In part 1 of a 2 part series, shoes takes you on a long journey through time and compares products from ATI and NVIDIA in order to debunk all of the fanboy myths.

Any gamer in their right mind knows at least something about computers. This knowledge could start with simply plugging in various cords, and stretch all the way to how each individual component works. There is, however, one little piece of information that every single gamer out there knows whether they are just beginners or experienced pros: when it comes to graphics, the only names that matter are NVIDIA and ATI. It seems nearly all hardware related arguments have something to do with the decision between these two companies unless it is an Intel vs. AMD argument, and even then it usually has an ATI vs. NVIDIA spin off. The question is: why? Why is it always about these two companies?

Whenever someone ‘knows’ something about something, they automatically form their own opinions about whatever it is they know. However, when someone’s knowledge on a subject is incomplete, there more often than not exists a nasty little thing called bias. Bias can be found just about everywhere in the gaming scene; someone likes Team 3D and ‘hates’ Complexity, someone likes Counter-Strike 1.6 and ‘hates’ CS: Source, someone likes GotFrag and ‘hates’ AmpedNews. And of course, we gamers have come up with our own little term for these types of people, the oh-so overused and yet amazingly fitting conjunction of a word: “fanboys”. As mentioned before, there is no more commonly occurring fanboy argument in all of computer gaming than that of ATI vs. NVIDIA. For an answer to the question of why this is the case, all I can offer is speculation that is based on my own logic and reasoning abilities. As gamers, we are absolutely obsessed with the way things look. As the entertainment industry as a whole progresses, the products it puts out become more and more graphic intensive, for proof of this look at the movie industry. The new King Kong movie that is coming out is a hell of a lot nicer looking than the 1933 version. You might say, “The only reason they could make it look better is because of the better technology that is out now.” Exactly. The technology used to make these products has increased as a result of demand from eye-candy loving consumers like gamers. Look at the original Half-Life vs. Half-Life 2. Although the game play is fundamentally the same, the visual effects and overall feel of the games are night and day. Computer graphics went from primitive chips playing games like StarCraft, to highly advanced and specialized chips capable of rendering some of the most true-to-life scenes and effects we have ever seen - in a little over 4 years. Compare this, if you will, to processors, where 4 years ago we already had the first versions of the Pentium 4, and now, 4 years later, Intel is still pumping out chips labeled as Pentium 4’s. Gamers obsess about graphics because that’s what games are: graphics.

So now I have just shot off two random paragraphs about fanboys, what is the point of this article? Well, simply put, the purpose of this article is to attempt to resolve a majority of the ATI vs. NVIDIA fanboy disputes by presenting all of the information, and not just some of it, and in the most objective possible manner.

But first, while I am kind of still on the topic of processors, I would like to point out one common misconception amongst fanboys that is really just too ridiculous to warrant its own section. NVIDIA graphics cards do not work better with AMD processors than they do with Intel processors. The same is true for ATI. I can only imagine that such a theory came about when some guy noticed that his video card worked better on an AMD system than on an Intel one when it was really the processor that was making the difference. You don’t “code” hardware, and any adjustments made in the design could potentially cost a huge amount of money, so why would you possibly want to make one piece of hardware work better with a specific kind of another?

Now then, let’s get some insight about the beginnings of each company.

ATI
Kwok Yuen Ho was born into a childhood of poverty. After selling vegetables from the family garden for many years, he eventually earned a spot in a top university in Taiwan. After graduating with a degree in Electrical Engineering, he landed a nice job at several top semiconductor firms in Hong Kong. Ho migrated to Canada in 1984, where he found it difficult to find a suitable position for someone with his amount of work experience and education. Frustrated, Ho collaborated with Benny Lau and Lee Lau to start up their own company: Array Technology Industry.

ATI was born, and after being forced, due to financial situations, into the relatively low-capital genre of computer components manufacturing called “graphics processing”, they were growing up nicely.

NVIDIA
In 1993, three graphics industry veterans got together and formed NVIDIA. Being a bit more than a startup company, NVIDIA was able to get off to a blazing start. They had some problems producing chips that were competitive with the competition from companies like S3 and 3Dfx until about 1997, when they released the RIVA 128. It was mostly uphill from there, and in December of 2000 they acquired the Intellectual assets of 3Dfx, which had by this point dropped to an afterthought of NVIDIA and ATI. With the rights to the technology behind 3Dfx’s previously successful product line, the stage was set for the fiercest rivalry the computer industry had ever seen.

{mospagebreak heading=An Introduction&title=Some Details}
Drivers

History

The debate all started back during the glory days of NVIDIA’s GeForce 2 GTS. ATI’s most recent release at the time, the Radeon 64MB, was thoroughly trouncing the GTS across the board in synthetic AND real world benchmarks. Faced with this dilemma of having their flagship card being one-upped by the competition, NVIDIA released the Detonator3 drivers for the GTS. These drivers boosted the performance of the card by amazing amounts, and were the first time we see NVIDIA’s now almost traditional dominance of OpenGL applications. The Radeon 64MB still beat the GTS in a majority of tests, but the advancements made in release of the Detonator3 drivers started off a continuous stream of outstandingly stable and well crafted drivers from NVIDIA.

So at this point we have NVIDIA producing very nice drivers, and ATI, for the most part, also producing stable and capable drivers. Then, along came an interesting part from ATI called the Radeon 8500. From the specifications to the benchmark rumors, this card had potential written all over it. It was built as a direct competitor to the as of then unreleased GeForce3 Ti500, and all signs suggested it would absolutely succeed at that task. However, when we first saw direct benchmarks of the cards going head-to-head, the ti-500 was almost always the victor, and by a pretty substantial margin too. The community was confused as to why this beastly card was being beat by its respectable but by no means equally powered counterpart. It turns out the problem was the drivers. ATI had a real problem with the drivers for the 8500, and it took them a few revisions to fix them. In fact, it wasn’t until several months later that the Radeon 8500 was actually up and running to its full potential, and expectedly, it was consistently wiping the board with the GeForce3 Ti500.

Verdict
Each company had its own share of driver problems, but for some reason nobody seems to be getting over ATI’s issue, which occurred about 4 years ago now. The reality of the situation with ATI’s drivers nowadays is that they are absolutely amazing. The Catalyst drivers are very stable and very capable. In addition, they are updated every month to fix minor bugs between game/application releases and previous driver revisions. BOTH companies make outstanding drivers now, so there are NO grounds on which to base your driver superiority arguments.
{mospagebreak title=GeForce3 Ti500 vs.}

Gen 1: Radeon 8500 vs. GeForce3 Ti500
The end of 2001 saw the increased development of more graphic intensive games. Quake III Arena, which had been out at this point for almost 2 years, was being updated repeatedly and becoming more and more resource hungry in the process. MadOnion (Futuremark) had produced its 3DMark 2001 (build 200) which was widely adopted as the synthetic benchmark of choice at the time. So with graphic intensive games like Serious Sam and Quake III arena, along with a glimpse of future graphics demands in 3DMark 2001, ATI and NVIDIA thought it would be appropriate to release some new hardware to power these games. NVIDIA chose to refresh its GeForce3 product line by slapping on “Titanium” in front and significantly boosting clock speeds, while ATI produced the next generation of their Radeon line with the 8500, based on an entirely new R200 core.

GeForce3 Ti500

Based on the same NV20 core that powered its GeForce3 predecessor, the Ti500 displayed some dramatically increased clock speeds that boosted performance by a fairly large amount. The NV20 core itself was a shrunken version of the NV10 core (GeForce2) with some very nice updates, including a fully programmable transforming and lighting engine. The GeForce3 Ti500 was released on October 1, 2001 at a price of $350 (MSRP) along with two other cards: the GeForce3 Ti200 and the GeForce2 Ti. After extensive benchmarks of these cards and their competition, it was eventually concluded that the Ti200, a slightly lower clocked version of the update, turned out to be the real winner in terms of price/performance for the generation. However, in this particular comparison we are matching best to best, so I chose to use the Ti500, the highest performing NVIDIA part at the time.

Specs

The specifications for the GeForce3 Ti500 are identical to those of the GeForce 3, with the exception of the clock speeds, which of course then changes the fill rates and memory bandwidth. With an AMAZING 4 pixel pipelines, this card could eat up any game at the time. It was NVIDIA’s flagship card, and at $350 dollars, that doesn’t seem too bad for the top of the line (at least by today’s standards).

{mospagebreak title=Radeon 8500}

The new R200 core used in the 8500 was received by massive fanfare, but ended up disappointing the ATI faithful who had waited so long for it. As mentioned earlier, this was the result of driver problems. After resolving those problems though, the 8500 quickly became the standout performer of this first generation. The Radeon 8500 was released on August 14, 2001 at a price of $299 (MSRP). The R200 was the first core in ATI’s lineup to accommodate for DirectX 8.1, and it also introduced the ATI technologies of SmartShader and Truform.

Specs

As you can see, the specifications of the Radeon 8500 are pretty similar to those of the GeForce3 Ti500. The R200 core on the 8500 is clocked significantly faster than The GeForce3’s NV20, which provided a very nice performance boost in the end. As ATI’s poster child for its new SmartShader and Truform technologies, the Radeon 8500 received a lot of attention from many tech sites. The Radeon 8500 was even adapted for use in Macintosh computers, not that that has anything to do with this article, because we hate Macs, yuck.

Analysis

Some other things to note about the two cards are their prices and release date. As you can see, the price of the 8500 is $50 less than that of the Ti500. Furthermore, the 8500 was “released” a month and half earlier than the Ti500. This does not mean the 8500 is totally the better card, because when I say it was “released” it really was released, but had a boat load of problems too. As mentioned earlier, it had driver problems that were not resolved for several months after the release date. On the other hand, the Ti500 was performing at its maximum potential straight out of the gates.

Verdict

As it turned out, the 8500 did wallop the Ti500 in pretty much every benchmark made after its final driver release. It even squeaked out a few wins with its crippled drivers. But the Ti500 was a more successful card in general. The bottom line in this generation is this: If you release a card, make sure it works right first.

ATI = NVIDIA
{mospagebreak title=GeFroce4 Ti4200 vs.}

GeForce4 Ti4200

This newfangled NV25 core had some pretty groundbreaking and interesting things going on with it, the most of which was something called a vertex shader. Vertex shaders opened the door for games with advanced lighting and highly complex shading operations, and were first put to really good use on the NVIDIA chip in Microsoft’s original Xbox. Another advancement on the NV25 core was its increased efficiency as a result of it being able to disregard calculations for unseen pixels. This allowed for each clock cycle to be more effective than previously seen.

Specs

Probably the most notable thing about the specifications of the Ti4200 besides the aforementioned vertex shaders is the increased amount of onboard memory. The general advantage of onboard memory is the performance increase it offers at higher resolutions. The overall increase in all of the specs, as well as keeping in mind the efficiency concept related above would lead us to believe this is one pretty nicely performing card. An interesting note about the Ti4200 is that it beat the Radeon 8500 at every single benchmark, except for some when Anisotropic Filtering and Anti Aliasing were turned up. But that is beside the point because the Radeon 8500 is an older card, so of course the Ti4200 should beat it. Ahh, I seem to have stumbled upon yet another fallacy amongst fanboy arguments: comparing cards from different generations is like comparing apples to oranges, there is no point.
{mospagebreak title=Radeon 9500 PRO}

As the budget friendly variant of the R300 based cards, the Radeon 9500 PRO was ATI’s fully featured and amazingly powerful little sister of the flagship 9700 PRO. When you see the specifications of this card, you’re going to think “wow, this is an unfair match up.” Well, pay attention to the release dates. The 9500 PRO was released almost 7 months after the Ti4200, which reveals the “different circumstances” I noted in the introduction to this generation. Even with this later release however, ATI still managed to offer the 9500 PRO at the same $200 (USD) price point as the Ti4200. In fact, the release of the whole R300 core lineup was kind of weird. The 9700 PRO was released a good 2 months before the memory-bandwidth-halved 9500 PRO and its launch mates. These secondary releases were supposedly targeted specifically at the GeForce4 line of cards: the Radeon 9500 against the MX series, the 9500 PRO against the Ti4200, and the 9700 against the Ti4600.

Specs

8 pixel pipelines. That is twice as much as anything else so far. 4 vertex shaders. This is also twice as much as anything else so far. By all accounts, at a mere $200 dollars, the Radeon 9500 PRO was an amazing piece of equipment at a very affordable price. Beyond its specs alone, one of the biggest stories about the 9500 PRO was its ability to unlock to become basically a 9700 PRO. Success rates for this modification were fairly decent, so among the enthusiast community, the 9500 PRO was obviously a standout choice.

Analysis
It’s almost not fair to compare these two cards because of the amount of time that passed in between their releases. However, it remains the most direct comparison of cards from each company for this generation. By specifications alone, you can clearly see that the 9500 PRO totally dominates every single category. However, the Ti4200 had a good 7 months head start on the 9500 PRO in terms of availability, which actually made it a more successful card in the sub $200 (USD) price point. One thing that was quite interesting about the 9500 PRO is that it was really the first ATI card that was released to target the mainstream market. The Radeon 9000 was chugging along quite well on the low end, and the 9700 PRO surpassed its closest competitor, the GeForce4 Ti4600, by a margin so large you could easily mistake it for a next generation card in comparison. In fact, it should be noted that at this point in time, ATI was by all accounts the leader in both graphic card technology AND performance, largely a result of the power of the 9700 PRO.

Verdict
Y

However, as I mentioned before, ATI already had a card on the market 2 months before the release of the 9500 PRO that performed so well that its nearest competitor was left very much in the dust. Even the card that was launched as a direct competitor to the GeForce4 Ti4600, the Radeon 9700, managed to beat the pants off everything with fairly lower clock speeds than the amazingly amazing 9700 PRO. If determining the winner of this round was based solely upon the two cards I chose to represent it, then I would readily hand NVIDIA the victory on the basis of its competitiveness with a card that was released 7 months later. However, the margin by which the 9700 PRO, ATI’s flagship, beat out the Ti4600, NVIDIA’s flagship, is so huge that I could not in my right mind say that NVIDIA’s lineup as a whole was superior to ATI’s. Throw in the fact that the R300 based chips were fully compatible with DirectX 9.0, and NVIDIA was pretty much begging for mercy. Bottom line: If you have a card that outperforms everything else by such a huge amount that nothing even comes close, you are the winner.
ATI > NVIDIA
{mospagebreak title=GeForce FX5950 Ultra vs.}

Before getting into the nitty gritty of the two chips I have selected for this generation, I would like to introduce a “mini match up” that took place before the real competitions began. The 9700 PRO that was part of that odd “in between” generation ATI release had remained unchallenged for a good 5 or 6 months before the NV30 manifested itself in the GeForce FX5800 Ultra. The first chip to utilize a memory type faster than traditional DDR, the FX 5800 Ultra had some fine looking specs and a revolutionary chip design that will be discussed in a moment. At the end of the last generation I mentioned the DirectX 9.0 capabilities of the R300 chip. Well, it is kind of hard to stress enough the importance of compliance with the latest DirectX standard. Sure, your new chip will run circles around the older competition in older DirectX standards, but if you do not have full support for the latest DirectX standard, then you are going to get eaten alive when bleeding-edge tests are carried out. This was the fate of the FX5800 Ultra and the NV30 core in general: It was primarily a DirectX 8.1 card, and its compatibility for DirectX 9.0 seemed like somewhat of an afterthought. However, the FX5800 Ultra possessed some groundbreaking features that allowed it to handle highly complex computations with a MUCH higher proficiency than the R300, and the benchmarks definitely showed it. The FX5800 Ultra gets ripped a new one in several benchmarks, but really shows what it is made of in the more complex and OpenGL tests.

GeForce FX5950 Ultra

The FX5950 Ultra, utilizing the now-quite-common-amongst-NVIDIA-cards 2-slot cooler, was NVIDIA’s final flagship card for this generation.

Specs

Although the specs (pipelines to be specific) might initially seem less impressive than other cards in the generation, the NV38 was definitely the most refined variant of the NV30. Incorporating the better pixel shader found in the NV35, and other core enhancements to eliminate excessive heat production that allowed for maximum clock speeds, the GeForce FX5950 Ultra was the strongest card in this generation from a pure power perspective. It required the use of an external power connector, which was a feature becoming more and more common amongst the chips that were pushing the envelope. The fact that many game designers were already kind of coding their game around NVIDIA chips didn’t hurt in the benchmarks either. One thing that might have hindered the 5950 Ultra was its memory latency, which was decidedly higher than on its predecessor, the 5900 Ultra.
{mospagebreak title=Radeon 9800XT}

It turned out that Valve was going to offer free copies of Half-Life 2 to anybody who purchased one of these new ATI cards, either the 9800 XT or 9600 XT, pretty neat stuff. This was right up ATI’s alley, as they probably felt some heat when NVIDIA decided to bundle Call of Duty with their 5900 chips. The 9800 XT chip itself was really just a refresh of the 9800 PRO, offering higher clock speeds and thus greater overall performance. Despite its new specs, the 9800 XT was still not as powerful in terms of raw performance as the 5950 Ultra, but we shall see how the head-to-head benchmarks turned out.

Specs

Those are some pretty impressive specs from the 9800 XT. You can see that the major differences between the two cards come from the memory bandwidth and pipeline/shader areas. This is pretty representative of the two different approaches that the companies took for this generation: NVIDIA attempted to use some new features combined with raw power, and ATI just stuck with the same successful approach it had been using for quite some time.

Analysis
A

It turns out that these two cards went back and forth with the benchmarks. Typically, the titles we saw the 9800 XT winning were the ones that put high emphasis on memory throughput, which was a result of the one problem I mentioned with the FX5950 Ultra: its high memory latencies. But in all honesty, the benchmarks were so close that you could not decidedly give the edge to either card. It would have to come down to overall availability, on which the 9800 XT had a nice edge.

Verdict
Everything in the generation was so close except for one match up that it is very difficult to determine a clear winner. Sure, NVIDIA took a long time getting the NV30 to its final NV38 revision, but had ATI not released the R360 based 9800 XT, the FX5950 Ultra would have had a pretty significant lead over the 9800 PRO. The FX architecture also introduced some nice new features that would be incorporated into many future products. So, both companies did what they needed to in order to stay competitive. I have no choice but to say this generation is pretty clearly a tie.
ATI = NVIDIA
Stay tuned for our second part of this series, when we’ll tackle the incredibly close 6-series vs. X-series battle, as well as the progress to date of the current generation.

Popularity: 3% [?]

You Should Also Check Out These Post:

• AMD launches Weekly Deals on Newegg
• No SLI for first batch of X58 boards
• New Gaming Peripherals from MS and Razer
• AMD Exec Reckons Graphics Throne There to Stay
• HYDRA renders SLI and CrossFire obsolete?

More Active Posts:

• Some Updates (8)
• NVIDIA has an overcomplicated lineup (8)
• Welcome To FPSLabs V3 With Added Awesome! (7)
• NVIDIA GeForce 9800GTX on April 1st (7)
• Razer Piranha Review (7)
• Microsoft to sell XP until 2010 (6)
• Weekly Hardware Recap (6)
• The Weekly Hardware Recap (5)
• Newbie's Guide to Vista: The Look (5)
• Arctic Cooling Freezer Xtreme is on the way (5)