Thursday, January 19, 2006

Opteron vs. Xeon

We put x2100 server with 1x Opteron 175 (2.2GHz) into production as an web server. Other servers in that service are running on IBM's x335 servers with 2x Intel Xeon 2.8GHz. Both server have 2GB of RAM. We were changing load balancer configuration to see how much given server can do in a production. The result is that x2100 is 30% faster here (30% more requests per second) than x335 server in a production serving real traffic (mostly dynamic web pages). Both servers are running Solaris 10.

Another test this time on IBM x336 with 2x Intel Xeon 2.8GHz, 8GB RAM and Sun X4100 2x Opteron 280 2.4GHz, 8GB RAM with MySQL 5. Just single query is issued (so one core is utilized). All data are cached so during query no IOs to disks are issued. The result is: 2.58s on IBM and 1.41s on Sun - this means that Xeon 2.8GHz is about 82% slower that single core of Opteron 280 in this test. Now Sun server has 2 times more cores than IBM. This is just simple test (however on real data) and it doesn't mean that such a difference will be on full production but it clearly shows potential in Opteron.

No comments: