i7 vs core2 quad
The title is not really correct it's not a comparison between two processors it's actually just a simple compilation test on two different boxes: one that was almost top of the line 2 years ago, and one that is the same today.
The test is simple: time the compilation of Zorba XQuery project on those boxes.
I used Zorba's svn revision 8144. Zorba is an XQuery processor that I've been working on lately.
The two boxes:
1. Core2 Quad Q9450 2.66GHz, 2GB RAM DDR2 1000Mhz, files on HDD
2. i7-980x 3.33GHz(6 core, 32nm), 6GB RAM DDR3 1333MHz, files on SSD
both running Ubuntu 9.10 Karmic Koala.
On box 1 Core2 Quad:
- single thread make:
user 9m 3.954s
sys 0m37.754s
- 4 threads make:
user 8m48.829s
sys 0m37.758s
On box 2 i7 (6 core):
- single thread make:
user 6m 4.360s
sys 0m31.060s
- 6 threads make:
user 6m32.280s
sys 0m32.200s
- 12 threads make:
user 9m36.400s
sys 0m44.480s
While the 4 core improves from 9m44s to 2m42 i.e. 3.6 times - expected 0.4 overhead, the 6 core improves from 6m35s to 1m17s i.e. 5.12 times for 6 threads and to 1m6s i.e. 5.98 times for 12 threads. Even if I ran it with 12 threads in reality there are only 6 real cores, just looking like 12 because of hyper-threading, the i7's overhead is very impressive just under 3.33% compared with 10% for Core2 Quad.
But what is more impressive, overall in less than 2 years and a half, the same problem was solved around 2.45 times faster.
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home