MIPS uses the term physical AI for AI that moves out of the data center to the broader edge.
“At at MIPS, we we call this physical AI, so as as AI sort of leaves the data center and and moves to the the broader edge.”
Person
1 episodes · 20 quoted claims
MIPS uses the term physical AI for AI that moves out of the data center to the broader edge.
“At at MIPS, we we call this physical AI, so as as AI sort of leaves the data center and and moves to the the broader edge.”
Drew Barbier of MIPS says his background is in CPUs and that he has worked on them for a long time.
“I'm a hardware guy, my my background is is CPUs, been in CPUs for a long time.”
MIPS has a tool called Atlas Explorer that gives detailed performance insights on large workloads and data sets.
“We have tools like Atlas Explorer that enable us to glean very detailed performance insights on larger workloads, larger data sets.”
Drew Barbier says MIPS often analyzes workloads before beginning to write any RTL, the hardware description code for a chip.
“a lot of times this is before we even begin designing or writing any RTL at all.”
The MIPS M8500 is a high-performance 32-bit microcontroller IP specialized for motor control applications.
“So we have a product called the M8500. It's a high performance 32-bit microcontroller IP that is specialized for motor control applications.”
The M8500 core includes real-time multi-threading, which MIPS says significantly lowers context-switching latency.
“this core has something that we call real-time multi-threading, which helps lower latency for context switching quite significantly.”
MIPS added dedicated ports to the M8500 to move data into and out of the core as fast as possible.
“we added specific ports to to move data into the core and out of the core as fast as possible”
MIPS added compute units for trigonometric functions to the M8500 because motor control loops rely on them.
“specifically, uh, trigonometric functions, and from there, we we were able to to add those to those type of compute units to the core.”
MIPS reports that the M8500 delivers 3 to 5 times uplift on some field-oriented control algorithms.
“in the case of M8500, we see uplifts from from 3 to 5X on some of these field-oriented control algorithms.”
MIPS says it is a big supporter of and contributor to the RISC-V open standard.
“We we're big supporters of of RISC-V, big contributors to RISC-V.”
Drew Barbier says RISC-V provides a foundation and ecosystem while letting MIPS add specialized functions on top.
“RISC-V is great because it gives us that foundation, that ecosystem to to start from, but still enables us to go implement more specialized functions on top of that”
Drew Barbier says the M8500 is about 20% better in raw instructions per cycle, but 3 to 5 times better on a motor control customer's own workload.
“not only are we gonna be, you know, 20% better in terms of raw IPC, but on your workload that you care about, it's actually 3 to 5X better, right?”
Drew Barbier says the M8500 combines the functions of a microcontroller and an accelerator, so customers need only one block.
“customers aren't having to put down a microcontroller and an accelerator. All of a sudden, those functions, both of those functions, get consumed by by the M8500.”
Drew Barbier says consolidating functions on the M8500 reduces data movement and lowers power consumption.
“the amount of data movement across the MCU is much less because everything's done on the M8500. And that leads to better power conversations, lower power consumption”
Drew Barbier says there is not yet much customer silicon using the M8500, so results are measured mainly by customer adoption and it is early days.
“there's not a lot of customer silicon that's out there that we can talk to in in terms of the M8500.”
Drew Barbier says MIPS is famous for multi-threading.
“So MIPS is famous for for multi-threading.”
Drew Barbier says MIPS has traditionally used multi-threading and has had success in networking and data movement applications.
“MIPS has traditionally been multi-threaded and has a lot of success in networking and and data movement applications and things like that.”
Drew Barbier says MIPS offers workload analysis tooling both internally and to its customers.
“we have a lot of tooling and capabilities both inside of MIPS as well as that we make available to to customers”
Drew Barbier says MIPS uses Atlas Explorer, virtual platforms and digital twins to engage customers earlier in the process.
“tools that we have, like Atlas Explorer and a lot of these virtual platform, um, tools, digital twins, those kinds of things”
Drew Barbier says mistakes in these systems are expensive for everyone involved when a promised system performance is not delivered.
“it's expensive to to make mistakes here, right? Um, if you, uh, don't deliver on what you're promising on a system, I, it's very expensive for for everybody involved.”