But again, gcc is optimised for x86 as well and PPC code will be slower in comparison. Of course you can use an open source benchmark that you compile yourself with gcc on all your test machines. The same is true for the benchmarking programs you use. Why? Because if you choose, let’s say, Linux, you will use an os that is optimised to run on x86 processors (simply because most of the developers own them: how and why do you optimize for a G5 if you will never own one?), so it will likely be faster. You are right that these benchmarks are not really unbiased, but the method you mentioned is biased as well. Really, is it too much to ask for? If these places have the resources to get their hands on systems like these, can’t they at least be bothered to set up a controlled benchmark environment? Applications should be compiled specifically for that platform. The benchmarks used should definitely not be tailored to a specific platform (as Photoshop is, for example). All should be using the same memory (or at least the same amount and speed, PC3200 is what should be used for all the above sysems), preferably the same video cards if applicable to the tests, and be running the same operating system, whether it’s FreeBSD, Linux, NetBSD, etc. How well can you compare benchmarks when the test machines aren’t even running the same operating system? Here’s what I’d like: and unbiased benchmark and comparison of a dual 2.5GHz G5 system, a dual 3.6GHz 64 bit Xeon system, and a dual 2.4GHz Opteron system.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |