Two very different analysts stress the importance of software in chip and server ecosystems.
Last week, veteran semiconductor electronic design automation (eda) analyst garysmith made the following observation about ecosystems in his annual Design Automation Conference (dac_2014) kick-off presentation: “With such a range of possible original equipment manufacturer (OEM) scenarios, you (chip designers) need to understand who is your customer and who is your competitor. And the relationships almost change day by day. The key is to develop an ecosystem as stable as ARM’s in this changing world of relationship.”
Smith’s comments were part of his discussion about the need for a larger system approach to chip design with a complementary need for a strong ecosystem. Both needs were driven by the changing roles of chip design tool vendors within the semiconductor (hardware and software) electronics supply chain. A traditional OEM would buy the platform design, manufacture the system-on-chip (SoC) at a foundry, wrap plastic around it and take it to market (e.g., low end cell phones). However, today’s OEM could also be a vertically integrated company that outsources manufacturing (e.g., high-end Apple cell phones).
Yesterday, Canaccord Genuity issued an equity report in which analyst Matthew Ramsay noted that, “ …the nascent ARM server ecosystem is gaining momentum and the eventual royalty opportunity for ARM will prove larger than consensus expectations, certainly larger than management conservatively has forecast to investors.”
Here, again, was another statement about the importance of ARM’s ecosystem. Taken together, Smith’s and Ramsay’s comments cover a wide portion of the semiconductor market, namely, System-on-Chip (SoC)/IP and cloud-based servers.
What are the common differentiators in each of these markets? Software! Both are industries where hardware is maturing and being commoditized. That’s why it is the software that plays the differentiating role in determining market share, from firmware to applications. To emphasize this point, Ramsay added this comment about ARM’s competitive and technological advantage (over Intel): “Operating system and software support - The role of and increasing momentum behind software/OS support for ARM servers including Linux/Windows and industry groups such as Linaro and ARM’s Server Base System Architecture (SBSA).”
Do you agree? I’d like to get your take on software’s part in this varied ecosystem. - JB
Does the ARM ecosystem have tiers? As you point out, the word itself seems to have gotten to big, to inclusive. I think you hint at the problem, namely, that meaning of "IP." In my opinion, this ambiguity reflects the consolidation occurring in the semiconductor hardware IP market. (I've written a lot about IP on another site, e.g.: Will Discrete IP Threaten the Subsystem Market?) Of course, software IP is a closely related but different matter.
It is truly amazing that chip vendors can tape out only once - with very careful simulation, sw and hw verification, emulation, prototypes and lots of planning. I also remember the days of adding jumper wires (often red as in "red lines") to fix problems on PCBs. Chip vendors do something similar in silicon via self-healing circuits and replication of critical circuits with spare transistors.
But the reason that a lot of new chips work is that very little changes from one chip version to another. Maybe a faster processor is added or a upgraded interface, but most other changes are incremental, limited to well tested blocks of hardware (and hopefully software). IMHO.
Great input! So, should the meaning of ARM's "ecosystem" be updated?
It is starting to get hard to know what ARM ecosystem means, it may include producing chips which perhaps have no IP they're paying ARM for at all! It involves cooperation using the IP of many companies which may be competing strongly with each other. What will ARM be doing in the future? Who can tell, everything changes - IBM now is not the same as IBM ten years ago or fifty years ago and will ARM be able to adapt like that?
As to software, who would have thought only ten years ago that design tools and libraries and emulation and verification would be such that companies have a reasonable chance that the first tape out of chips with a billion transistors will work to an acceptable standard? It is not so long ago that it was common for the first PCB to have tracks cut and extra wires soldered on to fix problems. Using great big FPGAs to simulate a system help of course too, but frankly I'm still amazed the stuff ever works. And we've applied the name ARM ecosystem to a lot of that even if it isn't quite deserved it still was the ARM IP model that started it all working properly.