Fiddling away with my recently acquired Fujitsu ICL Indiana.
Its got an ACER V12LC motherboard https://www.ultimateretro.net/en/motherboards/225
And I tried the various versions of BIOS available, they all seem to be very similar and no additional functionality as far as I can tell
Also I'm pondering if I wanna add the missing "JP7" jumper that allows disabling the internal graphics.
The included sound card (ct2800) is brilliant and fits perfectly, but the on board graphics don't really match the rest of the system.
Its just to damn slow 🤣, unless I'm missing some obvious tweaks for the CL5434

Today I have a remote job interview and I had an idea to try (again) using my old dual Atom computer with its micro iGPU just to say it could be possible with a low res camera and less encoding work but is not easy cause if using linux even with an external gpu the cpu isn't simply enough even with a fast GPU; but last time I didn't try Win where I've found the iGPU DXVA acceleration works better than I thought in both browser / app conference, offloading at least part of the decoding task even when not supported for modern codec. Interesting its iGPU should support H264 encoding too but not sure is enabled by the driver. So anyway I've built a better machine with a Athlon II X2 2,7Ghz/4GB DDR2/RV730 too which is a bit faster. I'll keep both machines near in any case of problems. 😉

I'm not sure if this counts as retro activity, but I have used cheap ATI Radeon X300 GPUs for a basic framebuffer in otherwise headless Linux machines for years. Today I upgraded a server to Fedora 36 and it shat a brick. Booting took forever and the system hung on shutdown.

The issue ended up being a bug in the Radeon driver:

12022-07-15T05:57:32.385213-04:00 fedora kernel: amdgpu 0000:0a:00.1: amdgpu: Fatal error during GPU init
22022-07-15T05:57:32.385322-04:00 fedora kernel: amdgpu: probe of 0000:0a:00.1 failed with error -12
32022-07-15T05:57:32.385334-04:00 fedora kernel: BUG: kernel NULL pointer dereference, address: 0000000000000000

Divide by zero at your own risk.

I'm getting my grubby paws on old SC MIDI files, so that I can enjoy the VSC soft-synth and the Edirol modules that I am going to be purchasing soon.

386SX wrote on 2022-07-15, 08:34:

Today I have a remote job interview and I had an idea to try (again) using my old dual Atom computer with its micro iGPU just to say it could be possible with a low res camera and less encoding work but is not easy cause if using linux even with an external gpu the cpu isn't simply enough even with a fast GPU; but last time I didn't try Win where I've found the iGPU DXVA acceleration works better than I thought in both browser / app conference, offloading at least part of the decoding task even when not supported for modern codec. Interesting its iGPU should support H264 encoding too but not sure is enabled by the driver. So anyway I've built a better machine with a Athlon II X2 2,7Ghz/4GB DDR2/RV730 too which is a bit faster. I'll keep both machines near in any case of problems. 😉

Is it with that GMA 3600 IGP?? Got one dual atom with that and it's a real "princess", pees through 17 mattresses on the daily. Whereas the GMA 3150 Atoms I've had no issues with apart from overall speed. Used to use those regularly up to 3ish years ago, for video chat with facebook or google groups. In non-integrated GPU, I think you need Radeon HD 6450 or Geforce GT610 or better for codec support in modern OS/apps. Even if you've got a "faster" GPU from an earlier series it won't do you any good for the codec support.

BitWrangler wrote on 2022-07-15, 13:50:

Is it with that GMA 3600 IGP?

Yes is the GMA3600 = PowerVR SGX545@400Mhz installed in the 32nm SoC with the dual Atom with SSSE3. It's "unique" like the last Cirrus Logic 3D accelerator or some Trident solutions. 😀

It's a patience stressing gpu but at the end not intended to compete in that notebook/desktop scenario imho mostly for simpler portable OpenGL ES devices o.s. while it "could" and worked in Directx9.0c and OpenGL 3.0 too. I imagine for a cost/power/SoC space/specifications reasons the idea to use it was ok. But with a different uncommon architecture where drivers were uncommon too and probably might have needed a big work into optimization, what is visible on benchmark numbers are the low fill rate (for its time and even lower in multitexturing) but good Pixel Shading numbers. On a common Direct3D9 application, without some ipotethical gpu specific optimization, the engine reach something like 25MT/s at best and around 700Mt/s single texturing fill rate value with something wrong into multitexturing that decrease at half. So even retrogames ran like in their times low end video card ran the same titles. Without an old o.s. driver support and too slow for modern o.s. sw, basically it had the specifications of a modern GPU but not the speed in that "3 watts" iGPU config, which was something to consider. The E-350 APU iGPU sure was much much faster but consume also many more watts (up to 10 watts more I'd say cpu included which is not much faster) with basically a lot of heat inside the already not easy to cool 40nm SoC for a big core size.

About the modern video conference requirements I think even when the codec isn't supported in hardware the DXVA generally offload something too compared to a OpenGL/VAAPI linux situation (when and if hw acceleration work or not at all) an in fact I was surprised to see low cpu usage in some test in Win with the GMA iGPU using Direct3D9. But I didn't try in a multiconference situation the cpu usage might be 100% easily. But the iGPU supported H264 1080p decoding (even 60fps work) and theorically the same IC used into the smartphone supported the same/similar encoding in hw. I don't think its Win driver had that but the hw should.

There are consumer webcams with video encoder hardware built in, which allows older/lesser computers to use high-res webcams that would otherwise use CPU or GPU resources for stream encoding. Logitech sells at least one model.

RetroGamer4Ever wrote on 2022-07-15, 16:53:

There are consumer webcams with video encoder hardware built in, which allows older/lesser computers to use high-res webcams that would otherwise use CPU or GPU resources for stream encoding. Logitech sells at least one model.

It's the first time I've heard about these products. But at a first thought I wonder how they work

EDIT: found this https://www.logitech.com/en-us/video-collabor … 4-encoding.html it looks like the sw needed to be compatible with the webcam encoder which I suppose might not be the case of any modern conference app. Still interesting and I didn't know these existed.

EDIT2: it looks like they are not discontinued after all. There's the Logitech C930e webcam that seems to have H264 encoding with UVC 1.5 specification for that which I didn't know about. They seems even certified for most know conference apps. I wonder how much less usb bandwidth is saved (which seems the point at the end) during the app<>camera communication compared to a common HD USB 2.0 camera.

Sometimes I cant help but be addicted to old cheese .. in this case its the old software variety of cheese, I saw this 12 pack for a tenner and thought it might make for a fun diversion for a few hours.

Discs 7 thru 9 look the best, but that Science Fiction disc looks entertaining too, I wonder what the quality of the Doom levels is going to be ...likely as good as the quality of this pack 😁

The first three discs are pretty much throw away, unless that Peter Norton disc has something interesting on it, I was thinking of how to archive these discs and I think there might be such a place where I can do this, Im sure there are others who like me just love digging through old DOS/Windows software and media.

386SX wrote on 2022-07-15, 17:04:

It's the first time I've heard about these products. But at a first thought I wonder how they work […]
Show full quote

RetroGamer4Ever wrote on 2022-07-15, 16:53:

There are consumer webcams with video encoder hardware built in, which allows older/lesser computers to use high-res webcams that would otherwise use CPU or GPU resources for stream encoding. Logitech sells at least one model.

It's the first time I've heard about these products. But at a first thought I wonder how they work

EDIT: found this https://www.logitech.com/en-us/video-collabor … 4-encoding.html it looks like the sw needed to be compatible with the webcam encoder which I suppose might not be the case of any modern conference app. Still interesting and I didn't know these existed.

EDIT2: it looks like they are not discontinued after all. There's the Logitech C930e webcam that seems to have H264 encoding with UVC 1.5 specification for that which I didn't know about. They seems even certified for most know conference apps. I wonder how much less usb bandwidth is saved (which seems the point at the end) during the app<>camera communication compared to a common HD USB 2.0 camera.

https://en.everybodywiki.com/List_of_cameras_ … deo_compression

So I started testing the GTS450 first with the problem games:

Doom 3 ran as poorly as with the 8800GT. I consulted the pcgamingwiki and found this graphics fix which isn't supposed to improve performance. It did. It's still a little slow, but with this it's much closer to a consistent 60fps and don't seem to drop below 45.

Quake 4 runs at rock solid 60fps.

Enemy Territory: Quake Wars remained at 30fps, but it didn't drop below 10 like with the 8800GT. As I've seen on pcgamingwiki, early versions were locked to 30 so I updated with patch 1.5. After that I could also unlock the frame rate and the game ran above 80fps.

As for temperatures, the GPU stayed around 73C° with factory fan control while stayed decently quiet. So far I'm satisfied.

RandomStranger wrote on 2022-07-15, 18:37:

So I started testing the GTS450 first with the problem games: […]
Show full quote

So I started testing the GTS450 first with the problem games:

Doom 3 ran as poorly as with the 8800GT. I consulted the pcgamingwiki and found this graphics fix which isn't supposed to improve performance. It did. It's still a little slow, but with this it's much closer to a consistent 60fps and don't seem to drop below 45.

Quake 4 runs at rock solid 60fps.

Enemy Territory: Quake Wars remained at 30fps, but it didn't drop below 10 like with the 8800GT. As I've seen on pcgamingwiki, early versions were locked to 30 so I updated with patch 1.5. After that I could also unlock the frame rate and the game ran above 80fps.

As for temperatures, the GPU stayed around 73C° with factory fan control while stayed decently quiet. So far I'm satisfied.

Its odd that Quake4 runs rock solid but Doom3 has issues .. they use the same engine and Quake 4 is newer ..very odd.

All 3 games use the same engine. Doom 3 wasn't well optimized when it released. Though this supposed to be an updated version since I installed Resurrection of Evil as well. I don't understand either. the 8800GT should have already been more than enough for both.

RandomStranger wrote on 2022-07-15, 19:25:

All 3 games use the same engine. Doom 3 wasn't well optimized when it released. Though this supposed to be an updated version since I installed Resurrection of Evil as well. I don't understand either. the 8800GT should have already been more than enough for both.

the 8800 GT should have been more than Doom3 needed, perhaps its using the wrong OpenGl library, though if Quake4 is fine then its likely an issues with one of the graphic settings in the ini files.

It's weird how Nvidia handles vertical sync. The driver leaves it to the game settings by default, but it doesn't always work. In Call of Duty (1) vsync is off by default, yet it's still locked to 60fps. If I force off from the driver, it's off. Maybe it works with Doom 3 and Quake 4 as well (though those are hard locekd to 62fps)?

With the 8800GT I also had similar experiences with Max Payne 2 which was frame rate locked locked to the screen refresh rate. Most games however were unlocked if I unlocked them in the game options and the driver left it to the game.

Got an auto switching KVM for my really old retro hardware but the KVM for my more "modern" hardware is not auto switching so you need to use buttons to switch to the correct pc. It has only 4 inputs but even then I can't seem to remember it so made some labels.

My KVS has labels saying 286, 386, 486, and Pentium.

In all seriousness, those are funky and cool labels.

I've written a thread on putting old PC's on the internet, but that's about it. I've been sick on and off for about 4-5 weeks now, including a bad fight with COVID despite doing everything right.....I wish I was better and had more time for stuff as of late.

I assembled my 645DX project yesturday (it's a LianLi's case MB drawer).
Obviously some wiring is missing in the picture (sound module cable to the board, USB card's connectors, etc).
I like how the card's end and the motherboard's form factor meet.
I like how the two Voodoo 5's fans are pleasanty supplied on two fan connectors that are "sticked" together. Just like it was designed for it.
I'll change some details over time (one more Corsair RAM module, another (thicker) fan on top of the Asus full copper heatsink (preferably with a Soyo sticker if I find one).

I found the Voodoo 5 to be very picky. Had to let the screw not very tightly secured and move it for some millimetters and try many times doing this to get it to work. Other AGP cards worked easily on this board (both 3.3V and 1.5V).
Is it something common with the Voodoo 5 ?

At a guess I would think that it doesnt like the AGP slot, the Voodoo5 5500 really saw the AGP slot as a souped up PCI slot and thus didn't use much if any of the advanced AGP features and was limited to AGP 2x (Really 1x). Not even sure how its running on a more modern AGP slot but I'm guessing its a Universal slot, you might want to consider a different GPU and move that old girl back to a true AGP 2x board where it'll likely be happier.

But to answer your direct question, Yes its the Voodoo5 5500 its a picky bitch even when working correctly and is well known for having AGP issues/conflicts.