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

[Ramses]

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

It does eventually, and I'm not seeing that happening anytime soon based on the last 3-5 years worth of CPU or GPU progress.

I hope they can pull some rabbit out of their collective hats and in a year or three I can buy a CPU with a 290x worth of GPU for $3-400, but I doubt it. And taking away revenue from dGPU outfits is only going to serve to screw those that still want them.
Other than making some money for people(intel) I don't see iGPU doing us (as in moderate to serious games, computer/hardware enthusiasts, etc) any good in the long run.

 

[Shivansps]

This is very simple, the igp already killed the low end sector, and they are starting now to get inside the more mainstream sector, yes dgpus gona improve, we know that, but as the market shinks, the costs increase.

Thats why igps does not need to beat the high end sector in order to kill it, how many whould be paying for a 980TI if it was x3 the cost?

It will happen, just not with current tech, Skylake GT4e at best will be able to reach a GTX750 TI with OC to clock and edram, and we are talking about a gpu already reemplaced by the GTX950.

As soon HBM and HMC gets intro a CPU die will the moment of igp killing the mainstream sector.

 

[Ramses]

Yep, thereby screwing those of us that aren't mainstream.

This is why I'm not a fan of the iGPU. There was nothing wrong with a person buying a $100-150 GPU if they wanted to limp wrist some games. They didn't have to spend $300+, and they could upgrade later if they outgrew it. Now it's integrated, great, the CPU cost more, so less great, but still great, for the masses, and Intel. dGPU sales slow, and slow, profits go down, prices go up, if they even bother to make anything better.
This is not an enthusiast-friendly direction and I'll not be supporting it with my monies.

Luckily, I guess, sorta, everything moves so slow now I can buy some stuff from a year or two ago and it'll still be relevant a year or two from now.

 

[cbn]

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.

Yeah, I think quad core with R7 250X level graphics isn't going to fly very well on desktop (except for certain niche applications), so I made a thread about Intel offering GT4e as a desktop dual core:

http://forums.anandtech.com/showthread.php?t=2444860

 

[cbn]

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.

Actually, APUs were not that great of a value. It is just that AMD appeared to limit supply of Oland.

For example, just take a look at the entry level hot deals thread (linked below) and you'll notice that during the time that thread was active not one R7 250 ever made it into the thread:

http://forums.anandtech.com/showthread.php?t=2403091

 

[Dribble]

This is very simple, the igp already killed the low end sector, and they are starting now to get inside the more mainstream sector, yes dgpus gona improve, we know that, but as the market shinks, the costs increase.
.

People have been saying this for years and it never actually happens, the igp gets faster but so does the mid range gpu and it stays comfortably ahead. The size, power requirements and memory bandwidth of a mid range gpu are just to much. Bottom line is they can't make igp to beat mid range without it really becoming a mid range gpu with a cpu attached (i.e. most of the cost and the transistors are in the gpu not the cpu). Outside of consoles there isn't much of a market for that.

 

[Head1985]

Nv and AMD logic:Wow we selling really low amount of GPUs because we overpriced them and doubled price from previous generation.
Sulotion:lets overprice them even more :biggrin:

 

[KingFatty]

What about history?

So the headline talks about "10-year" low, but how bad was it back in 2005? Weren't GPU sales pretty good then?

2005 was before the iPhone came out, before the whole "gaming on phones" thing put a dent in gaming, so I wonder why sales being the same as 2005 levels are bad?

 

[sm625]

I wonder how many GT210 and 5450 class cards were killed off versus how many mid and top tier cards. I'm willing to bet that the mid and top tiers arent contracting anywhere near as much as that chart suggests, and hell they may even still be growing.

 

[Shivansps]

I wonder how many GT210 and 5450 class cards were killed off versus how many mid and top tier cards. I'm willing to bet that the mid and top tiers arent contracting anywhere near as much as that chart suggests, and hell they may even still be growing.

Those gpus still sells today mostly to people that dont have a HDMI out on their mbs.

But it has nothing to do with 3D power, for example a 2nd gen Intel HD3000 is able to beat a HD5450, GT210, GT520, GT610... a 3rd gen HD4000 is able to beat a HD6450, and the low end sector was petty much killed after that.

And we now have Intel in a Pentium with a IGP pushing to rival a R7 240 DDR3...

 

[2is]

People have been saying this for years and it never actually happens, the igp gets faster but so does the mid range gpu and it stays comfortably ahead. The size, power requirements and memory bandwidth of a mid range gpu are just to much. Bottom line is they can't make igp to beat mid range without it really becoming a mid range gpu with a cpu attached (i.e. most of the cost and the transistors are in the gpu not the cpu). Outside of consoles there isn't much of a market for that.

Except for the fact that it is ACTUALLY happening

 

[SPBHM]

Except for the fact that it is ACTUALLY happening

they have huge IGPs with a huge cache but still can't do much, and the most popular IGPs are much slower anyway.

 

[Dribble]

Except for the fact that it is ACTUALLY happening

No it's not. AMD igp if anything is starting to fall behind a little, it was better vs mid range graphics about 3 years ago. Intel iris is no closer then AMD igp got in it's prime, and that has a huge amount of silicon dedicated too it so is very much a specialist solution - most intel igp doesn't use it. Current mid range (e.g. 950) is still as far ahead of current igp as it ever was.

 

[mrmt]

No it's not. AMD igp if anything is starting to fall behind a little, it was better vs mid range graphics about 3 years ago. Intel iris is no closer then AMD igp got in it's prime, and that has a huge amount of silicon dedicated too it so is very much a specialist solution - most intel igp doesn't use it. Current mid range (e.g. 950) is still as far ahead of current igp as it ever was.

In order to stay comfortably ahead the dGPUs are getting bigger, sporting wider memory interfaces and sport higher memory amounts. When you think this happened in the same node, e.g., without the benefits of the Moore's law new nodes can bring, what we see is that the midrange today has a much worse cost structure than it had a few years back when the iGPUs were a laughable joke, and that is forcing both Nvidia and AMD to raise up prices to keep up with the rising costs.

I think it's quite telling that the pedestrian Skylake iGPU supports DX 12.1 from day 0, and that 10nm SKUs will bring even more iGPU grunt to the Intel line up in a couple of years.

On the other hand, is Nvidia happy with the prospects of their GPU business? They certainly are, especially when they have the most incompetent and disoriented competitor of all times and they 100% of the profits on this market, but even so they certainly aren't going all hands on this market, they tried *hard* to diversify beyond it, first to mobile and then to the auto market. Were the dGPU business healthy and with huge prospects, they would be developing new GPUs and not trying to break into other markets.

AMD didn't crash the day Bulldozer launched, but everyone that knew a bit about business could see that the company was in line for a real banging in the next 3-4 years once the cards were disclosed. Today I see the same with the dGPU market, we're not going to see Nvidia business outright disappear, but the pieces are in place for it to shrink a lot in the next 5 years. Thankfully Nvidia is well managed enough to withstand this shift and survive.

 

[arandomguy]

One issue that seems to be ignored is the mainstream market doesn't place as much value on what is essentially paper performance compared to the the enthuasist market. The issue isn't how much faster a GTX 750ti or whatever is compared to the IGP but whether or not the IGP is good enough by itself. The GPU companies do recognize this, you'll notice how Nvidia had a strong focus with the GTX 950 in terms of conveying the benefits of the GTX 950 for MOBA gaming as opposed to just performance numbers in order to sell it as being worth the step up.

If the broader PC gaming market trends towards lower demanding games that perform adequately on the IGP, see Steam stats popular games as an example, then the IGP is adequate enough and the performance delta between discrete GPUs is not really relevant. This contrasts strongly with the past where there was problems just being able to play even the most basic games on IGPs. Even for non gaming consumer usage in the past IGPs were considered problematic which is why non gamers or professionals may have even bought discrete graphics ards.

Enthuasists on tech sites may consider r9 290 or GTX 970 class performance as a minimum for 1080p gaming and be fixated with 60fps at max settings with high AA but the mainstream market as whole is not all that interested.

This doesn't just apply to IGPs displacing the lower segments but also the lower rate of upgrades. Most people, especially the more price conscious, do not feel the cost of upgrading translates to worthwhile tangible results.

 

[Dribble]

In order to stay comfortably ahead the dGPUs are getting bigger, sporting wider memory interfaces and sport higher memory amounts. When you think this happened in the same node, e.g., without the benefits of the Moore's law new nodes can bring, what we see is that the midrange today has a much worse cost structure than it had a few years back when the iGPUs were a laughable joke, and that is forcing both Nvidia and AMD to raise up prices to keep up with the rising costs.

I think it's quite telling that the pedestrian Skylake iGPU supports DX 12.1 from day 0, and that 10nm SKUs will bring even more iGPU grunt to the Intel line up in a couple of years.

On the other hand, is Nvidia happy with the prospects of their GPU business? They certainly are, especially when they have the most incompetent and disoriented competitor of all times and they 100% of the profits on this market, but even so they certainly aren't going all hands on this market, they tried *hard* to diversify beyond it, first to mobile and then to the auto market. Were the dGPU business healthy and with huge prospects, they would be developing new GPUs and not trying to break into other markets.

AMD didn't crash the day Bulldozer launched, but everyone that knew a bit about business could see that the company was in line for a real banging in the next 3-4 years once the cards were disclosed. Today I see the same with the dGPU market, we're not going to see Nvidia business outright disappear, but the pieces are in place for it to shrink a lot in the next 5 years. Thankfully Nvidia is well managed enough to withstand this shift and survive.

Where I agree is Intel igp (the normal igp, not iris) is catching up with AMD igp, and their drivers are getting better. That has meant less low end cards as most machines now have much better igp. But AMD igp -> mid range gap has not got smaller. I don't agree that mid range cards have got larger either - if anything they've got smaller. Remember the many of the recent high end nvidia chips are using what would be considered a mid range sized chip a few years ago but got rebranded high end due to lack of competition, and the chips under that are smaller.

 

[2is]

No it's not. AMD igp if anything is starting to fall behind a little, it was better vs mid range graphics about 3 years ago. Intel iris is no closer then AMD igp got in it's prime, and that has a huge amount of silicon dedicated too it so is very much a specialist solution - most intel igp doesn't use it. Current mid range (e.g. 950) is still as far ahead of current igp as it ever was.

Yes, it is. You're comparing IGP progression to dGPU progression. You need to look at IGP progression compared to what's necessary to get decent gaming performance. You don't live in a box composed of PC gaming enthusiasts. It won't be long at all before your average joe can buy any off the shelf PC that will have steam preloaded on it and play any game he wants. The game will launch, auto-detect hardware and set the best settings. The person will start playing not knowing any differently. For many people, simply being able to play will be good enough meaning fewer people will have the need for a dGPU, meaning less of them sold, meaning those that are sold will be pricier. It's already happening, if you don't see it, you're either blind or in denial. Pick one.

 

[mrmt]

Remember the many of the recent high end nvidia chips are using what would be considered a mid range sized chip a few years ago but got rebranded high end due to lack of competition, and the chips under that are smaller.

That's because a middle range now has to bear a much heavier load in terms of R&D amortization than in previous generations.

 

[redzo]

they have huge IGPs with a huge cache but still can't do much, and the most popular IGPs are much slower anyway.

You're missing the fact that the 750 exists because the CPU IGP is cannibalizing the low end to mid. And CPU IGP's will keep doing so at a faster rate. The same way the mobile constantly eats the consumer PC for breakfast.
That CPU IGP comes for free at 0 costumer cost. It will keep doing so until the PC market will reach the final stable low time low. At that point, I have no idea how a company like nvidia will be able to keep a fair enough profit out of dGPU's. I suspect that most of nvidia's competent employees already cr*p themselves about this. They need to penetrate or invent other markets in order to survive. It has become more of a survival game.

People have been saying this for years and it never actually happens, the igp gets faster but so does the mid range gpu and it stays comfortably ahead. The size, power requirements and memory bandwidth of a mid range gpu are just to much. Bottom line is they can't make igp to beat mid range without it really becoming a mid range gpu with a cpu attached (i.e. most of the cost and the transistors are in the gpu not the cpu). Outside of consoles there isn't much of a market for that.

I was thinking the same some time ago, but moving the mid range higher on the price food chain is only a temporary solution to the problem. Companies like nvidia and AMD's GPU division need a constant profit nutrient otherwise they go under.

 

[stahlhart]

readers, 2is -- take your personal confrontation to PM and keep it out of this thread.
-- stahlhart

 

[zink77]

The reality is publishers no longer make PC specific games, when's the last time we had a real PC only videogame? Like more than a decade ago.

No modern games really push GPU's/CPU's so there's really no need to upgrade.

 

[cmdrdredd]

What about history?

So the headline talks about "10-year" low, but how bad was it back in 2005? Weren't GPU sales pretty good then?

2005 was before the iPhone came out, before the whole "gaming on phones" thing put a dent in gaming, so I wonder why sales being the same as 2005 levels are bad?

2005 also was the launch year for the xbox 360. I don't know how that plays into it but it's worth noting.

 

[monstercameron]

The reality is publishers no longer make PC specific games, when's the last time we had a real PC only videogame? Like more than a decade ago.

No modern games really push GPU's/CPU's so there's really no need to upgrade.

Xcom 2 is the next big game.

 

[cmdrdredd]

The reality is publishers no longer make PC specific games, when's the last time we had a real PC only videogame? Like more than a decade ago.

No modern games really push GPU's/CPU's so there's really no need to upgrade.

There's been a few but I would note that many of them are running older engines or aren't very resource intensive.

 

[Blitzvogel]

2005 also was the launch year for the xbox 360. I don't know how that plays into it but it's worth noting.

The Xbox 360 recruited a number of PC gamers into consoles. It offered pretty much the same experience as the PC at much lower cost in it's first couple of years. Also, many of the hit PC games of 2005 and even 2006 didn't require a high end system to get a relatively good experience. Many games like BF2, FEAR, CoD2 just leveraged bigger and better versions of feature sets and game designs that were already common place, and many gamers were still on 1024 x 768. 256 MB of VRAM was enough at the time of the 360's release, and something like a high speed Athlon 64 or Pentium 4 + Geforce 6600 or 6800 already could get you pretty far when it came to what you could run. So while the Geforce 7000s and Radeon X1xxxs were pretty good, I think the fervor for them was a bit downplayed by the 360's release and the success of the Geforce 6 and X800 series before them.