[Jacques]

Before my system disk threw a bearing and my V6 gave up (all in one day!), I managed to build a VFO file for SSI/SSE/MXI/MXE Octane boards for 1920x1200 @60Hz. However, whenever I ran the mode I experienced glitches (much like the O2 glitches when going close to the pixel clock range). I'm not sure if it's my SSE board throwing a wobbly or if the card is just being taxed too much.     

The build is based on 2RSS (so not SI/SE single raster boards) with single buffering only as there isn't enough framebuffer memory for double buffering. (it throws errors when trying to build for double buffering)

Furthermore, the build is also based on reduced blanking to try and stay <160Mhz pixel clock. I'm not sure whether it's the reduced blanking throwing the edge condition glitches.

So, is there anybody else out there with SSE/SSI/MXI/MXE boards who would be willing to test this? The VFS source file is based on EDID for a Dell 2407wfp panel. 

You'll need to copy the file to the location below for xsetmon to read it. 

Code:

/usr/gfx/ucode/MGRAS/vof/2RSS/

https://www.dropbox.com/s/0lndbq96uwmmz4...0.sdb?dl=0

[callahan]

Tried it today and worked pretty well on my EMXI Octane w/ a Dell U2412 monitor. I did notice a few stray lines on the edges of the screen (1 pixel wide by about 1 cm long) that very briefly appeared as I moved around windows.

Screenshot (distortion is from the camera): http://imgur.com/k4TWPzm

Care to paste the VFO you used to compile this?

[Jacques]

Thanks for trying it out. That sounds like the same distortion I get, pity.

I have to recreate the VFS file as it was on my drive that died. I'd exported the compiled file to my server before my drive gave up.

I'll try and work on increasing the front / back porch blanking a bit, not sure I'll be able to stay under 160Mhz though.

[Jacques]

Here you go.

https://www.dropbox.com/s/hys6lqx9hdtla2...0.vfs?dl=0

[callahan]

Alright, after a couple hours of messing around I think I have this fixed. The issue with artifacts to the left of the screen does appear to be reduced blanking-related, not a hardware defect. I was able to fix this by increasing the horizontal blanking incrementally. I also tried the different pre-compiler options (0,1,2), which appear change how the compiler treats blanking, but this did not have a noticeable effect.

The results of my work:

The 1920x1200 sdb file with marginally increased blanking (1.35 horizontal multiplier from my Dell U2412's EDID info): https://www.dropbox.com/s/wh9414jg91lmmf...2.sdb?dl=0
This should work on any 2RSS MGRAS (SSI/SSE/MXI/EMXI) system, but has only been tested with an EMXI card.

The 1920x1200 .vfs input file: https://www.dropbox.com/s/3p8v8oguvepjo7...d.vfs?dl=0

An excel spreadsheet I used to help calculate the appropriate data for a .vfs file both directly from EDID info and with increased blanking: https://www.dropbox.com/s/a5djb5ceicen2r....xlsx?dl=0

[Jacques]

Thanks for you efforts! I'll give it a spin later today / tomorrow.

Does your card run any hotter?

[eudatux23]

Just out of curiosity: Would this also be compatible with the Impact cards for Indigo² like the High Impact or are these Single Raster Boards?

[Jacques]

This is for 2RSS boards, so - Indigo2 MaxIMPACT, SSI, SSE, MXI, MXE.

1RSS boards include - Indigo2 SolidIMPACT, HighIMPACT, Octane SI & Octane SE.

How much framebuffer memory does the I2 SolidIMPACT & HighIMPACT have?

[Trippynet]

I would put money on this not working on Solid/High for Indigo2. I'm not sure how much framebuffer they have, but it won't be enough for 1920x1200. Heck, they struggle with 1600x1200. MaxIMPACT may be capable of supporting it as it does have double the framebuffer.

[callahan]

@Trippynet -- I'm ready to collect... It works!

.sdb file for 1920x1080 @ 60 FPS on 1RSS cards (w/ reduced blanking): https://www.dropbox.com/s/yq4srbu1s2xj6r...S.sdb?dl=0

This is confirmed to work on my Indigo2 Impact.

https://imgur.com/z5AZQbY

Note this does produce a few artifacts on the left edge when moving windows, but it is very tolerable/usable. It should be possible to eliminate this using an increased blanking setup (there are loads of pixel clocks to spare, this is only 138MHz), but I haven't done so.

Note that for some reason precompiler option 0 and 2 work, but .sdb files compiled with option 1 do not appear in my display manager.

Edit: This was compiled using DZBUF=0 and DDB32=0 out of concern for the SI's limited ram. It may work with either/both of those options enabled, but I can't confirm.