lkoop@pnet01.cts.com (Lamonte Koop) (01/11/91)
Here's an odd peculiarity of the CSA Mega-Midget Racer '030 board I have
come across that I hope someone may have an idea about before I'm forced to
pull the beast apart bit by bit to examine it. The MMR is a full '030
accelerator (this one is at 25MHz) with 32-bit RAM [512K SRAM for ROM
translation, 2 megs DRAM (80ns for both) currently installed. What I have
found is this: Under normal circumstances, the MMR does a ROM translation
WITHOUT the 030 MMU, using custom hardware. Now, running a WritePixel (draw a
box a pixel at a time, etc...) time test shows the board running about 40%
slower than an A2630...this is with the ROM image in the 80ns SRAM. Ditto
performance for doing a normal SetCPU MMU translation to the DRAM. Something
here is not kosher. [Especially in the light of running the ROM image out of
the SRAM]. Here's one more: Running the test with the Data cache turned OFF
actually speeds the test up (odd). A modification was done to the board
earlier to correct a timing problem with the system, but I don't see where it
could have induced this bizarre behavior. I don't wish to pull the board
apart, but curiosity is quickly getting the best of me, so if ANYONE (Dave ?)
has even the slightest idea what is going on with the beast, I'd appreciate a
line.
LaMonte Koop
Internet: lkoop@pnet01.cts.com ARPA: crash!pnet01!lkoop@nosc.mil
UUCP: {hplabs!hp-sdd ucsd nosc}!crash!pnet01!lkoop
A scientist is one who finds interest in the kinetic energy of Jell-O
moving at ridiculous velocities...an engineer is one who can find a
real-life application for such silliness.daveh@cbmvax.commodore.com (Dave Haynie) (01/15/91)
In article <6838@crash.cts.com> lkoop@pnet01.cts.com (Lamonte Koop) writes: >Here's an odd peculiarity of the CSA Mega-Midget Racer '030 board ... >What I have found is this: Under normal circumstances, the MMR does a ROM >translation WITHOUT the 030 MMU, using custom hardware. Now, running a >WritePixel (draw a box a pixel at a time, etc...) time test shows the board >running about 40% slower than an A2630...this is with the ROM image in the >80ns SRAM. Ditto performance for doing a normal SetCPU MMU translation to >the DRAM. Something here is not kosher. Ordinarily, you might find a couple of possible problems with such a system. There's not a whole lot CSA could have done to make the SRAM itself slower than A2630 DRAM. Though, with the ROM translated, there's not much SRAM left over for programs. While I would certainly expect the DRAM to be 32 bits wide, there's no guarantee its as fast as the DRAM on the A2630 or A3000. If CSA can tell you how fast their DRAM cycle runs, that might shed some light on your problem -- the A2630 and A3000 DRAM cycles in 5 clocks, or 200 ns (the 3000 of course supports faster burst cycles too, which run in 2 clocks, or 80ns, each). The other problem may be the efficiency of their 68030 <-> 68000 bus conversion logic. Something like WritePixel() certainly hits chip RAM alot. So even if they run the ROM code faster than an A2630 would, if the access to Chip RAM is slower, your WritePixel() test can easily go slower as well. The A2630 bus interface state machine uses both 7M and CDAC clock edges; CDAC isn't available on the 68000 socket these kinds of things plug into. You can make a CDAC from the 7MHz clock and a 35ns delay line, but if they tried to do the whole state machine with the 7MHz clock, that could account for the speed to 68000 bus resources. >A modification was done to the board earlier to correct a timing problem >with the system, but I don't see where it could have induced this bizarre >behavior. If "modification was done to the board earlier to correct a timing problem" really means "the original design was too fast for the supplied parts", that could easily give you such a result. I really don't know anything about the MMR board, so everything here is pure conjecture. I do know things that can give you the symptoms you're seeing, so I give them here to help you out. > LaMonte Koop -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "Don't worry, 'bout a thing. 'Cause every little thing, gonna be alright" -Bob Marley