Tuesday, September 9, 2014

Tablet processor shipment projections for Q3: Rockchip growth too good to be true?

Tablet chip growth in Q3, Rockchip expected to dominate

In a recent update on the Chinese tablet application processor market, DigiTimes reported projected growth of the tablet chip market in Q3 from seasonal demand, while chip companies are rolling out new chips with an improved cost structure.

The article projects that Rockchip's shipments will increase significantly in Q3, while its cooperation with Intel will bear fruit with plans to ship its first processors with integrated baseband processor in Q4 2014. Rockchip already was the number one provider of tablet processors in China in Q2 2014. It also mentions growth for MediaTek, a steep decline in market share for Allwinner, and growing shipments from Intel, which is likely to pass Allwinner and reach the number three position. The article is otherwise devoid of the more specific figures that DigiTimes tends to publish after the end of the quarter, and the article may be based on preliminary expectations and not reflect final numbers.

Rockchip's booming shipments too good to be true?

However, some of the information of the article appears to be based on very optimistic projections of sources very close to Rockchip. Expectations that the Intel collaboration (using an entirely new processor architecture, new and unfamiliar IP blocks, a different foundry, and a significantly more complex integrated SoC chip design than previous Rockchip products) will "bear fruit" in Q4 2014, several months ahead of the planning described in the Intel announcement press release, has to be met with skepticism. Rockchip's track record, which includes severe delays and problems bringing to market significantly less complex chips such as RK3168 and RK3288, puts further doubt on the optimistic projections.

Rockchip is continuing to rely on its existing products RK3188T, RK3168 and RK3026, which are not likely to be very cost-effective in terms of manufacturing cost. Information from the supply chain from sources such as ARMDevices does not show evidence of any new cost-effective chips from Rockchip shipping in new tablets. Only Allwinner (with their new A33 chip) and MediaTek are showing evidence of providing new chips with an improved cost structure.

It is not unlikely that Rockchip's relatively recent low-end RK3026 (a dual core 1.0 GHz Cortex-A9-based SoC manufactured using a trailing-edge 40nm process) is seeing traction in the less visible ultra-low-end market (shipping primarily to cost-sensitive end markets) and is shipping in high volume (taking over from bottom-end chips from companies such as Allwinner and Actions), partly explaining the increase in shipments. However, none of Rockchip's current line-up of chips is likely to be very cost-effective due to their design (especially the use of large, outdated Cortex-A9 CPU cores). The same design choices also make them less power-efficient (with average battery life or worse) than a number of competitive products.

Rockchip addressing cost concerns in upcoming chips

However, Rockchip appears to be finally addressing the high manufacturing cost and limited efficiency of its chips with new more cost-effective designs that will appear on the market in the near future. The quad-core Cortex-A7 RK3126/RK3128 with Mali-400 MP2 GPU, manufactured at 40nm, have recently appeared on Rockchip's website, closely matching the specifications of the recently announced A33 from Allwinner and in a more general sense SoCs that MediaTek has been shipping for some time. A roadmap of upcoming Rockchip SoCs also provides evidence of a high-performance octa-core Cortex-A53 "MayBach" SoC, which is likely to address the issues that makes the RK3288 virtually unsuitable for tablets.

Explanations for Rockchip's paradoxical market share dominance in Q3 2014

To explain to current market success of Rockchip in terms of volume shipments and market share despite chip cost concerns, one can speculate about possible explanations:
  • The combination of the utter failure by Allwinner to maintain a competitive product offering (their new A33 may be too late to be significantly reflected in Q3, and may not be free from the concerns that earlier caused the A20 and A23 to be utter failures), and the chip supply shortage faced by MediaTek, which otherwise may have been able to fill much of the gap because of its competitive and cost-effective products, may have conspired to make Rockchip (which has plenty of capacity available at struggling GlobalFoundries) the only source of sufficient numbers of tablet chips in the current quarter. In principle, this would cause in temporary rebound in tablet processor prices, making the cost structure of Rockchip's solutions somewhat less problematic.
  • Rockchip is buying market share, selling large numbers of chips for little profit or even losses, while taking advantage of significant debt financing, using up financing provided by partnership deals or other undisclosed sources of funds. In the Chinese technology world, similar irrational high-risk boom-and-bust schemes executed by a company are a fairly common occurrence. This may include putting in very large orders at foundries with little concern about whether the volume of chips can be sold, whether the chip design is free of defects and issues that could make it unusable, or whether a profit is actually made on sale of the chips (the only thing that counts is ambition and volume).
  • GlobalFoundries has been struggling, with Rockchip being one of the few significant clients for its 28nm HKMG process outside of AMD. It is possible that Rockchip is receiving preferential treatment and significant discounts for the manufacturing of chips such as the RK3188T and RK3168, because otherwise GlobalFoundries would see even more severe underutilization of its fabs. Additionally, tight capacity at most other foundries (especially TSMC) may not only have affected MediaTek, but also many of the other smaller players in the field.
The truth is probably a combination of some of the mentioned explanations and possibly other reasons that may become more clear in the future.

Update (November 2014)

Based on a recent DigiTimes report from November 11 on tablet processor market share in China in Q3 2014, it seems the estimates for Rockchip were indeed much too optimistic, with a unit shipment decline of 12% in Q3 instead of the strong growth implied by the statements from early September. Rockchip's shipments have been affected by lower than expected overall market growth, and also by Intel's penetration of the Chinese tablet market, whose chips to a large extent cover the segment of Rockchip's high-volume mid-range RK3188(T) chip, which is also becoming less competitive from a manufacturing cost standpoint.

Although based on information from DigiTimes Rockchip still held the leadership position in terms of unit shipments in China in Q3 2014 (while being passed by MediaTek in terms of worldwide shipments, including global brand-name manufacturers), for Q4 2014 its market share will come under pressure according to a forecast by DigiTimes that says that MediaTek will take over the number one position in the Chinese market in Q4 2014, probably because of lessened production capacity constraints. Rockchip is transitioning to the more cost-effective quad-core Cortex-A7-based RK3126 for the cost-sensitive market, but it appears that this will not prevent Rockchip from losing its leadership position in China to MediaTek.

Sources: DigiTimes

Updated September 28, 2014 (Add information about upcoming, more efficient Rockchip SoCs).
Updated November 26, 2014 (Update with recent information).

No comments: