AMD also was able to show a non-cash gain of $325 due to the GLOBALFOUNDRIES deconsolidation. While this does not affect their bottom line, it does reflect with the overall valuation of the company. AMD was also able to increase their cash on hand from $1.77 billion up to $1.93 billion, which is a welcome change for AMD as they were burning through cash for the past 12 quarters.
Desktop and server products held steady, but did see a decrease. Notebooks saw an increase in uptake due to the latest Turion platforms that are available to notebook makers. The introduction of 8 and 12 core server processors did result in some gains. While unit shipments were down, the richer mix of microprocessors lead to an actual ASP increase.
On the graphic’s side, AMD did very well. Q1 royalties from consoles were lower, but this is not surprising considering that Q4 is when most of the royalties come in due to the holiday buying season. AMD made significant gains in standalone graphics chips, primarily due to their top to bottom DX11 offerings. Unfortunately for AMD, they were unable to fully capitalize on demand due to the shortage of available 40 nm wafer starts at TSMC. AMD expects to be supply constrained for some time yet, so prices on these parts are expected to remain high. While the original HD 5850 and HD 5870 were released at MSRPs of $279 and $379 respectively, we are seeing actual prices $20 to $40 higher. Overclocked versions are significantly higher in price as well. This situation is also exacerbated by NVIDIA, who is using the majority of their 40 nm starts to manufacture the Fermi chip. This chip apparently is yielding quite poorly, so a lot of wafer space is wasted on defective chips.
AMD expects to see a graphics refresh late this Fall and early Winter. These will again be 40 nm parts, but with redesigned memory controllers, RBEs, texture units, and caches/GPGPU portions. It is said that the actual stream units will be unchanged from the current HD 5000 series of parts. Past 2010, AMD expects to use GF’s 28 nm HKMG process for graphics chips. GF is supposed to open that line around Q1 2011. Apparently TSMC had promised some years ago that it would have its 32 nm HKMG process open in Q3 2010 for production. Considering that their current 40 nm process is still not exactly mature, graphics chip manufacturers have had to radically change their plans. AMD was initially going to have a totally redesigned DX11 chip family for that process, but that design was not fully appropriate for the larger 40 nm process and a compromise on a refresh part had to be made.
AMD has already shipped 6 million+ DX11 chips to the market. With NVIDIA only now releasing a high end DX11 chip, AMD continues to expect strong sales up and down their range of parts. The HD 5800 series of parts, while not as fast as the GTX 470/480 chips, are cheaper overall, run cooler, pull less power, and are actually widely available. NVIDIA still does not have an adequate competitor to the 5500, 5600, and 5700 series of parts from AMD.
The “Llano” chip is comprised of four Phenom II based cores, no L3 cache, and integrated DX11 graphics. It has already been sampled to AMD’s partners… and apparently Apple as well.
AMD looks to only update their integrated graphics to DX11 with the upcoming Fusion parts. The primary desktop/notebook part is codenamed Llano, and it is designed around 4 Phenom II class cores with no L3 cache, but with a DX11 class integrated graphics portion. AMD announced yesterday that they have working samples of this part, and have distributed them to partners for testing and potential integration into desktop and notebook lines. AMD expects to begin mass production of these parts in 2H 2010, and is conservatively expecting a market release in early 2011. If the design is good enough, and GF can ramp 32 nm SOI production, then we might see a late 2010 release. These chips are rumored to hit the 3.0 GHz mark with a 25 watt power draw. Lower clocked variants will of course feature lower TDPs, down to the 2.5 watt range. The only question left unanswered is at what speed the graphics portions will be running at.
The other Fusion part is codenamed Ontario. This is based on a new micro-architecture called “Bobcat”. This is aimed squarely at the low power, high performance netbook/tablet section. AMD has samples that it is testing internally, but has not delivered samples to manufacturing partners. The tight integration of graphics on this chip, along with its brand new architecture with a focus on power savings, could give AMD a tremendous boost in thin and lights, tablets, and perhaps even handheld units.
The power saving and architectural features of Llano and Ontario have apparently piqued the interest of Apple. With tight supply of mobile i3/i5 parts from Intel, Apple already had to delay their new MacBooks. Furthermore, they are pairing most of these products with discrete graphics chips from NVIDIA and AMD. Apparently Apple is not particularly happy with Intel’s integrated graphics. If AMD can deliver an “all in one” part which achieves near similar processing performance and power consumption, along with advanced graphics features and performance, we might see next year’s MacBook models feature AMD processors and chipsets. Also consider that AMD already has SATA 6G support in current chipsets, while Intel does not.
We just saw the Magny-Cours introduced to the server market, and it has already had a very positive impact on AMD’s bottom line. Bulldozer will likely be initially introduced on the server side as “Interlagos” and the dual ALU per core architecture, combined with potentially on die graphics could make this a integer and floating point monster.
In the near future we can expect to see Athlon II X3 and X4 notebook platforms being released. These could very well be redesigned chips based on the revision C3 Phenom II core. Also consider that AMD could be manufacturing these with the extra Low-K dielectric is utilized in the upcoming Phenom II X6 processors. The combination of these advancements could drop the TDP of these parts dramatically enough to be a very interesting notebook chip. The Athlon II chips also do not possess the large L3 cache of the Phenom II processors, which can consume a larger percentage of power for the chip. AMD is also planning to release a slightly updated integrated graphics core with the 880G series. From my limited understanding it will not be much of a change from the current 785G/890GX graphics portion, and certainly will not be DX11.
AMD followed in Intel’s footsteps by having a stronger than usual Q1. They also have a solid product mix of desktop, notebook, and server chips. The release of the 6 core Istanbul processors last year has continued to improve AMD’s bottom line, and the recent release of the Magny-Cours 8 and 12 core products has further cemented them in the value server market. Add in AMD’s near dominance of the DirectX 11 market, and we can see that AMD is in good shape to continue to compete throughout this year. Next year’s introductions of the Bobcat and Bulldozer cores should help AMD compete more adequately against Intel and their next generation “Sandy Bridge” products for the desktop, and their next generation low power processors for netbook/tablet use.
AMD having a solid quarter in the black is a very positive building block for the company in 2010.
