Monitor suggestions

[Andi B.]

I find it strange that there is so much problems with OS/2 video.  I have a dell 30inch 2560x1600 monitor that runs without any problem.  No problem with video cards, I don't have one, it runs from the built in video of the Ryzen 3 processor.

I observe different problems with OS/2 video output nowadays.

1) The BIOS decides which output is to be uses or not. Not used are sometimes switched off here. F.i. when  at boot the HDMI output is attached on my RX550 (?) the Display Port output is switched off. And vice versa. When both are attached I think DP is used and HDMI is switched off. But when something is attached to the internal Intel GPU (DP or HDMI or VGA) in parallel to the outputs of the extra graphic card, things go to complicated to describe very fast.
 
2) Integrated Intel GPU on my system does work but not with 4k. So it's kind of useless on my 3840x2160 Monitor. Needed to put in a extra graphic card and needed a ticket and work from David A. to get this card to work at full resolution.

3) Ryzen 7 integrated GPU does not work in 4k with ArcaOS too. Of course it works with the other OSes I've installed. No clue if David can fix this, need to create a ticket for that.

4) Output (DP or HDMI) is switched off when I switch the KVM to another source. But it is not switched on again when I switch back to ArcaOS. So you don't see any output anymore from your running system until reboot. Of course the other OSes switches back on the output (tested with DP and HDMI on Ryzen 7) when the KVM switches to that input.

There are even more problems with video on our platform I don't wanna go into detail. Some of them are nice to have things which result from the way we are used to work nowadays. But the above are the most annoying.

[Eugene Tucker]

I just bought a 24 inch Asus VA24D and it is working fine. It has a display port and a HDMI port it also has a VGA port sorry no DVI port. I got the monitor from Amazon. I did get a Monitor from New Egg it was an LG and it only had HDMI ports. I sent it back as I use Display port. The video chip is in my AMD processor.

[Per E. Johannessen]

I'm now testing;

GPU: Asus Geforce RTX 4060 Ti 16GB, (Nvidia).
Monitor: Samsung Odyssey G8 G80SD 32" 4K OLED.
Connection: HDMI

Max resolution offered by Panorama is 1280x1024.

Is there a "safe" way to get a higher resolution or do I simply have to experiment.

[Dave Yeo]

I'm now testing;

GPU: Asus Geforce RTX 4060 Ti 16GB, (Nvidia).
Monitor: Samsung Odyssey G8 G80SD 32" 4K OLED.
Connection: HDMI

Max resolution offered by Panorama is 1280x1024.

Is there a "safe" way to get a higher resolution or do I simply have to experiment.

Change video cards? Nvidia is known to not be friendly with no way to patch its BIOS for higher video resolutions. You could open a ticket but I think you will be told the same thing.
Dave

[Andi B.]

Max resolution offered by Panorama is 1280x1024.

Is there a "safe" way to get a higher resolution or do I simply have to experiment.

My tests with a new system tought me to -
- better use AMD
- better use DisplayPort
- don't connect both (or 3) outputs at the same time (DP, HDMI, DVI VGA) especially from cold boot.
- don't disconnect output from monitor, don't use KVMs (no KVM works here reliable either DP nor HDMI)
- HDMI output here only offers 1400 x 1050 sometimes. When DP is connected too it magically offers full 3840x2160 but system traps when selected afterwards while boot.  Of course I use short high quality cables only.
- DP usually offers full resolution (or most complete list)
- Panorama with integrated AMD Ryzen only offers 2560x1440 altough the system runs 3840x2160 perfectly well with other OSes (except Win7 which only can do 2560x1440 cause Ryzen 7 is not supported anymore with AMD Win7 driver).
- AMD RX550 works better with Panorama (and Win7) then integrated GPU
- 3840x2160 on one system (UEFI) leads to trap 000e in VMAN while 2560x1440 works.

You can also look at the tickets in mantis to find the log file of video (gradxx?) which is written at boot up to see what Panorama driver has to say. Or look into 'testlog generic' output.

[Per E. Johannessen]

Now testing;

GPU: Radeon RX 7800 XT 16GB
(Same GPU I had a lot of trouble with when using an old Samsung monitor, SyncMaster SA950 27")
Monitor: Samsung Odyssey G8 G80SD 32" 4K OLED

Results;
HDMI: Offers same resolutions as the GeForce RTX 4060 Ti 16GB, max 1280x1024
DisplayPort: Offers lots of resolutions up to 3840x2160

Other issues;
When rebooting everything seems fine until the point where it normally displays the desktop, screen goes black with a blinking cursor in the upper left corner.
If I choose "Power off" the system shuts down as expected. When switching on again the system boots fine.

Just noticed another issue, don't know how/if it is related but it hasn't happened before.
After the cold reboot I wrote this message with Firefox and closed it when finished.
When trying to start Firefox again I get:
"Couldn't load xul.dll (OS/2 error 8, faulty module:
G:\programs\firefox\xul.dll"

[Dave Yeo]

So things work fine after a cold boot but not a warm boot? Strange and perhaps you should open a ticket.
Firefox failing to start after a warm boot is also strange and something I've never heard of. Unluckily the error message isn't very informative on exactly what didn't load. Have you looked in popuplog.os2 for anything related? What is your VIRTUALADDRESSLIMIT in config.sys? Perhaps test with 2560. Did you ever run ffturbo install? If not, try running it.

[Per E. Johannessen]

The system works fine after a cold boot except that Firefox can only be started one time only. Closing Firefox and then trying to start it again gives the "Couldn't load xul.dll". This is with VIRTUALADDRESSLIMT = 3072.
With VIRTUALADDRESSLIMIT=2560 it boots ok until it is about to start the desktop and then it goes black.

One thing came to my mind, earlier when trying the Radeon GPU with the old monitor something changed with the system. The shutdown dialog used to give me the possibility to choose which system I'd like to reboot to, I could choose any of the other OS installations on this system. At some point this has changed to the ordinary shutdown dialog with only the options "Power off" and "Reboot".

In order to change VIRTUALADDRESSLIMIT back to 3072 (from 2560) I cold booted another AOS 5.1 installation and edited config.sys for the troubled "G:\" installation which is AOS 5.0.6 and has all available updates installed.
From the AOS 5.1 installation, which has VIRTUALADDRESS=1536 and Firefox works fine, I was able to choose "Reboot to G:" and it actually worked fine.

So, obviously something has gone wrong on 5.0.6 on "G:\".

[ivan]

Regarding your problem on 'G', have you run CHKDSK against that drive?

[Andi B.]

it again gives the "Couldn't load xul.dll". This is with VIRTUALADDRESSLIMT = 3072.
With VIRTUALADDRESSLIMIT=2560 it boots ok until it is about to start the desktop and then it goes black.

Now I've here two systems which behave very strange when trying to use VIRTUALADDRESSLIMT = 3072. I've not a single system/installation which works okay with this setting. I've the same intermediate 'out of memory' errors loading some programs. That's what in my config.sys about -

Code: [Select]

REM 20210711 VIRTUALADDRESSLIMIT=2560 to 3072 because of qt browser
REM 20210711 neither simplebrowser nor Seamonkey nor ndpiso plugin works with 3072 -> 2900
REM 20211023 changed to 0xXX00 as Dave suggested in AN Testers from 2850 to 2816
VIRTUALADDRESSLIMIT=2816

I've to add OpenOffice and VirtualBox also give such errors sometimes with 3072.

[Dariusz Piatkowski]

Hi Andi,

Now I've here two systems which behave very strange when trying to use VIRTUALADDRESSLIMT = 3072. I've not a single system/installation which works okay with this setting. I've the same intermediate 'out of memory' errors loading some programs...

What other device drivers, especially the IFS stuff, are you loading? What size cache are you defining for these?

The symptoms you describe all look & feel like the "out of memory" stuff I was seeing when testing the very large (>1G) JFS cache sizes here.

On a running system, can you capture the output of mem.exe?

Here is what I see after 2 days of uptime:

Code: [Select]

[G:\util\misc]mem /v

Total physical memory:     12,287 MB
Accessible to system:       3,327 MB
Additional (PAE) memory:    8,960 MB

Resident memory:            1,187 MB
Available virtual memory:     430 MB

Available process memory:
  Private low memory:         179 MB
  Private high memory:      1,344 MB
  Shared low memory:           67 MB
  Shared high memory:         282 MB

Can you do this BEFORE and AFTER the issue appears?

[Per E. Johannessen]

Forgot to mention that chkdsk has been run several times without any errors and nothing in populog either.

Although the Radeon offers lots of resolutions it's not stable enough for the time being.
I put back the Nvidia RTX 4060 Ti and connected with DP it now offers a max resolution of 1920x1080 and so far it seems stable
with VIRTUALADDRESSLIMIT = 2816

[Andi B.]

What other device drivers, especially the IFS stuff, are you loading? What size cache are you defining for these?

The symptoms you describe all look & feel like the "out of memory" stuff I was seeing when testing the very large (>1G) JFS cache sizes here.

I usually let the IFS stuff as it is defined by ArcaOS installer. For older systems I also don't have big JFS cache. HPFS386 is remed out since long as it produces similar problems.

'mem -v' - my last tests on a pretty vanilla ArcaOS system I build up last week leads to 'out of memory' even when trying to open 4os2 or cmd windows. Sometimes afterwards it magically worked for one session but no more. -> very weird behavior with VIRTUALADDRESSLIMIT=3072. Reminds me on same experience a few years ago with another system.

So for me I can't get any system to work only near reliable with 3072. Wild guess, it may interfere with my 3840 x 2160 resolution. My observations lead me to guess there is a memory problem with the WPS or gradd or VMAN or ... with this high resolution too.

Administrator - this now runs very off topic. Maybe you wann split this thread. Thanks.

[Dave Yeo]

I think the problem is that IBM never finished the high memory support, or perhaps just buggy. PCI space, which is at the top of the 4GB address range, has grown with newer hardware. You put a video card in that uses address space for its large frame buffer and its BIOS and it takes up a bunch of address space. The JFS cache seems to be implemented up there too. So with 3072 virtualaddress space, it interferes with PCI address space and/or JFS address space and you get an unstable system. So you need a smaller virtualaddress space to be stable.
mem /v should tell you how much memory is accessible to the system, which should point to how much virtualaddress space you can use. Here, it is 3,502MB and using 3072 seems fine.

[Per E. Johannessen]

Below is the output of mem /v
With VIRTUALADDRESSLIMIT=2816 the system seems stable.
I'm on thin ice here so, as always, I welcome comments.

Total physical memory:     16,102 MB
Accessible to system:       1,766 MB
Additional (PAE) memory:   14,336 MB

Resident memory:              290 MB
Available virtual memory:   1,382 MB

Available process memory:
  Private low memory:         288 MB
  Private high memory:      2,012 MB
  Shared low memory:          234 MB
  Shared high memory:       1,729 MB