by Anton Shilov
11/01/2012 | 01:52 PM
Advanced Micro Devices is clearly very optimistic about the success of micro-servers in general and ARM-based Opteron processors for ultra-dense market segment in particular. However, financial analysts have mixed feelings about AMD’s plan to make ARM-powered products for servers.
The main concerns the analysts have is that the actual total available market of micro-servers will be significant over years from now, AMD ARM-powered Opteron microprocessors are still over a year away and given the current situation, the short-term business is more important for money-losing AMD. Moreover, the success of the company on the crowded market is not cast in stone to say at least.
The main advantage over competing chip design companies that AMD might have is its relationship with various server manufacturers. Additionally, AMD has strong capabilities in emerging heterogeneous processing trend thanks to its graphics chip division and its high-end graphics processing units, which have already found way into consumer accelerated processing units and are in route into server chips and will also help AMD to make relevant system-on-chips for tablets.
“While these new server processors are not expected to be in the market until 2014, we believe the directional change at least opens the door on a path for AMD to return to growth. […] In servers, the company’s historic relationships, architecture experience, and graphics strength should give the company a material advantage [...] In tablets, we expect the company will eventually also offer ARM-based alternatives to its current tablet platform. While tablets could be a material driver, we are less enthusiastic because of the customer concentration issues and the significant level of entrenched competition. With servers, all the players will be starting from about the same point,” wrote Cody Acree, an analyst with Williams Financial Group, in a note to clients, reports Tech Trader Daily blog.
In general, the Wall Street analysts agree that ARM architecture could have tangible advantages when it comes to ultra low-power server chips over x86 architecture.
“While the server ecosystem for ARM is still in early days, and ARM’s success in server compute is far from certain, ARM’s proven success in low power/mobile platforms could potentially translate into some compelling performance/watt cost reductions for server compute beyond what is currently available from x86 – likely the motivation behind previously announced HP/Dell development efforts based on ARM-based CPUs from Calxeda,” wrote John Pitzer, an analyst from Credit Suisse.
But also, analysts note that there will be loads of players on the market of server chips, including Applied Micro, Calxeda, Nvidia Corp. and other. Calxeda’s Lago ARMv8 chip with enhanced fabric able to connect hundreds of thousands of nodes, with quality of service features and the ability to allocate and control resources seems to be rather formidable rival for whatever AMD might offer. Nvidia’s project Boulder (which seems to be a predecessor for exascale project called Echelon) is designed to compete against current AMD Opteron and Intel Xeon in terms of performance and therefore represents even more threat for AMD’s low-power Opterons since it naturally can scale down in terms of power consumption.
Moreover, as the market will really be crowded, potentially that could lead to slow development of ARM-based micro-servers due to the battle of standards and incompatibilities.
“[The ARM server plans announcement] seems to be a mild positive for AMD, but the jury is still out, and likely will be for some time. [...] Standalone 64-bit ARM is going to be a crowded market, and it is not clear that there will be market demand for a combination product (if that is even in the roadmap),” said Daniel Berenbaum with MKM Partners.
But the main problem for AMD and its ARM server chips endeavor is not competition or uncertainties with the micro-server market in general. The main concern is whether AMD can survive before it rolls-out Opteron processors with ARM architecture. Ironically, we do know what AMD readies for launch sometimes in 2014, but we do not know what the company is getting ready to release next year.
“We do not expect to see any products coming out of this new partnership with ARM for at least 18 months, and management readily admitted that this would be a 3-5 year process. Given the state of their core business, however, we are not sure they have that much time – we forecast significant cash burn and balance sheet deterioration through much of the next five quarters,” wrote Stacy Rasgon, an analyst with Bernstein Research.