editor's blog
Subscribe Now

Netlogic Integrates Families

We looked at many-core processors recently, and one of the big issues with scaling up the processor count is memory access: if all of those cores need access to the same memory, then that bandwidth becomes the bottleneck. Which makes SMP with many cores very difficult without shared distributed memory structures.

Netlogic has just announced their XLP II family, following on the heels of their XLP processors that have been around for a while. XLP devices go up to 32 CPUs; XLP II goes up to 80 per device, clusterable to 640. And they explicitly claim SMP capability.

So I followed up with them to see how they manage to talk to memory fast enough to feed so many cores. And it bears saying that, assuming each core manages a different memory, it’s “only” 80 cores that have to vie for attention by a single memory manager. Their response was that they have plenty of headroom on their current 32-CPU devices, and the memory manager runs much faster on the new devices, so they believe that memory access will not get in the way.

As to running all 640 together, they have an inter-chip coherency interface to keep all processors and caches in sync. They have a tri-level caching system, although details aren’t available yet.

They are also claiming a “third-generation” inter-process messaging system to speed up the conversations that the CPUs will need to have with each other.

Above and beyond just the many-core aspects, they are also integrating a host of communication-related functions alongside, including their NETL7 knowledge-based processor, which we discussed recently.

Most technical details haven’t been made public, but there is more info on their press release.

Leave a Reply

featured blogs
Nov 14, 2019
In addition to playing retro games, THEC64 allows you to write your own programs in C64 or VIC 20 BASIC....
Nov 14, 2019
The Cadence Academic Network hosted an Academic Speaker Series event, in collaboration with the Shanghai Site Technical Talk series, in Cadence Shanghai Office. The talk attracted more than 150... [[ Click on the title to access the full blog on the Cadence Community site. ]...
Nov 14, 2019
Scientists, researchers, and data analysts from academia, industry and government agencies will be center stage at SC19 next week in Denver. SC19 is the International Conference for High Performance Computing, Networking, Storage, and Analysis. Next-generation high-performanc...
Nov 13, 2019
By Elven Huang – Mentor, A Siemens Business SRAM debugging at advanced nodes is challenging. With pattern matching and similarity checking, Calibre tools enable designers to more quickly and precisely locate SRAM modification errors and determine the correct fix. Static...
Nov 8, 2019
[From the last episode: we looked at the differences between computing at the edge and in the cloud.] We'€™ve looked at the differences between MCUs and SoCs, but the one major thing that they have in common is that they have a CPU. Now'€¦ anyone can define their own CPU ...