home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cbmvax!daveh
- From: daveh@cbmvax.commodore.com (Dave Haynie)
- Newsgroups: comp.sys.amiga.hardware
- Subject: Re: A4000(Speed) -=> FALCON (Speed)
- Message-ID: <37804@cbmvax.commodore.com>
- Date: 11 Dec 92 12:20:31 GMT
- References: <1992Nov24.153452.880@gserv1.dl.ac.uk>
- Reply-To: daveh@cbmvax.commodore.com (Dave Haynie)
- Organization: Commodore, West Chester, PA
- Lines: 57
-
- In article <1992Nov24.153452.880@gserv1.dl.ac.uk> g.coulter@daresbury.ac.uk writes:
- >Hi Everyone, I hope that one of you can put me straight on this:
-
- >I am trying to find out just how much faster the A4000 will be
- >the Falcon. I have heard that the Falcon is about 6 times faster
- >than the A500
-
- Don't count on it for most things. The Falcon runs a 16MHz 68030 on a
- 16-bit bus. Assuming this memory is reasonably fast w.r.t. this '030,
- you could average maybe 2x-2.5x the overall integer performance of the
- A500. Actually, a 68030 on a 16-bit bus with caches off is slower on
- the overall than a 68000 on the same 16-bit bus. The numbers I pulled
- out here assume a faster memory bus for the 68030 in the Falcon (a slower
- wide bus, as in the A1200, would give you the same boost).
-
- If you're comparing floating point math, a Falcon with an FPU would be
- considerably faster than the A500, maybe 5x-10x depending on the job.
-
- >and that the A4000 is about 39 times faster.
-
- I don't know where that number came from. The A3000 is roughly 8x faster
- than an A500/A1000/A2000 in integer math, but again that's very application
- dependent. A chip-bound operation may only be twice as fast, some very
- cachable code could be even faster. The 25MHz clock, fast 32-bit burst
- memory bus, and cache contribute to this speed. How do you want to make
- up the numbers? The best way is to run some kind of benchmark on each
- machine. I can tell you than A3000 Fast memory is 5.6x faster non-burst,
- 10.2x faster burst, than that of the 68000-based Amigas. The clock is 3.6x
- faster. But all of these factors apply to the A4000, yet we mark the
- A4000 as roughly 3x faster than the A3000, even though it uses the same
- basic system design. That's due to the much more efficient 68040 design and
- larger cache.
-
- >Also if a 68882 is added to the Falcon, how much of an overall
- >speed improvement is gained?
-
- It depends. If you're running code that uses floating point math, it may
- give you a 5x-10x boost on math, though I'm not sure what kind of penalty
- an FPU on a 16-bit bus takes, we've never run one. Most kinds of code
- don't get any improvement. Even code that's FPU intensive can be 70% or
- more actual integer operations.
-
- >I am not trying to start a war here but does anyone know of any
- >rumors regarding a possible DSP sound processor for the Amiga.
-
- There will be DSP on the Amiga in the future. It will handle sound, but
- it won't be restricted to sound (eg, it won't use the Motorola DSP 56001).
- The 56001 is a 24-bit fixed-point DSP. That's enough to go from the
- threshold of human hearing to a bit beyond the threshold of pain, but it's
- no use for most graphics processing or other scientific applications.
-
- --
- Dave Haynie / Commodore Technology, High-End Amiga Systems Design (cool stuff)
- "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh BIX: hazy
- SCIENCE: "I'll believe it when I see it"
- RELIGION: "I'll see it when I believe it"
-
-