P96 3.6.3 Released: major Amiga RTG graphics update adds new drivers and fixes

P96 3.6.3 has arrived as another focused update for one of the most important pieces of software in the modern Amiga graphics stack. It fixes awkward graphics issues, improves support for several classic and semi-modern graphics cards, adds new driver capabilities and smooths out some edge cases that only appear when old hardware, clever expansion cards and three decades of Amiga habits meet in the same machine. For anyone running RTG graphics on an Amiga, this is the sort of maintenance release that matters. It deals with blitting, overlays, screen dragging, Matrox behaviour, Radeon initialisation, VA2000 switching, Permedia2 support and even some colour-space problems that were hiding in the machinery like a sock behind a washing machine. None of that sounds glamorous at first glance, but RTG on the Amiga has always been about making the machine feel less trapped by its original chipset limits.

A short P96 history without the dust cover

P96 is an RTG software package for the Amiga. RTG stands for retargetable graphics, which in practical terms means the Amiga can use graphics hardware beyond the original native chipset modes. That is what allows Workbench and applications to run on graphics cards with higher resolutions, more colours and better performance than a stock Amiga display setup can offer. For many expanded Amiga systems, P96 is not an optional luxury. It is the bridge between classic software expectations and the graphics cards that helped keep the platform alive long after the mainstream market had moved on.

Older Amiga users may remember that P96 used to carry the name Picasso96. Because the family of Pablo Picasso does not like that name being used in this context, the package is now referred to as P96, and that is the name that should be used from here onwards. The software is commercial, copyrighted and still actively maintained. New package redistribution is not allowed, and users who want the current version need a proper licence. The old P96 V2.0 shareware archive keeps its special status, meaning redistribution remains legal as long as no money is charged and no registration claim is made, but the current commercial package is a different matter entirely.

There is also some history around AmigaOS 4. A pre-existing arrangement involving the original P96 authors Tobias Abt and Alexander Kneer and Hyperion became a co-ownership situation for the PPC port. Hyperion owns and maintains that PPC version as part of AmigaOS 4 and successors, so users of Hyperion’s PPC-based operating system do not need a separate P96 licence for that included port. For classic Amiga users and many emulator setups, however, licensing still matters. In a scene where copying old floppies was once treated as a neighbourhood sport, that distinction is worth making clearly.

What the P96 3.6.3 update is really about

The 3.6.3 update is mainly a practical release. It does not revolve around one giant headline feature. Instead, it gathers many smaller fixes and improvements that affect reliability, card support and visual correctness. The result is a broader and tidier RTG layer, particularly for users with Matrox, Radeon, RageXL, Voodoo, VA2000, 3DLabs Permedia2, S3Virge and newer niche hardware such as the PCMCIA-based pVision card.

One useful change concerns the AmigaVideo environment setting. P96 can now use an OFF setting to disable sync impulses from the Amiga video output. In plain English, this can help a DPMS-compatible monitor connected directly to the Amiga video output power down or switch to another input, such as the graphics card output. For users juggling native Amiga video and RTG output on the same desk, this is a welcome bit of housekeeping. It is not as flashy as a new graphics demo, but it may stop your monitor from acting like it is trying to solve a family dispute between two video signals.

The same AmigaVideo handling also recognises 31kHz, which keeps an active display at a 31kHz refresh and may help improve audio playback performance in certain situations. That is a very Amiga kind of improvement: a graphics setting that can matter to audio behaviour because the machine is a tightly packed collection of clever compromises.

Blitting fixes for fewer visual surprises

A sizeable part of P96 3.6.3 focuses on blitting. Blitting is the copying and manipulation of bitmap data, and on the Amiga it has always been central to how responsive and smooth the graphical environment feels. When blitting goes wrong, the symptoms can be ugly: broken rendering, incorrect masks, damaged screen areas or output that looks as though the machine briefly forgot what pixels are for.

One fix addresses BltMaskRastPortBitMap, where the offset into the mask could be computed incorrectly if the source was interleaved and the source Y position was not zero. That could cause bad rendering when interleaved bitmaps were blitted with non-zero offsets. This is exactly the sort of bug that only a graphics subsystem can produce: technically specific, visually annoying and almost impossible to explain to someone who just wants the window to look normal.

Another long-standing problem involved CPU blitting of overlapping bitmaps with different base addresses. The copy direction could previously be chosen only by coordinates instead of also considering the actual addresses. That could lead to incorrect results. The fix is described as addressing a very old defect, and in Amiga terms that could mean anything from last decade to somewhere shortly after dinosaurs discovered copper lists. Either way, old bugs being fixed is good news, especially in software that people still use on carefully maintained hardware.

There are also fixes for BlitTemplate in the Radeon and RageXL drivers, specifically when the bitplane width did not match its modulo value. The RageXL line drawer also now handles pattern offsets correctly, preventing line patterns from being aligned wrongly. These are not changes that every user will notice every morning, but they improve correctness in the places where graphics libraries and drivers do the quiet work.

Matrox support gets a careful polish

Matrox card support receives several improvements in P96 3.6.3, and this will matter to users who have built RTG systems around those boards. The update adds support for the Matrox G100 with SDRAM and fixes several areas of Matrox driver behaviour. Resolution changes, especially on MMS types, should behave more reliably. Tool type evaluation has been corrected, and the drivers now behave properly when interrupts are disabled with the INTERRUPT=NO tool type.

The update also improves how the VESA BIOS is evaluated on Matrox MMS cards. Previously, only the first of multiple chips was considered, which is not ideal when the hardware itself is more complex. Overlay image quality on Matrox cards has also been improved, giving users a better result where video overlay functions are involved.

These Matrox changes show the value of active maintenance. Amiga RTG is not just about supporting one perfect graphics card in one perfect setup. It is about dealing with the wonderfully untidy range of boards, bridges, chipsets and configurations that enthusiasts actually use. Somewhere in that chaos is a Matrox card that now behaves better, and its owner probably deserves a cup of coffee.

Radeon, RageXL and Voodoo colour handling gets cleaned up

P96 3.6.3 also tackles YUV-related problems across the RTG library and several drivers. The fallback functions for YUV422 rectangle fill were incorrect and could produce wrong colours. VUY422 modes were not mapped correctly for Voodoo4 and Voodoo5, while earlier Voodoo releases were already handled correctly. Documentation for 422 modes also had U and V interchanged, meaning the documentation and the actual behaviour did not line up.

The RGB-to-YUV colour conversion had the same U and V issue, but some driver implementations for Voodoo, Radeon and RageXL happened to swap the colours back into the correct order. That kind of accidental balance is charming in an old-computer way, but it is not ideal engineering. In 3.6.3 the core problem is fixed, and the affected drivers no longer need to compensate by swapping U and V themselves.

There is also a Radeon overlay fix for an off-by-one bounds-checking error when computing ideal filter coefficients in YUV mode. In cases where a YUV window was scaled, that error could cause distorted display. Off-by-one errors are the computer world’s equivalent of putting a shelf up one centimetre crooked. It might hold, but you will keep noticing it.

One particularly interesting Radeon fix concerns 68030 systems. Access to the Radeon VGA ROM could be unreliable on a 68030 when data-cache bursts were enabled. The card initialisation code now temporarily disables the data cache and data-cache bursting to improve reliability. That is a neat example of the kind of work involved in keeping RTG alive on real classic hardware, where old CPUs, caches and PCI-era cards may all have to cooperate despite never having signed a friendship agreement.

VA2000 users get cleaner switching and a new option

The VA2000 driver also receives attention. Switching between a native Amiga screen and an RTG screen did not always set the native screen clock correctly, and that has now been fixed. For users moving between chipset output and graphics-card-style output, this should improve reliability.

The update also adds a new CAPTUREPREFETCH tool type for the VA2000 driver. This allows the memory prefetch value for the flicker fixer screen to be set. Most users should not need to touch it, because the default is intended to work correctly for the screen size used by the flicker fixer output. Still, having the option available gives advanced users another tuning point when a particular setup demands it.

The VA2000 monitor switch function has also been corrected so it returns the previous switch state as it should. Again, this is a small but important correctness fix. Software ecosystems become stable not only because huge features work, but because small functions do exactly what they claim to do.

A new 3DLabs Permedia2 driver is the headline for many users

The most substantial driver news in P96 3.6.3 is the completely new 3DLabs Permedia2 driver. This is likely to attract attention from users with CyberVisionPPC-style hardware or PCI cards based on the same chip family. The new driver adds doublescan modes with a Workbench-friendly 2:1 aspect ratio, more careful memory testing and support for PCI cards based on the same chip.

It also supports bitmaps wider than 2048 pixels and allows blitting within them, with the hardware blitter operated line by line where necessary. Patterned line drawing is now supported, along with all possible line draw modes. That should make the driver more complete and useful across a wider range of software behaviour.

The driver has also been prepared for PCI cards using SDRAM rather than SGRAM. At this stage, SGRAM is supported, while SDRAM-related switches are present but untested. Clock handling has been cleaned up as well. Instead of hundreds of barely different clock settings, the new rates are spaced at least five percent apart, making them meaningfully distinguishable rather than a menu of almost identical choices. It is like replacing a restaurant menu with 600 kinds of toast with one that simply asks whether you want breakfast.

Perhaps the most welcome Permedia2 improvement is screen dragging support. For that, users need to make sure INTERRUPT=Yes is enabled. Screen dragging is one of those classic Amiga features that still feels distinctive, and seeing it supported on more RTG hardware adds a little more personality to expanded systems.

pVision joins the supported hardware list

P96 3.6.3 also adds support for Oliver Achten’s PCMCIA-based pVision card. This is an interesting addition because it broadens P96 support beyond the more familiar graphics-card routes. The pVision card includes a full blitter, a sprite and support for screen dragging with full mode mixing.

That combination is notable. It means the card is not treated as a simple framebuffer curiosity, but as hardware with meaningful acceleration and classic Amiga display behaviour. The Amiga community has a long tradition of turning unlikely expansion paths into serious tools, and PCMCIA graphics support fits that tradition nicely. Somewhere, an old laptop slot just stood up a little straighter.

Smaller fixes that still matter

Several additional changes round out the release. P96Prefs no longer warns about untested video modes if those modes are disabled. It now warns only about untested modes that are actually enabled, which is far more sensible. Users do not need software shouting about settings they are not using. We get enough of that from operating systems with update reminders.

P96ScreenCx is now more careful when fitting boards to icons and is aware of multiple boards of the same type. The maximum memory clock of the S3Virge has been raised to 85 MHz. There is also a fix for cases involving dragged screens on video cards without DAC switching or mode mixing, where background colours could be left uninitialised if the dragged foreground screen had a smaller colour map than the background screen.

These are the details that make P96 3.6.3 feel like a maintenance release built by people who know the messy reality of Amiga graphics hardware. It is not just one card, one driver, one machine. It is a web of native output, RTG output, overlays, screen dragging, interrupts, tool types, old CPUs, PCI bridges, Zorro boards and user preferences that sometimes look innocent until they ruin your afternoon.

Why P96 3.6.3 matters for Amiga users

P96 3.6.3 matters because classic Amiga computing is increasingly about stability, compatibility and preserving real usability on systems that have grown far beyond their original design. A modernised Amiga setup may include accelerators, graphics cards, scan doublers, PCI bridges, emulation layers or boutique new hardware. The more varied those setups become, the more important the RTG layer becomes.

This update improves that layer in practical ways. It expands hardware support, fixes rendering defects, improves driver correctness and adds options that make mixed native and RTG setups easier to manage. For some users, the headline will be the new Permedia2 driver. For others, it will be Matrox fixes, Radeon 68030 reliability, VA2000 improvements or pVision support. For many, it will simply mean fewer strange visual issues and a system that feels a bit more predictable.

That is a good result. The Amiga world does not always need fireworks. Sometimes it needs a driver update that fixes a very old blitting bug, tells a monitor when to stop listening to the wrong signal and gives a few graphics cards a better day at work. P96 3.6.3 does exactly that, and for RTG users, that is more than enough reason to pay attention.