Remix.run Logo
▲ kragen 3 hours ago

They're claiming better performance per watt, not better absolute performance, and out-of-order does worse on that metric than the superscalar in-order design they say they beat by 4.59×. That's why, for example, PULPino is in-order.

A third interesting metric is performance per transistor or per area, and generally the best designs there are the long-forgotten MISC designs like the MuP21 and the GA144.

▲adrian_b 27 minutes ago | parent | next [-]

The performance per watt of very slow CPUs is always very good, so that alone is not impressive unless much more data is provided to prove that is really better than what is typical.

On the other hand, the performance per area of the "traditional" CPU that is used for comparison, which I assume that it is some kind of RISC-V, is very poor, so that the performance per area of the new CPU is likely to be also poor, even if much better than that of the strawman.

Their "traditional" CPU has an area that is slightly bigger than that of Cortex-A720 made with a CMOS process almost twice as dense.

So if they were made with the same process, the "traditional" CPU and the new CPU would have an area somewhat bigger than a half of a Cortex-A720.

But Cortex-A720 is an out-of-order 4-wide superscalar core with advanced SIMD and scalable vector extensions, with a good branch predictor and with many other enhancements, for which the area ratio in the same process is expected to be much greater than 2 in comparison with the CPUs evaluated in TFA, while its performance must be several times greater than that of a 2-wide in-order CPU core.

So both the "traditional" CPU and the Frostbite CPU appear to have a much bigger area than expected for their performance.

Moreover, the performance of in-order CPUs is very sensitive to the benchmark selection and to the quality of the compiler optimizations.

So the performance obtained for the benchmarks selected by them, like FFT or LU decomposition, is expected to be much better than for benchmarks that would execute an application not dominated by such regular loops, but having lots of hard to predict branches, e.g. a compiler benchmark.

They say that the ISA will be published. At that time it will be easier to appreciate whether this CPU architecture is really innovative. For now, they have done a lot of work, by fully implementing the CPU core, but the data presented is not sufficiently convincing to demonstrate that it is indeed an architecture that will be able to compete with SOTA CPU cores, e.g. with those from the Arm company.

▲kragen 20 minutes ago | parent [-]

> The performance per watt of very slow CPUs is always very good

Generally this has not been my experience. Typically both very slow CPUs like the ATMega328 and high-performance CPUs like an Intel i7 are in the neighborhood of 1 nanojoule per instruction. (For operations more than 8 bits wide, the AVR is significantly worse.) Low-power CPUs like the STM32L line are commonly around 0.2nJ/insn. Ambiq's subthreshold logic gets them closer to 0.05nJ/insn.

I agree with the rest of your comment.

▲csande17 2 hours ago | parent | prev [-]

> They're claiming better performance per watt, not better absolute performance

I shouldn't have to read this far into the tea leaves to figure out what performance claims a company is making, but these facts combined sound like they're claiming much better absolute performance, too?

> Each Frostbite quad (a two-by-two grid of tiles) paired with four L3 nodes is 0.792 mm².

> This TCPU [the one they benchmarked against] is 0.895 mm² at the same process.

> Across all the workloads, we average an 8.77x performance-per-area

It got 8.77x performance per area, and was 88% of the area, so that's 7.76x actual performance, right?

Not that "a parallel processor outperformed a single-core processor on parallelizable tasks" is the most earth-shattering relevation in the world, especially when they won't say which single-core processor "for legal reasons".

▲kragen 23 minutes ago | parent [-]

The second paragraph of the linked page contains their first performance claim. It reads:

> On a set of 25 initially chosen workloads, including 15 held out until the final evaluation, we achieved a 4.59x performance-per-watt multiple when compared to a modern, silicon-proven superscalar (dual-issue, in-order) CPU core.

They do start claiming performance per area in the following paragraph, but that's not an interesting claim for a couple of different reasons:

- As I said above, MISC designs get dramatically better performance per area than any mainstream CPU.

- We're already in the dark-silicon era, where even the relatively low performance-per-area designs used in laptops and cellphones must thermally throttle on a regular basis to avoid burning themselves up, and contain many specialized accelerators used only a small fraction of the time (AES-NI, AVX-512, saturating 8-bit arithmetic, popcount, etc.) instead of additional small general-purpose cores that could run all the time.

Their claims would be a lot more interesting if they provided some kind of absolute numbers to compare against.