I've been thinking about writing a longer article on this subject, but I'm not sure I'll have time, so I'll just say something here. I have been reading commentary around the web lately about a number of recent graphics-related news stories like the GeForce FX pipeline controversy, the 3DMark03 controversy, and the latest NVIDIA cheating story. Most of these stories are part of a single larger story: the NV30's fizzle-and-die act. Much of this commentary has taken a particular tone, couching itself as a sensible and rational response to an overheated situation ("it doesn't matter how many pipelines at the end of day" or "the number of pixel shader units doesn't really matter"). And, in some cases, there is a fair amount of truth in such comments.
However, something more needs to be said. There are reasons we here at TR have followed these stories closely and considered them important to our task. Many of these stories have mattered because there was a fundamental issue of truthfulness involved. In the struggle to bring products to market in the midst of fierce competition and volatile stock prices, companies sometimes do things they ought not to do. For instance, NVIDIA arguably misled the press, analysts, and shareholders with its refusal to divulge the NV30's actual pipeline configuration. The relevant issue here isn't really how many pixels per clock the chip could produce when doing "color + Z" rendering; what makes this a big news story is the discovery that something isn't quite what we'd been led to believe. We may think it's no big deal that NV30 is a 4x2-pipe designheck, I'll tell you straight up it isn'tbut obviously someone inside NVIDIA thought it was a big enough deal to attempt to obscure this fact.
I apologize if I'm sounding rather obvious to many of you, but unfortunately, such things sometimes need to be said explicitly, and this is apparently one of those times. Over the past weeks and months, I've seen writers I respect write things that are uncharacterstic for them. I've seen people I respect defending plainly obvious benchmark cheats. I've seen prominent authors defend semiconductor companies' choices to obscure basic chip information from the public, supposedly because of a need to protect a competitive advantage in the face of short product cycles. (Sure, I can see that; but shouldn't we as journalists generally be inclined toward more, not less, public disclosure of information?) It's disheartening.
I don't want to get up on a high horse here myself. I am far from perfect, fallible beyond my ability to see it. But I will say this much. We at TR have tried to cover things as honestly as we can, even when the news is not good. And we have paid a price for doing so. Regardless, we will continue to pursue our tasks with an eye toward truthfulness, in our reporting, in our empirical testing, and in our editorial analyses. If you sometimes wonder why we dwell on seemingly small technical issues, please consider the context. We aren't always too concerned ourselves with how many angels can dance on the head of a pin, but we are definitely interested to learn that those dancers may not be angels after all.
|Biostar X370GTN is the first Ryzen Mini-ITX motherboard||4|
|MakeVR and Vive Tracker get HTC Vive ready for work and play||0|
|Intel gives hard drives a boost with Optane Memory||24|
|Starcraft Remastered constructs higher-fidelity pylons||28|
|Transcend steps into the NVMe arena with the MTE850 SSD||6|
|MSI GTX 1080 Ti Armor 11G is the first custom card on e-tail shelves||9|
|Gigabyte has two A320 boards for bread-and-butter Ryzen builds||31|
|Google points deep-learning machines at audio effect subtitles||5|
|Throw a Quadro card on Gigabyte's Z270X-Designare||14|
|Well, so much for Common Courtesy Day...||+32|