By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - PC Discussion - Monthly Steam hardware survey analysis

BasilZero said:
JEMC said:

Yes, gaming laptops exist, but it's the first time a laoptop only chip has appeared on the survey, which is what makes it odd.

Maybe there was a deal with several laptops with that specific GPU.

I don't think so. You'd need several thousands of units to make it up to the list.



Please excuse my bad English.

Currently gaming on a PC with an i5-4670k@stock (for now), 16Gb RAM 1600 MHz and a GTX 1070

Steam / Live / NNID : jonxiquet    Add me if you want, but I'm a single player gamer.

Around the Network
JEMC said:
BasilZero said:

Laptop gaming

I guess people want to play games portable.

Yes, gaming laptops exist, but it's the first time a laoptop only chip has appeared on the survey, which is what makes it odd.

GPUs are in "other" until they reach 0.15%. So with little steps it finally reached that threshhold:

GPUs also disappear from the "all video cards" list when they fall under that threshhold... but if they ever reached these 0.15%, they still show up in the DX12-systems list. And since most GPUs are DX12-compatible, thats close enough to still track them, f.e. the AMD Fury series or the GTX 780 and 780 Ti. I mark these in italic in my Excel table:

Oh, and my new PC got surveyed today for next months data:

(and yes, that 66 TB storage are correct)

Last edited by Conina - on 03 December 2020

I see, thanks for the explanation. It's not often that a mobile chip manages to cross that threshold.



Please excuse my bad English.

Currently gaming on a PC with an i5-4670k@stock (for now), 16Gb RAM 1600 MHz and a GTX 1070

Steam / Live / NNID : jonxiquet    Add me if you want, but I'm a single player gamer.

JEMC said:

I see, thanks for the explanation. It's not often that a mobile chip manages to cross that threshold.

There are a lot that cross them (f. e. many Intel chips and GeForce MX110, 130, 150, 250)... but I don't track them (or GeForce GT cards without the "X"), since most of them are even slower than a GTX 750.

I only track the better ones, f. e. GTX 970M and GTX 980M.



Valve has updated the Steam hardware survey with December numbers:

As you can see, huge gains for RTX Turing GPUs, RTX 3080 also doubled the percentage, but no other Ampere or RDNA2 cards yet in the statistic.

The gain of the RTX20x0-GPUs in itself ain't surprising:

- due to many new PC game releases in October - December with higher hardware requirements many PC gamers want to upgrade their systems
- due to the colder season and Corona lockdowns more PC gamers have to stay at home
- due to canceled vacations many also have money to spent for home entertainment instead for travelling/cinema/theatre...
- more and more PC games support raytracing, so for RT-interested gamers the options are RTX Turing, RTX Ampere and Big Navi (RDNA2)
- most RTX Ampere and RDNA2 graphic cards are out of stock or currently overpriced
- RTX Turing cards are available and their prices have fallen since the Ampere/RDNA2 price announcements

So even without RTX 3060Ti/3070/3090 + RX 6800/6800XT/6900XT numbers the percentage of raytracing compatible systems on Steam took a huge step from 12 to 17 percent. Depending on the number of estimated active Steam users (100 - 200 million) these 17% would be 17 to 34 million raytracing-compatible gaming PCs already:

If the RDNA2 and other Ampere cards were already visible in the hardware survey, 20% RT-compatibility (so 20 - 40 million PCs) are probable. Just take a look in this data:

Every Nvidia/AMD/Intel GPU currently sold is (of course) DX12-compatible, but the DX12-GPUs have fallen 3 percentage points over the last months while the DX8-GPUs (and older) have risen 4 percentage points.

So probably the gained 4 percentage points "DX 8 and older" are new DX12-GPUs not properly tagged in the survey data, very probably Ampere and RDNA2 GPUs.

Unspecified Nvidia GPUs are shown as "NVIDIA Graphics Device" in the data. For a long time this was around 0.5% and probably mainly Nvidia Quadro cards, but it has grown the last months 2.2 percentage points... these are probably Ampere cards. The remaining 1 - 2 percentage points gain of "DX 8 and older" could be RDNA2.

The next months will show if my theory can hold up, when more and more Ampere and RDNA2 cards show up in the hardware survey and get correctly tagged as "DX 12 compatible", which would result in less than 3% "DX8 and older" GPUs and at least 93% DX12-GPUs.

Last edited by Conina - on 03 January 2021

Around the Network

I think your theory is very plausible. Valve always goes late when it comes to updating its hardware databases, and so it's possible that part of the new Turing cards aren't properly identified. And when it comes to AMD... well, I was going to say what for the first gen of Navi cards because they're capable of mixing the new ones with them but, since they're down, they haven't done it (yet).

We'll see if they've updated it next month.



Please excuse my bad English.

Currently gaming on a PC with an i5-4670k@stock (for now), 16Gb RAM 1600 MHz and a GTX 1070

Steam / Live / NNID : jonxiquet    Add me if you want, but I'm a single player gamer.

Conina said:

Every Nvidia/AMD/Intel GPU currently sold is (of course) DX12-compatible, but the DX12-GPUs have fallen 3 percentage points over the last months while the DX8-GPUs (and older) have risen 4 percentage points.

So probably the gained 4 percentage points "DX 8 and older" are new DX12-GPUs not properly tagged in the survey data, very probably Ampere and RDNA2 GPUs.

Unspecified Nvidia GPUs are shown as "NVIDIA Graphics Device" in the data. For a long time this was around 0.5% and probably mainly Nvidia Quadro cards, but it has grown the last months 2.2 percentage points... these are probably Ampere cards. The remaining 1 - 2 percentage points gain of "DX 8 and older" could be RDNA2.

The next months will show if my theory can hold up, when more and more Ampere and RDNA2 cards show up in the hardware survey and get correctly tagged as "DX 12 compatible", which would result in less than 3% "DX8 and older" GPUs and at least 93% DX12-GPUs.

Something is very off with their charts right now. Just look at the CPU charts: The landing page gives AMD 26.91% market share, which would be a plus of 0.4 percentage points over last month.

However, if you click on the chart, it brings up something where AMD lost 2.18 percentage points instead:

Of note is that Intel would have increased in the 2.7-3Ghz category by a whooping 5.33 percentage points, while the 3-3.3Ghz range would raise by 1.8 percentage points, and everything else, both from Intel and AMD, is down.

So my thought is that some Asian Cybercafés got counted again, and their older hardware and software is what made made all the more modern systems drop down 



Bofferbrauer2 said:

Something is very off with their charts right now. Just look at the CPU charts: The landing page gives AMD 26.91% market share, which would be a plus of 0.4 percentage points over last month.

However, if you click on the chart, it brings up something where AMD lost 2.18 percentage points instead:

Of note is that Intel would have increased in the 2.7-3Ghz category by a whooping 5.33 percentage points, while the 3-3.3Ghz range would raise by 1.8 percentage points, and everything else, both from Intel and AMD, is down.

So my thought is that some Asian Cybercafés got counted again, and their older hardware and software is what made made all the more modern systems drop down 

So after some strange December data (f.e. the CPU charts and the huge RTX Turing jump from 12 to 17 percentage points), most charts are back to November levels (plus expected Ryzen gains):

They also changed some December data afterwards (f. e. Turing now shows 16% instead of 17% for December):

RDNA2 and RTX 3070 GPUs are still not shown... but the later released RTX 3060 Ti???

Even if RTX 3060 Ti and RTX 3090 numbers were added, there are probably a lot of new GPUs (RDNA 2 + RTX 3070) hidden in the "other" group, falsely counted as "DirectX 8 GPUs and below":  

Probably even some of the RTX 3080 cards, since 1.15% of the surveyed cards have 10 GB VRAM, but only 0.66% are counted as RTX 3080 GPUs:

As a VR fan I'm happy that the downwards trend in the December data seems to be and exception. Now Steam shows a new record: 2.13% of the surveyed PCs have a VR headset connected.

More and more of the hugely popular "Oculus Quest 2"-headsets get connected to PCs for more demanding VR games:

A look to the OS versions shows the still growing dominance of Windows 10 64-bit.

Even Windows 7 64-bit is now down to 2.28%:

Last edited by Conina - on 03 February 2021

I agree that the lack of 3070 or any Navi 21 cards is strange, but we know Steam doesn't always make a good job with these surveys, as proven by the December results.



Please excuse my bad English.

Currently gaming on a PC with an i5-4670k@stock (for now), 16Gb RAM 1600 MHz and a GTX 1070

Steam / Live / NNID : jonxiquet    Add me if you want, but I'm a single player gamer.

Glad to see VR still on the rise. And yes, the Oculus Quest 2 users popped out of nowhere. Seeing lot of them in Pavlov VR.



Intel Core i7 8700K | 32 GB DDR 4 PC 3200 | ROG STRIX Z370-F Gaming | RTX 3090 FE| Crappy Monitor| HTC Vive Pro :3