View Single Post
Old 25 February 2012, 20:00   #2350
Registered User
SpeedGeek's Avatar
Join Date: Dec 2010
Location: Wisconsin USA
Age: 54
Posts: 380
Originally Posted by Schoenfeld View Post
Any 020/030 accelerator makes a faster access to the A1200 than a PPC/060 accel. Don't look at the MHz number of that card, but at the bus interface: The PPC and 060 processors don't have dynamic bus sizing, which is why they lose big time if the logic is as limited as it was back in the days when the PPC card was new.

Here are my Zorro2 (cpu nodatacache) bustest results from my A3000 + A3640:
New Shell process 3
3.System3.9:> cpu nodatacache
System: 68040 68882 (INST: Cache Burst) (DATA: NoCache NoBurst)
3.System3.9:> bustest addr f00000
BusSpeedTest 0.19 (mlelstv) Buffer: 262144 Bytes, Alignment: 32768
memtype addr op cycle calib bandwidth
user $00F00000 readw 563.6 ns normal 3.5 * 10^6 byte/s
user $00F00000 readl 1127.4 ns normal 3.5 * 10^6 byte/s
user $00F00000 readm 1127.6 ns normal 3.5 * 10^6 byte/s
user $00F00000 writew 563.7 ns normal 3.5 * 10^6 byte/s
user $00F00000 writel 1127.6 ns normal 3.5 * 10^6 byte/s
user $00F00000 writem 1127.8 ns normal 3.5 * 10^6 byte/s

Note: I have an A3640 running @ 50 MHz with modified 68030 state machine logic and a Super Buster Zorro2 speed up mod. The purpose of this post is to demonstrate that dynamic bus sizing need not slow the performance of any 040 or 060 card. However, regardless of how efficient the state machine logic is with respect to dynamic bus sizing what was overlooked here is the fact that most 3rd party 040 and 060 cards use Asynchronous state machine logic which always adds 1 or 2 wait states to every bus cycle.

Last edited by SpeedGeek; 25 February 2012 at 20:09.
SpeedGeek is offline  
Page generated in 0.04453 seconds with 10 queries