New Steam survey results

It seems it can't detect crossfire either. When it gathers my system information, it sees everything but my 2nd GPU and states that no SLI or crossfire detected.

 

Latest numbers are up. Win7 is now above 50%. Same with 64bit.

 

64bit was above 50% last month as well.

 

Are you including Mac numbers in that? On the Windows side I get the number to be 48.98% for last month.

 

Yes. of all OS's, it was already over 50%.
Something else to note, 1920x1080 is now the most popular resolution. 1680x1050 has been dethroned.

GPU and CPU market share apparently has not been updated yet. However, I already calculated AMD's and NVidia's DX10 and DX11 market share respectively. Overall, AMD should have gained a bit of market share over Nvidia which should lose a bit. AMD's gain was mostly in DX10 due to the overall size of DX10. They kept strong in DX11 only losing 0.46% on an API that is less than 20%. They probably are going to continue losing that much every month.

 

Good news as Nvidia breaks 33% of the DX11 marketshare for the first time in June.

 

Valve Removes NVIDIA SLI Data From Steam Hardware Survey

 

Last time I knew NVidia provides proprietary API (NVAPI) that allows developer to query SLI status pretty accurately.

ATI/AMD on the other hand, only let developers guess whether Crossfire is exists/running based on the number of GPUs found in the system via Windows/DX API. Which is hardly reliable for anything useful like Steam statistics. (Check AMD Jon Story/H Gruen GDC 2008 Presentation on MGPU optimization, or something like that.)

So congratulations for AMD for succesfully [strike]dragging down the industry[/strike] making Valve comply with their low standard of developer support.

 

Do your really think this qualifies as dragging the industry down ? it's just an annoying thing at most.

 

Ha, that's funny. So AMD is upset that the stats are inaccurate so instead of fixing their shit they have Valve just delete the category?

 

So you think Steam claiming that SLI consists of 93% of all multi-gpu solutions (ridiculous, at best) would be better than not presenting those statistics at all?

I think it's for the best interest of Steam to provide statistics that are as accurate as possible. The statistics shown for multi-gpu solutions were way off to be taken seriously.


Sure, blame AMD all you want for not providing a reliable method to detect CrossfireX, but you can't blame Steam for taking out a statistic that was otherwise unreliable.

 

I'm stand corrected regarding Crossfire detection method. I'm at home now, and after some proper digging I found a tool from ATI examples to detect MGPU count: http://developer.amd.com/downloads/CrossFireDetect.rar

For comparison, SLI detection: http://developer.download.nvidia.com/whitepapers/2011/SLI_Best_Practices_2011_Feb.pdf.

So it's either Crossfire detection library isn't reliable, or Steam doesn't know/use it. Either way it's sign of poor developer support.

While it's useless for comparing both vendors multi GPU install base, it still can be useful to be able to access the trend of SLI among NV users -- for example. Steam just need to make it clear that one of their detection method is not reliable enough so nobody could interpret the numbers in the wrong way.

 

Last time i checked, AMD has better multi-gpu drivers than Nvidia now, so claims how AMD dragging down the industry when some highly unreliable stat from some page was removed, is 100% ridiculous

As far as I'm concerned, as long as they are (both GPU makers) improving multi-gpu solutions, its all good. Less propaganda based on unreliable stat for either company, good as well.

 

The detection mechanism changed between some revisions of Crossfire.

 

Which shouldn't be a problem provided that the Crossfire detection communicates directly to the driver instead of only "best-guessing" it by calling some functions in WDK, gdi32 & user32 libraries. Is it not?

 

shouldn't dx be responsible for sli/xfire detection
its not like its new and amd/nv are a part of the dx consortium (or whatever its called) surely they must of lobbied ms for it

 

Having the competitor's solution shown as a having a 93% marketshare should be a great incentive to fix whatever problem is stopping steam from correctly detecting their own tech. ...Not a great incentive to remove the competitor's results entirely from display.

Now where's the incentive?

The lazy way is indeed the lazy way to do it, not the good way.

 

Considering Steam doesn't even tell their statistics methodology and that their results seem to change even within a month even though when they're out they shouldn't anymore, I think more important factor than thinking if AMD or nVidia should help them less/more to detect cfx/sli, would be actually getting the data out on how they get their percents, why they change, how many machines are included, how long one stays in statistics without re-sending the data etc.

 

The detection mechanism is available, and requests can be made to change it, but that doesn't mean they will be changed.

 

What does this mean in practice? Does the old detection way not work anymore? What's the new mechanism?