The world of GPUs has always been a cutthroat one, and while no GPU designer ever intends to release a flop, there are instances under the high pressure environment of this market where mistakes are inevitable. A GPU maker has to survive by timely launches of competitive graphics solutions, straining to anticipate their rival’s intentions and often stretched for time against ever-present problems. It’s a complex undertaking, where design takes place four or five years before it will be powering whatever crop of demanding games currently hold our attention, having to understand what the capabilities of manufacturing will be at the time, and the state of software standards like Microsoft’s DirectX. Then when launch time comes, you have to be ready with enough chips hitting close enough to performance goals, and feature-complete drivers that are polished and ready to spit out images error-free.
It isn’t easy, and it’s why so few have managed to stay in the game, with even once revered hardware slingers too often destroyed by the masticating forces of a harsh technological death race. Even the graves of these former greats are scarcely left in peace, with many of their remnants scattered and bartered for so they can be added the warchest of the next towering juggernaut, whether truly useful or not.
While the primordial soup of the early days has undoubtedly settled its churn in the last ten or fifteen years, the two companies that are left still can’t escape the occasional flub, and it’s unlikely we’ll ever see an end to the controversies and disappointments until such a time as imperfect beings no longer govern them.
So below just for funsies I’ve compiled a list of what I feel are the Top 10—from crap to crappiest—infamous, ill-reputable, inimical cockups in graphics card history. Feel free to disagree, form your own opinions, and comment to the contrary or concurrence to my conclusions afterward.
(A side note: Some graphics cards have aged better (Voodoo5), or worse (GeForce FX) over time. This is largely an evaluation of how they were initially received, and in some cases some short-term consideration for their future capability.)
#10
ATI Radeon X1800XT Platinum Edition
Released: October 2005 - Price: $549
This card will be the first example on our list of a solution to a problem that only existed in the short-term. When the design of your architecture is highly focused on only the present demands of software, you’re left selling something with very little longevity, and potentially a major weakness that only grows worse in the future.
The X1800XT, while a powerful card, and mostly a suitable answer to NVIDIA’s GeForce 7800GTX at the time, ran into such a situation. It was three months late in answering its competitor, likely due to utilizing a brand new 90 nanometer manufacturing process. It finally brought support for Shader Model 3.0, which was a sore spot on the otherwise well-loved X800 series. But it was barely adequate for the task, largely bringing the same 16-wide pipeline layout of its predecessor into a looming era of complex shader-heavy games for which it was ultimately too weak. The mid-generation refresh card, the X1900XTX, fixed this issue by fitting more shaders processors into the pixel pipes of the chip, but with the side-effect of immediately obsoleting the X1800 flagship after only a few months.
#9
NVIDIA GeForce GTX 970
Released: September 2014 - Price: $329
NVIDIA has had a long string of successful second-rung highend cards, offering most of the performance of the top-of-the-line at a drastically reduced price. The GTX 970 was no exception. It was a great performer often trading blows with AMD’s best at the time, and with its attractive MSRP it was the favored choice of many Master Race members that didn’t have rolls of cash coming out their ears. It did so by taking the beefiest Maxwell 2 chip at the time and just moving it down a few pegs in the number of stream processors and core clock, while keeping the rest of the awesomeness the same. Or so everyone thought.
Then it was discovered that NVIDIA had misled the press and public by publishing the wrong numbers for some of the details. In what they characterized as a simple accident, an incorrect ROP count and L2 cache size was reported on the specifications sheet, which had ramifications for how the graphics memory was addressed. In short, the GTX 970 could only access 3.5GB at normal speed, with 512MB placed at lower priority, and as a result, making it way slower. While this didn’t affect end-performance that much, the deception nonetheless hurt NVIDIA’s reputation among PC gamers, and forced them to take steps to improve the vetting process for published information. Whether you think this is a real issue or not in the end, the widespread controversy alone earns it a spot on this list.
#8
ATI Rage Fury Maxx
Released: December 1999 - Price: $299
It’s honestly hard to hate the Rage Fury Maxx. While it probably earns a spot on this list just for that obnoxious extra X, it was an ambitious card, and set the groundwork for technology we’d end up using a lot several years later. Plus it actually performed kind of OK, when everything was going well.
ATI was a rather humble competitor at the time, dreaming big of standing alongside the frontrunners but as a smaller, less funded group, somehow always falling short. The Rage Fury Pro came out too late to really make a dent, and the easiest solution to catching up to the big boys was to simply put two of them on the same card. Hey, it worked for 3dfx!
And that was also the problem. While NVIDIA was using a newer manufacturing process and extolling the virtues of hardware Transform-and-Lighting, ATI was stuck with last-generation features and technology. It cut costs every way it could (even sacrificing a second VGA port which could have proved advantageous) to position the Maxx against the GeForce SDR, but against the GeForce DDR it was no match. Worse, it rarely met the performance of the slower GeForce, and was deemed too expensive for what it offered. But it also introduced the world to Alternate Frame Rendering (AFR), and that became the basis for multi-GPU rendering to this day. It stings a bit to put this underdog at #8 on this list.
#7
Hyper Memory/TurboCache Cards
Released: 2005 - Price: Cheap
Here’s an idea! Let’s take all the fun of integrated motherboard graphics with its shared slow-ass system memory and put it on a graphics card! We’ll put a fraction of the usual amount of VRAM onboard but only charge marginally less money, and reap greater profits! People really don’t care about performance anyway!
#6
ATI Radeon 9800SE
Released: Sometime 2003 - Price: ~$199
ATI was riding high on the success and accolades of its 9700 and 9800 series cards. They sported a futuristic architecture with mad perf (yes, PERF!) and wicked features and suffered very little competition from NVIDIA.
Then they got a little greedy. While the 9800 Pro was the fastest card available at the time, and the regular 9800 was a wonderfully economical runner-up, the 9800SE (or “Special Edition”) emerged to muddle up everything. Although nothing in the name suggested it, it had half the pixel pipes as the two better cards, and yet cost the same or very close to the 9800. Admittedly this card was mostly seen from OEMs, with only a few board partners offering retail versions, but the confusion still duped a lot of uneducated consumers who only knew of the enthusiasm behind the ‘9800’ name. Worse, some were even equipped with half the memory bandwidth, which was even harder to discern. It was a rightfully hated product in the community among an otherwise heralded line-up.
#5
ATI Radeon HD 2900XT
Released: May 2007 - Price: $399
Any card that comes out six months after its rival is never off to a good start. Any card that’s so slow it must concede the performance crown and price itself accordingly, is doing even worse. The HD 2900XT was ATI’s first DX10 card, and had the misfortune of having to do battle against the NVIDIA’s Sherman tank: the 8800GTX. Going against one of the most forward-looking and brilliant GPU designs of all time probably foretells an ill fate, but that doesn’t mean ATI was going to throw in the towel. They just….probably should have.
In addition to being too slow, it also consumed too much power and ran way too hot. It wasn’t likely to steal any sales away from the still-brilliant 8800GTS, and it was quickly excised when the better-refined 3870 came to market. Being too quick to jump on 80nm and massively overreaching with a crazy 512-bit memory bus probably didn’t help things.
#4
NVIDIA GeForce GTX 480
Released: March 2010 - Price: $499
Then we move on to DX11, and the tables have turned. People really wanted this card to rock. Delay after delay, however, put a damper on people’s spirits when the Radeon HD 5870 was doing dances on the showfloor all by itself. Six miserable (fucking) months went by, and I….er, the community was presented with the now infamous George Foreman Grill card. A card so half-baked, it was still baking when you put it in your computer. Along with everything else that happened to be in the room.
Yes, it outperformed the 5870. Somewhat. I mean, nobody cared, because for $100 more the dual-GPU 5970 defecated on its sandwich in every way, even consuming less power! It was overpriced on top of every other problem with it, but hey. At least it made the cold winters more bearable.
#3
3dfx Voodoo5 5500
Released: June 2000 - Price: <$300
Objectively, the Voodoo5 isn’t any worse than the GTX 480. On its own merits, aside from being slower than the GeForce 2 GTS, barely matching the GeForce 256 DDR, being disgustingly huge and being unconventionally power hungry (perhaps the first to have a molex connector on the board), it didn’t do too badly. It wasn’t terribly overpriced either, and it fixed the wrongs of the Voodoo3 (which damn-near made the list) by supporting 32-bit color during 3D rendering. And actually looking back, many modern system builders of retro PCs have noted the performance scales beautifully with CPU speed, indicating it was probably held back in its day.
So why did it place above the GTX 480? Well, many credit this card with destroying 3dfx. It was their swan song, as the Voodoo6 never made it to consumer hands (probably for the best), and the sheer expense of those newfangled VSA-100 chips constrained their resources as their yields proved disappointingly low. 3dfx died after this card, being swallowed up by the Big Green Behemoth, but it’s fondly remembered nowadays by obsolete hardware enthusiasts for its speed, features, and well-rounded support for multiple APIs. It was also late to the market, a little dated in feature support (no T&L hurt), and yeah... It killed a former giant.
#2
NVIDIA GeForce4 MX400 Series
Released: 2002 - Price: <$199
It’s one thing to unintentionally deceive people. It’s another to deliberately brand a series of cards so that your customers are confused about its features. The GeForce4 series was great, and brought advancements to the first programmable pixel-shading engine of the previous generation with a monstrous improvement in performance. And the GeForce4 Ti4200, one of the most well-renowned cards ever released, held the standard for best value for the money for a very long time. But for those consumers that needed a solution under $199, NVIDIA turned to the MX sub-series to facilitate their needs.
Except, they did not support the same features of the proper GeForce4 cards. Or even the GeForce3 generation. Aw no, they were actually refinements of the GeForce2 hardware, limited to DX7 support and possessing no programmable pixel-shading whatsoever. As a result, this was a hated line of cards, and won NVIDIA a lot of flak for implementing completely different capabilities within the same generation. Worse, it sold really damn well, meaning the shameless ploy actually worked.
Perhaps the backlash had an effect, however, as all future generations supported the same feature sets all the way down the product stack.
#1
NVIDIA GeForce FX 5800 Ultra
Released: January 2003 - Price: $399
Ah yes, the Dust Buster. The Leaf Blower. The Flame Thrower. The King of All Disappointments. The GeForce FX 5800 Ultra.
All the things that could have gone wrong, did. All the hype and anticipation mounted on top of what could only be described as a massive shipwreck on the rocks of NVIDIA’s lofty pride. A champion forced to run a race with a broken leg, amongst a legion of fans that expected only the best. NVIDIA, the undefeated, had shit their silky pants.
But no, it couldn’t be, could it? All the blustering about a CineFX engine, the “128-bit studio precision pixel processing”, the blistering fast DDR2 memory… NVIDIA couldn’t take the 9700 Pro’s brazen assault lying down! This was the company that coined the term ‘GPU’, the one that defeated and consumed the mighty 3dfx, and the one that stood unparalleled in graphics performance for the last four years running!
It was worse. Not only did it not beat the 9700 Pro in most cases, it often needed to sacrifice image quality to do so. Its claims for high color precision and processing were all just words, as in most games the card was forced to run at sub-standard pixel precision in any game that utilized Shader Model 2.0 effects. Effectively it supported DX9 in theory only, as it couldn’t back up its features with enough performance. NVIDIA resorted to underhanded shader-replacement algorithms in its drivers, while also lying about the practice. The shoddier anti-aliasing results were another unneeded blight.
The ramifications of the poorly designed FX 5800 didn’t stop at one card. The entire line was affected by image quality and performance issues, even when they tried to correct some of the problems with the 5900 refresh. As time went on and more support for DX9 emerged, the problems only got worse for those who owned them. The FX 5200, the lowest rung of the product line, is one of the worst remembered cards in existence, and its problems were also the blame of the series’ originator.
NVIDIA bore so much hate, underwent irreparable damage to their reputation, incited endless flamewars online from fanboys who trusted them, and lost valuable market share as a result of a generation of horrendous graphics cards that never seemed to go away. Even today, while they often prove the kings once again in raw performance, there still remains that small strand of conspiracy theory following them wherever they go, casting aspersions on their tactics. They still fight, and often fail, to regain some appearance of integrity. And it all started with one, timelessly infamous card.
No comments:
Post a Comment