'Cirrus'ly NASty

"Late to market….feature-deficient….why'd they even bother introducing this thing"? That was my gut reaction after getting briefed by Cirrus Logic last week on their NAS reference design, which launched today. When I found out during the pre-briefing arrangements that the company would be talking to me about a NAS-related product, I was quite excited. As many of you know from my past writeups, I'm a strong NAS advocate, and the NAS market is currently cost-hampered by a reliance on overly complex PC and Mac-reminiscent hardware designs, coupled with a dearth of application-specific integrated silicon offerings. Cirrus's reference design could, if it (ideally) exceeded or (minimally) matched the capabilities of products already in the market, rapidly become a serious contender. Unfortunately, my hopes were dashed once I got on the phone with Cirrus's product representative, John Delfeld.

Cirrus's approach is based on a geriatric sub-$12 (200,000) CPU, the EP9312, introduced nearly six years ago, a product which the company admitted to me also only went into production two years back (yes, that's a four-year announcement-to-production delay). It doesn't optionally support RAID, either for the multiple integrated HDDs that the Cirrus reference design does tackle via the CPU's integrated IDE controller, or for the combination of a single integrated drive paired with a USB-tethered external drive, as Broadcom's two-year-old NASoC manages to pull off. It doesn't comprehend UPnP or the follow-on DLNA protocols, nor does it support Apple's alternative Bonjour zero-configuration protocol; instead, Cirrus offers a Windows-only 'auto-detect' utility. It doesn't natively support SATA drives, either for performance, interface pincount or (soon, driven by PC trends) cost savings reasons. And although it contains an integrated 100 Mbit Ethernet MAC, it doesn't natively support Gigabit Ethernet; if you want to include this feature, you need to somehow bolt a GbE transceiver up to the processor's SRAM bus.

I'm not particularly hung up on that last point; with the exception of home and SoHo power-users like me, GbE is predominantly found in large corporate environments right now (although products such as Buffalo's Gigabit LinkStation, which has been in the market for quite some time, target the consumer market). But it's the capper on a generally underwhelming offering. Cirrus's performance benchmark claims look impressive at first glance; Delfeld indicated that the competitors were a Broadcom-based Western Digital Netcenter and two Buffalo Technologies NASs, one IDT-based and the other Freescale-based. Dig into the next level of detail, however, and you'll discover that the PC connected to the NASs had a 100 Mbit Ethernet controller inside. Was the Broadcom and IDT performance bottleneck the NAS, or the network? Based only on the data Cirrus provided, I can't tell. And while it's certainly possible for a motivated design engineer to append UPnP to the reference design via the software offerings of a company like Mediabolic or TwonkyVision, Cirrus has not assembled any formal partnership program with a UPnP provider which'd simplify your enhancement process. Cirrus points out the audio and image decoding capabilities of the EP9312; while it's conceptually interesting to think of a multimedia NAS with native audio and video outputs, I don't see much market demand for such a product. And Cirrus wasn't able to point me towards any data that proved the EP9312 was simultaneously capable of juggling network management, HDD management and multimedia decoding tasks.

Cirrus indicates that the reference design schematics and software are available free of charge. The company also promises that enhancements to resolve some of the shortcomings I've earlier noted are under development; I look forward to revisiting the reference design once this has occurred. Bigger-picture, though, my point in crafting this writeup isn't just to alert you to an underwhelming product. Compare this coverage to that coming from other journalists at other publications (specifically, folks who haven't seen my blog post and therefore 'borrowed' from it). Have they gone into this much analysis detail? Do they understand the NAS market's history, current status and near-future trends, and therefore the requirements placed on NAS building block suppliers? Or did they simply regurgitate Cirrus's press release?

I've long believed, first as an EDN reader and now as an EDN technical editor for over nine years, that the staff's combination of engineering background and targeted per-editor application focus gives EDN a notable differentiation versus other publications staffed by journalists without technical backgrounds and with extremely broad semiconductor 'beat' responsibilities. Do you agree or disagree, and in either case, why?