Our GPU Flashback Archive series continues today with what can only be described as a pivotal moment in GPU history. The NVIDIA NV10 was in fact the first chip to be called a GPU, a term coined by NVIDIA themselves back in 1999. Let’s take a look at the chip itself and the two cards that were produced using it, plus a few of the more notable score submissions that have been made using first ever generation of NVIDIA GeForce branded cards.
NVIDIA GeForce 256: Overview
The NVIDIA RIVA series (covered last week) put the company firmly on the graphics card map, proving that the silicon they were producing could compete with offerings from other companies. It’s important to remember also that at the end of the nineties, you could purchase a card from one of several companies including ATI, S3, Matrox and Voodoo. NVIDIA as we all know would go on to become leader of the GPU market and one of the most successful companies in the industry. The direction taken with the first GeForce-branded GPUs, the GeForce 256, reflects NVIDIA’s bold and ambitious approach as a company generally. The GeForce 256 was unique, offloading geometric calculations to a specific engine while also increasing the amount of fixed pixel pipelines. The outcome was the first Direct3D 7-compliant card, one that offered a genuine leap in 3D gaming performance.
In terms of historical pedigree, the NVIDIA GeForce 256, codenamed NV10, was the first chip sold by NVIDIA that was marketed as a Graphics Processing Unit, or GPU. There are obvious similarities with the term CPU, a household term familiar to all in 1999, but again NVIDIA’s bold approach was doubtlessly inspired by Intel’s rise to prominence in an x86 dominated PC market. In August 1999 NVIDIA (to use the original capitalization) defined a Graphics Processing Unit
as “a single-chip processor with integrated transform, lighting, triangle setup/clipping, and rendering engines that is capable of processing a minimum of 10 million polygons per second.”
One major aspect of the new NV10 GPU was its integrated T&L (Transform and Lighting), a task usually implemented in software and performed by the CPU. The NV10 was the first GPU to offer hardware support for T&L. This resulted in faster vertex processing, a task which was further aided by caching that helped avoid duplicate vertex processing. Direct3D 7.0 was in fact the first release to support hardware T&L.
The NVIDIA NV10 GPU featured 4 Pixel Shaders, 4 Texture Mapping Units (TMUs), 2 Render Output Units (ROPs) and could manage a pixel rate of 480 MPixels/sec. In all of these metrics we see the numbers doubled from the previous generation, a major factor in producing a chip that could suddenly top the charts of any graphics card review. The only caveat to that statement involves graphics memory.
The new GeForce 256 card will historically be remembered as having two distinct variants; the GeForce 256 SDR and the GeForce 256 DDR. The former arrived in October of 1999 supporting up to 32MB of standard SG/DRAM memory and potential bandwidth of 1.144 GB/s. The card cost around $250 USD and performed at similar levels to competing cards such as the ATI Fury MAXX and the S3 Savage 2000. NVIDIA recognized that the main bottleneck preventing the card from fulfilling its full potential was memory bandwidth.
In February 2000 NVIDIA released an updated design that used 32MB of DDR memory, a very new form of memory technology at that time. The Double Data Rate memory installed on the card was configured at 150MHz (300MHz effective). The memory bus used the same 128-bit width but the doubled memory bandwidth meant a massive boost to the tune of 4.8GB/s.
The outcome was a card that effectively destroyed the competition. Check out the benchmark results above which come from Anand Lal Shimpi himself. The GeForce 256 occupies a space all of its own at higher resolutions of 1024 x 768 (a reasonably high pixel count in 1999). Graphics card vendors brought the GeForce 256 DDR to market with a steep asking price of $300, a premium which at the time was considered pretty steep for a graphics card.
This is what Anand had to say back in December 1999:
The price was and still is quite steep for a single peripheral, especially considering that we are currently in a period where a relatively powerful gaming system can be had for around $1000 – $1500. But then again, everything is relative, right? So the actual severity of the price depends on the overall experience though, correct? If you pay $300 for a graphics card and barely get an extra 5 fps under your favorite game then the card wasn’t really worth $300 to you in the first place, now was it? Then again, if you purchase a $300 card and, immediately after firing up your favorite game, it feels like you’ve never even seen the title before this moment, the card quickly begins to earn its value.
You can find the GeForce 256 DDR review from Anandtech here.
Most Popular NVIDIA GeForce 256 Graphics Card, the GeForce 256 DDR
As outlined above, although several vendors produced and sold graphics cards sporting the new NVIDIA NV10 GPUs, most cards followed the standard specifications. The only two models were the SDR and DDR versions. At this time, mid-range and budget segments were amply served by NVIDIA’s RIVA and VANTA branded cards.
- -GeForce 256 DDR – 56.36%
- -GeForce 256 SDR – 43.64%
As you can see, in terms of submissions, the NVIDIA GeForce 256 DDR remains the more popular card historically. It’s relevant to note that the GeForce 256 series actually predates the existence of HWBOT by several years which means that all submissions were made retrospectively. Let’s take a look at some of the outstanding scores made with a GeForce 256 DDR card.
NVIDIA GeForce 256 DDR: Record Scores
We can now take a look at some of the highest scores posted on HWBOT using NVIDIA the GeForce 256 DDR.
Looking at older generation graphics cards invariably means looking at older benchmark scores too. We start with the classic 3DMark 2000 benchmark from Futuremark. The highest score submitted to HWBOT using an NVIDIA GeForce 256 card was made by German overclocker Masterchief79. He pushed a ASUS GeForce 256 DDR card to 165MHz (+37.50%) on the GPU side with graphics memory at 195MHz (+30.00%) to make a hardware first place score of 8,949 marks. The rig he used also featured an Intel Core 2 Extreme QX6700 ‘Kentsfield’ processor clocked at 3,518MHz (+31.96%).
You can find the submission from Masterchief79 here on HWBOT: http://hwbot.org/submission/2477953_masterchief79_3dmark2000_GeForce_256_ddr_8949_marks
3DMark 99 MAX
Looking at the even older 3DMark 99 MAX benchmark we find that the fastest score submitted using an NVIDIA GeForce 256 card comes from Romanian pasatoiutd who managed a score of 20,650 marks using a Creative Labs GeForce 256 DDR card with the GPU pushed to 155MHz (+29.17%) and graphics memory at 190MHz (+26.67%). The rig also used an Intel Core 2 Extreme QX9650 ‘Yorkfield’ clocked at 4,305MHz (+43.50)
You can find the submission from pasatoiutd here on HWBOT: http://hwbot.org/submission/2369405_pasatoiutd_3dmark_99_max_GeForce_256_ddr_20650_marks
In the classic Aquamark benchmark we find the legendary havli from the Czech Republic is also the man to beat. His score of 9,510 marks was made with an ASUS GeForce 256 DDR card with the GPU clocked at 145MHz (+20.83%) and graphics card memory at 175MHz (+16.67%). The CPU used was an Intel Pentium E5700 ‘Wolfdale’ chip clocked at 3.9GHz (+30%) mounted on an ASRock 4CoreDual-SATA2 motherboard.
You can find the submission from havli here on HWBOT: http://hwbot.org/submission/2307351_havli_aquamark_GeForce_256_ddr_9510_marks
Thanks for joining us for our first ever GPU Flashback Archive article. Next week we will return with a look at the NVIDIA GeForce 2 series of graphics processors and cards.