Hardware test results

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

SpiceWare

Shows 60. The rolling seems worse than before.


Andrew Davie

#16
Quote from: SpiceWare on 08 Jun 2026, 06:32 AMShows 60. The rolling seems worse than before.

TY.  At least we know for sure the DEV carts are 60 MHz ARM-based.

I begin to suspect a malformed TV-frame VBLANK/BSYNC etc. Will look into that.

Andrew Davie

Quote from: Andrew Davie on 08 Jun 2026, 09:27 AMTY.  At least we know for sure the DEV carts are 60 MHz ARM-based.


Actually the way the detect works, this is not strictly true.
We know that it's not a 70MHz ARM. By inference I assumed 60MHz. But the detect code will find the closest it knows about. Technically NTSC 60MHz is the "lowest" value. If the cartridge were running, for example, a 50MHz ARM it would still detect as 60.

Also, a slower ARM than 60MHz on-board would explain all the rolling.

SpiceWare

We ran tests in 2017 that show the # of ticks the ARM did over 262 scanlines. The results I had on my 2600 ranged from 11d326 to 11d32e and my 7800 ranged from 11d311 to 11d313.

I loaded the NTSCvsPAL2.bin on my Dev cart and it returned 11d3e3, so the Dev cart seems slightly faster than my 70 MHz Harmony did back in 2017 (or my 2600's clock is running slight slower, don't know if that's possible). I'm still putting my office back together, will run the test again on my original Harmony once I locate it.

If I recall correctly the driver sets the clock rate.  I see this in the source for the CDFJ+ driver:
  /* Set PLL multiplier and divider values for 70MHz */
  ldr     r0, =PLLBASE
  mov     r1, #0x26
  str     r1, [r0, #+0x04]
...
  /* Set peripheral clock divider to 1 (APB is 70MHz) and enable fast GPIO */
  ldr     r0, =APBDIV
  /* mov     r5, #0x01 */
  strb    r5, [r0]
  ldr     r0, =SCS
  str     r5, [r0]

  /* Enable memory accelerator and set timings for 70MHz */
  ldr     r0, =MAMBASE
  mov     r1, #0x04
  str     r1, [r0, #+0x04]
  mov     r1, #0x02
  str     r1, [r0]

It's possible the CDFJ driver for large ROM sizes sets it for 60 MHz as I recall batari had an issue sourcing 70 MHz parts for a while.

My workday has started so I'll check back in later.

SpiceWare

Found my original Harmony on my lunch break, results were 11d32d to 11d337 which overlap my 2017 results.

I tested both when flashed onto the Harmony (just like the Dev cart) as well as when loaded from the SD card via menu (not possible on Dev cart, it does not support SD cards).

11de3e / 11d332 = 1,168,355 / 1,168,178 = 1.00015

The Dev cart is insignificantly faster than the Harmony.

Bomberman94

If it is of any help i can run your program with my Dev cartridge on my PAL 2600 (light sixer) and see if it differs.

I use a development cartridge that can handle CDFJ+ and holds up to 127K of data (yes, not completely 128K).

SpiceWare

Quote from: Bomberman94 on 09 Jun 2026, 08:19 AMI use a development cartridge that can handle CDFJ+ and holds up to 127K of data (yes, not completely 128K).

Are there different Dev carts?

I did some tests with the 128K build of Turbo Arcade and the limit I hit was 120K.  Basically I ran the dd command in the terminal to truncate the ROM:
dd bs=1024 count=120 if=Turbo-Arcade_48pin_128K.bin of=test_120K.bin
bs = block size
count = how many blocks to copy
if = input file
of = output file

The game worked with a count=120, but nothing higher.

Bomberman94

Thanks for clarifying this - it's only 120K then 👍