Qualcomm server CPUs battle Intel for IoT data center growth

Qualcomm, already a leader in the mobile processor space of phones and tablets, made a splash late last week with the announcement and shipment of its first server-grade discrete processor targeted at disrupting the cloud enterprise ecosystem. With more than four years of development behind it, the Qualcomm Centriq 2400 (pronounced “centric”) family is a high-core count processor for cloud applications that have seen and will see drastic expansion and growth. SVP and GM of Qualcomm Datacenter Technologies Anand Chandrasekher took stage in San Jose to showcase the capability and partners for the Centriq 2400.

As consumer demand for cloud-based applications and data increase, which is forecast to be unstoppable for the next 5-10 years, Qualcomm has a sizeable market it can address. With more than 75 billion attached devices expected by 2025 in the IoT field, the need for more advanced server architectures and systems will increase along with it.

A key to any enterprise launch is the acceptance of big players that make up the majority of the market. Qualcomm brought several of them on stage at its Silicon Valley launch last week. Microsoft discussed the processor capabilities for its Azure platform, Alibaba talked about its recognition of Qualcomm’s long-term interest in the field, and Cloudflare’s CEO called potential deployment of Centriq a “no brainer.”

The technical advantages that Qualcomm offers surround its power efficiency and the ability to offer similar per-thread performance to the flagship Intel Xeon Scalable processors. Built on Samsung’s 10nm process technology, arguably the most advanced in the industry, Centriq 2400 is based on the Arm architecture. This is the same base design as the processors that power the world’s top smart phones, but altered and designed specifically for server-class workloads.

For Arm, this marks the most important of its server partners coming to market with data center-ready solutions. The move from Intel (and AMD) compatible x86 platforms to ARMv8-based CPUs will require software rewrites and platform adaptation. The enterprise markets are more willing to integrate these hard shifts in system design (compared to general consumers) since they can control the software in its entirety.

Centriq 2400 will combine up to 48 cores in a 120-watt processor package, compared to over 200 watts for the top Intel Xeon processors, and at less than half the cost. Though performance metrics were minimal at this stage, early results show the Qualcomm option matching Intel in SPECint, an industry standard test for processor potential. Combining that with the ability to run at lower power, the Centriq 2400 family can offer decreased total costs to power-hungry server farms.

Intel has recognized the threat and started building and promoting processors of its own that combine smaller cores in more dense packages. But its emphasis has remained on the higher end, leaving the door open for Qualcomm to have an impact on the market in 2018.

Intel is the clear market share leader in the data center, and Qualcomm will have to offer compelling reasons for customers to switch. AMD is facing the same dilemma – incumbents have every advantage. Change takes time and financial investment, so competitors have to offer compelling advantages in power, performance, cost, or all the above, to make headway. Initial results for Qualcomm’s Centriq 2400 prove that it understands the task at hand, and has developed a technology and strategy appropriate to take on Intel.

Though Qualcomm’s announcement last week promoted the initial shipments for revenue this quarter, it has not yet announced a full-scale deployment partner. Instead, interested parties like Microsoft and Alibaba are in evaluation and bring-up stages. This puts mass adoption and revenue at least 6 months out and as far as 18 months depending on data center refresh cycles.