Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
I'll take a stab at it, but I'm only hypothesizing (aka no proof). I would think the quad duals would be faster because of it being easier to dissipate the heat (should be less per chip). However, if you get a really industrial strength liquid cooling system for the two quad's then they'll probably be faster because of the faster clock speed of the cores.
Geoff
Sent from my BlackBerry wireless handheld.
-----Original Message----- From: "Erick Perez" eaperezh@gmail.com
Date: Fri, 31 Aug 2007 21:51:17 To:"CentOS mailing list" centos@centos.org Subject: [CentOS] OT: 4 dual cores agains 2 quad cores
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
-- ------------------------------------------------------------ Erick Perez Panama Sistemas Integradores de Telefonia IP y Soluciones Para Centros de Datos Panama, Republica de Panama Cel Panama. +(507) 6694-4780 ------------------------------------------------------------ _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Hi,
performance measurement is always strongly dependent on the kind of computation you are performing. So to really find out, you need to do benchmarks with the programs you want to run.
If the price is not important, I would buy four quad cores with each 3GHz per server.
Regards, Sebastian
Erick Perez wrote:
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
Hello,
to my mind the most impacting on a SMP server is the synchronisation time between dies/cores. the time to synchronize 2 cores on different dies is higher than the time to synchronize 2 cores on the same die. So to achieve best performances you have to limit number of dies. So a dual quad core should be more powerful than a 4 dual core.
You can verify this here : http://www.spec.org/cpu2006/results/cpu2006.html
results on a 2 chips server are ~20/30% higher than a 4 chips using the same processor
regards,
David Amiel
Le Sam 1 septembre 2007 04:51, Erick Perez a écrit :
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
--
Erick Perez Panama Sistemas Integradores de Telefonia IP y Soluciones Para Centros de Datos Panama, Republica de Panama Cel Panama. +(507) 6694-4780
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Fundamentally it comes down to I/O performance between the CPU cores and memory bus.
Quad core CPUs suffer from performance bottlenecks with regard to Dual cores which leads to a less of an improvement than one might expect from clock speed alone
There are articles here worth a read:
http://www.tomshardware.com/2006/09/10/four_cores_on_the_rampage/
As has already been mentioned, if you have four sockets to fill you might as well fill it with Quad cores anyway....
You might find this useful too:
http://www.tomshardware.com/2007/08/08/extreme_fsb_2/
Regards
Pete
David Amiel wrote:
Hello,
to my mind the most impacting on a SMP server is the synchronisation time between dies/cores. the time to synchronize 2 cores on different dies is higher than the time to synchronize 2 cores on the same die. So to achieve best performances you have to limit number of dies. So a dual quad core should be more powerful than a 4 dual core.
You can verify this here : http://www.spec.org/cpu2006/results/cpu2006.html
results on a 2 chips server are ~20/30% higher than a 4 chips using the same processor
regards,
David Amiel
Le Sam 1 septembre 2007 04:51, Erick Perez a écrit :
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
--
Erick Perez Panama Sistemas Integradores de Telefonia IP y Soluciones Para Centros de Datos Panama, Republica de Panama Cel Panama. +(507) 6694-4780
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
On Friday 31 August 2007, Erick Perez wrote:
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
If you do pure IO workloads, the 4 dual cores are probably going to be as fast as the 2 quads because of the clock differences.
For CPU bound workloads, the quad cores will beat the duals easily because of the higher clock speed (and more efficient caching in case of AMD).
The only other things I would worry about is the number of memory slots. Usually boards that have 4 cpu sockets have a larger number of memory slots too. So if you need lots of ram, you're better off on that.
Peter.
On 9/1/07, Peter Arremann loony@loonybin.org wrote:
On Friday 31 August 2007, Erick Perez wrote:
Hi people, Do you have pointers to web documents that help me make comparisons between buying a server with two quad core 2.33 ghz or buying a 4 dual core 2ghz server? I am trying to answer a question of performance. It is not important the redundancy/failover or the price of the server. Just the performance. obviously all the hardware specs are the same, the question is the CPU.
If you do pure IO workloads, the 4 dual cores are probably going to be as fast as the 2 quads because of the clock differences.
For CPU bound workloads, the quad cores will beat the duals easily because of the higher clock speed (and more efficient caching in case of AMD).
The only other things I would worry about is the number of memory slots. Usually boards that have 4 cpu sockets have a larger number of memory slots too. So if you need lots of ram, you're better off on that.
Peter. _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
Thanks Peter and thanks to all for the information. It turns out that the several HP Proliant DL380G5 the company is about to buy, will run SQL Server 2000, RHEL 5 w/Tomcat and Exchange 2003.
So it seems that after reading several documents linked here and on the net, Two Xeon Quad Core at a little lower speed will be more efficient that 4 Dual Core Xeons at a little higher speed. And not to mention the benefit of using only two sockets instead of four.
So, I guess i'll go for quad cores. If anyone is interested in benchmarks, please let me know offline at : eaperezh ((at)) gmail ((dot)) com
Thanks,
-
David Amiel -
Erick Perez -
gjgowey＠tmo.blackberry.net -
Peter Arremann -
Peter Farrow -
Sebastian Walter