Loongson’s effort is the 3A6000 processor, which uses its own LoongArch CPU instruction set that has characteristics of both the MIPS and RISC-V architectures. The chip has four physical cores and can run eight hardware threads, includes a pair of DDR4 controllers, and runs at between 2.0GHz and 2.5GHz, consuming 38 watts when running at the latter speed.
Loongson has cited benchmark results that it claims place the 3A6000 on par with a comparable product from Intel’s 10th-generation Core family, circa 2020.
The chip shop has not revealed which foundry made the processor, but has revealed it’s built on a 12/14nm process.
China still have ways to go, but this show that they seems to make at least some progress. I wonder how long it will be until western benchmarker start doing test on it, would be interesting to see!
however it seems to be heavily constrain by only having 4 core and relatively low clock speed.
So their attitude towards serious cpu development seems entirely nailing the uArch rather than rapid deployment towards mass enterprise adoption. Sure some super sensitive state and military machines that don’t need to be fast will likely use it, as Russians use Elbrus… but it seems money is last of their concerns.
Make sense as they likely have piles somewhat recent of Epyc and Xeon chips for their data centres till that tap goes off.
they are all basically skylake, just with some more cache(but not the type of increase that amd’s 3d cache has for example) and out of the box over clocking.
Comparing a 6700K to a 10100 you’ll have at best a 10% increase, and most of the time they are roughly even. If you can OC your 6700K 10% which isn’t difficult you’ll likely beat it in every situation.
Are we speaking IPC or IPS here? Since IPC (also I/C → Instructions per (Clock-) Circle) doesn’t change depending on the clock-speed. Meanwhile IPS (also I/S → Instructions per Second) increases the higher the clocks (basically IPC×time).
The impressive part is that it has IPC on par with Raptor Cove, what’s holding it to Skylake performance is the low max clocks.
It literally has higher IPC than Zen4, which is an impressive showing for Loongson. I’m guessing it’s a very wide design that has to stay at low clocks because it is hampered by the relatively ancient 14nm process it’s manufactured on.
Nothing is impressive until there’s 3rd party testing. Wouldn’t be the first time a company made big claims only for the arch to shit the bed in other workloads.
The impressive part is that it has IPC on par with Raptor Cove,
But isn’t that in only one single benchmark that they’ve shown? I’d wait for a range of benchmarks before judging it, power efficiency could also be a huge factor too.
China still have ways to go, but this show that they seems to make at least some progress. I wonder how long it will be until western benchmarker start doing test on it, would be interesting to see!
https://www.tomshardware.com/news/loongson-launches-3a6000-cpu-matches-14600k-ipc
Tomshardware mentioned that the IPC match Intel 14600k, however it seems to be heavily constrain by only having 4 core and relatively low clock speed.
All the interesting Chinese chips are by other companies. Huawei, Alibaba, etc. all have CPUs that blow this out of the water.
So their attitude towards serious cpu development seems entirely nailing the uArch rather than rapid deployment towards mass enterprise adoption. Sure some super sensitive state and military machines that don’t need to be fast will likely use it, as Russians use Elbrus… but it seems money is last of their concerns.
Make sense as they likely have piles somewhat recent of Epyc and Xeon chips for their data centres till that tap goes off.
look at the SRAM specs on the loongson.
there is no way it is getting those performance figures unless they’ve created the fastest SRAM on the planet.
That’s a bit of a loaded statement…
4c/8t Comet Lake? So that’s also circa 2015 Skylake.
So there were no IPC-increases from Sky Lake to Comet Lake?
they are all basically skylake, just with some more cache(but not the type of increase that amd’s 3d cache has for example) and out of the box over clocking.
Comparing a 6700K to a 10100 you’ll have at best a 10% increase, and most of the time they are roughly even. If you can OC your 6700K 10% which isn’t difficult you’ll likely beat it in every situation.
Are we speaking IPC or IPS here? Since IPC (also I/C → Instructions per (Clock-) Circle) doesn’t change depending on the clock-speed. Meanwhile IPS (also I/S → Instructions per Second) increases the higher the clocks (basically IPC×time).
The impressive part is that it has IPC on par with Raptor Cove, what’s holding it to Skylake performance is the low max clocks.
It literally has higher IPC than Zen4, which is an impressive showing for Loongson. I’m guessing it’s a very wide design that has to stay at low clocks because it is hampered by the relatively ancient 14nm process it’s manufactured on.
You’ve gotta give credit where credit is due.
Is the IPC normalized between RISC and CISC? Otherwise comparing it is pointless.
Nothing is impressive until there’s 3rd party testing. Wouldn’t be the first time a company made big claims only for the arch to shit the bed in other workloads.
But isn’t that in only one single benchmark that they’ve shown? I’d wait for a range of benchmarks before judging it, power efficiency could also be a huge factor too.