Even at a sample size of 30+M, is the Steam GPU survey accurate?

[-Slacker-]

http://store.steampowered.com/hwsurvey/videocard/

As an example taken from the table, the gtx 580 is reported as owned by 2.49% of the dx 11 user base, whereas the hd 6850, a card that came over a month earlier and cost almost 3 times less, represents only 1.85% of the dx11 video cards.

I find it hard to believe that this is an accurate representation of the gtx 580 / hd 6850 market ration across the world, so, if my suspicions are right, I'd like to know what makes surveys that are literally thousands of times smaller in sample sizes more credible and accurate sources for assessing all kinds of statistics.
​

 

[OCGuy]

GTX580 is the fastest GPU out there. 6850 was pointless as soon as it came out. Isnt it slower than the 5850?

I don't know anyone who even considered the 68XX series...it seemed a little lost in the mix sitting between 58XX and 69XX.

 

[OILFIELDTRASH]

GTX 580 owners tend to be smarter, better looking, and drive nice cars.

 

[Arkadrel]

the 6850 is like 20% slower than a 560 ti.

But you can get them down to like 137$ on newegg after MIR.
That alone should probably make it more sold than the 580.

The 6850 is priced competitivly... its slightly faster than a 460 1gb.
I dont see why it wouldnt sell.

 

[happy medium]

I find it hard to believe that this is an accurate representation of the gtx 580 / hd 6850 market

Its not, I have learned of a bug that cannot detect some AMD cards in the Steam survey.
This was covered in a thread a while back.

 

[notty22]

The 6 series in general appears to not be selling well, or showing up in Steam's survey anyway. The gtx 570 came out about a week before the 6970/50 and shows a higher % also.
Maybe thats why the whispers about a AMD 7 series lately.

 

[Mr. Pedantic]

GTX580 is the fastest GPU out there. 6850 was pointless as soon as it came out. Isnt it slower than the 5850?

I don't know anyone who even considered the 68XX series...it seemed a little lost in the mix sitting between 58XX and 69XX.

Lower cost cards tend to sell better than higher cost cards. The 68xx series wasn't as well received as it probably could have because of the 5850 and 5870, but it still should be selling better than a flagship card like the 580.

 

[notty22]

Its not, I have learned of a bug that cannot detect some AMD cards in the Steam survey.
This was covered in a thread a while back.

That is a issue about crossfire I believe. There are only 3 vendors that can show up/ or be detected. Most certainly AMD cards are not showing up as Nvidia , P

 

[-Slacker-]

Its not, I have learned of a bug that cannot detect some AMD cards in the Steam survey.
This was covered in a thread a while back.

That is a issue about crossfire I believe. There are only 3 vendors that can show up/ or be detected. Most certainly AMD cards are not showing up as Nvidia , P

Hmm so, then .... uh, yeah, I'm still confused...

 

[notty22]

Some gameres refuse to use digital d/l's and don't use Steam.

There is a new issue I just learned about, that Dirt 3 will have no physical box launch in North America.

 

[Lonyo]

No it's not accurate in any way, shape or form.
We've been through this many times and there are many different ways in which it is inaccurate, and many flaws that show up in its reporting over time.
Search for any thread which has Steam in the title from this specific subforum and at least some of the problems with it will be posted.

 

[-Slacker-]

Some gameres refuse to use digital d/l's and don't use Steam.

Yes, but we're talking about a survey with a figure of 30 to 40 million ... if surveys this big can have such a large variance, then, how is it that the much smaller ones (with participants that are only in the thousands, if not hundreds) get so much credit, and what methodologies are used that make them so much more reliable? Unless it's like happy says, and there's a bug that prevents the site from cataloging it's user's hardware accurately.

edit

No it's not accurate in any way, shape or form.
We've been through this many times and there are many different ways in which it is inaccurate, and many flaws that show up in its reporting over time.

Ah ok, that's that, then.

 

[OCGuy]

No it's not accurate in any way, shape or form.
We've been through this many times and there are many different ways in which it is inaccurate, and many flaws that show up in its reporting over time.
Search for any thread which has Steam in the title from this specific subforum and at least some of the problems with it will be posted.

I agree with this. One poster on this forum likes to post the survey as proof of marketshare, and I always question the scientific nature of that.

 

[Fox5]

I think nvidia has bigger mindshare among "hardcore" gamers. Or at least most people I've met in person who are into PC games are like "ATI? Don't they suck? NVIDIA 4 LYFE!"
Nvidia's been much better at marketing to online gamers I think, what with almost every multiplayer game proudly showing their logo on boot.

 

[drizek]

GTX580 is the fastest GPU out there. 6850 was pointless as soon as it came out. Isnt it slower than the 5850?

I don't know anyone who even considered the 68XX series...it seemed a little lost in the mix sitting between 58XX and 69XX.

THe 6850 gives you good performance/watt, pretty useful for people with sub 400W PSUs. If it was $20-30 cheaper, it would be very compelling against the GTX 460 and all the other sub-560Ti cards that nvidia has(an ungodly number, tbh)

 

[Nox51]

Yes, but we're talking about a survey with a figure of 30 to 40 million ... if surveys this big can have such a large variance, then, how is it that the much smaller ones (with participants that are only in the thousands, if not hundreds) get so much credit, and what methodologies are used that make them so much more reliable? Unless it's like happy says, and there's a bug that prevents the site from cataloging it's user's hardware accurately.

edit



Ah ok, that's that, then.

You have to relate the sample size to the actual population size. Just because the sample size is a large numeric number doesn't mean it is appropriate for a really large population.

Also just because you have a large sample size doesn't mean you are capturing a non biased representation of the population. The survey nature doesn't rely on random selection of participants and, as previously mentioned, has encountered several technical difficulties.

Going outside of the steam population and trying to relate the findings to the whole computer user population is an exercise that will encounter many errors.

 

[Athadeus]

the 6850 is like 20% slower than a 560 ti.

But you can get them down to like 137$ on newegg after MIR.
That alone should probably make it more sold than the 580.

The 6850 is priced competitivly... its slightly faster than a 460 1gb.
I dont see why it wouldnt sell.

I've been looking at that MSi one all weekend, but I'm still afraid it's a mislabeled open box or something. MSi wouldn't sell a failed OC version of that card at stock as new, would they?

 

[busydude]

Its not, I have learned of a bug that cannot detect some AMD cards in the Steam survey. This was covered in a thread a while back.

Yup, steam could not detect my 5750 and 5770.

Although it has a huge base, I do not consider this as an actual representation.. way to much variation each month. That said, this can be a powerful tool for fanbois, on either side, to sling mud at each other.. .

 

[Dribble]

I have found that people find it completely accurate if it proves their point, and completely inaccurate if it doesn't.

My personal opinion is it's probably a pretty good indication of gamers as steam users come from all walks of PC gaming life and the survey is large enough to avoid most statistical anomolies.

 

[scooterlibby]

The sample size is very good, but I wouldn't trust it because of possible selection bias. The same thing that you in the way see Newegg reviews skew - the type of people who respond to surveys or write reviews are not random.

 

[Mopetar]

Lower cost cards tend to sell better than higher cost cards. The 68xx series wasn't as well received as it probably could have because of the 5850 and 5870, but it still should be selling better than a flagship card like the 580.

They probably would have gone over a lot better if they were branded as 67XX instead of 68XX. Especially considering that the actual eventual 6700 series turned out to be rebrands or 6850 rejects.

 

[dualsmp]

The numbers probably are not that far off. There is a fairly large group of gamers that buy nothing but the fastest available and won't settle for less. The GTX 480 has an even bigger user base than the GTX 580.

The 68xx wasn't priced aggressively enough, and think many of those sales were lost to the GTX 460.

 

[gramboh]

If there are underlying issues with detection of certain hardware and reporting, then the information might not be accurate.

But to claim it is not an accurate sample of PC gamers is insane. 30-40M is a pretty large chunk of active gamers, and Steam now has games from all the major publishing houses including EA, it's not just CS players anymore. There are 2+M people using it at a given time, that is huge.

The people still resisting digital distribution are a minority.

 

[Tempered81]

GTX580 is the fastest GPU out there. 6850 was pointless as soon as it came out. Isnt it slower than the 5850?

I don't know anyone who even considered the 68XX series...it seemed a little lost in the mix sitting between 58XX and 69XX.

As often as I disagree with ocguy, he's absolutely right on this. 6850 is a POS card. People who do their homework would buy a 580. 6850 is crap.

5850 to 6850 naming & marketing was the dumbest thing amd/ati has done in recent memory.

I completely agree that the steam survey is accurate. No offense to 6850 owners, the card has value in it's price and features (morph aa), but it's stupid name that does not fit in where you think it should performance wise.

Steam survey is accurate.

The 460 was a bomb deal, fast card for cheap and it scooped up share

amd had 90% of the market with the 57xx and 58xx for the longest time since the first fermi was a flop.

Steam is the most correct representation you can find, IMO. Just watch out since it goes by the name in the driver string/firmware seen by the os

4800 series

9800 series

those two groups categorize like 30 individual products, whereas 580 & 6850 are 1 part

 

[alcoholbob]

Nvidia is like Intel.

If they have a product that's simultaneously slower AND costs more, people would still buy it and ignore the competition. Didn't AMD have a faster AND cheaper CPU lineup for like 3 straight years over Intel and their market share barely made a move?

Marketing > Performance. Bose 101.