The G4 is also known as the first supercomputer on a chip. But, what's a supercomputer? What makes a supercomputer "super" is its ability to execute at least one billion floating-point operations per second (1 gigaflop). In fact, the dual 1GHz PowerPC G4 processor hit speeds of 15 gigaflops!!! How can it be so fast?
1. They use a special technology called RISC (reduced instruction set computing). It means that it uses a relatively small set of instructions for faster operation. This type of architecture makes PowerPC chips smaller and cooler.
2. The G4 processors have Velocity Engine (also called AltiVec). The Velocity Engine is a new set of instructions used to speed up arithmetic operations commonly used in 3D graphics, image and sound manipulation among other things. The Velocity Engine processes data in huge 128-bit chunks, instead of the smaller 32-bit chunks used in Pentium 3, Pentium 4 and Athlon processors. This means that the PowerPC G4 can perform four (in some cases eight) 32-bit floating-point calculations in a single cycle. And since Athlon processors are 32-bit, you would need about two 2GHz K7 processors to accomplish the same tasks as the dual 1GHz G4.
3. The G4 processor floating-point operations use 64-bit registers, wich means that the G4 is 2x faster than a Pentium 3 or Athlon at the same "speed".
4. The G4 processors have less pipeline stages (4 for integer calculations and 7 for floating-point operations) than Pentium 4 (20 pipeline stages for both). This means that G4 starts to do it's job faster than Pentium 4 (G4 takes 4 or 7 cycles while Pentium 4 takes 20).
5. The G4 processors are faster than Pentiums and Athlon in multiprocessing (while a Dual G4 1000 would be equivalent to a G4 1.8GHz, a Dual Pentium 3 1000 is equivalent to a 1.5 GHz Pentium 4).
6. The G4 processors have a bigger L1 cache (G4 has 64KB - 32 for data and 32 for instructions - while P4 has 20KB - 8 for data and 12 for instructions).
2. How much faster than Pentium processors?
The G4 processor is 3 to 4 times faster than Pentium processors in graphics, sound manipulation, etc. Some tests point to 3 times faster (like Apple tests), others point to 3,5 times faster (like Intel tests!!!). In floating-point operations G4 is about 2.5 times faster than Pentium and Athlon processors. In integer calculations, since both G4 and Athlon/Pentium integer calculations are 32-bit, the G4 is +/- equivalent to a Pentium/Athlon processor in looooooong integer calculations, but since the G4 has less pipeline stages, the G4 is faster for small calculations. It's not possible to find an exact number, because they use different sets of instructions (Pentium use complex instruction set while G4 and other Mac processors use reduced instruction sets).
But this is not everything. Both G3 and G4 chips are smaller than Pentium or Athlon chips, and require less power than Pentium or Athlon chips (this is a big advantage for Apple laptops in terms of battery life and screen brightness).
3. Don't trust benchmarks
Pentium processors are not really CISC processors (Complex Instruction Set Computing) but CRISC (Complex Reduced Instruction Set Computing). This because Intel is "injecting" RISC instructions into their CISC processors. This makes Pentium processors bigger, hotter, harder to design and use more power than RISC. Pentium processors have more bugs since they are much more complex, but Intel is hiding the symptoms of not being RISC by improving their microcode to be more RISC like and by throwing 10 times more designers at the problem. They can get nearly the same performance, but they require more transistors, design time, heat, power, and cost to do it.
With the Pentiums (or Athlon) there are no RISC advantages for modernizing the instruction set. Programmers still have problems with a register starved architecture, and an ugly outdated instruction set that can't do many things programmer want (easily). They still have this butt-ugly stack based floating point unit, and so on. The compilers (and programmers) can't access the RISC Operations directly. The implementation of that RISC-like core is hidden from programmers (and changes implementation to implementation). Because of all this, it is harder to properly schedule things (with psuedo-RISC) and more difficult to optimize well, and when you do you are dependent on that implementation of the chip. Which brings up another point - you need more specialization in your compiler optimization to get the same results - and there can be more variation processor to processor (Pentium, PentiumPro, AMDs, CYRIX and so on). So one of the negatives of RISC (requiring smarter compilers) is magnified by this psuedo-RISC implementation - it takes an even smarter compiler (or more hardware) than RISC to achieve the same results. And the other negative of this (that code was more specifically optimized to a particular flavor of chip) is magnified as well. This is one of the reasons why Pentiums do better in special case specifications and benchmarks than they do in most real world tests.
Great News!!!
IBM's Power4 Desktop chip will be detailed on October 15th at the Microprocessor Forum. Current speculation is that the Power4 Desktop chip, "...designed for desktops and entry-level servers," will be used in Apple Macintosh computers starting in 2003, instead of a Motorola G5 processor.
Other details of the Power4 Desktop chip that are known are a vector processing unit handling over 160 vector instructions (AltiVec?), and a system interface that can transmit data at 6.4 GB/second. Rumors suggest that the chip could arrive at speeds of up to 2GHz. It will be capable of handling 8 Instructions Per Clock (IPC) and will come with 96KB of L1 cache (64 for data and 32 for instructions).
Send me more comments, sugestions, jokes, quotes, critics,...
