Now that'll be impressive!

Now don't start about stability or heat production. Stability will be at least the same, and the Opteron will run cooler.

Join team BOINCstats
ID: 178359 · 举报违规帖子
Profile Sir Ulli
志愿者测试人员
Avatar

发送消息
已加入:21 Oct 99
贴子:2246
积分:6,136,250
近期平均积分:0
Germany
消息 178358 - 发表于:14 Oct 2005, 21:53:23 UTC - 回复消息 178345.  

50% of the workstations on my small farm are dual processor, based on the WinTel alliance. The last I had built was a dual 2.8 Xeon with HT active. Not bad, but should have a higher RAC I believe. Currently 405, and I think the RAC should be around 450 - 480.

For my next machine, likely to be WinTel again, I am thinking of going for dual core 64 bit Xeon MPs. These would have at least 1Mb L2 cache. The system would probably be 4 way, with 2 CPUs in place and adding the other 2 when I can afford them. As I understand, the dual core Xeons have HT active on each core, so a 4 way (physical CPU) dual core system would have 8 cores, each with HT, giving an apparant 16 way system.

The lowest spec I would entertain is a 2.0 GHz MP system, with Dell quoting me about 5,200 UK pounds for a 4 Gb of RAM server.

Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John


I'd be happy to answer these questions for you. All that's required is you ship me the server and let me test it for a period of time. I will then report the statistics back to you. :)



i am also glad tho help you


I'd be happy to answer these questions for you. All that's required is you ship me the server and let me test it for a period of time.

i think i can handle this also...

Greetings from Germany NRW
Ulli

ID: 178358 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178356 - 发表于:14 Oct 2005, 21:49:15 UTC - 回复消息 178348.  

[quote][quote]
Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

Hi John!

I'd advice against a 4-way system. The 4 CPUs will share a single front side bus.
Since seti is memory hungry, the CPUs would be competing for memory bandwith.
You'd get more WUs per invested pound if you set up several dual core crunchers
instead.

Hi Hans

Space is limited to one of my bedrooms converted to an office and stuffed with PCs. I hink I can put one of my old P3 systems on a platform by a desk, and use the space for the machine I seek to introduce sometime in the next 12 months.

Yes, both my dual P3 and dual Xeon exhibit the shared FSB and RAM slowdown problems. The Xeon particularly demonstrates this ...

When using 1 processor only with 1 WU is is cleared in about 2.5 hours.
Using 2 CPUs with a WU on each, each result takes about 3.5 hours.
Using both CPUs and HT on each, all WUs average 5.5 hours.

However, the average output per day is best with all physical and virtual processors occupied.

I think I will probably use the same money as a 4 way Xeon MP system would cost to buy 2 dual core dual Xeon systems. This way I can buy the higher L3 cache and clock speeds.

Thanks

John

It's good to be back amongst friends and colleagues



ID: 178356 · 举报违规帖子
Profile Landroval

发送消息
已加入:7 Oct 01
贴子:188
积分:2,098,881
近期平均积分:1
United States
消息 178353 - 发表于:14 Oct 2005, 21:41:02 UTC - 回复消息 178350.  


Also, is the optimised client (assuming there is one) related to the BOINC Manager or THE SETI@hOME project I attached to?


There are optimized versions of both the science application (which processes the work units) and the BOINC manager (which does scheduling, benchmarking, and calculates claimed credit).

Regards,

Brian

If you think education is expensive, try ignorance.
ID: 178353 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178351 - 发表于:14 Oct 2005, 21:38:34 UTC - 回复消息 178348.  


Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John


Hi John!

I'd advice against a 4-way system. The 4 CPUs will share a single front side bus.
Since seti is memory hungry, the CPUs would be competing for memory bandwith.
You'd get more WUs per invested pound if you set up several dual core crunchers
instead.
(Provided, you have the space, that is.)


Regards Hans


It's good to be back amongst friends and colleagues



ID: 178351 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178350 - 发表于:14 Oct 2005, 21:37:18 UTC

Sorry to post to my own thread, but is there an optimised SETI-BOINC client for a dual Xeon on SETI-BOINC? I run them on a WinXP Pro server, with the HT active, and 1 Gb or RAM.

Also, is the optimised client (assuming there is one) related to the BOINC Manager or THE SETI@hOME project I attached to?

I am currently using the basic BOINC Manager V4.45.
It's good to be back amongst friends and colleagues



ID: 178350 · 举报违规帖子
Hans Dorn
志愿者开发人员
志愿者测试人员
Avatar

发送消息
已加入:3 Apr 99
贴子:2262
积分:26,448,570
近期平均积分:0
Germany
消息 178348 - 发表于:14 Oct 2005, 21:36:43 UTC - 回复消息 178339.  
最近的修改日期:14 Oct 2005, 21:37:39 UTC


Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John


Hi John!

I'd advice against a 4-way system. The 4 CPUs will share a single front side bus.
Since seti is memory hungry, the CPUs would be competing for memory bandwith.
You'd get more WUs per invested pound if you set up several dual core crunchers
instead.
(Provided, you have the space, that is.)


Regards Hans
ID: 178348 · 举报违规帖子
Profile Borgholio
Avatar

发送消息
已加入:2 Aug 99
贴子:653
积分:18,623,738
近期平均积分:45
United States
消息 178345 - 发表于:14 Oct 2005, 21:32:02 UTC - 回复消息 178339.  

50% of the workstations on my small farm are dual processor, based on the WinTel alliance. The last I had built was a dual 2.8 Xeon with HT active. Not bad, but should have a higher RAC I believe. Currently 405, and I think the RAC should be around 450 - 480.

For my next machine, likely to be WinTel again, I am thinking of going for dual core 64 bit Xeon MPs. These would have at least 1Mb L2 cache. The system would probably be 4 way, with 2 CPUs in place and adding the other 2 when I can afford them. As I understand, the dual core Xeons have HT active on each core, so a 4 way (physical CPU) dual core system would have 8 cores, each with HT, giving an apparant 16 way system.

The lowest spec I would entertain is a 2.0 GHz MP system, with Dell quoting me about 5,200 UK pounds for a 4 Gb of RAM server.

Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John


I'd be happy to answer these questions for you. All that's required is you ship me the server and let me test it for a period of time. I will then report the statistics back to you. :)


You will be assimilated...bunghole!

ID: 178345 · 举报违规帖子
Profile John Clark
志愿者测试人员
Avatar

发送消息
已加入:29 Sep 99
贴子:16515
积分:4,418,829
近期平均积分:0
United Kingdom
消息 178339 - 发表于:14 Oct 2005, 21:24:53 UTC

50% of the workstations on my small farm are dual processor, based on the WinTel alliance. The last I had built was a dual 2.8 Xeon with HT active. Not bad, but should have a higher RAC I believe. Currently 405, and I think the RAC should be around 450 - 480.

For my next machine, likely to be WinTel again, I am thinking of going for dual core 64 bit Xeon MPs. These would have at least 1Mb L2 cache. The system would probably be 4 way, with 2 CPUs in place and adding the other 2 when I can afford them. As I understand, the dual core Xeons have HT active on each core, so a 4 way (physical CPU) dual core system would have 8 cores, each with HT, giving an apparant 16 way system.

The lowest spec I would entertain is a 2.0 GHz MP system, with Dell quoting me about 5,200 UK pounds for a 4 Gb of RAM server.

Anyone got any experience/used a 4 way Xeon MP system? Especially the new dual core systems?

How long would the system take to do a WU using all cores and HT?

What would be the expected output per day, or the expected RAC when bedded down?

John
It's good to be back amongst friends and colleagues



ID: 178339 · 举报违规帖子
前 · 1 · 2

留言板 : Number crunching : Using a 4 way dual core server for SETI-BOINC


 
©2020 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.