Alibaba Crafts A 16-Core RISC-V Chip @ 2.5GHz

[phoronix]

Phoronix: Alibaba Crafts A 16-Core RISC-V Chip @ 2.5GHz

To date there haven't been any really compelling RISC-V processors from a performance perspective but it's looking like we could soon be crossing that threshold...

http://www.phoronix.com/scan.php?page=news_item&px=Alibaba-RISC-V-16-Core-2.5GHz

[polarathene]

Is limiting it down to 16-bit just to cut costs to scale the design elsewhere(cores/frequency) within the same budget(or other constraint)?

I wouldn't imagine you'd have much of a benchmark happening with 16-bit CPU? Not likely to run much with Linux would it?(not the kind we usually see benchmarked here anyhow), sounds more like it'd be running their code only without OS, or something more light like RTOS?

Still an interesting design, I don't think there is many <32-bit micro-controllers that have many cores or high frequencies?

[polarathene]

Oh... nevermind. Michael made a typo, it's 64-bit. That makes a lot more sense :P Makes my first post irrelevant(didn't see a strike-through option in the editor?).

[programmerjake]

phoronix maybe you could run coremark on some modern x86 processors as a comparison, in multi-threaded and single-threaded modes, maybe also at a locked clock frequency to more accurately measure coremark/mhz. I would assume coremark is small enough to run entirely in the L1 (or L2?) cache, due to not using much memory since it's a embedded benchmark.
That sounds like it would make an interesting comparison -- seeing how far risc-v or arm is behind ryzen 3000 or intel 9th gen in this admittedly limited fashion.

[dreich]

Maybe Amazon will follow.

[Guest]

The fact that this design is proprietary is a perfect example of why permissive licenses are bad. If RISC-V was copylefted, people would be going through the publicly available design right now, and could consider grouping together to order these chips from a fab.

[lilunxm12]

The fact that this design is proprietary is a perfect example of why permissive licenses are bad. If RISC-V was copylefted, people would be going through the publicly available design right now, and could consider grouping together to order these chips from a fab.

If it's only used privately within alibaba, copyleft licenses wouldn't make a difference.

[pkese]

The fact that this design is proprietary is a perfect example of why permissive licenses are bad. If RISC-V was copylefted, people would be going through the publicly available design right now, and could consider grouping together to order these chips from a fab.

Anyone willing to analyze this two sentences and write down on how many levels this is wrong (or clueless)?

[pkese]

phoronix maybe you could run coremark on some modern x86 processors as a comparison...

According to CoreMark score table, an Intel Core i5-8500 with 6 threads running at 3 GHz is scoring 57.21 CoreMarks/MHz. That would be 9.53 CoreMarks/MHz/core.

Coremarks are often used in RISC-V community to compare their architectures, but there's much more to CPU than just the core and I'm not sure how well does the Coremark reflect the other aspects (quality of branch predictors, cache & memory latencies, etc.)

[programmerjake]

According to CoreMark score table, an Intel Core i5-8500 with 6 threads running at 3 GHz is scoring 57.21 CoreMarks/MHz. That would be 9.53 CoreMarks/MHz/core.

Coremarks are often used in RISC-V community to compare their architectures, but there's much more to CPU than just the core and I'm not sure how well does the Coremark reflect the other aspects (quality of branch predictors, cache & memory latencies, etc.)

Coremark is not a very good benchmark for the reasons you mentioned, however, I still think it would make for an interesting comparison. That particular benchmark result (assuming it's the same one I found) doesn't state if the processor had the clock frequency fixed or if turbo was enabled (which would invalidate the coremark/mhz result), it just listed the base frequency from the spec. Another result in the same source I found listed the Ryzen 1700X tested using all 16 threads at once and listed the frequency as the rated max boost clock, which does not seem plausible unless they were overclocking. If Michael does test it, we can be pretty sure that he will correctly measure the performance, allowing determining a much more reliable coremark/mhz rating.