[AkiraFactor]

The Pentium 4 is fast, fast enough to keep up on most tasks. Maybe not fast enough to beat the G5 on many benchmarks and programs, but fast enough. As far as benchmarks check IBM's Power4 achitecture SPEC benchmarks to back up Apple's G5(IBM PowerPC 970) SPEC benchmark claims.

However undeniable facts I will list.

The P4 is single processor only and 1 Pentium 4 can at least keep up with the G5 at some tasks. The G5 however has another PPC 970 G5 processor sitting right next to it ready to smoke the P4 running at say 5 GHz even though that doesn't make main memory any faster.

If the P4 is roughly as fast as 1 G5 then you would need a P4 nearly twice as fast to equal a dual G5.

If you want to come close to the G5 you need Xeons and they come at a premium price. No cheapie Dell compares in performance and really is a +2GHz Celeron or P4 even in the same league versus a dual G5?

HYPErthreading is just that, hype. On our 3.06 machines we keep HT off because it only increases performance very rarely in only special circumstances and can't even tell the difference without benchmarking and then finding out in that special case its slightly faster. It hurts performance the rest of the time and hence it stays off. Also Hypertheading seems to help the P4 a little with SPEC, but hurts the Xeon.

G5 wins hands down on any test with in memory data more that 4GB against the P4. The P4 will spend all day swapping back and forth to disk in the Page File.

As far as price comparisons compile the price of the parts and then include which vendor to contact when you spend countless hours tracking down a glitch/bug in a homemade system. Then again who would put a mission critical task on a homemade system? Would you bet your career on a homemade system or just Quake 3?

[Thain Esh Kelch]

Remember, not all tasks a multi-processor usable!

And on top of that, 2x2 Ghz doesnt mean 4.0 Ghz...

[AkiraFactor]

It cuts both ways though. Tasks that do not work well with multiprocessing do not work well with hyperthreading.

[nvaughan3]

However undeniable facts I will list.

I must be missing where the "facts" are. All I see are theoretical comparisons and a bunch of nonsense. I'll wait for the G5 to be publically released before judging just who is faster in real world applications. Funny how the G5 wins "hands down" when perhaps only a dozen people have had their "hands on" a G5, and then not even for full testing.

hen again who would put a mission critical task on a homemade system? Would you bet your career on a homemade system or just Quake 3?

I would (and do) put trust in the stability and reliability of my (multiple) custom machines. I (and many others) know enough about to be able to diagnose most anything that comes up. If something does go wrong, you can be damm sure that I can get the problem identified and fixed before *any* consumer prebuilt's customer support would be able to.