Catalyst 10.6 - Video Quality Improvements

Couple of AnandTech anecdotes here that I have never forgotten that sort of illustrate the trends you are talking about with respect to the attitude of the publication:

I think AnandTech is just plain sloppy at times--not maliciously--but just sloppy in a dopey kind of way. I will never forget Anand Le Shrimp's (pardon my French) initial review of the 3dfx Voodoo 3 card a few years ago. The screen shots AT posted that purported to show the IQ of the V3 were, in a word, awful looking, with an image quality so sub par that the Voodoo 2 looked like a million $ comparatively. Yet, AT actually published those pictures with only one small sentence in the article as a caveat, which I'll paraphrase thusly: Anand wrote, "These screen shots look really bad and that's what I couldn't understand, because these screen shots look nothing like the image quality I saw on the screen while testing this card." And, in characteristically enigmatic fashion, that's exactly where AT let the issue lie.

A few days later when somebody bothered to take the time to inquire of 3dfx what the problem might be with screen captures from the V3, it was revealed that due to the post-processing effects new to the V3 you *could not use the screen-grabbing software used commonly at the time* in order to grab frames from the V3 that looked remotely like the images the V3 displayed on the screen. 3dfx supplied a screen-capturing program that included the post-processing data, and suddenly the screen grabs looked just like what people saw on the screen and their IQ was excellent. Anand never to my knowledge corrected his nearly unforgivable error in his initial V3 review, even though his screen grabs were so distorted and inaccurate that nobody in his right mind would have bought a V3 after reading that AT review of the product and looking at those uuuuu-u-u-u-ugly screen shots.

Then there was the months-long "AGP texturing" saga/debacle in which AT and a few other web sites at the time were heralding as the Second Coming of 3D Graphics. Guess who wasn't on the AGP texturing bandwagon at the time? Yep, it was 3dfx, and AT let 3dfx have it with both barrels for being so stupid and stubborn that they couldn't see the incredible importance that AGP texturing would hold for the future of 3d gaming. No matter how many times and in how many ways 3dfx tried to explain that texturing out of local ram on a gpu would always be multiples of time faster than texturing out of system ram across the AGP bus--AnandTech just couldn't get it. Flash forward to today and AGP texturing, not to mention PCIe texturing, is all but dead and buried as a performance solution, because the fastest PCIe texturing from system ram is multiples of times slower than texturing from local ram on a 3d card. And that's why, of course, today's 3d cards offer anywhere from 512mbs to upwards of 2 gigabytes of *local* ram--and that's the ram they texture from--the system buses are far too slow to texture from. This was another instance that when the facts came out in the wash AT never bothered to correct or amend them. My guess is that getting the facts straight, or at least non-prejudicial, just isn't a priority at AnandTech.

I really felt for 3dfx during those last days--the company was being pummeled from left and right by so-called self-appointed web technocrats who not only didn't listen to much of anything 3dfx had to say, but failed as well to even understand what 3dfx was trying to say, and as a result self-righteously reached the wrong conclusions and ran with them. The exquisite irony of the day came out of nVidia, seeing how 3dfx was getting a bad rap over avoiding AGP texturing, nVidia jumped on the bandwagon and claimed that its products were taking "full advantage of AGP texturing"--and AnandTech dutifully lapped it up and like a mindless parrot spun it to the heavens.

When it was later revealed that not only did the nVidia cards of the day make use of the exact same amount of onboard ram as the non-AGP texturing 3dfx cards, but that they also relied on local texturing instead of AGP texturing for their 3d frame-rate performance--uh, just like 3dfx which had at least been honest about the whole issue from the start--AnandTech never acknowledged it or made the corrections in previous articles in which such corrections most assuredly should have been made.

In most technical issues of this nature, 3dfx had it right--the web pundits had it wrong. I sometimes wonder if the web techies have learned much of anything in the interim. It's for sure they've become masters at procuring marketing dollars for themselves by pumping some products at the expense of others, but one hopes that their technological acumen has improved enormously since the days of 3dfx. I hope so, but I cannot swear to it, unfortunately.