sharangad wrote on 2026-09-04, 18:32:

vetz wrote on 2026-09-04, 18:08:

Speedy3D SDK, V1000 + V2200 BIOS source among more:
Massive archive of late-90s GPU stuff (NVIDIA, S3, 3DFX, Rendition)

Holy! Is this safe to use?

As far as I understand this comes from backup tapes connected to STB. The SDK's and source files are material that would typical be under NDA 30 years ago. I do not see much difference between this material and all the other SDKs, reference material and other similar things that already floats on the web and have been discussed here on VOGONS previously. The historical and educational value is very high for us interested in this hardware.

Tried to test: Scorched Planet
Failing:
installed game can't produce any sounds.
trying to reinstall it, fails to copy files, checked CD seems ok

nfm fixed somehow. will explain later

vetz wrote on 2026-09-04, 18:43:

sharangad wrote on 2026-09-04, 18:32:

vetz wrote on 2026-09-04, 18:08:

Speedy3D SDK, V1000 + V2200 BIOS source among more:
Massive archive of late-90s GPU stuff (NVIDIA, S3, 3DFX, Rendition)

Holy! Is this safe to use?

As far as I understand this comes from backup tapes connected to STB. The SDK's and source files are material that would typical be under NDA 30 years ago. I do not see much difference between this material and all the other SDKs, reference material and other similar things that already floats on the web and have been discussed here on VOGONS previously. The historical and educational value is very high for us interested in this hardware.

Several Vogons have said the same thing, so I guess it's safe.

RaVeN-05 wrote on 2026-09-08, 17:31:

Tried to test: Scorched Planet Failing: installed game can't produce any sounds. trying to reinstall it, fails to copy files, c […]
Show full quote

Tried to test: Scorched Planet
Failing:
installed game can't produce any sounds.
trying to reinstall it, fails to copy files, checked CD seems ok

nfm fixed somehow. will explain later

Ok. Scorched Planet doesn't work in DMA mode, only FIFO mode. Also, FIFO performance should be a bit better than before.

I'm testing the next release and RaVeN-05, you were right. With vsync turned off in Rebel Moon, mouse and keyboard input is ridiculously fast.

********************************************************
RReady Alpha (September 12, 2026, 1.0.50) Fully tested release
********************************************************
https://nirvtek.com/downloads/RReady.Alpha.20260912.001.7z
MD5: 146d38150392844b6bed7ad65431e74f
- vQuake 1 Radeon performance fix. This should improve front buffer/console rendering on Nvidia GPUs as well.
- Dosbox-Staging-Rendition initial CGL support. Currently restricted to the 3DBPCI game BATosh, 21stCDB demo and partial support for Actual Soccer 1 and Euro 96.
- Dosbox-Staging-Rendition ICR2 widescreen support
- Faster and cleaner Dosclient shutdown.
- Battle Arena Toshinden should render perfectly with both version of Dosbox
- Grand Prix Legends will shutdown cleanly after the August Windows updates with this release. Note: Nvidia users will have to set the “Vulkan/OpenGL present method” to “Prefer layer on DXGI compatible swapchain” in a custom game profile for GPL.
Should be live on the store soon.
******************************************************

ICR2 performance issues with RReady:

It looks like using high-poly carsets with 32/32 (cars ahead/behind visible) rendering causes the RReady to chock on a massive number of simple polygons. In the video rendering 32/32 with and without polygon rendering is compared on both the Australia and Detroit Downtown tracks.

https://youtu.be/BrNMx8kjqWE
(https://www.patreon.com/Rready/posts/icr2-ren … ntent=join_link)

A tweaked version of RReady with slightly higher framerates is available here:
https://youtube.com/shorts/dGHrX2SLaJs?feature=share
Also available on my Patreon:
https://www.patreon.com/Rready/posts/icr2-per … ntent=join_link

Feedback from laptops would be nice. It's not a massive uplift, so you've been warned.

LGR mentioned RReady:
https://youtu.be/o0wDkCp3Eu4?si=ouhpeLOA0_guY2mu

BEEN_Nath_58 wrote on 2026-10-07, 17:53:

LGR mentioned RReady:
https://youtu.be/o0wDkCp3Eu4?si=ouhpeLOA0_guY2mu

Thanks @been_nath58! I saw that last night. I never realized these cards were that rare, so rare that @lgr doesn't have one. So once again, many many thanks to @robertmo3 for gifting me a v1000 and v2100.

According to @lgr there's another rredline game, the knockoff of rise of the machines: cyber gladiators, which came with the Sierra demo disc. That might just be D3D though.

In other news there may be other ways to speed up rready with dos, especially vquake. The quake span renderer I think would work really well with a vertex buffer object. Icr2 has another round of optimizations coming up as well.

BTW, I should've mentioned that being featured on LGR is the highest point of this project!

I have another build for ICR2:
https://nirvtek.com/downloads/RReady.Alpha.20 … misationTest.7z
MD5: e5d507ecf225333ebb9793dd0c938bc7

The rendering defects should mostly be gone, but it should run faster with custom carsets. I'm not sure what the actual performance gain is either 5 fps on a Ryzen 5700G or in the case of a i7 2600/Radeon R7 200 25%. The audio pops should be gone as well, either that or my hearing isn't what it used to be.

Another ICR2 performance video, this time on more realistic hardware i7 2600/Radeon R7 250:
https://youtu.be/fgUwpnWb88A
https://www.patreon.com/Rready/posts/icr2-per … ntent=join_link

It's still not that great, but minimum frame rates have gone up.

00:00 Cars 95 - Australia Practice Run
00:24 Cars 95 - Detroit Downtown Replay
00:45 Cars 95 - Houston Practice Run
01:07 Cars95 - Houston Race
01:22 Indy2k08 - Australia Practice Run
01:56 Indy2k08 - Detroit Downtown Replay
02:19 Indy2k08 - Houston Practice Run
02:40 Indy2k08 - Houston Race
03:02 LOLA 1995 - Australia Practice Run
03:24 LOLA 1995 - Detroit Downtown Race
03:44 LOLA 1995 - Houston Practice Run
04:06 LOLA 1995 - Houston Race

Finally managed to get interrupts going with 3DBVLB Test3d.exe.

The next stage sends FIFO commands to the VLB.

(Update with video)
https://youtu.be/pKFTmHGkY20
Also available on my Patreon:
https://www.patreon.com/Rready/posts/3d-blast … ntent=join_link

Resolution change and mem writes have been detected.

Could someone with a VLB board post a video of Test3D.exe? It looks like it might be rendering a greyscale image and then possibly cycling the palette to switch to an RGB ramp.

https://youtu.be/18W_LBNg2_8?t=802

robertmo3 wrote on Today, 14:04:

https://youtu.be/18W_LBNg2_8?t=802

Thanks @robertmo3! Good old @LGR to the rescue.

I saw that video quite a while back but didn't remember that bit.

i have noticed you are not only not using default resources but also not even from the lists:
https://youtu.be/18W_LBNg2_8?t=723
i wonder if that won't make a problem later.
btw isn't 204 used by joystick?
0x200 - 0x207 Game Port / Joystick
isn't irq 9 (2) used by either intelligent mpu-401 or acpi?

robertmo3 wrote on Today, 17:48:

i have noticed you are not only not using default resources but also not even from the lists: https://youtu.be/18W_LBNg2_8?t=723 […]
Show full quote

i have noticed you are not only not using default resources but also not even from the lists:
https://youtu.be/18W_LBNg2_8?t=723
i wonder if that won't make a problem later.
btw isn't 204 used by joystick?
0x200 - 0x207 Game Port / Joystick
isn't irq 9 (2) used by either intelligent mpu-401 or acpi?

The mpu-401 shouldn't be a problem (I hope) since a Google Ai summary of two vogons posts says this :
MIDI IRQ Usage in DOS: In practice, many standard DOS games and MIDI applications actually do not require an active IRQ for intelligent-mode MPU-401 operation (relying on polled I/O at port 330h instead). However, if a driver or software explicitly binds to the IRQ, it must remain dedicated.

But this could be wrong.

I based the IO resources on what one of goneau's 486 machines reported (though I might've changed the irq) . Speedy3D support also configures irq 9 under dosboxen rendition (based on RaVeN-05's logs), though I'm not actually using it. The irq can be set to trigger when the verité FIFO queue runs low. I took out support for this because I wasn't sure how to correctly trigger an interrupt. I do know how to do it now.

VLB devices report their IO ports via device detection, so I should be able to change that quite easily.

Dosbox uses port 0x201 for its joystick emulation. So port 0x204 should be fine. I used it because it was the base port for PnP queries and I think it's in the allowed range, as in the Pnp spec allows it to be 0x204. Test3d doesn't perform direct port io, using memory mapped io instead.