Difference between revisions of "Performance Testing with NIB3"

From Jay's Cafe' Wiki
m
m
 
(3 intermediate revisions by the same user not shown)
Line 1: Line 1:
[[NIB Stress Testing #2]]
Build:
Build:
*Manjaro ARM, RPI4B/4GB
*Manjaro ARM, RPI4B/4GB
Line 42: Line 43:
*Time Test - allowing board to idle while running, no network transmissions for about 3 hours to witness time drift.
*Time Test - allowing board to idle while running, no network transmissions for about 3 hours to witness time drift.
**Start 3:33 am
**Start 3:33 am
**Stop at 5:05 am.
*Test was stopped as there was no time drift (difference between the NIB clock and the host clock for over 1.5 hours. Additionally, all witnessed cpu speeds were a constant 100%.
*Configuration and addition to network.
*Running with a configuration, system is between 80-100%. Not acceptable. Going to retry with a lower resolution.
* set to 640x480, took aspect true off, set rendering to nearest neighbor. Flickered to 99% during startup temporarily and then picked back up to 100% and 59.8 fps.
***I think I may have a winner. Need to set up subs and get it going for the next test of a live setup. I see no point in testing x64sc until the subs are setup.  I wanted to do prelims to see if I could get the board to idle at 100% using xscpu64. x64sc historically has been less resource intensive than xscpu64.
*'''Testing with Openbox and not treatments
*resolution 640x480, no aspect, nearest neighbor, no sound, fullscreen with menu decorations.
*Note: I plan on running production without the menu and status bars. They are strictly being run to enable monitoring of the %CPU and fps of VICE-Emu

Latest revision as of 11:59, 20 March 2023

NIB Stress Testing #2 Build:

  • Manjaro ARM, RPI4B/4GB
  • XFCE
  • Resolution 1024x768
  • VICE 3.7.1 r43529
  • tcpser 1.15a (Jim Brain)
  • Disable power settings, screensaver and lock.

===========================
Trial 1, Preliminary

  • removed e.data and bd.data to force reconfiguration of bpc.
  • configured "Bayou Black" BBS
  • No sound
  • Full Screen (alt-D) with decorations (alt-B)
  • config file below
~/.config/vice/vicerc 
[Version]
ConfigVersion=3.7.1

[SCPU64]
FullscreenDecorations=1
Window0Height=768
Window0Width=1024
Window0Xpos=0
Window0Ypos=0
Sound=0
SCPU64Name="/home/jay/JiffyDOS/scpu64.bin"
MachineVideoStandard=2
VICIIFullscreen=1
VICIIModel=4
RsDevice3ip232=1
RsDevice3Baud=38400
Acia1Dev=2
Acia1Enable=1
Drive8Type=4844
Drive9Type=1581
Drive10Type=1542
DosName1541ii="/home/jay/JiffyDOS/JiffyDOS_1541-II.bin"
DosName1581="/home/jay/JiffyDOS/JiffyDOS_1581.bin"
DosNameCMDHD="/home/jay/JiffyDOS/CMD HD BOOTROM v280.bin"
  • Time Test - allowing board to idle while running, no network transmissions for about 3 hours to witness time drift.
    • Start 3:33 am
    • Stop at 5:05 am.
  • Test was stopped as there was no time drift (difference between the NIB clock and the host clock for over 1.5 hours. Additionally, all witnessed cpu speeds were a constant 100%.
  • Configuration and addition to network.
  • Running with a configuration, system is between 80-100%. Not acceptable. Going to retry with a lower resolution.
  • set to 640x480, took aspect true off, set rendering to nearest neighbor. Flickered to 99% during startup temporarily and then picked back up to 100% and 59.8 fps.
      • I think I may have a winner. Need to set up subs and get it going for the next test of a live setup. I see no point in testing x64sc until the subs are setup. I wanted to do prelims to see if I could get the board to idle at 100% using xscpu64. x64sc historically has been less resource intensive than xscpu64.
  • Testing with Openbox and not treatments
  • resolution 640x480, no aspect, nearest neighbor, no sound, fullscreen with menu decorations.
  • Note: I plan on running production without the menu and status bars. They are strictly being run to enable monitoring of the %CPU and fps of VICE-Emu