|
Vintage Computers Any vintage computer systems, calculators, video games etc., but with an emphasis on 1980s and earlier equipment. |
|
Thread Tools |
5th Dec 2023, 5:54 pm | #1 |
Hexode
Join Date: Jan 2021
Location: Ashford, Kent, UK
Posts: 318
|
MK14 VDU Modules Bench-marking
In developing programs for MK14 I've realised there can be quite a performance difference depending which VDU solution is connected. To that end I've been bench-marking the VDU modules using a test version of Invaders that removes the game delays and toggles FLAG0 each game cycle. I've used invaders as I wanted something with a program cycle much longer than the VDU frame time (and hence much larger than the NENIN interrupt time). This results in 2 benchmark timings as shown in the photo.
I've made measurements for the VDU solutions I have access to and Coolsnaz2 kindly ran the benchmark on his original SoC VDU. In all cases, the MK14 has a 4MHz crystal installed. Code:
Burst Cycle Performance VDU variant time time (SoC/Variant) Comment ===================================================================================================================================== SoC VDU Card 10.8ms 129ms 1.0 This is the standard against which all others are compared RealView (PAL) 7.8ms 98ms 1.3 Looks like I reduced the NENIN interupt time on this one RealView (VGA) 9.4ms 112ms 1.1 Blanking period is shorter than for PAL, reducing its performance PiView 3.9ms 45ms 2.8 !! There is hardly any NENIN activity so the processor runs almost full speed OrtonView TBA P.E. VDU TBA Chris Oddy Replica TBA Others ????? (Note all timings have been averaged over a few readings) The main drive behind this is to be able to include specific delays in games so that the game performance could be made broadly the same across all VDU module variants (plus it's just interesting to know!) Thanks in advance |