Steam Hardware & Software Survey: July 2014

Zarathustra[H];1041006167 said:
So I'm reading this as XP use GROWING since its EOL?

Are people morons?

could just be peolple upgrading graphics cards on windows xp from dx 8-9 to hire...
 
could just be peolple upgrading graphics cards on windows xp from dx 8-9 to hire...

I don't think so, because the DX9 portions of the chart haven't shrunk (or at least haven't shrunk as much as the "XP and DX11 hardware" section has grown.

Possibly people thinking: XP is now free, I'm going to use it, paying no attention at all to the fact that it is a security swiss cheese that is no longer patched, and that they are more or less guaranteed to become part of a botnet in no time...

It will get worse before it gets better...
 
hmm weird... steam hasn't asked me to perform hardware survey in at least 3 years.
 
And I always end up in the "Other" category... yay.
 
This is something interesting:

HyperThreading 48.11% +0.84%

Since people are continuing to repeat it, no XP didn't grow, it shrank (again) in this month's Steam stats:
Windows XP 32 bit 4.69% -0.21%
Windows XP 64 bit 0.25% -0.02%

That's about 5x smaller than the estimated usage percentage desktop XP still has. That's a good thing.
 
Yeah, not sure what he's looking at. They only Windows versions that are showing growth in this survey are interestingly 8 and 8.1.

Not sure why its "interesting" when Win8 is pretty much the only choice for people - especially non tech savvy ones - when buying a new PC.

Its little more than coasting on inertia, and a fraction of the growth that MS could have seen with Win8 had they not been so pigheaded with trying to ram a tablet UI onto desktops.
 
Considering how Intel dominates the CPU share, 2 core CPUs edging out quad core CPUs (48% vs 44%) isn't too surprising. It takes 4 AMD cores to match even pretty low end Intel dual core CPUs. "Two core" performance (or equivalent using twice as many cores) is acceptable. :D

And the highest item having a 48% share makes it a plurality, not a majority.
 
so most gamers are still using 1GB VRAM...and 1080p and below resolutions...
 
sounds like steam gamers (the majority anyway) are no better than console gamers, gag
 
lawl, if you think controller > keyboard + mouse, gtfo :)
 
Are we looking at the same chart? I don't see that.

Yeah, not sure what he's looking at. They only Windows versions that are showing growth in this survey are interestingly 8 and 8.1.

14661479409_f8af08ca57_o.jpg


Well, either my chart reading skills are failing me, there is a bug in the chart generation code (has happened before on steam) or this is growth, especially since the green areas below are mostly unchanged...
 
Zarathustra[H];1041006878 said:
14661479409_f8af08ca57_o.jpg


Well, either my chart reading skills are failing me, there is a bug in the chart generation code (has happened before on steam) or this is growth, especially since the green areas below are mostly unchanged...

What I find funny about that chart is, what are those 66.79% not using a dx capable gpu using?
 
Yeah, that DX chart is confusing the hell out of me as well. Must be something wrong with it, same as the other one xIronCrossx linked to which is clearly AFU.
 
wtf majority of gamers are on 2cpus?!

Lots of people have laptops, and lots of laptops only have 2 cores. A lot of people are running Steam on older PCs too. It's not like you need a particularly new PC to run any of Valve's games for example, not to mention the countless indie games.
 
I am still amazed at how many people are running 2-4 cpu's.
Would like to see a breakdown of desktops vs laptops.
Maybe some day we will see D3D vs Mantle on there too :D
 
Zarathustra[H];1041006167 said:
So I'm reading this as XP use GROWING since its EOL?

Are people morons?

Well I just added a +1 to Steam's XP statistics with my classic games machine. :p
 
I am still amazed at how many people are running 2-4 cpu's.
Would like to see a breakdown of desktops vs laptops.
Maybe some day we will see D3D vs Mantle on there too :D

How is that amazing? It's no surprise to me people are using 2-4 cores.
 
Not sure why its "interesting" when Win8 is pretty much the only choice for people - especially non tech savvy ones - when buying a new PC.

Its little more than coasting on inertia, and a fraction of the growth that MS could have seen with Win8 had they not been so pigheaded with trying to ram a tablet UI onto desktops.

Not really. I'm typing this on a notebook that I bought less than a month ago and it came with Win 7 Home Pro x64. And it's not an older model, considering it sports the i7-4710HQ CPU.
 
Not really. I'm typing this on a notebook that I bought less than a month ago and it came with Win 7 Home Pro x64. And it's not an older model, considering it sports the i7-4710HQ CPU.

You got lucky. I was at OfficeMax a few months ago and when I talked to the rep about whether or not they would custom install Windows 7 in place of 8 they said:

"No sorry, Microsoft themselves bought back all of our Windows 7 licenses."

Whether or not that's true I don't know, but I found it interesting regardless.
 
I am still amazed at how many people are running 2-4 cpu's.
Would like to see a breakdown of desktops vs laptops.
Maybe some day we will see D3D vs Mantle on there too :D

Whats wrong with my 4 core 4770k???
 
Whats wrong with my 4 core 4770k???

Yeah, keep in mind they are measuring real cores, not logical HT cores.

Even a core i7-4790k will show up as a 4 core CPU, and for most games it is overkill.

To get to the above 4 core category, you need to be using a SandyBridge-E, IvyBridge-E, Xeon or one of AMD's CPU's.

I'm actually surprised the "above 4 cores" category is as large as it is.
 
Whats wrong with my 4 core 4770k???

Nothing wrong with it at all, especially gaming wise, just thought more would be in the 6-8 core range, but im kind of looking through a altered view, because i always think AMD, and the 6-8 cores are relatively inexpensive.
 
And they say PC gamers are elitist snobs.

Exactly what I thought. Funny those two think that all PC gamers will spend $1k - $3k on multi-head multi-card display setups for slight noticeable visual improvements.

May make their mind explode that I use 1 24" TN panel by choice, not because of financial restrictions.
 
Nothing wrong with it at all, especially gaming wise, just thought more would be in the 6-8 core range, but im kind of looking through a altered view, because i always think AMD, and the 6-8 cores are relatively inexpensive.

AMD cheats because their cores aren't really full cores in that they don't have FPUs for each core. Rather, they share FPUs between two cores so if you have an 8 core AMD processor, it really only has 4 cores worth of floating-point capability. It is basically AMD's answer to Hyperthreading (which allows two threads to execute on the same core) but if you have floating-point heavy apps (like games), you are gong to bottleneck.
 
AMD cheats because their cores aren't really full cores in that they don't have FPUs for each core. Rather, they share FPUs between two cores so if you have an 8 core AMD processor, it really only has 4 cores worth of floating-point capability. It is basically AMD's answer to Hyperthreading (which allows two threads to execute on the same core) but if you have floating-point heavy apps (like games), you are gong to bottleneck.

wow, really? what in the hell does my post have to do with AMD vs intel, how did you read that much into my first post? put down the coffee/monster and take a breath.
This is EXACTLY why they call us PC snobs.
 
AMD cheats because their cores aren't really full cores in that they don't have FPUs for each core. Rather, they share FPUs between two cores so if you have an 8 core AMD processor, it really only has 4 cores worth of floating-point capability. It is basically AMD's answer to Hyperthreading (which allows two threads to execute on the same core) but if you have floating-point heavy apps (like games), you are gong to bottleneck.

Two things.

1.) He is using a Phenom II Thuban core, so he actually has 6 integer cores AND 6 fpu's.

2.) The comparison of the bullozer two core module to HT is a bit flawed. The oversimplified description of HT is that it has two sets of decodes per core, presented as logical cores to make more efficient use of a single core and fpu.

AMD on the other hand has two of everything per module, except the FPU, where one 256bit FPU is shares for the module. The FPU is also splittable into two 128bit FPU's so when operating in that mode it's like each core has its own FPU.

AS far as scaling goes, the VMWare guide is that HT scales at about 30% of adding a real core on single socket systems and about 15% of adding a real core on dual-socket systems. (Not sure why it goes down as you add sockets).

AMD's Bulldozer/Piledriver cores - on the other hand - scale almost indistinguishably from traditional non-shares cores.

The reason AMD's chips are slower is primarily due to worse performance of each individual core, not because of module resource sharing.

Again, this is based on virtualized server workloads, as games and desktop applications are usually a poor benchmark for how well a system scales with added cores (except Cinebench, which is actually pretty good at testing this)
 
Zarathustra[H];1041007904 said:
Two things.

1.) He is using a Phenom II Thuban core, so he actually has 6 integer cores AND 6 fpu's.

2.) The comparison of the bullozer two core module to HT is a bit flawed. The oversimplified description of HT is that it has two sets of decodes per core, presented as logical cores to make more efficient use of a single core and fpu.

AMD on the other hand has two of everything per module, except the FPU, where one 256bit FPU is shares for the module. The FPU is also splittable into two 128bit FPU's so when operating in that mode it's like each core has its own FPU.

AS far as scaling goes, the VMWare guide is that HT scales at about 30% of adding a real core on single socket systems and about 15% of adding a real core on dual-socket systems. (Not sure why it goes down as you add sockets).

AMD's Bulldozer/Piledriver cores - on the other hand - scale almost indistinguishably from traditional non-shares cores.

The reason AMD's chips are slower is primarily due to worse performance of each individual core, not because of module resource sharing.

Again, this is based on virtualized server workloads, as games and desktop applications are usually a poor benchmark for how well a system scales with added cores (except Cinebench, which is actually pretty good at testing this)

To add to this, I've been using an FX-8120 AND (more recently) an FX-8350 in my basement ESXi server for some time, and have been very happy with core scaling. I only recently bought the Intel based HP server in my sig because I needed more ram, and the FX chips can only handle 4 UDIMMS which effectively limit me to 32GB until 16GB udimms are available.

(I could technically have bought 16gig udimms on special order, but they were going to be over $400 per piece, meaning the entire dual Xeon server and 64gb of registered ecc ram cost half used of what I would have spent in ram alone :p )
 
AMD cheats because their cores aren't really full cores in that they don't have FPUs for each core. Rather, they share FPUs between two cores so if you have an 8 core AMD processor, it really only has 4 cores worth of floating-point capability. It is basically AMD's answer to Hyperthreading (which allows two threads to execute on the same core) but if you have floating-point heavy apps (like games), you are gong to bottleneck.

Considering that most workloads are integer it's actually a pretty clever compromise that helps with TDP. AMD's 16-core Opteron 6376 at 32nm is 115W TDP while Intel 12-core Xeon E5-2697v2 at 22nm is 140W. I had 80W Xeons that ran hot so I can't imagine 140W with Intel's signature tiny surface area for dissipating heat to heat sink. You're also talking $2500 for Intel and $700 for AMD.
 
Back
Top