First post, by leileilol
- Rank
- l33t++
Has anyone else here attempted such feat? Is it even feasible? Are third-party drivers a bad idea for any possible modern CPU function overhead taking over?
Has anyone else here attempted such feat? Is it even feasible? Are third-party drivers a bad idea for any possible modern CPU function overhead taking over?
I've used a Voodoo card in a 486 and it worked just fine.
i'm using a voodoo2 in one right now and i'm experiencing slow performance, but i blame my sound card
(CS4248)
I still wonder why everyone advised a Pentium was required for a voodoo card to function, though.
Many years ago I used an Orchid Righteous Voodoo1 card in a 486 dx2-66 machine without any problems. I don't remember it being particularly fast, but I didn't think (at the time) that there was anything wrong with it.
The pre-GPU cards accelerated only a part of the rendering pipeline and they benefit a lot from faster CPU's. This is true for the Voodoo - I just checked a review of the Monster 3D and they recommend at least P100.
Ya the same goes with some games that I played on My old Amd 486 DX120 that required a Pentium 90 or better the game ran fine but it was really slow.
really slow? i'm having happy times playing quake2 on 486 😀
can't play ff7 though because ff7config uses a tls instruction to get cpu speed and because of that it crashes (wcpuid, cpuz and dxdiag also crash for doing that)
I used a nVidia RIVA 128 PCI card and it worked just fine on a 486. Was VESA 3.0 compatible too. I miss that card and I am sorry I gave it away.
Y2K box: AMD Athlon K75 (second generation slot A)@700, ASUS K7M motherboard, 256 MB SDRAM, ATI Radeon 7500+2xVoodoo2 in SLI, SB Live! 5.1, VIA USB 2.0 PCI card, 40 GB Seagate HDD.
WIP: external midi module based on NEC wavetable (Yamaha clone)
wrote:Has anyone else here attempted such feat? Is it even feasible? Are third-party drivers a bad idea for any possible modern CPU function overhead taking over?
Ran a Voodoo3 in a 486 once. I had to go to some older drivers because they'd apparently started using Pentium instructions in later versions. Proved this by swapping in a Pentium Overdrive and seeing BSODs go away.
Even a 160 Mhz 486 is too slow to really move that 3D code though so even Jedi Knight was barely playable. It was doing probably 10-15 fps. The Pentium Overdrive overclocked to 100 did the trick though! Well, as much as it can with a 32-bit, 40MHz bus with 70ns FPM RAM and retro direct mapped L2 cache that doesn't work well with the PODP5V because it was a compatibility nightmare.
May I ask something?Why can't the 486's do justice with 3D?
Pentium was required for 3D but why?Is it because of the different instructions?
wrote:May I ask something?Why can't the 486's do justice with 3D?
Pentium was required for 3D but why?Is it because of the different instructions?
486s tend to have poor FPUs, which are pretty damn essential for transforming of vertices in real time.
Though they're technically in the same boat as Cyrix6x86s and K5s having 3d cards
The Pentium "requirement" was only a mere guide for the sake of technical support so no stupid calls like
"Hi i've got a 486 66MHz your 3dfx Voodoo2 isn't giving me super performance like your ad said"
"*LAUGHS* HAAHHAHA NO CLOSED"
wrote:May I ask something?Why can't the 486's do justice with 3D?
Pentium was required for 3D but why?Is it because of the different instructions?
In early accelerated 3D games the instruction set did not play an important role, because Pentium instructions did not get widely used in games until much later.
There are other reasons why 486 class CPUs are generally too slow for high-resolution 3D games:
Don't forget that the 486 L2 cache, if there even is one, shares the processor bus with everything else. It's the same with a Pentium, but that 64-bit, 66MHz front-side bus gives a lot more headroom than the usual 25/33 MHz 32-bit 486 bus. And Pentiums usually have pipelined burst cache vs. plain direct mapped SRAM leading to better efficiency.
Part of the reason 486 PCI designs don't have much PCI bandwidth is because the entire system bus has (at most) slightly more bandwidth than 32-bit, 33 MHz PCI. But they also usually don't do DMA very well and that's really bad for a lot of PCI cards. Don't try a Verite in a 486 for sure.
Quake is actually designed around the pipelined Pentium FPU. I don't recall the details, but they coded the game to really leverage the Pentium architecture. It puts chips without pipelined FPUs at a major disadvantage (486, K5, K6, 6x86, Winchip, etc). Lots of people were up in arms over that, but it made sense for id to do it because Intel owned the market and they were pushing 3D tech to the limit and needed the speed. Assuming a 486 and Pentium are both equipped with fast vid cards and good mobos, the Pentium will be something like 2x faster per clock.
Of course, if you were really serious about Quake, you wanted a Pentium Pro. 😀
I ran Voodoo Banshee cards in DX4100 and DX266 486 machines. Both were fine. I also tried to run an Intel i740 with the DX4100, not so great. Driver errors and many render errors. Not surprising as the i740 was designed with the P2 in mind.
i740 was designed around AGP. It's a strange chip. The PCI version has to use a bridge chip to map the chip to PCI. And they have extra RAM onboard to make up for the lack of AGP texture memory.
Yup and the bridge chip was made by Real3D the only company that was making the PCI version themselves. It's really funny to read Intel interviews of the day where they basically deny any involvement in PCI versions of i740. It was a strange beast for sure, but when it did work some games, especially SiN, Quake 2 and Half Life, looked particularly beautiful.
half-life and 486s aren't a good combination btw
For sure, but by then I was i740'ing a P200MMX