JPR Graphics Add-in-board-market Q2 2014 sales

[ShintaiDK]

Its rather a myth to claim that core gamers are the ones paying. If we look on the CPU side, do you think the desktop LGA2011 platform would even exist without Xeons? The answer is obviously no.

As volume shrinks, while cost and R&D goes up. Prices will keep rising and each time it will also fuel to the negative feedback cyclus that dGPUs are in. Cores gamers are simply not willing to pay the price for a product without someone else to share the cost with.

I bet you nVidia and AMD makes more money on sub 200$ cards than they do at the cards above when applying volume.

And when we see volume go from something like 65m units in 2013 to 50-55m units in 2014. Then it all gets clear.

 

[3DVagabond]

OEM's are the ones footing the bills. In order to make these chips "affordable" it requires volume.

 

[Meghan54]

GMA900 was released in 2014, it was fine for the masses at the time.

Where did you see GPU sales dropped by half ??

First, I think you meant GMA900 was released in 2004, not 2014....typo.

Second, while the article the OP linked talks a lot about y-t-y and q-t-q sales, it also speaks to the last 10 years, if you'd bothered to actually read the article.

And if you'd have read the article, you'd have noted this line (last sentence in article):

...the volume of those boards (AIBs) peaked in 1999, reaching 114 million units, in 2013 65 million shipped.

And while that's not a true 50% drop in sales of AIB's (Add-In-Boards, or dGPUs), it's damned close.

 

[DominionSeraph]

Sorry but it's just a downward slope for desktpo GPUs.

Until either AMD or Intel figures out that wasting die space on a GPU that is 5000% more powerful than 99% of their customers need is kinda stupid and they massively undercut their competition by releasing CPUs without the baggage.

 

[NTMBK]

Until either AMD or Intel figures out that wasting die space on a GPU that is 5000% more powerful than 99% of their customers need is kinda stupid and they massively undercut their competition by releasing CPUs without the baggage.

Intel produce plenty of CPUs like that. They're called "Xeons". If you care about CPU performance, buy moar cores.

The consumer "CPUs" are just overclocked laptop parts.

 

[AtenRa]

First, I think you meant GMA900 was released in 2004, not 2014....typo.

Second, while the article the OP linked talks a lot about y-t-y and q-t-q sales, it also speaks to the last 10 years, if you'd bothered to actually read the article.

And if you'd have read the article, you'd have noted this line (last sentence in article):



And while that's not a true 50% drop in sales of AIB's (Add-In-Boards, or dGPUs), it's damned close.

Ehm yea 2004 for the GMA900

Now about discrete GPUs,
The 114M units are not discrete ONLY, there is an error in the report.
From another JPR pdf,

As we can see in 1999 TOTAL Graphics Units (Includes both Discrete and iGPUs) were close to 120M. There is no way discrete GPUs alone were 114M units in 1999. Total PC market was way too small for that number anyway.

https://jonpeddie.com/download/media/slides/Dynamics_in_GPU_market.pdf

 

[blackened23]

What the HELL are you talking about? Every PC had discrete graphics in 1999, there was no buy an intel CPU with an IGP and certainly not on the AMD/cyrix/whatever else either. There was no IGP or APU. Some motherboards had discrete graphics on the motherboard but they were still discrete graphics made by ATI/NV/Tseng/S3/etcetcetc. Similar to mobile discrete GPUs with the mobile dGPU hardwired on the motherboard. However most PCs had an actual PCI dGPU. There was no CPU that you could buy a Windows PC with integrated graphics. Intel IGP didn't exist. I don't even get what you're saying here. Seriously what are you talking about? Since you're a reseller I assume you sold stuff back then in 1999?

If there were 114M total graphics for PCs sold in 1999, that was 114M discrete graphics. Period. I have NO idea where you get the idea about intel or AMD PCs having iGPU or APUs. [Redacted]
As far as desktop PCs go, the market was definitely larger back then, now you walk into a Best buy and you only find ultrabooks and macbooks. The overall PC market might be larger but it isn't as large of a percentage of desktops as it was in 1999, not even close. The only computing devices that realistically existed in 1999 were desktops. Now you have a ton of options such as tablet, laptop, ultrabook, smartphone, etc.


I caught your ninja edit and next time it will be an infraction. Watch the language.

-Rvenger

 

[Silverforce11]

I bet you nVidia and AMD makes more money on sub 200$ cards than they do at the cards above when applying volume.

Then find a financial report that breaks it down. You will be surprised! In recent times, the mid-range as steadily risen in price, it used to be the $200 mark a few years ago, now its the $300-400 mark.

 

[ShintaiDK]

Then find a financial report that breaks it down. You will be surprised! In recent times, the mid-range as steadily risen in price, it used to be the $200 mark a few years ago, now its the $300-400 mark.

300-400$ is performance/enthusiast price segment. Anything but midrange.

 

[AtenRa]

For those that dont remember, prior to 2004 and Intel GMA900 there were integrated graphics on motherboard chipsets even in 1999 by S3 and Via, SiS, AMD, ALi, etc etc.

Also there were more Discrete GPU designers like 3Dfx, S3, SIS, ALi, ATI, NVIDIA, Matrox also Intel with 740 etc etc.

 

[PPB]

You also need to account that people build PCs considering only CPU performance and will never slab a dGPU. For those, that iGP will be the best thing they will have. Having a strong iGP makes sense, specially when you build a feature set around them and the industry is pushing creative ways for using your iGP, even when you have a dGPU. Compute calculations offloading to your iGP being one of them (Havok recently submitted physics offloading to GPU, for example).

People need to stop the denial and realize that pure CPUs in the mainstream space are long goners, and that the desktop space has been living off Mobile and Server reject SKUs for a long while now.

 

[sontin]

Its rather a myth to claim that core gamers are the ones paying. If we look on the CPU side, do you think the desktop LGA2011 platform would even exist without Xeons? The answer is obviously no.

As volume shrinks, while cost and R&D goes up. Prices will keep rising and each time it will also fuel to the negative feedback cyclus that dGPUs are in. Cores gamers are simply not willing to pay the price for a product without someone else to share the cost with.

I bet you nVidia and AMD makes more money on sub 200$ cards than they do at the cards above when applying volume.

And when we see volume go from something like 65m units in 2013 to 50-55m units in 2014. Then it all gets clear.

nVidia's GPU business did their best Q2 in history.

 

[ShintaiDK]

nVidia's GPU business did their best Q2 in history.

Their GPU business was their best ever in Q1. And it was up 2% YoY in Q2. And it includes Quadro and Tesla cards. The highlight for the PC desktop is the 750 cards. Not exactly the "core gamers willing to pay" statement.

And volume is completely excluded and all numbers are only revenue based.

But I guess you assume people will be happy to pay plenty? Else it looks dark for nVidias future, unless Tegra can save them.

 

[sontin]

Oh, their gross margins hit a new record, too.

Revenue up, gross margins up.

BTW:
Who cares about "Quadro, Tesla and GTX750TI"? Either you include the whole discrete gpu market or nothing.

Fact is nVidia's GPU business is on a new record course. Fact is also that the overall discrete shipment is down from last year. And nVidia did this without increasing their market share.

So they are selling more and more GPUs to gamers instead of OEMs. Their Quadro, Tesla and vGPU business is getting stronger and stronger.

People still underestimate nVidia.

 

[blackened23]

I think the answer lies somewhere in between. dGPUs aren't dying tomorrow, but I also don't think dGPUs will be around 15 years from now.

As things are it is still a profitable industry. As such, if there is demand, NV and AMD will see fit to create them.

And when there isn't demand, NV and AMD have hedged their bets to expand beyond dGPUs. NV is big on the mobile market with a focus on performance per watt; they sell a ton of mobile dGPUs and use their same technology in tegra/quadro. The same R+D dollars spent on Maxwell is also spent on Tegra/Quadro/Tesla, their entire product stack.

While AMD has hedged their bets on Windows portables with their APUs. I still feel AMD needs a breakthrough in performance per watt to better compete with intel, CPU performance IS important for mobile form factors and OEMs have made this abundantly clear. Be that as it may.....this is AMD's to expand past dGPUs once it becomes unsustainable. Which isn't anytime soon.

Regardless, like I said, answer is somewhere in the middle. Last forever? Probably not. Dying tomorrow? Heck no. I see people in this thread talking like dGPUs will disappear 6 months from now. Not the case.

 

[AtenRa]

Both AMD and NVIDIA can sell APUs (x86/ARM) to substitute the entry level dGPU volume decline. Since both of them use the same architecture from bottom to highest GPGPU cards, then there is no problem even if we only have $200 and up dGPUs in desktop.