Future of my support for ATI

[Porter]

The Infoweek statistics are totally skewed. The reason why they're skewed is that they are collected by a company called Net Applications... and Net Applications collects the stats only from THEIR customers who use the Net Applications HitsLink tool on their small-business e-commerce websites, to track and analyze traffic.

Therefore, these browser/OS stats are only being collected from people who shop via e-commerce from small/medium businesses that happen coincidentally to use the HitsLink software on their web sites.


That's a pretty crappy way to sample the "Internet" in terms of real users. In fact, in my opinion the targeted sample pool is completely inappropriate for getting an idea of real-world industry-wide market share, so the numeric results in general shouldn't be considered significant. If this thing has less than a 6-7% margin of real-world error, I'd be very surprised.

[rbmorse]

This is no doubt true, but it's not any worse than "me and my friends," or, "I talked to a couple of guys," or, "I just have to believe that it's more than 1% because....because....well, because I want it to be" which are the other predominate metrics we've seen here.

[Dragonlord]

At least those sources you "deny" state that it's a "guess" while the so called "statistical sources" claim to be representative but obviously fail. I prefer an honest "i guess" over blind statistics by a long shot.

[duby229]

double post sorry guys...

[duby229]

At least those sources you "deny" state that it's a "guess" while the so called "statistical sources" claim to be representative but obviously fail. I prefer an honest "i guess" over blind statistics by a long shot.

I have to agree with this sentiment, though, at some point guess just turn into blind speculation.

Really I'd like to see numbers that have been collected by some other means then web hits, and that is not motivated by someones greed for profit... And frankly I just dont think it is possible. I dont think we'll ever see accurate numbers, and we should all just take a break, and say that the numbers are certainly higher then what these biased statistics show, but we dont know what they actually are.

I think that is a reasonable compromise for both sides to take...

First of all it benefits use in that becosue we can assume higher numbers, we can expect better support. And it benefits the other side in that becouse they assume higher numbers they can expect better income.

In the end the bottom line is that we are stuck in a catch22.... If we dont have the support and drivers we need, then you'll never get the sales that you need from this market segment. It's a self fulfilling prophesy. If the drivers and support are provided then the sales will increase as a result.

Take my case as an example, If you just looked at page hits from unique computers I'd probably count for hundreds of windows users just by myself. I fix Windows computers for a living, and so those numbers have absolutely no value... And after the HD4850 was benchmarked, I was totally stoked, and would have bought one right then and there. The problem was that it is stuck on a piece of shit proprietary driver, and no hardware acceleration from the open drivers.... So instead I am stuck with a 1950Pro that was given to me from my dad. You You didn't make a single dime off of me, because my father already payed for that card 2 years ago. Instead of getting that 4850 sale, you now have to worry about supporting an older product that I didn't even pay for that you didn't make a single dime off of. Wouldn't it have been so much better to have proper drivers from the beginning and gained that extra 4850 sale? Dont you think the benefits of having a proper driver outweigh the drawbacks?

Theres an old saying that I'm sure everybody has heard once or twice... If you build it they will come..... Thats the catch22 I'm talking about.

[rbmorse]

At least those sources you "deny" state that it's a "guess" while the so called "statistical sources" claim to be representative but obviously fail. I prefer an honest "i guess" over blind statistics by a long shot.

I don't deny anything, but in general I prefer to work with data even if I know the data is incomplete...so long as the limitations and omissions from the dataset are understood...over pure, rank, baseless speculation, if that's what you mean.

But, then again, I don't work in climate science or marketing.

[Redeeman]

Geez, even I can answer that one

The "expensive cards" you mentioned are generally expensive because they sell in relatively small quantity. GPUs sell for a lower price but sell in much higher volumes, and the R&D cost (and hence the IP investment) for a GPU is at least an order of magnitude higher than any of the products you mentioned.

In addition, the programming interface for most of those products is relatively far removed from most of the actual IP investment -- in a raid card the proprietary stuff is typically the firmware on the card which a driver developer doesn't get to see, and in a high speed network card (a) you typically have big firmware again and (b) most of the "secret sauce" is in the analog bits not the digital bits.

Finally, none of the products you mention need to strike a balance between open-ness (for Linux) and regulation (FCC cert, DRM robustness etc.. ) the way that GPUs and modems do. Not sure what the deal is with printers but I suspect the margins are just so thin that the cost of providing programming documentation would be the final nail in an already desperately competitive business.

yet the most commonly sold devices, which far outshadow the numbers of sold that gpu's have, i dont know, say... chipsets, ide controllers, CPUs have no problem with documentation?

the simple fact is, nvidia, amd, everyone, wouldnt loose a DIME if they just shelled out complete docs, well, except in the case of ATI where the management were complete and utter morons which seemingly didnt HAVE documentation..

your problems are problems you create for yourself, such as paying for insanely expensive lawyers to put a "ok" stamp on a document which you can bet your ass they didnt even understand ANYTHING of..

and as far as fcc goes.. give me a break, that may be the answer you get should you call some fcc lawyer, but it wont EVER stick anywere.. its not as if its hard to get around anything, source/documentation form or not, and its not as if the workings of a radio is top classified material, hell, you can even order premade kits available for children to assemble and play with..

[energyman]

eh, Redeeman, creating Documentation is a costly process - and you HAVE to check with lawyers because of all the contracts with external 'ip sources' and Hollywood.

Also there is one thing to make a short ranged, low power radio set to play on your property or connecting to a land line and maybe killing all other modems/telephones in the range of a kilometre.

[Porter]

It would be interesting to see the distros like Ubuntu publish their download figures... or use the package update servers to track the total number of deployed, working systems.

[Redeeman]

eh, Redeeman, creating Documentation is a costly process - and you HAVE to check with lawyers because of all the contracts with external 'ip sources' and Hollywood.

Also there is one thing to make a short ranged, low power radio set to play on your property or connecting to a land line and maybe killing all other modems/telephones in the range of a kilometre.

well... you may say that the radio thing isnt an issue then, i agree, yet this is the excuse intel initially shelled out for creating the abomination of a "regulatory daemon" in binary only format..

and this is the same kind of crappy excuses everyone else comes out with..

Yes, creating documentation is costly, but its the only decent thing to do, and it ought to be the law.