P4 does not have an arithmetic unit == low FPU?

[SuperDune]

Is it really true that P4 doesn't have arithmetic unit?? anyone?

 

[imgod2u]

No......
The FPU and ALU are two types of execution units and operating on 2 different types of data. The Arithmetic Logic Unit works on integers while the Floating Point Unit works on FP data. The P4's low FP performance is because it has weak FPU's. It's ALU is actually quite impressive. One complex 32-bit ALU and 2 double-pumped, skewable simple ALU's. The next revision, Prescott, will have double-pumped 32-bit ALU's (don't know if it's skewable or not). I'm not sure whether the complex ALU will be double-pumped or not.

 

[jiffylube1024]

Also, the P4's high clockspeeds ( >2.0 Ghz) make up for its relatively "weak" FPU performance. Not to mention SSE2 (which is all floating point).

 

[imgod2u]

I believe SSE2 added the ability to do 4x 32-bit Integer data types per pass as well.

 

[Rainsford]

Originally posted by: jiffylube1024
Also, the P4's high clockspeeds ( >2.0 Ghz) make up for its relatively "weak" FPU performance. Not to mention SSE2 (which is all floating point).

Yes, the P4 approach is clockspeed scaling vs FPU power. This approach is working right now, but if/when AMD releases a chip that does both, Intel could have a little more competition. There is no reason that I can see why the two have to be mutually exclusive, and I'm not sure why Intel picked this approach with the P4.

As far as SSE2, it helps a lot. But it only helps if the app is written for SSE2, and I don't see a ton of those apps floating around though. But only time will tell if betting on SSE2 and higher-than-AMD clockspeeds was a good idea or not.