2010: GPU Market Outlook
As promised in my article analyzing the roadmaps of major CPU vendors where predicting the market trends for rest of the year, I’m here today with the next article in the series – the 2010 GPU Market Outlook. The article today is based on the roadmaps, announcements, (some solid) rumors and past trends shown in the graphics industry. Usually the objective of such an analysis is to cover the breadth of the market rather then to go deep and pin point the exact dates (or shader cores for that matter) of a graphics card.
Like I mentioned in my previous article, 2010 could have very well been the year of the GPU, had Intel actually made some breakthrough with its Larrabee architecture. Nevertheless there is still a lot of exciting stuff by AMD and NVIDIA to make this article interesting.
Unlike the CPU market, the Graphics market still only has two major players in almost all the segments. For those who don’t really know much about it all – they are NVIDIA and AMD (who bought ATI back in 2006, though the brand name is still alive). Surprisingly, most market share statistics would show a third player that overshadows both of these. Unfortunately, that third player – Intel, is only present because of its integrated GPU solutions in most of its chipsets which actually don’t have the potential for powerful 3D graphics.
While the market for dedicated graphics doesn’t span a vast array of devices like we found in our CPU, we still do have a difference in target audience that can very well put today’s article at the same magnitude. So without further delay, here’s a glimpse of what we’ll be going through today:
- Desktop: Enthusiast
- Desktop: Performance
- Desktop: Mainstream
- Desktop: Budget
- Mobile: Enthusiast
- Mobile: Performance
- Mobile: Mainstream
- Mobile: Budget
While the desktop and mobile (laptop/notebook/netbook) markets look increasingly similar, the actual product range offered there vary a great deal both in terms of specs and performance.
I decided to include a brief overview of the handheld industry as well thanks to all the craze of MIDs, Tablets and Smartphones.
The enthusiast segment of the desktop market marks the territory in which the actual consumers are rarely bold enough to venture into. These chips demand a high premium for their performance, but are guaranteed to be the fastest thing on the entire planet.
While usually represented by only a couple of Dual-GPU solutions – this year, the increasingly high price of the Radeon HD 5800 series puts it well into the enthusiast class even though AMD’s intentions were to market the Radeon HD 5850 as a Performance chip. For most part of the year Radeon HD 5870 and Radeon HD 5790 would be AMD’s offering for this market segment carrying a price tag north of $300.
As for NVIDIA, the only offering they have left for this segment is the Dual-GPU GeForce GTX 295 and the older GeForce GTX 285.
But exciting stuff in this segment starts at the end of March when NVIDIA would unveil the full details behind their upcoming DirectX 11 based Fermi processors – the GeForce GTX 480 and GeForce GTX 470. Both of these cards would support NVIDIA’s PhysX along with its upcoming Eyefinity competitor – the 3D Vision Surround View.
From what’s known up till now, all the GF100 based GPUs from NVIDIA would boast GDDR5 memory and put special emphasis on its CUDA parallel processing technology.
Later down the year – probably around Q4, AMD are rumored to unveil their next generation GPU architecture which is said to be something totally different from their current Evergreen chips which were just a refresh of the R700 architecture. Expect the current generation enthusiast chips to venture way down into the performance and mainstream markets around that time.
The performance market usually spans between $200 and $300, and as of now there had been a huge vacuum of new chips in this segment. The reason was that AMD was aggressively phasing out its Radeon HD 4890, the Dual GPU 4870 X2 and 4850 X2. Further to increase AMD’s woes, low yields of TSMC’s 40nm manufacturing processes hiked the prices of AMD’s intended performance chip – the Radeon HD 5850 well into the Enthusiast level budgets. The resulting gap proved to be a problem for AMD up till now.
Thanks to the recent release of the Radeon HD 5830, AMD again has a chip in the Performance segment. At around $250, the Radeon HD 5830 paks similar performance to a Radeon HD 4890 it intended to replace though still has about $40 premium over its price. And even with features like Eyefinity and DirectX 11, the $250 tag makes it a hard sell right now.
AMD’s Performance lineup would actually look more attractive once NVIDIA rolls out their GF100 cards into the market in high volumes – which is expected to happen around May or June. AMD would probably cut the prices of its Radeon HD 5850 and Radeon HD 5870 to bring them in the more affordable $200 to $300 segment.
Ironically, the Radeon HD 5830 is enjoying almost no competition from NVIDIA either thanks to the low supply of its GeForce GTX 275. The only NVIDIA card that manages to get in this segment right now is the GeForce GTX 260 which still retails for a bit over $220 but even they are getting scarce.
Though the available options in the performance segment are almost none right now, the upcoming release of Fermi GPUs by NVIDIA would pretty much fill this gap – both with new chips around late Q2/early Q3 and price cuts on the existing boards.
Mainstream graphics are where all the action is at right now. You’ll find a lot of higher end old generation cards here which received some healthy price cuts as well as new offerings by AMD.
The Radeon HD 5770 and Radeon HD 5750 are AMD’s premier offerings for the mainstream market right now. And these are a couple of awesome pieces or hardware. While they do have the ability to post impressive numbers on their own, pairing up two 5750s in Crossfire (total cost around $290) would result is performance pretty close to – and sometimes even higher than a single Radeon 5850 (cost $320-$340).
Similarly, a Crossfire combination of Radeon HD 5770 (cost around $340) would result in a performance that can challenge a single Radeon HD 5870 (goes around $410). These two chips have an impressive potential to scale and I won’t be surprised if AMD plans against reducing their prices even after the release of GeForce 400 series.
On NVIDIA’s front, their Mainstream lineup would likely be filled with new parts from the GeForce GTS 400 platform. While nothing has been announced regarding it yet, we can be assured that it would offer the bare minimums – DirectX 11, PhysX, Surround View and CUDA parallel processing technology.
While current generation NVIDIA cards like the GTS 250 and GTX 260 would still be around, I wouldn’t recommend either as a viable option because of their lack of support of DirectX 11.
Don’t have more than a $100 to spend on a new GPU? Don’t worry because there are a number of options available for you as well.
First and foremost if you want the new technology like DirectX 11 and Eyefinity, then AMD is the way to go. For less than a $100, they have three separately priced solutions which pack both of this options across the board. The Radeon HD 5670, the 5570 and the 5450 retail at $100, $80 and $40 respectively. There’s also a planned Radeon HD 5550 which may land around $60.
On NVIDIA’s front you can find everything from current generation GeForce 200/300 variants and all the way down to even GeForce FX series. All the cards are randomly priced and therefore, I wouldn’t recommend getting any one of them besides the GeForce 200/300 variants.
NVIDIA would populate the sub $100 range with at least one DirectX 11 chip this year so there is hope for NVIDIA fans who are on a tight budget.
The mobile graphics segment is very different from its desktop counterparts. Even though most of the mobile graphics chips are named the same, their prices and performance vary greatly.
AMD released all its mobile graphic chips in one go at CES in January. Unlike its desktop lineup, the Enthusiast chips (HD 5800 series) for notebooks are based on the Juniper architecture instead of Cypress. Similarly, the Performance chips (HD 5700 series) - while still retaining the same naming schemes are based on Redwood instead of Juniper as in the desktop. And continuing the same trend, the Mainstream lineup (HD 5400) would be based on Cedar instead of Redwood.
In the mobile space, AMD has named its new architecture as Broadway, Madison and Park for Enthusiast, Performance and Mainstream respectively. And each would have three variants being XT, Pro and LP. All of the new lineup would feature DirectX 11, GDDR5 memory and a 40nm manufacturing process across the board.
As for NVIDIA, they would be introducing new chips in their GeForce 300 lineup in the first half of 2010 as well as an entirely new GeForce 400 lineup in the second half of 2010. While the the GeForce 300 series will only support DirectX 2.1 and OpenGL 2.1, the GeForce 400 series will be going next generation by supporting DirectX 11 as well as OpenGL 3.x.
Though by the time NVIDIA releases its DirectX 11 solution for notebooks, AMD would have released its Danube and Nile platform boasting integrated DirectX 10.1 with along with discrete DirecrX 11 GPUs for more demanding applications.
While it looks like AMD has grabbed a commanding share of DirectX 11 in the Desktop and Mobile GPU segment, there is still one market which is still pretty much alien to the company. AMD’s lack of interest in the Handheld graphics market has given its arch nemesis NVIDIA an open hand to attempt a hostile takeover.
You would think that this means handhelds would prove to be a cinch for NVIDIA to dominate, but that’s not the case. There is one more player here – someone who has a far more commanding lead compared to NVIDIA which is still seen as a newbie in this market. And that player is PowerVR – with its SGX Series5 graphics chip line.
SGX enjoys wide market adoption, thanks to being utilized by the likes of iPhone 3GS, iPod Touch, Nokia N900, Palm Pre, Motorola Droid and Intel’s GMA500 platform. It is also present in the Apple A4 system-on-a-chip powering the iPad.
NVIDIA’s offerings in this segment include last year’s Tegra platform which only was prominently featured in Microsoft’s Zune HD. This year, NVIDIA plans to use Tegra 2 for higher end MIDs and Tablets while Tegra would still be used for Smartphone and other low power demanding platforms. The GoForce processors in the Tegra platform, while being technically more powerful compared to SGX, are still a dwarf in the market because of being tied exclusively to the Tegra platform.
Other notable players in the segment include Imageon (Qualcomm Snapdragon) and Vivante (Marvel Armada 500/600).
The next five years (if there are any) are clearly going to deviate from the current trend of raw graphics muscle and put more emphasis into secondary features. NVIDIA is already touting its CUDA and PhysX as the future of gaming, and now AMD/ATI itself has come up with its answer which is Eyefinity.
But both of these desktop graphic giants are yet to prove that General Purpose GPU computing is finally here. NVIDIA says that its Fermi architecture would revolutionize GPU computing – but that comes at a high premium of cost, die size, chip voltages and temperatures – something that matters a lot to the average consumers.
Meanwhile Intel’s status on the graphics front remains unknown. Whether the company would decide to dive into the Larrabee project again? Its hard to tell but they really should figure something out before AMD gets serious with its Fusion technology next year.