Steam Hardware Survey updated with GPUs

Why would the AMD fanboys want to skew the survey? What could they possibly gain by providing fake numbers?
Mind share? Brand recognition? Consumer confidence?
The idea is that if AMD appears to make large gains on the Steam survey, people will see that and be more likely to buy AMD products. It's also just confirmation bias, 'AMD is great and their numbers going up on Steam prove it.'

With miners sucking up all the cards these days, the Steam figures will probably trend down over the next few months, but AMD is still selling a lot of GPUs.
 
Last edited:
Since they are polling randomly then it isn't much different. But again you have any factual basis or information as to accuracy then by all means share.
If you look at the original Steam Surveys between 2004-2007, they were polling ~1,000,000 users when their concurrent daily users were less than 4 million. Today Steam has ~13.5 million concurrent daily users, while you only need ~10% polled for >99% confidence at +/- 0.005% statistical accuracy. It seems rather unlikely they would be polling any less than they were originally, even as their Steam users have skyrocketed. The major difference between then and now, is Steam is now capable of gathering enough surveys for statistical significance monthly, rather than needing to wait 3-4 months.


If you take a 10-year view of Steam Surveys, you can see that NVIDIA is only beginning to show long-term growth on Steam this year (Pascal vs Polaris), while AMD has been showing a trend of long-term loss since the beginning of 2013 (Kepler vs Tahiti/Hawaii):


Steam May-Aug 2007:
57.74% NVIDIA, 35.20% AMD, 2.98% INTEL, 4.08% Other


Steam August 2012:
46.68% NVIDIA, 35.58% AMD, 11.18% INTEL, 6.56% Other


Steam August 2014:
51.06% NVIDIA, 29.65% AMD, 18.92% INTEL, 0.37% Other


Steam August 2016:
57.57% NVIDIA, 24.44% AMD, 17.56% INTEL, 0.43% Other


Steam May 2017:
62.41% NVIDIA, 21.53% AMD, 15.65% INTEL, 0.41% Other


You can kind of see the story here, and how NVIDIA really had a major fuckup with Fermi, and similarly how AMD's GCN has been poorly received to Steam users when compared against NVIDIA's change in direction with Kepler/Maxwell/Pascal.

We would hope that Vega will halt AMD's downward spiral on Steam, but even for NVIDIA, regaining former marketshare took an excess of four additional years when they were doing well.
 
It doesn't really reflect on these charts but AMD is selling every polaris (Ellesmere) die they can produce at the moment and we all know why...

Expect another price crash on used gpu's in couple of months.
 
It doesn't really reflect on these charts but AMD is selling every polaris (Ellesmere) die they can produce at the moment and we all know why...

Expect another price crash on used gpu's in couple of months.


Its going to be next year some time, Eth mining isn't going to stop till early next year, then their are 2 or 3 other coins those miners can switch to and keep the money coming in. So its going to be a while. (or until the next gen cards show up). Unless Vega's price and ROI is similar to the rx 580/70, but that probably won't happen.
 
It doesn't really reflect on these charts but AMD is selling every polaris (Ellesmere) die they can produce at the moment and we all know why...

Expect another price crash on used gpu's in couple of months.

Because they messed up logistics after going from Polaris 1x to 2x?

Mining exist on NVidia too, see zcash for example.

Remember the last mining craze? Ye, AMD didn't sell much there again due to logistics. Remember how people thought Polaris originally sold like crazy? ye that wasn't the case either.

Seems RX 5xx cards will come in stock again in in around 10 days.
 
I like that you have no facts...nor it seems any clues to polling.
Try again.

cyberbeing said they were polling 1 million out of 4 million users in 2004-2007, that's something you could have said instead of "no facts, try again".
 
Steam June

DX12 systems. This requires Windows 10 with a DX12 GPU.
GTX 1080TI unlisted or below threshold.
GTX 1080 1.24% (+0.05%)
GTX 1070 2.50% (+0.20%)
GTX 1060 3.23% (+0.39%)
GTX 1050TI 1.20% (+0.29%)
GTX 1050 0.76% (+0.19%)
GTX 1030 unlisted or below threshold.
RX 580 unlisted or below threshold.
RX 570 unlisted or below threshold.
RX 560 unlisted or below threshold.
RX 550 unlisted or below threshold.
RX 480 0.70% (-0.03%)
RX 470 0.24% (+0.01%)
RX 460 0.26% (+0.02%)

Installed base.
AMD loses 1.0% in graphics.
Nvidia gains 1.2% in graphics.
Intel loses 0.1% in graphics.
Intel gains 0.8% in CPU.
AMD loses 0.8% in CPU.

0.01% change in VR headsets up to 0.23%.

DX12 GPUs alone without OS requirement.
GTX 1080TI unlisted or below threshold.
GTX 1080 1.78% (+0.05%)
GTX 1070 3.60% (+0.26%)
GTX 1060 5.19% (+0.17%)
GTX 1050TI 2.80% (+0.76%)
GTX 1050 1.74%(+0.20%)
GTX 1030 unlisted or below threshold.
RX 580 unlisted or below threshold.
RX 570 unlisted or below threshold.
RX 560 unlisted or below threshold.
RX 550 unlisted or below threshold.
RX 480 1.02% (-0.04%)
RX 470 0.39% (+0.01%)
RX 460 0.51% (0.00%)

8 core CPUs grew 0.06%. 6 cores grew 0.01%.
 
so, according to Steam's data the most popular gpu is the GTX1060 ?
I was rather surprised, expecting the GTX1070 to be honest. (*or RX480 from AMD's side)
 
so, according to Steam's data the most popular gpu is the GTX1060 ?
I was rather surprised, expecting the GTX1070 to be honest. (*or RX480 from AMD's side)

Yes, the problem tho is that 1060 is spanned over 2 cards. The GTX 1070 at least for now is still outselling the individual 1060 cards I guess.
 
lol in the process of Troubleshooting an issue with my 390X, now installed in my personal machine was asked by steam a third time this year to do the survey... weird..

Hardware survey.png
 
lol in the process of Troubleshooting an issue with my 390X, now installed in my personal machine was asked by steam a third time this year to do the survey... weird..

View attachment 30228
Never have on this Machine 290/8350 but multiple time on my livingroom machine 7770/965BE (retired) now a 7870K only but not surveyed on this one yet but only been 2-weeks.
 
Steam July

DX12 systems. This requires Windows 10 with a DX12 GPU.
GTX 1080TI 0.40% (+0.08%)
GTX 1080 1.27% (+0.03%)
GTX 1070 2.40% (-0.10%)
GTX 1060 3.32% (+0.09%)
GTX 1050TI 1.46% (+0.26%)
GTX 1050 0.96% (+0.20%)
GTX 1030 0.01% (+0.01%)
RX 580 0.07% (0.00%)
RX 570 0.01% (0.00%)
RX 560 0.02% (0.02%)
RX 550 0.01% (0.00%)
RX 480 0.64% (-0.03%)
RX 470 0.21% (-0.03%)
RX 460 0.26% (0.00%)

Installed base.
AMD loses 0.3% in graphics.
Nvidia gains 0.3% in graphics.
Intel stays flat in graphics.
Intel loses 0.3% in CPU.
AMD gains 0.3% in CPU.

Flat but slightly negative change in VR headsets.

DX12 GPUs alone without OS requirement.
GTX 1080TI 0.53% (+0.11%)
GTX 1080 1.74% (+0.04%)
GTX 1070 3.41% (-0.14%)
GTX 1060 6.37% (+0.17%)
GTX 1050TI 3.30% (+0.54%)
GTX 1050 2.17% (+0.45%)
GTX 1030 below threshold.
RX 580 below threshold.
RX 570 below threshold.
RX 560 below threshold.
RX 550 below threshold.
RX 480 0.92% (-0.08%)
RX 470 0.34% (-0.04%)
RX 460 0.53% (+0.03%)

8 core CPUs grew 0.03%. 6 cores grew 0.02%.
 
So the "treshold" is below 0.34% (RX470 is listed with 0.34%)
Lets make it simple and say ~0.33%

That means that the RX580, RX570, RX560 and RX550 (~0.33% + ~0.33% + ~0.33% + 0.33% = ~1,32%) at BEST have less market share than the 1080 (1,74%).

Just to put things in perspective...at BEST case senario.
 
Technically the threshold is around 0.31% for DX12 GPUs without OS, but you are right. One could also say all the RX4xx cards combined have sold around what the GTX1080 did alone :)

However you just have to look at the DX12 Systems with Windows 10 to see the trend. And there is no mining excuse for the 550 and 560. The 1030 is also pretty much DOA.
 
Last edited:
The gains for AMD CPUs and >4 core processors seem to be attributable to Ryzen.

8 and 6 cores only grew 0.05% combined for Intel and AMD. So AMD have had to sell quite a lot of 4 cores and below to make the change. When you look at graphics its easier to see why, its APU sales.

Remember over half of all gaming systems sold today are mobile.
 
That means that the RX5*** ....at BEST have less market share than the 1080 (1,74%).

Just to put things in perspective...at BEST case senario.


Amongst Steam users (and only amongst those responding to the survey).
 
So the "treshold" is below 0.34% (RX470 is listed with 0.34%)
Lets make it simple and say ~0.33%

That means that the RX580, RX570, RX560 and RX550 (~0.33% + ~0.33% + ~0.33% + 0.33% = ~1,32%) at BEST have less market share than the 1080 (1,74%).

Just to put things in perspective...at BEST case senario.

Steam survey is pretty worthless man. DX12 with windows 10 if you add it all up only shows 11.04% total owning any of those cards, which means 88.96% is using something else. So if I take this survey literally Nvidia and AMD both have terrible market penetration. Hell without anything more then just a dx12 gpu it shows only around 17% owning a card that can do that, if you think that is correct amongst gamers then your crazy.
 
I love when facts are dismissed, based on nothing.

#EchoChamber

Reject the data all you want...I won' loose any sleep...but I will ignore you like I ignore flat-earthers, chem-trailers, creationists and other people in the sae category...
 
The 1030 is also pretty much DOA.

The 1030 is intended for people who want two 4k @60hz at low-power on the cheap, or want a passive-cooled half-height single-slot gaming card. Small market, but it's been awhile since Nvidia released such a card (GT 640, which sucked), so there's some pent-up demand.

This thing is almost as fast as a GTX 750. Still overpriced for the gaming performance, but how many of those SFF systems from DELL/HP have any other choice for an upgrade?
 
Last edited:
I just got asked to do the Steam survey a few days ago for the first time in 4-5 years, so my Titan X was added to the insignificant percentage of users on Steam who have one.
 
It's not accurate for AMD, some of the 200-series still poll as their HD7000 counterparts. I imagine the same is true for 300 vs 200 series.
The only accurate data is to sum all of them together for a total representation, you can't look at individual GPUs.

Case in point: According to Steam, there are 24x as many GTX 970s as R9 390s. Even if we're generous to Nvidia and assume 4:1 sales over the same time period, the GTX 970 is 2 years old vs 1 year on the R9 390, which makes 8x.

Also according to them, the 980 Ti outnumbers the 380 & 390 combined and doubled (or 4x as many as the 380 by itself). Doubt it.
The Fury is also being detected incorrectly.
 
Steam survey is pretty worthless man. DX12 with windows 10 if you add it all up only shows 11.04% total owning any of those cards, which means 88.96% is using something else. So if I take this survey literally Nvidia and AMD both have terrible market penetration. Hell without anything more then just a dx12 gpu it shows only around 17% owning a card that can do that, if you think that is correct amongst gamers then your crazy.

Both DX12 GPUs alone with ANY OS and DX12 GPU with Windows 10 are counted separately. And almost 78% of all steam users got a DX12 GPU.
 
so, according to Steam's data the most popular gpu is the GTX1060 ?
I was rather surprised, expecting the GTX1070 to be honest. (*or RX480 from AMD's side)

The low to mid market are always the most popular. Not surprising at all.
 
Steam survey is pretty worthless man. DX12 with windows 10 if you add it all up only shows 11.04% total owning any of those cards, which means 88.96% is using something else. So if I take this survey literally Nvidia and AMD both have terrible market penetration. Hell without anything more then just a dx12 gpu it shows only around 17% owning a card that can do that, if you think that is correct amongst gamers then your crazy.

Please don't be an analyst.
 
The low to mid market are always the most popular. Not surprising at all.

Based on how much popular the GTX970 had been among the 9xx line of GPUs, that's why i thought that the GTX1070 would be the great hit of the 1xxx line of GPUs. (*although i 'm not sure if the GTX970 had been more popular than the GTX960 )
 
Remember the 1060 consist of 2 GPUs. But its still something people buy in droves.
 
I just got asked to do the Steam survey a few days ago for the first time in 4-5 years, so my Titan X was added to the insignificant percentage of users on Steam who have one.

Ha, I never get the survey when I'm actually on my gaming machine. :D I always get asked when I'm on a laptop, and once it asks you, it doesn't ask again on another machine. I'm probably part of the mobile CPU, and integrated graphics numbers. :D
 
Remember the 1060 consist of 2 GPUs. But its still something people buy in droves.

Which is pretty encouraging really. It's not the top of the line, but it's a serviceable gaming card. The fact that so many people own one, is good for PC gaming in general. (at least to a small degree)
 
Based on how much popular the GTX970 had been among the 9xx line of GPUs, that's why i thought that the GTX1070 would be the great hit of the 1xxx line of GPUs. (*although i 'm not sure if the GTX970 had been more popular than the GTX960 )

It's because AMD forced Nvidia's hand with the RX 480. The 970 seemed like a much better value when it was the cheapest performance Maxwell part for six months. But with the full lineup available after just 3 months, it was an easier choice to go down this generation.

Also, the 1070 was DOA compared to the 970 thanks to the $120 higher launch price. Nvidia hasn't launched a cut high-end card priced that high since the 8800GTS. And that only sold well because it was hideously powerful, and had no competition.
 
Last edited:
It's because AMD forced Nvidia's hand with the RX 480. The 970 seemed like a much better value when it was the cheapest performance Maxwell part for six months. But with the full lineup available after just 3 months, it was an easier choice to go down this generation.

Also, the 1070 was DOA compared to the 970 thanks to the $120 higher launch price. Nvidia hasn't launched a cut high-end card priced that high since the 8800GTS.


According to JPR numbers that isn't true. The performance segments since its inception have been the highest selling volume over all. So where are all those going? We need to factor in the gtx 1080 into the figures too. You look at 1070 + 1080 figures which is the entire performance segment its higher than the 1060 figures which are the only cards in the midrange segment for nV. So they tally up to JPR figures.

And this is why Steam isn't that bad, its actually fairly good, granted you can't take it as empirical figures, but trends hell ya it fits right in.
 
According to JPR numbers that isn't true. The performance segments since its inception have been the highest selling volume over all. So where are all those going? We need to factor in the gtx 1080 into the figures too. You look at 1070 + 1080 figures which is the entire performance segment its higher than the 1060 figures which are the only cards in the midrange segment for nV. So they tally up to JPR figures.

And this is why Steam isn't that bad, its actually fairly good, granted you can't take it as empirical figures, but trends hell ya it fits right in.

I'm just saying there isn't the massive gap like in the Maxwell generation. They're still the most popular cards, but higher price and easier availability of midrange dulled the dominance somewhat,
 
I'm just saying there isn't the massive gap like in the Maxwell generation. They're still the most popular cards, but higher price and easier availability of midrange dulled the dominance somewhat,

Well maybe, hard to say from those figures on Steam. The 970 + 980 figures have to look at it that way to have similar looks. Also midrange back then the 960 wasn't the greatest of cards either in the Maxwell line, there was a big swing in perf/watt on 960 vs the 970.
 
I'm one of those people with a fury nano.

I love the 970 being king after all the fuss.
People were being ridiculous.
 
Ha, I never get the survey when I'm actually on my gaming machine. :D I always get asked when I'm on a laptop, and once it asks you, it doesn't ask again on another machine. I'm probably part of the mobile CPU, and integrated graphics numbers. :D

as has been the case the last few surveys I've been asked as well but I mostly game on my main rig which never gets polled. I did get polled on my secondary machine once when I was browsing the steam store but it had only a GTX750 in it at the time and an older Q6600 CPU which died shortly thereafter. Definitely not getting correct stats for my Steam usage.
 
as has been the case the last few surveys I've been asked as well but I mostly game on my main rig which never gets polled. I did get polled on my secondary machine once when I was browsing the steam store but it had only a GTX750 in it at the time and an older Q6600 CPU which died shortly thereafter. Definitely not getting correct stats for my Steam usage.

What they should do is allow you to take the hardware survey at any time. I'm sure they have some form of hashed up system ID for a given install, so they could remove duplicates automatically. Every time I build a new gaming setup I'd pop on there, and run the survey really quick.
 
Back
Top