Q1 2013 discrete GPU market share?

[boxleitnerb]

I've seen the numbers somewhere, but for the life of it, I cannot remember where. Could someone please help me out?

 

[sontin]

I found that:
http://www.investorvillage.com/smbd.asp?mb=476&mn=263502&pt=msg&mid=12790483

 

[Genx87]

I found that:
http://www.investorvillage.com/smbd.asp?mb=476&mn=263502&pt=msg&mid=12790483

Fascinating infor on discrete vs integrated marketshare. I would had guessed 10 years ago discrete would be the majority of total market. The erosion of discrete marketshare has happened. But not at nearly the rate I believed.

Depressing because it shows AMD has managed to piss away a lot of marketshare.

 

[RussianSensation]

Depressing because it shows AMD has managed to piss away a lot of notebook/laptop marketshare.

That's key right there. If you look at the years 2004 to 2013, other than the Q3 2004 to Q2 2005 where AMD made in-roads in desktop dGPU market share, the overall desktop dGPU market share was roughly split 60%/40% (NV/AMD) for the last 9 years.

As has been stated ad-naseum on the forums, AMD's GPU market share losses came almost predominantly from their poor mobile execution on HD7000 series, which they explained as them not having enough $ to execute on those design wins. People tend to mix mobile and desktop dGPU market share as one, on many occasions pointing out how AMD's desktop cards can't compete, they are inefficient, etc. It shows the opposite. AMD is able to charge more $ for HD7000 series without losing market share compared to HD4000-6000 cards.

Looking at desktop market share from Q1 2008 to Q1 2012 also shows that the small die strategy and aggressive prices of $299-369 for HD4870/5870/6970 didn't work as AMD didn't gain desktop dGPU market share much beyond 40% using that strategy.

Where NV has been killing AMD is the mobile sector: execution, Optimus, etc. That's not surprising since they got 300+ design wins with Kepler and AMD threw in the towel on HD7000 due to lack of resources.

 

[sontin]

AMD won market share in Q1 and Q2 2012 because of 28nm shortage. And it looks like they won a little bit again because of the game bundles.

And the small die strategy helped them because they did not needed to sell a 370mm^2 die for less than $300.

 

[f1sherman]

NV said notebooks will be down in Q2, but GTX will smash Desktop

having to bundle several AAA games in order to compete is far from ideal for AMD

 

[boxleitnerb]

As has been stated ad-naseum on the forums, AMD's GPU market share losses came almost predominantly from their poor mobile execution on HD7000 series, which they explained as them not having enough $ to execute on those design wins. People tend to mix mobile and desktop dGPU market share as one, on many occasions pointing out how AMD's desktop cards can't compete, they are inefficient, etc. It shows the opposite. AMD is able to charge more $ for HD7000 series without losing market share compared to HD4000-6000 cards.

Ehm, no.
Compared to HD4000 days, AMD charges more, but so does Nvidia. Back then, all GPUs were dirt cheap.

Let's break it down with the 7970:
Q1 2012: market share increase at $549. No wonder, no competition.
Q2 2012: market share increase but price cut of $80 and 3 free games from mid-April on: http://www.tomshardware.com/news/Radeon-HD_7970-HD_7970-HD_7770-GTX_680,15340.html
Q3 2012: market share loss and price cuts in July and September
Q4 2012: market share loss and price cuts in the form of Never Settle
Q1 2013: market share increase but still low effective prices due to Never Settle 2

So AMD couldn't charge more than with past generations and win share at the same time (when competition was around). Either they lost market share or they had to make their products more attractive with price cuts and bundles.

 

[SirPauly]

To AMD's credit: They did take over-all discrete leadership away from nVidia in 2010. Maybe it was too much focus on the hardware and software for the APU -- why they're sliding!

 

[RussianSensation]

AMD won market share in Q1 and Q2 2012 because of 28nm shortage. And it looks like they won a little bit again because of the game bundles.

The argument was that AMD is getting killed in market share but this is not true. Their desktop market share has remained nearly flat for the last 9 years, outside of small outliers. Where AMD took a dump is notebook sector, not desktop. Something that was repeated over and over on forums for the last 1.5 years and yet people ignored; many instead extrapolated overall declining dGPU market share with declining desktop discrete GPU market share -- yet this didn't happen. When looking at market share, desktop and notebook sectors should really be separated if data is available.

And the small die strategy helped them because they did not needed to sell a 370mm^2 die for more than $300.

I disagree. It helped consumers but it didn't benefit AMD much. They didn't gain market share and neither did they make a lot of $. That's why it was abandoned by new mgmt. For 3 generations in a row they were unable to gain much market share or make a lot of $ with aggressive pricing. NV users weren't switching anyway. That's why that strategy was abandoned and GTX580's price of $450 helped AMD execute the price increase. After that, HD7970's price increase helped NV sell mid-range GTX680 at $500. And here we are in the most overpriced GPU generation of all time.

I think you confused less with more. HD6970 had a larger die size than HD7970 and sold for less. That's a losing strategy since during HD4870/5870/6970 generations, they barely made a dent in NV's desktop market share. And when they did, that was NV's fault with Fermi's fabric issue that delayed that series by 6 months. This generation, AMD is able to sell a 365mm2 die for $400-550 and the desktop market share is staying around the same 40% mark. With HD6970 they were selling a 389mm2 die for $350-380.

Of course this wouldn't be possible if consumers weren't willing to pay $400-550 for mid-range GPUs, but that's another story entirely. The point is AMD abandoning the $299-369 pricing didn't hurt them on the desktop since the desktop market share is really no worse than during HD4870-6970 days.

Ehm, no..

I don't know what you are saying no to. AMD's market share losses came predominantly from notebook sector in the last 1.5 years, due to poor execution of HD7000 mobile parts, not desktop. The data is right there linked by the OP and it shows AMD's desktop market share barely changed from HD4000 to HD7000 series. Game bundles and subsequent price drops is a normal course of action for GPUs. The point is HD7000 desktop parts did no worse than HD4000-6000 parts on the desktop despite large price increase from those $299-369 price levels. Therefore, from AMD's point of view, raising prices was the way to go since they maintained market share without suffering losses for their GPU division. For us consumers, this results in HD7970-680 the most overpriced generation.

The 1.5 year old argument that AMD raising prices was a bad move doesn't hold water anymore in light of the desktop market share data. If raising prices was a bad move, AMD's desktop market share would have eroded from 40% to 20-30%, etc. It barely moved. On the contrary, Rory Read mentioned that wafer prices will rise with lower nodes. That means if AMD didn't raise prices, then the more expensive 28nm wafers would have transferred into losses for their division if they had to sell HD7970 at $299-369. With NV things are a little different since they have a very profitable professional GPU division which helps amortize their R&D and boosts profitability of their graphics unit.

AMD started at $550 and 1.5 years later the same GPU is selling for $390 on Newegg, a price higher than HD4870/5870/6970 debuted at. That means AMD was able to maintain higher ASP and for longer this generation without suffering desktop GPU market share losses. It stands to reason they will want to repeat the same strategy again - launch early at high prices by beating Maxwell's launch date and then as NV responds, lower prices to maintain market share should Maxwell beat VI.

 

[SirPauly]

I don't know what you are saying no to. AMD's market share losses came predominantly from notebook sector in the last 1.5 years, due to poor execution of HD7000 mobile parts, not desktop. The data is right there linked by the OP and it shows AMD's desktop market share barely changed from HD4000 to HD7000 series. Game bundles and subsequent price drops is a normal course of action for GPUs. The point is HD7000 desktop parts did no worse than HD4000-6000 parts on the desktop despite large price increase from those $299-369 price levels. Therefore, from AMD's point of view, raising prices was the way to go since they maintained market share without suffering losses for their GPU division. For us consumers, this results in HD7970-680 the most overpriced generation.

What is the break down of their Graphic Business?

Discrete, console, and professional?

 

[SirPauly]

They didn't gain market share and neither did they make a lot of $

AMD did over-take nvidia on over-all discrete leadership and had nice quarters!

 

[UaVaj]

majority of pc are a utility (surfing web, desktop publishing, youtube, etc). for that - integrated GPU is plenty.

 

[SirPauly]

One can make the point that traditional x86 proprietary integration has to worry about mobile devices and its open nature! Will be interesting to watch!

 

[Siberian]

With no answer to the 7xx series AMD could dip well below 25%

 

[Rvenger]

With no answer to the 7xx series AMD could dip well below 25%

700 series is a glorified rebadge and has no upgrade incentive if you already have a 79xx or 600 series. Titan doesn't count since it is branded as a 600 series card.

 

[BallaTheFeared]

Luckily there are more products than just the 770, and the flash to 770 rumor was debunked, but keep going it's good humor.

 

[Siberian]

Luckily there are more products than just the 770, and the flash to 770 rumor was debunked, but keep going it's good humor.

They will probably price the old 6xx series more agressivley. When your enemy goes to ground, leave no ground to go to.

 

[Rvenger]

They will probably price the old 6xx series more agressivley. When your enemy goes to ground, leave no ground to go to.

6xx series will be discontinued Skus shortly after 700 release besides low end and maybe some midrange including the 650ti boost.

 

[tviceman]

deleted

 

[blackened23]

They will probably price the old 6xx series more agressivley. When your enemy goes to ground, leave no ground to go to.

Nah, they won't lower in price. Having watched pricing of prior generation parts over several years, they will just be EOL and slowly be discontinued among all major retailers. This happened with every prior generation GPU series that I can think of.

Intel does the same with their CPUs. They just go EOL and slowly evaporate.

 

[blackened23]

Luckily there are more products than just the 770, and the flash to 770 rumor was debunked, but keep going it's good humor.

Well, I see nothing wrong with the 770. The performance differential between the 670 and 770 is the same as was the 470 to 570, so who cares?

That said, it can certainly be considered a rebadge. It's a GK104 with GPU Boost 2.0 and slightly higher clocks. Sounds like a rebadge to me. By the way, from what I understand, the BIOS floating around actually does enable GPU boost 2.0 on the GTX 680. So yeah.

There's really nothing wrong with a re-badge, the 770 should be a 400$ card and perform appreciably better than the 670.

 

[BallaTheFeared]

I don't either, so long as it isn't accompanied by another price hike of the x70 bracket.

We don't know this yet, until it's shown that that is all it is we're just discussing rumors. If that is the case, it's just GK104 clock pumped with boost 2.0 than it's a rebadge for sure. A video of that would go a long ways.

I don't disagree, it's not a bad thing, but an actual uarch refresh is worlds better at the same time.

 

[RussianSensation]

There's really nothing wrong with a re-badge, the 770 should be a 400$ card and perform appreciably better than the 670.

We don't really know how much faster it will be to conclude that it will be appreciably better. Your comparison of GTX470 to 570 and GTX670 to GTX770 seems odd since GTX570 is about 15% faster and GTX770 won't be a replacement for GTX670 but 680 unless NV prices GTX780 at $499. The price cut of GTX670 to $299 in the form of GTX760Ti should have a much bigger impact on the market than GTX680 improving in performance by 5-10% and dropping to $400 in the form of GTX770. We already have such a card and it sells for $410 and it will come with a superior game bundle than just Metro LL. Dropping GTX670 level of performance from $360-380 to $299 would put significant pressure on HD7950s though since most people do not overclock. Once that happens, HD7950 would need to drop to $240-250.

 

[boxleitnerb]

I don't know what you are saying no to. AMD's market share losses came predominantly from notebook sector in the last 1.5 years, due to poor execution of HD7000 mobile parts, not desktop. The data is right there linked by the OP and it shows AMD's desktop market share barely changed from HD4000 to HD7000 series. Game bundles and subsequent price drops is a normal course of action for GPUs. The point is HD7000 desktop parts did no worse than HD4000-6000 parts on the desktop despite large price increase from those $299-369 price levels. Therefore, from AMD's point of view, raising prices was the way to go since they maintained market share without suffering losses for their GPU division. For us consumers, this results in HD7970-680 the most overpriced generation.

The 1.5 year old argument that AMD raising prices was a bad move doesn't hold water anymore in light of the desktop market share data. If raising prices was a bad move, AMD's desktop market share would have eroded from 40% to 20-30%, etc. It barely moved. On the contrary, Rory Read mentioned that wafer prices will rise with lower nodes. That means if AMD didn't raise prices, then the more expensive 28nm wafers would have transferred into losses for their division if they had to sell HD7970 at $299-369. With NV things are a little different since they have a very profitable professional GPU division which helps amortize their R&D and boosts profitability of their graphics unit.

AMD started at $550 and 1.5 years later the same GPU is selling for $390 on Newegg, a price higher than HD4870/5870/6970 debuted at. That means AMD was able to maintain higher ASP and for longer this generation without suffering desktop GPU market share losses. It stands to reason they will want to repeat the same strategy again - launch early at high prices by beating Maxwell's launch date and then as NV responds, lower prices to maintain market share should Maxwell beat VI.

I'm saying no to your claim that AMD could sell for higher prices and (at the same time) don't lose marketshare. Prices were higher for ALL market participants at the time of HD4000 vs GTX200. So that says nothing. HD5870 was launched at a $379 MSRP, HD6970 was sold for $369 and although HD7970 launched at $549, it reached less than $400 quickly (in April) after competition arrived (price cut to $479 + 3 free games which I estimated at $100 value); which puts it at HD6970 levels. In July, the 7970 GHz launched at $499, but again bundled with 3 games, so the effective price was rather $400 - again at HD5000/6000 levels. So it is absolutely false that AMD could sell at higher prices. You cannot just ignore the bundles.

As for your "bad move" part: AMD raised prices, but only in the beginning when there was no competition at all or when availability of said competition was bad. After that, they quickly backwatered with price cuts and bundles. Had they not done that, their market share would definitely have plummeted the way you describe. A 7970 GHz @$499 without free games would not work out for them.

And the mobile segment has nothing to do with that - the numbers in sontins link which I used were for the desktop segment only. Top left graph. Discrete notebook is a different graph, right next to it.

 

[SirPauly]

AMD could sell higher with no 28nm competition! I'm curious if Radeon is actually making profit when their Graphic Business also is workstation, where they're hitting record territories, and revenue from consoles?

This higher price discussion from AMD doesn't make any sense based on they have been lowering their prices virtually with every Kepler launch and only can be sustained when there was no nVidia competition.