WEBVTT

00:00:00.000 --> 00:00:08.460
the owner of this GPU was recently offered fifteen thousand dollars by a

00:00:05.400 --> 00:00:10.980
collector why I'll tell you why because

00:00:08.460 --> 00:00:17.160
it's a ridiculously rare artifact from the early days of 3D gaming the bad news

00:00:13.320 --> 00:00:20.460
is the collector wasn't me

00:00:17.160 --> 00:00:22.699
good news is that I got my hands on it

00:00:20.460 --> 00:00:22.699
anyway

00:00:23.939 --> 00:00:31.199
the deal fell through the owner Ross agreed to drive it up to Canada for us

00:00:28.380 --> 00:00:39.000
to show it off and as luck would have it it actually works this is the real deal

00:00:35.940 --> 00:00:41.640
guys a very late stage prototype of what

00:00:39.000 --> 00:00:47.820
would have been 3dfx's final Flagship card the voodoo 5 6000 and this one has

00:00:45.420 --> 00:00:53.219
even been reworked by a former 3dfx engineer to reach what we believe would

00:00:49.739 --> 00:00:56.280
have been the final spec could this

00:00:53.219 --> 00:00:58.140
Beast of a quad GPU card of saved 3dfx

00:00:56.280 --> 00:01:03.180
and changed the face of gaming as we know it probably not but that doesn't mean that

00:01:01.440 --> 00:01:07.380
we can't fire it up and relive our wildest dreams from a parallel universe

00:01:05.280 --> 00:01:12.900
and fire up this message from our sponsor iFixit learn more about the

00:01:10.140 --> 00:01:17.280
extremely portable and affordable minnow and moire sets and how they can make

00:01:14.939 --> 00:01:21.360
your repair life easier at the end of this video what makes this card really

00:01:19.140 --> 00:01:26.280
special is that it's straight from 3dfx themselves making it the best Glide

00:01:23.820 --> 00:01:31.320
gaming experience on the planet I mean yes we showed recently that you can

00:01:28.200 --> 00:01:33.479
emulate 3dfx Hardware on a modern PC but

00:01:31.320 --> 00:01:39.200
so far you can only go up to the voodoo 3. you want to go faster

00:01:35.820 --> 00:01:39.200
you're gonna need bare metal

00:01:39.420 --> 00:01:45.540
Let's Ride not being a 3dfx historian

00:01:42.659 --> 00:01:53.040
myself I wasn't 100 sure what to expect I can tell you that I didn't expect an

00:01:48.240 --> 00:01:53.040
external DC Barrel Jack though okay

00:01:53.460 --> 00:02:00.299
sure rather than just putting a bigger cooler on the front you just add these little

00:01:58.320 --> 00:02:04.439
heatsinks to the back of where the dies are soldered I guess

00:02:02.640 --> 00:02:08.280
man I wish we'd done this video even like six months ago you guys would have

00:02:06.659 --> 00:02:11.220
been so impressed by the size of this card but now that the 490 exists it's

00:02:10.080 --> 00:02:18.780
like make up for it with your personality

00:02:14.700 --> 00:02:21.319
here we go it's loud it's right next to

00:02:18.780 --> 00:02:29.239
my mic wish I'd thought of that hey nice Athlon XP 2200 plus oh no

00:02:29.580 --> 00:02:34.860
is everything working right now like I it captures working and everything

00:02:32.940 --> 00:02:39.900
capture is working it's so exciting World of Warcraft

00:02:37.200 --> 00:02:46.019
1.12.1 I mean surely we can't play WoW on this

00:02:43.860 --> 00:02:49.500
yeah I didn't like that what should we play half-life

00:02:47.459 --> 00:02:56.340
if you have a 3D card that incorporates the voodoo 2 or Banshee chip sets you

00:02:52.620 --> 00:02:59.879
will want 3dfx mini driver let's go

00:02:56.340 --> 00:03:04.200
that is some mighty fine screen tearing

00:02:59.879 --> 00:03:06.900
the latency is great and this is Butters

00:03:04.200 --> 00:03:13.080
move if I had to guess and this is going to be pretty tough because I'm on a 60hz

00:03:08.580 --> 00:03:16.260
display I would say probably 100 to 150

00:03:13.080 --> 00:03:21.620
frames per second like it feels very

00:03:16.260 --> 00:03:21.620
very low oh get back here

00:03:22.560 --> 00:03:28.080
man imagine having this kind of money for a computer back then and like six

00:03:26.760 --> 00:03:35.879
hundred dollars for one of these things 600 GPU but it runs like

00:03:33.120 --> 00:03:40.319
I just at now it's worth noting that this is a game from 1998. yeah right so

00:03:38.700 --> 00:03:45.720
this game came out a couple of years before this even had reached the

00:03:42.780 --> 00:03:51.480
Prototype whoa hey hey stop that yeah hence the voodoo 2 and Voodoo Banshee

00:03:48.000 --> 00:03:54.080
notes in the option screen there ah okay

00:03:51.480 --> 00:03:58.980
this thing absolutely blows away what I would have been using

00:03:56.580 --> 00:04:03.060
at the time I know obviously it looks like crap by today's standards but I

00:04:01.500 --> 00:04:08.580
wouldn't have been able to believe my eyes if I had seen this back in 2000.

00:04:05.700 --> 00:04:12.540
can I just say how painless this was for a prototype graphics card I don't think

00:04:11.099 --> 00:04:19.620
it was very painless to get it working to begin with well yeah but it's painless for me now

00:04:15.919 --> 00:04:22.199
yeah so go to console and then time demo

00:04:19.620 --> 00:04:27.660
as one word space blowout okay here we go

00:04:24.960 --> 00:04:31.740
oh look at it go we're gonna find out what the FPS was and this is a much more

00:04:30.360 --> 00:04:38.400
demanding scenario than what I was looking at and I feel like I

00:04:33.479 --> 00:04:40.380
underestimated it 119 frames per second

00:04:38.400 --> 00:04:45.419
oh okay let's try tribes a voodoo card has

00:04:43.380 --> 00:04:50.520
been detected If you experience Oddities you should try the wicked GL drivers at

00:04:47.520 --> 00:04:50.520
wicked3d.com

00:04:53.040 --> 00:05:01.560
oh so first of all 10 24 by 768 right

00:04:56.940 --> 00:05:04.320
yep bit depth 16 right can be 16 or 32

00:05:01.560 --> 00:05:08.639
well let's go 32 let's go okay I mean obviously we're just gonna

00:05:06.479 --> 00:05:12.620
crank everything to high right what could go wrong let's see how many FPS's

00:05:11.040 --> 00:05:17.240
we get here plenty we're getting shot down boys

00:05:17.639 --> 00:05:22.740
oh wait did I actually die seriously

00:05:21.120 --> 00:05:27.419
okay how do I how do I make this go away I

00:05:25.979 --> 00:05:33.000
think it's just paused you while it tells you about everything ah

00:05:30.419 --> 00:05:39.360
come on oh hey here we go we got something to fight play Master let's go

00:05:35.960 --> 00:05:41.840
what the what the crap am I even firing

00:05:39.360 --> 00:05:47.160
it's a bouncy ball oh my goodness bouncing up the thing oh I got one it's

00:05:44.580 --> 00:05:51.720
running shockingly well that is to say the performance is great on that era

00:05:49.680 --> 00:05:58.080
games and kind of problem free

00:05:53.880 --> 00:06:00.800
as far as I can tell for flipping gpus

00:05:58.080 --> 00:06:06.960
uh um Benchmark I mean she's running it

00:06:04.680 --> 00:06:10.620
Yep this is about what I remember this looking like

00:06:08.699 --> 00:06:16.320
I remember this second run through oops uh-oh okay mostly Problem free

00:06:13.860 --> 00:06:19.800
I thought this was pulled from an actual Matrix game at the time

00:06:18.000 --> 00:06:24.240
I don't think are you sure I don't you're making me doubt myself

00:06:22.560 --> 00:06:31.199
I mean the reflections are all looking okay kinda nope you're totally right

00:06:27.840 --> 00:06:32.819
it's just a matrix inspired demo so why

00:06:31.199 --> 00:06:37.259
Shazam you okay hold on a second did we

00:06:35.100 --> 00:06:43.020
faster and faster than a Ferrari and more reliable that that PC World ad in

00:06:40.560 --> 00:06:48.180
the bottom love it I remember this running a lot better on

00:06:45.539 --> 00:06:52.139
Hardware a couple of years later oh boy yikes I think software rendering would

00:06:50.460 --> 00:06:58.380
be actually faster oops no Hardware support rip

00:06:55.620 --> 00:07:02.039
we'll talk about that later oops no Hardware support just doesn't do bump

00:07:00.180 --> 00:07:10.319
mapping at all we'll talk about that later or not yet pixel Shader no

00:07:05.100 --> 00:07:11.940
Hardware support rip hey 26 47 3D marks

00:07:10.319 --> 00:07:16.919
let's go there it is yep that's why it was so

00:07:14.819 --> 00:07:21.720
slow if you ran 2000 it would actually like probably be pretty good basically

00:07:19.560 --> 00:07:26.220
NVIDIA having that was a killer app for them what a cool trip to 3D gaming

00:07:23.940 --> 00:07:31.860
Christmas past but believe it or not this isn't the only time machine there

00:07:29.160 --> 00:07:37.020
are other brand new voodoo 5 6000s that are available today for a fraction of

00:07:34.500 --> 00:07:42.180
the cost of this one Russian modern Anthony zx64 has made a name for himself

00:07:39.539 --> 00:07:47.400
refurbishing and modding vintage gpus and the voodoo 5 6000 rebuild is his

00:07:45.000 --> 00:07:51.780
crowning achievement not only did he build his reproduction with only a

00:07:49.440 --> 00:07:57.000
non-working prototype for reference he actually improved the design by doubling

00:07:53.880 --> 00:07:58.680
the vram to 256 megabytes though it

00:07:57.000 --> 00:08:03.180
should be noted that Anthony recommends using the onboard jumper to fall back to

00:08:00.660 --> 00:08:06.900
128 Megs to improve compatibility with older games

00:08:04.380 --> 00:08:11.580
genuine 3dfx Hardware or not though this thing was wicked fast at the time even

00:08:09.419 --> 00:08:16.259
if the truth is that the voodoo 56000 was destined for failure before it even

00:08:13.500 --> 00:08:20.759
got to this prototyping phase so here's the thing 3dfx had a wealth of

00:08:18.599 --> 00:08:26.340
experience using multiple gpus together to make FPS number go up but the fact is

00:08:23.819 --> 00:08:32.820
the single GPU competition was getting good really good and in 2001 the GeForce

00:08:30.180 --> 00:08:36.899
2 and Radeon 8000 showed up with both strong performance and a new technology

00:08:35.459 --> 00:08:43.260
that would radically change the graphical Fidelity of video games shaders

00:08:40.080 --> 00:08:46.140
this meanwhile lack shaders and was

00:08:43.260 --> 00:08:50.339
expected to cost an estimated 600 US dollars at retail Unthinkable at the

00:08:48.600 --> 00:08:56.880
time even if it sounds pretty good today so as amazing as it was that's a turd

00:08:53.640 --> 00:08:59.519
that absolutely will not polish

00:08:56.880 --> 00:09:04.260
how did 3dfx get to this point they were the Undisputed Kings of the 3D gaming

00:09:01.680 --> 00:09:08.640
space from 1996 onward to understand that we have to understand why they were

00:09:06.899 --> 00:09:14.940
the Kings because it's not like they were the first incumbents like matrox

00:09:11.220 --> 00:09:17.040
ATI and S3 had their own 3D boards and

00:09:14.940 --> 00:09:23.100
Renditions Veritas series came out at around the same time and unlike 3dfx's

00:09:20.279 --> 00:09:29.600
early cards none of those required a separate video card just to run the 2D

00:09:25.620 --> 00:09:32.399
desktop two gpus one for 3D one for 2D

00:09:29.600 --> 00:09:37.680
3dfx actually came into this Market with a huge disadvantage but what they had

00:09:34.920 --> 00:09:41.940
was as perfect a blend of features and performance as you could possibly expect

00:09:39.540 --> 00:09:46.019
at the time not only did the voodoo 1 sport a faster fill rate than its

00:09:43.800 --> 00:09:51.060
competitors it had Hardware support for Z buffer depth information bilinear

00:09:48.300 --> 00:09:55.920
texture filtering anti-aliasing and it boasted higher resolution higher color

00:09:53.279 --> 00:10:01.260
output these are all things that other cards either relied on the CPU to do for

00:09:58.800 --> 00:10:05.700
them or just straight up couldn't do at all which meant that you could run a

00:10:03.360 --> 00:10:11.100
much lower tier processor like even a 486 and still get an enjoyable 3D

00:10:08.580 --> 00:10:16.740
experience I mean to be clear unlike today a dedicated 3D accelerator wasn't

00:10:13.620 --> 00:10:18.540
necessary to play most games but if you

00:10:16.740 --> 00:10:25.800
used full software rendering you'd be stuck at super low video quality I'm

00:10:21.000 --> 00:10:27.000
talking 320 by 240 256 colors if you

00:10:25.800 --> 00:10:31.440
were lucky Voodoo to the rescue then and a big part

00:10:29.580 --> 00:10:35.820
of their special sauce was the way that they wrapped these advantages into a

00:10:33.600 --> 00:10:41.459
relatively easy to program package with the Glide API remember guys this is in

00:10:38.459 --> 00:10:44.339
the days before direct 3D or widespread

00:10:41.459 --> 00:10:49.440
opengl support so each vendor had their own API that game developers needed to

00:10:46.800 --> 00:10:54.660
support but Glide was surprisingly similar to opengl in both syntax and

00:10:52.260 --> 00:10:59.040
feature set limited though it was so if a game developer had to program in

00:10:56.459 --> 00:11:04.740
support for each vendor's API well they were far more likely to support Glide

00:11:00.779 --> 00:11:07.440
over say s 3D or Redline this almost

00:11:04.740 --> 00:11:12.480
immediately gave 3dfx a special place in the hearts of Gamers even if the API was

00:11:10.079 --> 00:11:16.920
proprietary 3dfx then subsequently enjoyed a virtual monopoly in the 3D

00:11:15.120 --> 00:11:22.200
gaming space while the competition either made expensive gambles or sat

00:11:20.339 --> 00:11:26.640
with their development effectively on pause while they waited for the first

00:11:23.760 --> 00:11:30.959
direct 3D specification the nv1 for example from NVIDIA used a wildly

00:11:29.160 --> 00:11:35.760
different rendering mode than direct 3D would eventually use forcing NVIDIA to

00:11:33.600 --> 00:11:40.980
scrap their nv2 and then start playing catch-up but while 3dfx became well

00:11:39.000 --> 00:11:46.680
known for their robust feature set easy development and their multi-GPU scanline

00:11:44.399 --> 00:11:51.180
interleave or SLI which allowed for higher frame rates and resolutions than

00:11:48.600 --> 00:11:55.200
anyone else success seemed like a paralyzing Force for them and by the

00:11:53.100 --> 00:12:00.839
time of the voodoo 3 things started to take a turn with direct 3D and clearer

00:11:58.500 --> 00:12:06.360
opengl specifications to Target the competition was catching up and fast and

00:12:03.600 --> 00:12:12.660
in some cases even surpassing 3dfx's offerings for example NVIDIA's Riva tnt2

00:12:09.899 --> 00:12:17.399
brought competitive performance 32-bit color rendering and maximum texture

00:12:15.120 --> 00:12:22.260
sizes that were eight times larger than the voodoo 3 which could still only

00:12:19.620 --> 00:12:27.959
display 16-bit color in game meanwhile 3dfx rather than investing in GPU r d

00:12:25.320 --> 00:12:31.440
invested in a Litany of side projects and attempts to Stave off the

00:12:29.279 --> 00:12:35.940
competition the failed bid to provide chips for the Sega Dreamcast was a major

00:12:33.720 --> 00:12:39.899
blow that ended in a lawsuit and while the terms weren't disclosed Sega had

00:12:38.160 --> 00:12:45.000
intended to purchase the rights to use 3dfx's technology so as to prevent the

00:12:42.720 --> 00:12:48.540
competition from being able to and it's likely they got their way you can get

00:12:46.920 --> 00:12:53.160
your way by the way with a desk pad from ltdstore.com same price any size and now

00:12:51.720 --> 00:12:58.079
available in our stealthy circuit pattern and while they didn't know it

00:12:55.200 --> 00:13:03.060
then 3dfx's 1998 purchase of their biggest board partner stb systems ended

00:13:00.779 --> 00:13:07.500
up being not the monopolistic takeover of the market that many feared but

00:13:04.980 --> 00:13:12.300
rather the beginning of their end in effect it changed 3dfx's relationship

00:13:10.019 --> 00:13:17.040
with their board Partners from chipset provider partner to competitor which

00:13:15.480 --> 00:13:22.680
incentivized their Partners to support the real competition primarily NVIDIA

00:13:19.980 --> 00:13:27.240
and ATI whose rapid Hardware development Cycles became popular with system

00:13:24.240 --> 00:13:29.160
integrators like HP and Dell which 3 3D

00:13:27.240 --> 00:13:33.360
effects had no interest in catering to so without chips going into game

00:13:31.380 --> 00:13:37.980
consoles and with their board partners mutinying and taking over the system

00:13:35.100 --> 00:13:41.899
space 3dfx was left in the unenviable position of relying on only the

00:13:40.079 --> 00:13:46.860
Enthusiast segment as their customers and better options were now beginning to

00:13:44.700 --> 00:13:52.019
emerge for them as well perhaps seeing the writing on the wall 3dfx purchased

00:13:48.899 --> 00:13:53.940
gigapixel in mid-2000 after that company

00:13:52.019 --> 00:13:59.040
lost its bid to provide chips for the original Xbox the idea it seems was to

00:13:56.940 --> 00:14:04.500
finally get their next-gen GPU Rampage across the finish line that GPU had been

00:14:01.560 --> 00:14:08.100
in development since 1998 and had it released when 3dfx would have liked it

00:14:06.180 --> 00:14:12.120
to it's possible that they could have stayed competitive with the GeForce 2

00:14:09.839 --> 00:14:16.260
that they were staring down in 2000 unfortunately for them their creditors

00:14:14.279 --> 00:14:21.060
were fed up by Year's End voting for bankruptcy and liquidating the company

00:14:18.380 --> 00:14:24.720
NVIDIA would go on to acquire all of their intellectual property pretty

00:14:22.860 --> 00:14:29.160
bitter pill to swallow the parallels between 3dfx's downfall and NVIDIA's

00:14:27.240 --> 00:14:33.540
behavior lately is kind of ironic if you think about it because they've been going down a very similar path

00:14:31.820 --> 00:14:37.440
antagonizing their board Partners diverting attention to side businesses

00:14:35.579 --> 00:14:41.339
like crypto mining autonomous vehicles machine learning you get it but

00:14:39.480 --> 00:14:44.880
fortunately for NVIDIA it seems like they've got a wide enough mode around

00:14:42.779 --> 00:14:49.260
them and enough cash on hand to whether pretty much an apocalypse and there are

00:14:47.399 --> 00:14:52.920
no existential threats to them like 3D effects faced during their decline at

00:14:51.600 --> 00:14:57.600
least not yet this my friends is why competition is so

00:14:55.740 --> 00:15:02.579
important in the industry because it can and will light a fire under a company

00:14:59.459 --> 00:15:04.440
that's grown complacent like 3dfx did if

00:15:02.579 --> 00:15:08.279
they'd had less debt if they'd gotten the Sega contract or simply made

00:15:06.300 --> 00:15:13.680
different decisions who knows what might have happened when the competition heated up a 3dfx forced to compete well

00:15:12.000 --> 00:15:18.600
they still could might have survived unfortunately for this timeline they

00:15:16.079 --> 00:15:24.120
chose the voodoo 5. and I choose this segue to our sponsor iFixit if you're

00:15:21.839 --> 00:15:28.920
constantly on the go ifixit's Moray and minnow kits are for you I fix its

00:15:26.519 --> 00:15:33.420
pocket-sized tool kit The Minnow is just 14.99 it features a magnetized case

00:15:31.680 --> 00:15:37.199
that's easily opened and doubles as a built-in sorting tray it has 16

00:15:35.040 --> 00:15:40.680
different bits and the handle has a built-in Sim eject tool and if you're in

00:15:38.880 --> 00:15:46.260
the need of something bigger the moire driver kit is only 19.99 it's got 32

00:15:44.040 --> 00:15:50.220
different bits with extended reach necks and best of all every iFixit kit

00:15:48.600 --> 00:15:54.300
regardless of the price comes with a lifetime warranty so check out our links

00:15:52.320 --> 00:15:58.860
in the description and buy yours today if you enjoyed this video go check out

00:15:56.459 --> 00:16:03.300
our video on the 3dfx voodoo 2 SLI gaming setup for a little more on just

00:16:00.660 --> 00:16:07.160
how good 3dfx used to be and how much they allowed to slip through their

00:16:04.800 --> 00:16:07.160
fingers
