New Steam survey results

[Squilliam]

And if you add all % changes for April, you get:

ATI: +0.72%
NV: -0.67%

(Somehow the "Video card usage by manufacturer" pie chart on the Steam Hardware Survey front page has not been updated for quite some time).

Thats good, so whats the overall percentages now?

True but IMO they still havent managed to sell as well as they would have liked, esp considering they have no competition.

Im sure they've made it up with higher margins to a good extent. They do get more chips per wafer for the same performance tier which doesn't hurt any. Besides, what would they have liked to have sold? They probably expected an Nvidia response by now!

Look at NV's Geforce 8 and 9 series of cards. The Geforce 9800+8800+9600+8600 have a combined share of around 22%(and this dosent count the various G1xx and G2xx renamed parts). The 4800 series in comparison has 8.3%. The Radeon 3xxx parts also have very poor marketshare. In the list of top 20 video cards, Nvidia cards occupy 16 slots, ATI 3 and Intel 1. Stats taken from here http://store.steampowered.com/hwsurvey/videocard/

Thats because you're comparing a range of chips from when ATI was relatively uncompetitive and Nvidia gained and even more disproportionate share of the overall market. In addition to this they have been on the market for an unprecedented length of time whereas ATI has gone through the 3xxx, 4xxx, 5xxxx chips in the same time. If you threw everything into the same pot the whole equation looks a lot more balanced considering the GTS 250 is selling against the Radeon 57xx cards.

I guess they're probably still suffering from short supply given that TSMC hasnt increased 40nm production yet. If i remember correctly, the current supply is 40k wafers per quarter which is going to be doubled to 80k sometime in Q3. Yields should have increased a fair bit, considering its now 7 months since launch

Yes, definitely. 40k wafers per quarter isn't much really. However if my calculations are correct they can still get 300 viable Juniper dies and 150 Cypress off a single wafer so 15,000 of those yields at a 1:1 split is 2.25M Juniper and 1.125M Cypress per quarter. Though I've probably screwed up in several places!

 

[Richard]

Once again, Valve is retrofiting marketshare values and I'm not. Also, W2K drops back into "Other".

 

[ferro]

Thats good, so whats the overall percentages now?

I don't know, not all cards are included in the "Video card usage details" table. We can only determine a trend from the ones that are included.

 

[Silent_Buddha]

Once again, Valve is retrofiting marketshare values and I'm not. Also, W2K drops back into "Other".

Nice a couple more months and Win 5.x will be under 30% share. Hopefully, that'll prod devs to base their rendering pipes on dx10/11 rather than basing it on dx9 and then retro-fitting it to dx10/11.

Regards,
SB

 

[Squilliam]

If Steam userbase numbers are ~25M then according to Ferros numbers Nvidia lost ~170,000 Steam users and ATI gained ~180,000 users. Does that make sense? Effectively that means that 170k people migrated from an Nvidia system to an ATI one.

ATI: +0.72%
NV: -0.67%

 

[Erinyes]

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

It would appear to be proof AMD is doing pretty well with the RV770.

"Radeon HD4800 series" sits at 6.53% DX10 GPU share already. It's unfortunate they just pool all 4800 together rather than break it out. However I suspect the vast majority of those are 4850/70 (aren't the lower end models much newer to market anyway?). The 260 has 1.48% and the 280 1.22%, or 2.70% combined.

You can see just how unsuccessful AMD's prior gen was there. As well as Nvidias prior domination, combined 8800+8600=43.75% of all DX10 GPU's.

I added up all the previous AMD DX10 GPU's, excluding 4800 series (thats 2600+3870+2400+3600+3470+mobility2600+2900+3450+3200) and they only amounted to 13.05 % share!

Im sure they've made it up with higher margins to a good extent. They do get more chips per wafer for the same performance tier which doesn't hurt any. Besides, what would they have liked to have sold? They probably expected an Nvidia response by now!

Thats because you're comparing a range of chips from when ATI was relatively uncompetitive and Nvidia gained and even more disproportionate share of the overall market. In addition to this they have been on the market for an unprecedented length of time whereas ATI has gone through the 3xxx, 4xxx, 5xxxx chips in the same time. If you threw everything into the same pot the whole equation looks a lot more balanced considering the GTS 250 is selling against the Radeon 57xx cards.

Yes, definitely. 40k wafers per quarter isn't much really. However if my calculations are correct they can still get 300 viable Juniper dies and 150 Cypress off a single wafer so 15,000 of those yields at a 1:1 split is 2.25M Juniper and 1.125M Cypress per quarter. Though I've probably screwed up in several places!

Yea they probably had good margins, which is of course good for them. As to how much they would have like to have sold, i wanted to compare with 8800 GT and 8600 GT but unfortunately steam dosent have an archive(or i cant seem to find it if they do). But ill quote the first post in this thread, after five months, the 4800 series had a market share of 6.53% of DX10 GPU's and the 8600+8800 had 43.75% at the same time(roughly 18 months since 8600 released and 12 months since 8800). The Geforce 8600 had the HD 2600 series as competition wihle the 8800 series had the HD 3800 and HD 4800 later on as competition. In the case of Juniper, it barely has any competition so in my opinion just isnt selling too well.

If Steam userbase numbers are ~25M then according to Ferros numbers Nvidia lost ~170,000 Steam users and ATI gained ~180,000 users. Does that make sense? Effectively that means that 170k people migrated from an Nvidia system to an ATI one.

Nvidia may not have lost all that many users, if ATI gained 180,000 users then then NV's share automatically drops by about 0.4%, assuming that the userbase is now 25.18 million. There may be people upgrading from NV->ATI, or ATI->ATI or Intel->ATI/NV or even ATI->NV or NV->NV so i guess we cant really quantify how many users ATI or NV gained or lost

 

[Dave Baumann]

In the case of Juniper, it barely has any competition so in my opinion just isnt selling too well.

Yet it is the highest increase in the stats.

 

[Squilliam]

Yet it is the highest increase in the stats.

About these stats:

If I have a Steam account attached to a computers which have: HD 5750, HD 5770 and one with an HD 5780 of my own and one which is owned by my friend with the HD 5870 you sent so very graciously, how is that counted into the statistics?

Nvidia may not have lost all that many users, if ATI gained 180,000 users then then NV's share automatically drops by about 0.4%, assuming that the userbase is now 25.18 million. There may be people upgrading from NV->ATI, or ATI->ATI or Intel->ATI/NV or even ATI->NV or NV->NV so i guess we cant really quantify how many users ATI or NV gained or lost

Well we can infer that new Steam users will probably be proportionally similar to existing user groups already on Steam. If someone goes from ATI->ATI it wouldn't change that stat, ditto for Nv -> Nv however as Intel doesn't have discrete hardware the majority of the changes in market share has to be from people upgrading from NV -> ATI.

As to how much they would have like to have sold, i wanted to compare with 8800 GT and 8600 GT but unfortunately steam dosent have an archive(or i cant seem to find it if they do). But ill quote the first post in this thread, after five months, the 4800 series had a market share of 6.53% of DX10 GPU's and the 8600+8800 had 43.75% at the same time(roughly 18 months since 8600 released and 12 months since 8800). The Geforce 8600 had the HD 2600 series as competition wihle the 8800 series had the HD 3800 and HD 4800 later on as competition

Thats hardly a fair comparison given the fact that the 8800GTX, GT, GTS 320, 640 and 512 as well as the 8600 are being compared to only 5 months of AMD sales, especially as many of the 8800GT series were still selling well when the 48xx series were released. Its not good enough data to draw any meaningful conclusions.

 

[Andrew Lauritzen]

Did anyone figure out/resolve the problem of steam "sampling" specific PCs every year rather than each month and thus biasing away from new products? Or are they extrapolating the samples for each month across the entire user base? Anyone know how this works?

 

[Malo]

I believe it was posted a few pages back that Steam basically samples once a year unless your hardware changes. So if you put a new video card it, Steam should update its statistics.

 

[Florin]

Thats hardly a fair comparison given the fact that the 8800GTX, GT, GTS 320, 640 and 512 as well as the 8600 are being compared to only 5 months of AMD sales, especially as many of the 8800GT series were still selling well when the 48xx series were released. Its not good enough data to draw any meaningful conclusions.

It's just a snapshot in time, how is that not 'fair' data?

Only seems a problem if you're tyring to fit it to prove something you'd like to be true.

 

[Dave Baumann]

I believe it was posted a few pages back that Steam basically samples once a year unless your hardware changes. So if you put a new video card it, Steam should update its statistics.

I wouldn't necessarily view that as the case. Do you see a pop-up requsting to submit data when you put in a new card?

 

[Erinyes]

Yet it is the highest increase in the stats.

Im not denying that its currently the best selling card, but the increase for the 5700 series was 0.44%, while for the 5800 series was 0.37%. The selling price for the 5700 series is 120-170 USD while for the 5800 series is 250-400 USD. Surely you expected proportinately more sales given the difference in price?

 

[Malo]

I wouldn't necessarily view that as the case. Do you see a pop-up requsting to submit data when you put in a new card?

Really not sure Dave. I quote neliz:

I found a new install will also trigger the HWS opt-in.. I also have multiple machines running on the same account.. it seems to me it should trigger more often as it would only allow one machine to be surveyed that way. Same would go fo Cyber Cafés right?

Interesting though.. I can't find a thing on the frequency of the survey except what valve describes on the page itself "Each month, Steam collects data about what kinds of computer hardware our customers are using."

I don't remember doing another opt-in when I put in my 5770 though.

 

[brain_stew]

Nice a couple more months and Win 5.x will be under 30% share. Hopefully, that'll prod devs to base their rendering pipes on dx10/11 rather than basing it on dx9 and then retro-fitting it to dx10/11.

Regards,
SB

If Just Cause 2's performance in the Steam "top sellers list" is anything to go by (its been there since preorders went up a couple of months ago and has never left) being DX10 exclusive doesn't seem to have had any adverse effects on our first real "test case."

 

[Silent_Buddha]

I don't remember doing another opt-in when I put in my 5770 though.

I always thought the opt-in was a once or once a year question. If you opt-in you consent to sending Steam info on your computers once a month without the need to be prompted each time.

Regards,
SB

 

[Squilliam]

I always thought the opt-in was a once or once a year question. If you opt-in you consent to sending Steam info on your computers once a month without the need to be prompted each time.

Regards,
SB

How do you think having multiple different graphics cards on the same Steam account would effect things? Does it mean that whichever Steam account is logged in during the reporting period is the reported configuration or would they measure multiple cards for the same account? Whats your take on this? Its likely that many people have multiple computers attached to a Steam account.

OT: I wonder how many Dave has saved over this past month? As the prophet of the holy trinity of AMD CPU, chipset and GPU. Could it have been over 100 thousand like I suggested?

 

[homerdog]

Is there a way to manually submit your specs to Steam? I don't recall ever being polled after I went from the 8800GT to the GTX260 and that was almost a year ago.

 

[Richard]

FWIW I got a new HW req when I updated to the new Steam client, no changes in hardware.

 

[Humus]

Its likely that many people have multiple computers attached to a Steam account.

As an example I've been logging in to my personal account a number of times at work for testing, and then of course used it at home too, which are machines with different hardware, or using the dev account on the test machine and my main box too. I'd say I've seen the steam survey pop up probably every month during the time spent I spent working on Steam integration. Could also have something to do with random sampling and me logging in a hundred times a day too.