Isn’t it more the case of software development needing to catch up so hardware acceleration and optimizations can utilize the processor to its fullest capacity?
Not really. That’s sort of the whole point of benchmarks: that they do not need to account for optimizations. They show raw power. Unfortunately they chose to highlight some really old benchmarks. Spec cpu 2006 is from 2006. Unixbenchmark is even older. These tools don’t really portray the type of tasks that are going to be needed to support modern use cases. There’s also SpecCPU 2017 benchmarks, but those aren’t as close, about 10% slower than the slightly less underclocked i3-10100 at 3.6 GHz rather than at 2.5 GHz. The 10100 normally runs at 4.3 GHz out of the box though and is significantly faster.
All in all, really poor benchmarking, not really showing any usable information aside from the generation bump from the 3A5000.
I really want to see Loongson and Zhaoxin do some true market-crushing magic at least enough to put the fear of God Intel and AMD
It still seems RISC-V and RISC-like CPUs are making a lot of progress, though. I still find 3A6000 to be quite impressive compared to a few years ago, and I expect exponential gains to be made in performance soon.
Take my words with a grain of salt as I do not have a lot of experience with low-level software and hardware, but from my understanding, traditional software will run poorly on these new architectures as they are designed to run on CISC instruction sets, and a new ecosystem of software needs to be developed to work with the new instruction sets. I am assuming the binary translations required to run current software is what hurts the performance of these machines, and a lot of performance could be squeezed out if software was built according to 3A6000’s architecture from the ground up.
A lot can be done with 2-3 GHz, and I find this to be an impressive speed considering the current stage of reduced instruction set architectures. Having the performance of an HP EliteDesk/ProDesk Mini and a similar efficient power usage is a great milestone achieved, imo.
Isn’t it more the case of software development needing to catch up so hardware acceleration and optimizations can utilize the processor to its fullest capacity?
https://yewtu.be/watch?v=fB9vItIDhnw
https://www.laitimes.com/en/article/63eyz_6js04.html
Not really. That’s sort of the whole point of benchmarks: that they do not need to account for optimizations. They show raw power. Unfortunately they chose to highlight some really old benchmarks. Spec cpu 2006 is from 2006. Unixbenchmark is even older. These tools don’t really portray the type of tasks that are going to be needed to support modern use cases. There’s also SpecCPU 2017 benchmarks, but those aren’t as close, about 10% slower than the slightly less underclocked i3-10100 at 3.6 GHz rather than at 2.5 GHz. The 10100 normally runs at 4.3 GHz out of the box though and is significantly faster.
All in all, really poor benchmarking, not really showing any usable information aside from the generation bump from the 3A5000.
I really want to see Loongson and Zhaoxin do some true market-crushing magic at least enough to put the fear of God Intel and AMD
It still seems RISC-V and RISC-like CPUs are making a lot of progress, though. I still find 3A6000 to be quite impressive compared to a few years ago, and I expect exponential gains to be made in performance soon.
Take my words with a grain of salt as I do not have a lot of experience with low-level software and hardware, but from my understanding, traditional software will run poorly on these new architectures as they are designed to run on CISC instruction sets, and a new ecosystem of software needs to be developed to work with the new instruction sets. I am assuming the binary translations required to run current software is what hurts the performance of these machines, and a lot of performance could be squeezed out if software was built according to 3A6000’s architecture from the ground up.
A lot can be done with 2-3 GHz, and I find this to be an impressive speed considering the current stage of reduced instruction set architectures. Having the performance of an HP EliteDesk/ProDesk Mini and a similar efficient power usage is a great milestone achieved, imo.
If a RISC-like computer can do this, I’m happy: https://yewtu.be/watch?v=amVP96OYfUg