Posted by Richard Fichera on September 26, 2012
[For some reason this has been unpublished since April — so here it is well after AMD announced its next spin of the SeaMicro product.]
At its recent financial analyst day, AMD indicated that it intended to differentiate itself by creating products that were advantaged in niche markets, with specific mention, among other segments, of servers, and to generally shake up the trench warfare that has had it on the losing side of its lifelong battle with Intel (my interpretation, not AMD management’s words). Today, at least for the server side of the business, it made a move that can potentially offer it visibility and differentiation by acquiring innovative server startup SeaMicro.
SeaMicro has attracted our attention since its appearance (blog post 1, blog post 2) with its innovative architecture that dramatically reduces power and improves density by sharing components like I/O adapters, disks, and even BIOS over a proprietary fabric. The irony here is that SeaMicro came to market with a tight alignment with Intel, who at one point even introduced a special dual-core packaging of its Atom CPU to allow SeaMicro to improve its density and power efficiency. Most recently SeaMicro and Intel announced a new model that featured Xeon CPUs to address the more mainstream segments that were not a part of SeaMicro’s original Atom-based offering.
So what does AMD hope to gain from this acquisition? The strategic goal is a differentiated server architecture into which it can insert AMD silicon and IP. The SeaMicro architecture shares many of the characteristics of the Calxeda/HP Redstone architecture and can potentially allow AMD to produce reference architecture for its system partners to leverage that gives them significant density advantages over any conventional x86 server offering. [Looks like I was wrong about this.] Since this architecture is initially targeted at dense computing environments such as those found in Web 2.0, cloud service providers, and similar workloads, as well as potentially at conventional HPC, the prospects for AMD are interesting. Long-term success will hinge on several interlocking factors:
- How much will AMD focus this offering on the ultra-dense space that SeaMicro started addressing versus trying to also take a bite out of the enterprise space?
- How aggressively will it scale it? The SeaMicro fabric does not extend beyond the single enclosure boundary, but that appears to be an implementation decision, not a fundamental architectural issue. My personal belief is that fabric-based servers are the wave of the future, with SeaMicro Calxeda/HP serving as very significant guidelines for future development. Scaling the fabric first for management and then for actual data traffic for shared I/O and eventually for coherent memory operations if needed is a natural progression.
- What other IP will AMD add? AMD has an interesting portfolio of GPU technology as well as CPU, and the incorporation of GPU technology could provide further differentiation in niche markets. AMD has a choice of ways to monetize this IP, from selling systems based on its own processors (possibly as early as late this year) to providing reference designs and/or licensing the architecture or selling components like the fabric ASICs.
- Will it also support ARM processors? If AMD becomes the source of licensed IP for vendors wanting to chase HP in ultra-dense servers, it can also make money by licensing its technology and selling components for non-x86 architectures.
- What will Intel do? At least one of Intel’s partners, HP, already has a fabric-based dense server design that can support x86. Will Intel develop a similar reference architecture in the hopes that its major systems partners will adopt it?
All in all, many ways this could play out, but a bold move and one that may serve to trigger a rapid change in the types of servers available to users over the next 24 months.
[In its most recent announcement, AMD announced a Xeon and AMD CU based upgrade, the SeaMicro 15000, which has the ability to connect to a total of 16 I/O trays for a total of over 1000 spindles of disk.]
Search Forrester's Blogs
Free Webinar Series
The Top Emerging Technologies To Watch »
- Amy DeMartine (4)
- Andre Kindness (28)
- Bryan Wang (16)
- Christian Kane (5)
- Christopher Voce (8)
- Dave Bartoletti (26)
- David Johnson (48)
- Doug Washburn (37)
- Eveline Oehrlich (12)
- Frank Liu (8)
- Glenn O'Donnell (28)
- Henry Baltazar (8)
- James Staten (116)
- Jean-Pierre Garbani (13)
- John Rakowski (27)
- JP Gownder (95)
- Katyayan Gupta (17)
- Laura Koetzle (1)
- Lauren Nelson (10)
- Michele Pelino (5)
- Nupur Singh Andley (7)
- Richard Fichera (134)
- Sophia Vargas (5)
- Stephanie Balaouras (1)