gdjacobs wrote:

Even if popularity was a useful metric, we have no data regarding what was actually sold, just some data regarding what was available. Mining Usenet for mentions is nothing more than a statistical wank session. It's not a representative sample of ownership, just a sample of Usenet users.

Even the data on what was potentially available wasn't much to go by. I was always interested in what people chose to buy though since success in the market tends to accumulate on few things. Forums are one way to find trends in adoption since people love to talk about what they have or consider to have. The dedicated thread on vogons for saying you bought a set of things takes up 1170 or so pages, and you see the same effect on other forums. Associations farmed from these discussions will encode an element of 'what goes together'. There's noise in the data of course, and it depends on your level of enthusiasm how much effort you want to put into reducing it. In the end, as with any approach, you'll note the weaknesses and expect to better on them at some point.

In no way does that chatter capture a representative sweep of the market. How many people brag about their boring office computer or the machines they have to use in the department or class computer lab?

gdjacobs wrote:

In no way does that chatter capture a representative sweep of the market. How many people brag about their boring office computer or the machines they have to use in the department or class computer lab?

Obviously the cards that people could brag about their framerate e-penises with would get more conversation, this has not changed in the last 25 years.

gdjacobs wrote:

In no way does that chatter capture a representative sweep of the market. How many people brag about their boring office computer or the machines they have to use in the department or class computer lab?

It doesn't need to though since there isn't a requirement to learn everything in one go. Knowing what it does represent and to what extent would be useful but it's not clear here. We only know that it's likely a group of people so the output won't be random. The words chatter and brag are somewhat caricaturish here though and in themselves unlikely to be sweepingly accurate.

vvbee wrote:

I was always interested in what people chose to buy though since success in the market tends to accumulate on few things. Forums are one way to find trends in adoption since people love to talk about what they have or consider to have.

What about the choices they don't make?
For example, in the past 10-15 years, Intel has been by far the biggest supplier of GPUs.
Most people with an Intel GPU however didn't specifically choose to buy one. It just came with the machine.
Do we see much talk about Intel GPUs here, or on any forum at all? Not in the least. So if you were mining forum discussions for GPU usage, you'd probably get a very skewed measurement for Intel GPUs.

To be fair the Intel GPUs do get talked about quite a bit in the context of how bad they are, but alas, you are correct that the study is not really representative of anything other than how much each card was talked about. You probably can't prove any correllation between discussion and ownership.

appiah4 wrote:

To be fair the Intel GPUs do get talked about quite a bit in the context of how bad they are

They are?
I see an occasional remark here and there, but nothing that would indicate the massive marketshare these GPUs have (I believe it's around 60%?).
I think most talk is concentrated around GeForces and Radeons. And the Radeons are probably over-represented there. Fanboys of both sides are about as loud, but the majority just buys a GeForce, but never engages in conversation (I believe the marketshare is about 80-20 currently, forum talk is more 50-50).
With linux vs Windows it's even worse. The linux community is very vocal online, but the amount of linux users is marginal (somewhere around 1-5%, where about 90% uses Windows, and most others use Apple).

Well, we have to define discussion perhaps. Discussions do not senter around Intel GPUs for sure, but they are almost inevitably mentioned with regards to how much they suck; so if the above study was conducted in a similar fashion, looking into how many times they are mentioned, it would not come up with very low figures, I would think.

appiah4 wrote:

Well, we have to define discussion perhaps. Discussions do not senter around Intel GPUs for sure, but they are almost inevitably mentioned with regards to how much they suck; so if the above study was conducted in a similar fashion, looking into how many times they are mentioned, it would not come up with very low figures, I would think.

Well, my point is that it won't be representative of the market share.
I think we can both agree that considerably less than 60% of all GPU-talk is about Intel GPUs, right?

Scali wrote:

appiah4 wrote:

Well, we have to define discussion perhaps. Discussions do not senter around Intel GPUs for sure, but they are almost inevitably mentioned with regards to how much they suck; so if the above study was conducted in a similar fashion, looking into how many times they are mentioned, it would not come up with very low figures, I would think.

Well, my point is that it won't be representative of the market share.
I think we can both agree that considerably less than 60% of all GPU-talk is about Intel GPUs, right?

Of course, we agree on both points. It would certainly be hugely underrepresented.

Scali wrote:

What about the choices they don't make? For example, in the past 10-15 years, Intel has been by far the biggest supplier of GPUs […]
Show full quote

What about the choices they don't make?
For example, in the past 10-15 years, Intel has been by far the biggest supplier of GPUs.
Most people with an Intel GPU however didn't specifically choose to buy one. It just came with the machine.
Do we see much talk about Intel GPUs here, or on any forum at all? Not in the least. So if you were mining forum discussions for GPU usage, you'd probably get a very skewed measurement for Intel GPUs.

No doubt you'll find varied discussions about intel's solutions, but how well it tells you about the extent of adoption relative to others you'll need to test. It won't be fully accurate nor will figures on availability since a notable number don't use the intel graphics hardware they buy. Likely too you'll find distinctions between the groups of owners who are more likely to actually use them and those that are less likely, so it may not be ideal to try to force the same method to infer about all of them.

I know you're making the case that the millennium was common in the (wealthier) office and people just didn't know, but I don't think the intel gpu is in the same league anyway. It's generic hardware while the millennium wasn't - the latter would've doubled the price over your already expensive cpu in many cases.

vvbee wrote:

I know you're making the case that the millennium was common in the (wealthier) office and people just didn't know

I'm not, just saying that the amount of talk on a given product is not necessarily representative (also gave two examples of the opposite side: Linux and AMD are over-represented in terms of forum talk compared to marketshare).

vvbee wrote:

It's generic hardware while the millennium wasn't - the latter would've doubled the price over your already expensive cpu in many cases.

This is what I already addressed earlier...
The Millennium in 2 MB configuration was only around $330-$340 even shortly after its introduction in mid-1995.
Now, we've also seen that high-end 486 and Pentium 60 systems went for $3000+ around that time (even though you could get a budget 486 for under $1500, and a budget Pentium under $2000).
A Millennium really isn't that big of a deal in that sort of configuration (even a cheap video solution would easily set you back $150-$200, so the premium you pay for a Millennium is not that big of a deal).

Also, why are you concentrating on only the CPU? Back in those days, the rest of the system was considerably more expensive relative to the CPU than it is today. Especially the high-end machines, which came with network cards, SCSI HDD controllers, and a lot of memory.

It's not a way to measure adoption that you can rely on without criticism, I'll agree and have criticized it. Only, you need to be critical of your own approach as well if you want the pursuit to go somewhere.

Why the cpu? Because that's what the intel gpu comes with. You can look at the total cost of the system though and say, you're paying a couple thousand so what's $200 more for an even better video card. Yet the video card isn't the only thing you could put it toward, if you could. Some would look at it and say, where am I going to pinch the $200 to afford this system, and strike out the pimped-out millennium. It depends on what they were wanting to do with the machine.

vvbee wrote:

Only, you need to be critical of your own approach as well if you want the pursuit to go somewhere.

No, we don't, since I merely tried to answer a question.
I don't need nor want this pursuit to go anywhere.
I am just being critical of the 'data' you keep bringing up and the 'argumentation' you are using.

vvbee wrote:

Why the cpu? Because that's what the intel gpu comes with.

Wait what?
I thought we were talking about the Millennium and the Pentium 60 here. Intel GPUs and CPUs have nothing to do with that (aside from the fact that you can't put a real price tag on Intel's GPUs anyway, since they are not sold separately).

vvbee wrote:

Some would look at it and say, where am I going to pinch the $200 to afford this system, and strike out the pimped-out millennium. It depends on what they were wanting to do with the machine.

That's the thing, we are talking about Dell and Compaq machines that are twice as expensive as the 'bargain basement' competition anyway.
If you are shopping for these brands, you are clearly not trying to pinch off a few hundred bucks. By definition you are going for 'the best that money can buy', or something in that general direction.
And in that clientele, either the OEM will make the choice for you: "We're just putting Millenniums in all our high-end machines, period"... or, a reasonable share of clients will say: "I'm spending this much money on a machine I will be using for many hours every day for years to come, I want to have the best possible image quality and performance as well, I'm not going to cheap out on the video card and monitor".

Why are we having this argument anyway? I mean, what could it possibly lead to? It was already established that Pentium 60s with Matrox Millenniums were indeed sold.
I still don't get why you have to keep beating this dead horse. Why this cognitive dissonance? How difficult is it to grasp that $3000+ Pentium 60 system + $340 Matrox Millennium + $1500 high end CRT (Eizo or such) makes sense? The Millennium isn't the thing that's going to make or break the bank.
These people aren't penny pinchers. Penny pinchers would never buy the type of 'brand machine' that Millenniums would come in by definition.

I had a penny-wise, dollar stupid university professor who likely would have tried to buy such a high-end system without the video card, however, I don't think he was the norm.

Scali wrote:

vvbee wrote:

Only, you need to be critical of your own approach as well if you want the pursuit to go somewhere.

No, we don't, since I merely tried to answer a question. I don't need nor want this pursuit to go anywhere. I am just being critical of the 'data' you keep bringing up and the 'argumentation' you are using.

Thanks.

vvbee wrote:

Knowing what it does represent and to what extent would be useful but it's not clear here. We only know that it's likely a group of people so the output won't be random.

That's the point. We have no way of applying your usenet data as a proxy for the overall consumer market, so it's simply a waste of time.

gdjacobs wrote:

vvbee wrote:

Knowing what it does represent and to what extent would be useful but it's not clear here. We only know that it's likely a group of people so the output won't be random.

That's the point. We have no way of applying your usenet data as a proxy for the overall consumer market, so it's simply a waste of time.

It's a survey of usenet users, not a very reliable one, but it is what it is. I wouldn't call it a waste of time, but it doesn't prove any points he previously made.

By the way, here is the Matrox Mystique 220 Driver CD image that I had promised earlier.

Unfortunately my CD does not seem to include that video or I have been unable to locate it.. I believe that video must be from the original Mystique's driver CD.

gdjacobs wrote:

vvbee wrote:

Knowing what it does represent and to what extent would be useful but it's not clear here. We only know that it's likely a group of people so the output won't be random.

That's the point. We have no way of applying your usenet data as a proxy for the overall consumer market, so it's simply a waste of time.

You missed the point. The point is that you shouldn't try to apply it that way or it's a waste of time.