IBM, Micron Technology Planning to Build 3D ‘Memory Cube’
IBM, in a partnership with Micron Technology, plans to develop a very special chip it calls the “Hybrid Memory Cube,” or HMC. IBM says this 3D chip will deliver memory speeds 15 times faster than what’s already on the market. How does it work and more importantly, should you care? Read on …
The technology of a 3D memory cube includes “through-silicon vias,” or, more simply, vertical aqueducts inside the chip. The first HMC will be built in a fairly standard 32nm core (for reference, Apple’s A5 CPU sits in at 45nm), but thanks to the “3D” layers of the chip and the vias, it can achieve super-fast processing in a smaller footprint and with less energy consumption:
HMC prototypes, for example, clock in with bandwidth of 128 gigabytes per second (GB/s). By comparison, current state-of-the-art devices deliver 12.8 GB/s. HMC also requires 70 percent less energy to transfer data while offering a small form factor — just 10 percent of the footprint of conventional memory. HMC will enable a new generation of performance in applications ranging from large-scale networking and high-performance computing, to industrial automation and, eventually, consumer products.
Before anyone gets too excited, remember, this is all fresh, new and untested technology, and most definitely not cheap. It’s also specifically targeting the RAM space, not CPUs. But 3D-memory chip technology definitely has the potential to build dense and fast CPUs that do more than simply crunch numbers. We already know that from the various system-on-a-chip CPUs we use every day in our smartphones.
If you’re wondering if Intel recently did something similar, you’re right. Intel actually worked with Micron Technology developing the actual blueprints for the Hybrid Memory Cube, but IBM actually will be fabricating the physical chip with Micron Technology.
What does this all mean for the channel? Like IBM said, high performance computing is likely where these chips are headed. For VARs, SMBs and enterprises alike, that means the server space is about to get a lot more interesting. As virtualization, cloud computing and mobile services become more common and more demanding, the fabrication of these chips will likely accelerate and with it, so too will the complexity of cloud and virtualization technology. Perhaps only 10 years from now, we’ll have 3D RAM and CPUs, powering entire clouds with one tiny blade server. How’s that for data center consolidation?
Here’s a weekend reminder: We’re living in the future. Let’s get excited.