Debatable. I always preferred the crisp look of the software renderer to the washed out GLQuake. Same with Quake 2. I think it because textures back then were too low resolution so filtering just makes them look muddy.
It’s also because the VGA signal quality from the 3dfx Voodoo wasn’t very good.
It didn’t have a traditional 2D graphics core at all, so it needed another graphics card for rendering the desktop (any non-accelerated apps really), and this was connected to the Voodoo using VGA passthrough. There was a noticeable image quality loss from this arrangement.
A Matrox card would give you crisp VGA with nice saturation, but the 3D acceleration was nearly worthless. Choices…
I really disagree. There were some nice Matrox cards. They weren't as good at 3d as 3DFX but for the time they really improved gaming. I developed Battlezone on G200. In those days we tried to have everyone have a different graphics card because the companies would just give them to us and we wanted to work with every card.
Matrox had great hardware, but the software drivers took too long to catch up. I was on the OpenGL team and my life's mission was to get Quake running as fast as the G200 and G400 was capable of. We finally caught up and got parity with Nvidia's TNT2, and then bam, they released the GeForce 256 series, and it was curtains for Matrox because their next gen hardware wasn't ready yet.
I agree that the washed-out textures haven’t aged well.
But at the time, not having pixelated textures was the first thing people looked at when judging graphics quality. I remember that being a HUGE selling point of the N64 and something that made the console look almost a generation ahead of the PlayStation and Sega Saturn to kids back then.
Today, I think I prefer the PSX look, thoug. Maybe with Z-buffer correction to avoid the warped textures of the PlayStation.
Might have also been one of those things that looked better on the 14-15" CRTs of the time vs crisp high-res flat panels of today. They were blurry enough that 640x480 was "high resolution" (I remember not being able to easily see the pixels at 800x600 on a 14" CRT unless I came super close to the monitor).
Even today I think a lot of Doom clones look better (or more nostalgic) with software rendering and texture mapping rather than OpenGL. There's an intensity of saturation to the colors that's lost. Fireblu is never quite so eye burning as when it's in software.
You can 'fix' the texture filtering to nearest neighbour in hl by adding the following to userconfig.cfg (should be in a directory called 'valve' in the game's root directory):
gl_texturemode GL_NEAREST_MIPMAP_LINEAR
gl_ansio "0"
gl_round_down "0"
Or just entering those lines in the consoe, preceded by 'set'
I came here to comment similarly, the lower pixelated software rendered Quake seems to work well with the textures. They have a bumpmappy fuzzy feel that gets lost with the sharp corners everything is super flat texture mapped and filtered version that one got from the 3d accelerators of the time. I guess my brain just adds context to the low res images.
Before unreal, I had a s3-virge for 2d and a powerVR 3d accelerator pair, and I was always flipping between software, virge and powerVR depending on game. Which at the time were largely hexen/heretic. The powerVR was higher resolution and clean/sterile but never seemed like a lot better experience.
But then there was unreal, the first game I think was absolutely better on an accelerator (voodoo2 in my case). Its also pretty much the last of the serious software renderers and outside of the voodoo's definitely did a better job with software lighting /texture mapping/etc than any of the previous (affordable) accelerators. Which is why I ended up finally replacing the powerVR with the voodoo2. The results were 'unreal'. Some of that might just be bias, I played insane amounts of doom/etc but never really got into quake. Quake just seemed like doom rehashed to me, so I was off playing warcraft/diablo/hexen/etc.
And frankly, outside of FEAR, I stopped playing 1st person shooter games for 20 years, the graphics improvements were so incremental, I just kept seeing flat low polygon models everywhere. And I don't think that looks good. Even after all the tessellation/bump mapping/endless tricks I kept seeing frames where I could play "count how many polygons are onscreen right now" games. Its gotten better the past few years, particularly some of the lighting, at least the screenshots/cut scenes are no longer obviously not in game rendering. The UE5 demo is slowly becoming reality in actual games, so maybe its time to revisit a few of them.
Debatable. I always preferred the crisp look of the software renderer to the washed out GLQuake. Same with Quake 2. I think it because textures back then were too low resolution so filtering just makes them look muddy.