It’s testomony to the Nvidia advertising and marketing crew that one among its buzzwords has now slipped into widespread parlance. Not solely did the 1st-gen Nvidia GeForce 256 introduce us to its now well-known GeForce gaming graphics model, nevertheless it additionally introduced the time period “GPU” to the gaming PC with it. Now, two and a half many years later, Nvidia is celebrating the GeForce 256’s twenty fifth birthday with a video and inexperienced popcorn. Right here, we glance again on what made a real fashionable traditional of the PC period tick.
Again to that time period GPU, it’s an initialism that we now use as shorthand for any PC graphics chip, as present in the perfect graphics playing cards, nevertheless it began life on the PC as an Nvidia advertising and marketing slogan. To offer you an concept of how way back this was, I used to be launched to the time period “GPU” by a paper press launch the identical week I began my first tech journalism job in September 1999.
We didn’t get press releases through e mail then – they have been bodily posted to us, and the editorial assistant sorted all of them right into a field for the crew to peruse. “In an occasion that ushers in a brand new period of interactivity for the PC, Nvidia unveiled at present the GeForce 256, the world’s first graphics processing unit (GPU)”, it stated.
On the time, I believed it appeared pompous – how might this relative newcomer to the 3D graphics scene have the nerve to suppose it might change the language of PC graphics? However I now see that it was a bit of promoting genius. Not solely did “GPU” stick for many years to come back, nevertheless it additionally meant Nvidia was the one firm with a PC GPU at this level.
GeForce 256 rework and lighting
Nvidia’s first GPU did certainly deal with 3D graphics fairly in another way from its friends on the time, so it’s time for a little bit historical past lesson. If we wish to perceive what made the primary GeForce GPU so particular, we first have to check out 3D pipelines of the time.
It was October 1999, and the primary 3D accelerators had solely been doing the rounds for a number of years. Up till the mid-Nineties, 3D video games akin to Doom and later Quake have been rendered totally in software program by the CPU, with the latter being one of many first video games to require a floating level unit.
If you wish to show a 3D mannequin, it has to undergo the graphics pipeline, which at this stage was all dealt with by the CPU. The primary stage is the geometry, the place the CPU works out the positioning (the place polygons and vertices sit in relation to the digital camera) and lighting (how polygons will look beneath the lighting within the scene). The previous entails mathematical transformations, and is often known as “rework” with the 2 processes collectively referred to as “rework and lighting” or T&L for brief.
As soon as the geometry is nailed, the following step is to fill within the areas between the vertices, which is known as rasterization, and pixel processing operations, akin to depth examine and texture look-up. That is, in fact, an enormous oversimplification of the 3D graphics pipeline of the time, nevertheless it provides you an concept. We began with the CPU dealing with the entire graphics pipeline from begin to end, which resulted in low-resolution, chunky graphics and poor efficiency.
We then had the primary 3D accelerators, such because the 3dfx Voodoo and VideoLogic PowerVR playing cards, which dealt with the final levels of the pipeline (rasterization and pixel processing), and massively improved the way in which video games regarded and carried out, whereas additionally ushering within the broad use of triangles slightly than polygons for 3D rendering.
With the CPU now not having to deal with all these operations, and devoted {hardware} doing the job, you could possibly render 3D video games at larger resolutions with extra element and quicker body charges. At this level, the CPU was nonetheless doing a good quantity of labor although. Should you wished to play 3D video games, you continue to wanted a good CPU.
Nvidia aimed to vary this case with its first “GPU”, which might course of your entire 3D graphics pipeline, together with the preliminary geometry levels for rework and lighting, in {hardware}. The CPU’s solely job then was to work out what needs to be rendered and the place it goes.
GeForce 256 vs 3dfx Voodoo 3
As with all new graphics tech, in fact, the business didn’t immediately transfer towards Nvidia’s {hardware} T&L mannequin. At this level, the one actual method to see it in motion in DirectX 7 was to run the helicopter take a look at initially of 3DMark2000, though some video games utilizing OpenGL 1.2 additionally supported it.
The latter included Quake III Enviornment, however the undemanding nature of this recreation meant it virtually ran simply as properly with software program T&L. DirectX 7 additionally didn’t require hardware-accelerated T&L to run – you could possibly nonetheless run DirectX 7 video games utilizing software program T&L calculated by the CPU, it simply wasn’t as fast.
The GeForce was nonetheless a formidable graphics chip whether or not you have been utilizing {hardware} T&L or not although. Not like the 3dfx Voodoo 3, it might render in 32-bit colour in addition to 16-bit (as might Nvidia’s Riva TNT2 earlier than it), it had 32MB of reminiscence in comparison with the extra common 16MB, and it additionally outperformed its opponents in most recreation exams by a considerable margin.
The response from ATi (now AMD) on the time was a brute-force method, placing two of its Rage 128 Professional chips onto one PCB to make the Rage Fury Maxx, utilizing alternate body rendering (every graphics chip dealt with alternate frames in sequence – word how I’m not utilizing the time period “GPU” right here) to hurry up efficiency. I examined it shortly after the discharge of the GeForce 256 and it might certainly sustain.
The GPU wins
The Rage Fury Maxx’s limelight was minimize shortly afterward, although, when Nvidia launched the DDR model of the GeForce in December 1999, which swapped the SDRAM used on the unique GeForce 256 with high-speed DDR reminiscence. At that time, Nvidia had received the efficiency battle – nothing else might compete.
It additionally took some time for everybody else to catch up, and at this level, numerous individuals within the business have been nonetheless swearing that the ever-increasing velocity of CPUs (we’d simply handed the 1GHz barrier) meant that software program T&L could be wonderful – we might simply keep it up with {a partially} accelerated 3D pipeline.
When 3dfx was constructing as much as the launch of the Voodoo 5 in 2000, I bear in mind it having an FAQ on the web site. Requested whether or not the Voodoo 5 would have software program T&L help, 3dfx stated, “Voodoo4 and Voodoo5 have software program T&L help.” It’s not intentionally dishonest, however each 3D graphics card might help software program T&L presently – it was accomplished by the CPU – it regarded as if the reply was there to sneakily recommend characteristic parity with the GeForce 256.
In actual fact, the one different graphics agency to give you a good competitor in an affordable time was ATi, which launched the primary Radeon half a yr later, full with {hardware} T&L help. In the meantime, the 3dfx Voodoo and VideoLogic PowerVR traces by no means managed to get {hardware} T&L help on the PC desktop, with the Voodoo 5 and Kyro II chips nonetheless operating T&L in software program.
However 3dfx was nonetheless taking a brute-force method – chaining VSA-100 chips collectively in SLI configuration on its forthcoming 3dfx Voodoo 5 vary. The Voodoo 5 5500 lastly got here out in the summertime of 2000, with two chips, gradual SDRAM reminiscence, and no T&L {hardware}. It might sustain with the unique GeForce in some exams, however by that point Nvidia had already refined its DirectX 7 {hardware} additional and launched the GeForce 2 GTS.
By the tip of the yr, and following a sequence of authorized battles, 3dfx went bust, and its belongings have been purchased up by Nvidia. GeForce, and the idea of the GPU, had received. It’s honest to say Nvidia deserves its inexperienced popcorn.
We hope you’ve loved this private retrospective in regards to the very first Nvidia GeForce GPU. For extra articles in regards to the PC’s classic historical past, try our PC retro tech web page, in addition to our information on the best way to construct a retro gaming PC, the place we take you thru all of the {hardware} you want.
