[KitGuru] Sales of desktop graphics cards hit 10-year low in Q2 2015

[sam_816]

What do you people think about VR as a factor to reignite the charm of(I am assuming here) pc gaming and hence the graphic cards' sales? While games in general are being consolized people get decent experience on consoles but I doubt consoles will be able to push decent VR experience.

With so many companies investing in VR I am sure they would want returns on these investments.

Morpheus on ps4 will push 720p to each eye(correct me if i m wrong) that's like ps3 level game experience n having tried oculus rift(dk2) I believe vr has potential to give the market the push it requires...

Thoughts?

 

[Ramses]

I'm eager, but have been told vr was just around the corner for so long that I'm not holding my breath. Nor am I buying the first thing that comes to market right away.

 

[RampantAndroid]

I toyed with the HTC/Valve VR headset at PAX on Monday and while it seemed to work just fine and DID look cool...the demo was also pretty "meh" - pick up spider bombs and throw them while tanks and mechs battle each other in front of you. And you had hand controls you needed - no walking around.

VR looks cool...but I fear it's the new gimmick. At least at first.

 

[AtenRa]

Total Combined PlayStation 4 and Xbox One Sales: 37,342,738 (+54%)***
just asking
where do these 37.mil chips get added to amd's market share ?

Both Jon Peddie Research (JPR) and Mercury Research numbers are for PC graphics, consoles sales are not included in to the PC market share.

And here comes the wrong impression people have about Developers would prefer NVIDIA over AMD hardware because of the higher market share.
Every developer is coding for the AMD GCN architecture used on both the MS XBone and Sony PS4.

Not only that, but the big difference here that nobody takes in to consideration is that, AMD GCN based hardware market share will only grow as time goes by because both XBone and PS4 will continue to increase in sales the coming years, when NVIDIA Kepler and MAXWELL will diminish.
So no developer would actually would like to specifically code its games for MAXWELL only, because the market share of Maxwell hardware in use by Gamers in two years from now, will be way lower than that of the Consoles alone, without even counting the PC GCN hardware found in dGPUs and APUs on both desktop and Laptops.

Actually if you thing about it, if NVIDIA Pascal is way different than MAXWELL, that is with huge Async Compute performance over Maxwell, then developers would like to make and optimized their DX-12 games for GCN and Pascal, not Kepler and Maxwell.
And not only that but in that instance, NVIDIA Pascal hardware market share will be extremely small in 2016 vs GCN based hardware market share.

 

[2is]

Both Jon Peddie Research (JPR) and Mercury Research numbers are for PC graphics, consoles sales are not included in to the PC market share.

And here comes the wrong impression people have about Developers would prefer NVIDIA over AMD hardware because of the higher market share.
Every developer is coding for the AMD GCN architecture used on both the MS XBone and Sony PS4.

Not only that, but the big difference here that nobody takes in to consideration is that, AMD GCN based hardware market share will only grow as time goes by because both XBone and PS4 will continue to increase in sales the coming years, when NVIDIA Kepler and MAXWELL will diminish.
So no developer would actually would like to specifically code its games for MAXWELL only, because the market share of Maxwell hardware in use by Gamers in two years from now, will be way lower than that of the Consoles alone, without even counting the PC GCN hardware found in dGPUs and APUs on both desktop and Laptops.

Actually if you thing about it, if NVIDIA Pascal is way different than MAXWELL, that is with huge Async Compute performance over Maxwell, then developers would like to make and optimized their DX-12 games for GCN and Pascal, not Kepler and Maxwell.
And not only that but in that instance, NVIDIA Pascal hardware market share will be extremely small in 2016 vs GCN based hardware market share.

That's a nice theory, just like all the other nice "light at the end of the tunnel" theories that haven't panned out for AMD over the last decade. If we look at reality though, I'm not seeing any shortage of developers implementing things like GameWorks. If anything, it has proliferated since PS4/XB1 release, not lessened, so I'm not entirely sure what you're basing this new found optimism on.

 

[sam_816]

I toyed with the HTC/Valve VR headset at PAX on Monday and while it seemed to work just fine and DID look cool...the demo was also pretty "meh" - pick up spider bombs and throw them while tanks and mechs battle each other in front of you. And you had hand controls you needed - no walking around.

VR looks cool...but I fear it's the new gimmick. At least at first.

Hardware is as cool as the software that runs on it. I wasn't impressed with demos available at Oculus rift's site but far cry 4, dishonored, borderlands2, euro truck sim2 & elite dangerous demo were really something else. Couldn't make dying light and mirrors edge work properly but they were breathtaking in the jittery form they ran on my PC. Facebook, HTC, valve, Razer n sony are investing a lot and both AMD & NVIDIA realize that vr is not going to fizzle out quickly. I hope ppl show some enthusiasm to early products as well. I think it will be good for the industry as a whole..

 

[redzo]

PC market shrinks => dGPU market shrinks even more than the PC market=> dGPU prices are being moved up the ladder by both AMD and nvidia. dGPU's are becoming more expensive.

nvidia wins this round => AMD bleeds gpu market share.
nvidia builds the entire 9XX stack upon great/positive product launch reviews across the web(i mean better than the competition). Can't blame them here. They did a better job and they deserve it.

The geforce 960 is probably selling like hot cakes and it is probably the main contributor to the market share. But it is such a bad price/performer. People are buying it because it spells 9XX and because it is cheap. It is quite ironic that nvidia gains market share over such a bad performer.
960 class gpu's, I believe this is where most volume sales are achieved, sadly I think this is where most customers get screwed.

 

[ShintaiDK]

PC market shrinks => dGPU market shrinks even more than the PC market=> dGPU prices are being moved up the ladder by both AMD and nvidia. dGPU's are becoming more expensive.

nvidia wins this round => AMD bleeds gpu market share.
nvidia builds the entire 9XX stack upon great/positive product launch reviews across the web(i mean better than the competition). Can't blame them here. They did a better job and they deserve it.

The geforce 960 is probably selling like hot cakes and it is probably the main contributor to the market share. But it is such a bad price/performer. People are buying it because it spells 9XX and because it is cheap. It is quite ironic that nvidia gains market share over such a bad performer.
960 class gpu's, I believe this is where most volume sales are achieved, sadly I think this is where most customers get screwed.

Looking on steam the GTX970 is outsells the GTX960 at 3:2(July). And the GTX970 is the most used GPU after the HD 4000 IGP.

But the dGPU segment is deep in its death spiral. Some people move up in price to get an upgrade. Others fall of the ladder and stick to IGP.

Personally I think I would go down the ladder as well with the next GPU upgrade. Mainly because I dont a need much higher than the current. Give me a 125W or less 14/16nm GPU that is 50% faster than my GTX980 with 8GB or more and I am set. Perhaps for life in terms of dGPU.

 

[tential]

Looking on steam the GTX970 is outsells the GTX960 at 3:2(July). And the GTX970 is the most used GPU after the HD 4000 IGP.

But the dGPU segment is deep in its death spiral. Some people move up in price to get an upgrade. Others fall of the ladder and stick to IGP.

Personally I think I would go down the ladder as well with the next GPU upgrade. Mainly because I dont a need much higher than the current. Give me a 125W or less 14/16nm GPU that is 50% faster than my GTX980 with 8GB or more and I am set. Perhaps for life in terms of dGPU.

I see that being 2 gens out. But ya, I think igp will eat up more of the market share but there will always be a place for pc gaming.

I'm on the opposite spectrum though, I want more gpu horsepower for 4k. Next gen I'll probably spend close to 700-1000 to get a good gpu that can handle 4k well.

 

[RampantAndroid]

Personally I think I would go down the ladder as well with the next GPU upgrade. Mainly because I dont a need much higher than the current. Give me a 125W or less 14/16nm GPU that is 50% faster than my GTX980 with 8GB or more and I am set. Perhaps for life in terms of dGPU.

This is my sentiment as well - I think Pascal will end up being a GPU that people will buy and stick with for a long time. It's why I plan to sell off my GTX980 used (I bought it used...) and get Pascal, if the price is right. I do PC gaming, but the 980 does everything I need. I don't care anymore about amazing graphics. It's cool, but not what I play for. Now, maybe that'll change with VR...but I don't see me doing VR for a number of years.

 

[AtenRa]

That's a nice theory, just like all the other nice "light at the end of the tunnel" theories that haven't panned out for AMD over the last decade. If we look at reality though, I'm not seeing any shortage of developers implementing things like GameWorks. If anything, it has proliferated since PS4/XB1 release, not lessened, so I'm not entirely sure what you're basing this new found optimism on.

How many games are coded for the GCN based XBone and PS4 and how many do they have Gameworks ??

Not only that, but those that have gameworks run better on the consoles vs the PCs with those GameWorks features.

ps. Numbers dont lie, GCN based hardware is increasing while Maxwell hardware will start to decrease once Pascal arrive. Nobody will care about Maxwell GPUs, not even NVIDIA at that point.

 

[2is]

How many games are coded for the GCN based XBone and PS4 and how many do they have Gameworks ??

Not only that, but those that have gameworks run better on the consoles vs the PCs with those GameWorks features.

ps. Numbers dont lie, GCN based hardware is increasing while Maxwell hardware will start to decrease once Pascal arrive. Nobody will care about Maxwell GPUs, not even NVIDIA at that point.

What are these numbers you speak of? They sound made up. All I see is AMD losing market share to NVidia and Gameworks games increasing while more AMD fans cry foul about it. Your inclusion of console hardware isn't spilling over to the PC market as much as you'd like to pretend it is. I've already given you examples of that (market share/gameworks)

Do you have anything besides your own imagination to back up what you're saying as far as PC gaming and hardware is concerned? Like I said, your "silver lining" well, you've been touting that for a decade now I'll bet. What makes you think that THIS time it will actually happen when all the ACTUAL numbers are suggesting the complete opposite?

 

[zelachang]

I'm part of the problem. Still running a 2500k and a 5870. Got both when brand new expecting to replace them in ~3 years but so far I've seen no reason to. Sure I struggled to get Witcher 3 at 1920x1080 at acceptable framerates but for most other games I play my 5870 still runs great. Was going to buy a 980 ti or a Fury X but price and numbers just didn't add up to me. Waiting next year for more mature HBM and 16mm/14mm node.

 

[crisium]

Pretty rare to skip the whole 28nm generation when it has been so long. There is wisdom in waiting, but you could also find a used 7950 for ~$100 and get a noticeable upgrade to tide you over. Of course, that doesn't help these anemic sales since it is used.

 

[Shivansps]

How many games are coded for the GCN based XBone and PS4 and how many do they have Gameworks ??

Not only that, but those that have gameworks run better on the consoles vs the PCs with those GameWorks features.

ps. Numbers dont lie, GCN based hardware is increasing while Maxwell hardware will start to decrease once Pascal arrive. Nobody will care about Maxwell GPUs, not even NVIDIA at that point.

Did you know that Gameworks and Physx(CPU) is on PS4 as well right? not sure about XONE though, i really dont think Nvidia is trying to shoot themselves in the foot with Gameworks crippling PC perf, that just unfounded BS.

EDIT: Metal gear solid: Phantom pain seems to be running great with GameWorks, on AMD as well as they where able to close perf gap.

 

[cbn]

Broadwell's Iris Pro 6200 is shown in one review to be slightly behind GT 740. GTX 750 is ~50% faster than GT 740. Skylake GT4e is said to be "up to 50%" faster. At best, we might have it equal to a GTX 750. Notice that its not even up to a Ti, which is 20% faster in addition. 950 is 40% on top of that. By the time SKL GT4e is out, it'll be x40 level. Intel is process-constrained like everyone else.

GT 740 (aka GTX 650) is about equal to a HD 7750 GDDR5.

HD 7770 (aka R7 250X) has around 50% more GPU core (in the form of stream processors/clockspeed, 640sp @ 1000 Mhz vs 512sp @ 800 Mhz) compared to HD7750 so I expect GT4e (which has 50% more EUs compared to GT3e) to be around the level of HD7770/R7 250X.

Also I did some more calculations/estimates based on the Skylake Gen 9 and R7 240 here.

 

[ShintaiDK]

Steam August numbers shows the same tendency. Intel grew 0.5% in the established base for august.

Windows 10 is up to 17.02%.

Quadcores lost share to dualcores. Dualcores goes up 0.86%, quadcores down 0.68%. Octocores also declines 0.03% down to 0.27%. Hexcores decline as well with 0.06% down to 1.62%.

 

[Qwertilot]

I'm definitely also in the wait until the die shrink, get something midish range (to cover 4k/early VR etc reasonably, which seems very realistic) then maybe nothing else camp.

I do think that at least NV will achieve one more die shrunk generation past 16nm. Very profitable and plenty of money in hand I think?

So once there's a huge performance advantage to be had from shrinking - probably past 10nm - the refresh sales alone might well justify doing it. Or just the simple fact that its shrink or die.

Maybe cleaning up AMD's remaining market share at that point too, or maybe AMD will pick up

Could also always be something really disruptive like VR second life/video messaging getting huge, and the market booming. Future prediction not a science.

 

[Ramses]

I can see it now in a couple years, Intel/AMD announce a breakthrough technology allowing them to place the graphics hardware on it's own card to achieve never before seen performance at new low prices...

 

[ShintaiDK]

I can see it now in a couple years, Intel/AMD announce a breakthrough technology allowing them to place the graphics hardware on it's own card to achieve never before seen performance at new low prices...

Not gonna happen

More like they announce fixed main memory on the same package as the CPU to boost IGP performance.

 

[ShintaiDK]

I'm definitely also in the wait until the die shrink, get something midish range (to cover 4k/early VR etc reasonably, which seems very realistic) then maybe nothing else camp.

I do think that at least NV will achieve one more die shrunk generation past 16nm. Very profitable and plenty of money in hand I think?

So once there's a huge performance advantage to be had from shrinking - probably past 10nm - the refresh sales alone might well justify doing it. Or just the simple fact that its shrink or die.

Maybe cleaning up AMD's remaining market share at that point too, or maybe AMD will pick up

Could also always be something really disruptive like VR second life/video messaging getting huge, and the market booming. Future prediction not a science.

With the rate of decline there would be no ROI past 14/16nm for nVidia. Even with 100% dGPU share. As you say something disruptive is needed. But I cant see it happening. Specially not because it needs to happen within a relatively fixed timeframe.

 

[Ramses]

Not gonna happen

More like they announce fixed main memory on the same package as the CPU to boost IGP performance.

Were gonna end up with cpu's the size of slot 2 xeons again lol...

 

[Blitzvogel]

GT 740 (aka GTX 650) is about equal to a HD 7750 GDDR5.

HD 7770 (aka R7 250X) has around 50% more GPU core (in the form of stream processors/clockspeed, 640sp @ 1000 Mhz vs 512sp @ 800 Mhz) compared to HD7750 so I expect GT4e (which has 50% more EUs compared to GT3e) to be around the level of HD7770/R7 250X.

Also I did some more calculations/estimates based on the Skylake Gen 9 and R7 240 here.

And how much is it going to cost us to get our hands on a desktop Intel processor (a quad core in particular) with GT4e? :whiste:

I think Intel is going to have difficulty challenging the value of discreet graphics in desktops. Laptops are another story.

 

[Shivansps]

It does not need to, if GT4e gets anywhere near a GTX750TI the market will suffer a lot, and already sufering a lot with that A8-7600 behing faster than a R7 240. and the rest of cpus just killed the low end dgpus.

The IGPs does not need to have 980TI perf in order to kill dgpus.

 

[Blitzvogel]

It does not need to, if GT4e gets anywhere near a GTX750TI the market will suffer a lot, and already sufering a lot with that A8-7600 behing faster than a R7 240. and the rest of cpus just killed the low end dgpus.

The IGPs does not need to have 980TI perf in order to kill dgpus.

Let me rephrase that: "I don't think Intel is going to be able to challenge the middle and higher end desktop discreet graphics market anytime soon".

I'm not surprised at the killing-off of what was historically the low end segment. Had it not been the past couple years of GPU stagnation, GTX 750 and Radeon 7770 level performance would be in the low end of the performance spectrum. Hell, they are depending on your point of view.

AMD APUs were such a damn good value in how much graphics and CPU performance they provided in combination, it's not hard to see how they helped to kill off low end discreet. Intel IGPs were just the logical integration of a basic feature into the CPU, much like the FPU in the 90s. It was good enough to supplant low end discreet GPUs, especially in entry desktops and a good majority of laptops out there. Most buyers don't know, and don't care as long as it works for basic use. But it made laptop production quite a bit cheaper which is good for consumers.

Also, getting into Skylake with GT4e means building a system with it in mind (is it BGA like GT3e?). That means limited upgrade path if it's of concern. On the flip side, that doesn't mean as much anymore until some major breakthrough in IC fabrication and software necessitates moving onto new software and hardware paradigms. Plenty of people only need to upgrade their graphics performance and RAM intermittenly, so you have a poop-ton of people with decent quad cores already who have no use side grading to a GT4e system just for the CPU performance. That still leaves a rather big base of gamers and workstation-like systems that will need graphics upgrades as the years roll on. I am one of those people. Depending on what AMD comes out with in the next generation of cards, I may upgrade from my R9 270.

When 16/14 nm GPUs come out (next year hopefully), I think we'll see a nice boost in performance and value of discreet GPUs. 4K actually seems to be a popular thing, and that will require a very good dGPU for the time being.

You know what, I could be totally wrong and the whole dGPU market could go to hell anytime soon. I just don't see the end of it anytime soon, since Intel's process lead is about to not be so special anymore, and Intel can't just subsidize large iGPs. They need to make money like anyone else, especially in this post-PC market.