Recent content by juanrga

Entropy is a physical property. The 'lie' (really a misconception) is to associate entropy with disorder https://aapt.scitation.org/doi/10.1119/1.5126822

Both matter and matter+radiation verify the second law.

From the links: A couple of nonsensical statements from people who don't know a bit (PUN intended) about thermodynamics. This new 'violation' of the second law will end up in the trash like the thousands of previous reports of violation of this venerable law.

The engine using more cores does not always translate into higher performance. If the main thread is bottlenecked, the rest of the cores will be underutilized.

The reason why engineers are moving to heterogeneous designs is because a single kind of core is not optimal for all the workloads and situations. No single core design can be optimal because the microarchitecture requirements for each situation are antagonistic: high frequency vs low, deep...

Because TDP for Ampere is the peak power and for AMD it is the average power. So the Ampere chip is a ~200W chip if you want to compare it with AMD EPYC.

It will down costs, but x86-64 cannot really compete in the long run: Where Ampere and the Altra definitely is beating AMD in is TCO, or total cost of ownership. Taking the flagship models as comparison points – the Q80-33 costs only $4050 which generally matching the performance of AMD’s EPYC...

https://fuse.wikichip.org/news/4795/arm-launches-new-neoverse-n2-and-v1-server-cpus-1-4x-1-5x-ipc-sve-and-armv9/ New cores with better microarchitectures, new instructions (including SVE), and adapted to new process nodes. The N2 is an evolution of the N1 used by Amazon and others. The V1 is a...

This is a rewrite of the history. Check the talks, the roadmaps, and the official plans when Rory Read was CEO. Not only the x86 core was unnamed then, but the ARM core was officially called K12, because it was the successor to K5, K6, K7, K8, K9, and K10.

You are entirely right. The x86 tax has a fixed part and a variable part. The larger and more complex the design, the less weights the fixed portion of the tax and the ARM-x86 efficiency gap is reduced, but the gap is still significant at the ~ 200W level.

We have no x86 legacy, like 32-bit support and things like that,” said Hegde. “We are able to optimize our code, and our core area is significantly smaller [as a result]. Just to give you an idea, in the previous generation, if you look at ThunderX2, compared to AMD or Skylake, for the same...

Microsoft and Sony both wanted ARM for their consoles, but used x86 because ARM was only 32bit then. Carmack also publicly liked the idea of ARM-based consoles. I believe developers are open to the nice idea of using a single modern architecture (ARM) for the entire gaming ecosystem. x86 has...

ARM isn't low power. ARM is more efficient. So you can use that superior efficiency for providing (i) the same performance within smaller power envelope, or (ii) higher performance for the same power envelope, or (iii) a combination of both. There are 250W ARM SKUs, and there is no...

The full thing is rendered in the post by Red Falcon. There are no slots, because slots are slow. It is based in a future microarchitecture to be announced by ARM. But Nvidia has already announced over 300 points on SPECrate2017_int_base. The Swiss National Supercomputing Center and the Los...

Extra details and the historical perspective behind Nvidia ARM/HPC aspirations https://www.nextplatform.com/2021/04/12/nvidia-enters-the-arms-race-with-homegrown-grace-cpus/