through put speeds… fuck how fast the clock is… how much data can you process through your transistor package ? thats my only care thats real world…
size doesn’t matter my g5 quad (970mp2.5ghz) on 90nm vs my 14nm ryzen 3 1200 3.1ghz
my g5 quad is better at video production, ffmpeg mp4 h264/h65 conversions
but its shit at video playback its shit at audio capture w/o a interface and digital art it suffers because of the way the chip handled the io and communicated with the pcie bus
leave everything to the cpu and its a golden experience plug something in (thats not fw) its gonna be a nightmare
“intel adding things that could be external to their shit is shit” is just good constructive criticism
Ineed AVX512 and all of the SIMD instructions have a bit of a problem where you need extended instruction sets, and use intrinsics in your code that change w/ every new iteration.
The vector engine approach of RISC-V does look like it could be quite a bit more stable over time, while working with chips of different internal capabilities.
I love true risc computers while io has suffered from my experience processor through put is amazing
I have so much more use out of my g5 quad 970mp dual power pc (power 4-5 based) then a over clocked intel i5 7th gen quad core (razor laptop) a ryzen 3 1200 (tower)
Ffmpg h264-h265 it’s faster
I wish we had more true risc consumer products where it’s screamed about being risc
Riscv is going to fuck over the way we look at computers
It’s interesting that Apple chose to go from x86 to ARM and not RISC-V. Why give ARM all that licensing money? It truly could have been a paradigm change in computing if they went with Risc-v.
Here, the thing is the you’re looking at it from a general performance perspective, rather than a “how do we establish some proprietary lock in to intel” perspective.
They have a well established ARM based ecosystem (10+ years of development of their own custom arm-based designs), and RISC-V is not real world proven.
ARM has a massive number of developers and has a 35+ year history for the kinks to be worked out. It was available on end user computers in 1987.
Apple going RISC-V would be absolutely insane. The licensing fee(s)? Not a big deal (I mean, ARM based processors are used in tiny cheap devices, so it isn’t much), and Apple’s customers aren’t super cost sensitive.
They’d be throwing away decades of base-level ARM refinement, and at least 1 decade of their own development. For what? What’s the net win for Apple here? An open platform anyone else can more easily clone? Why would they even want that?
If RISC-V was around 15 years ago, maybe they would have been on board. But that ship sailed in 2005 or so, and Apple committed to what was available when they needed a new arch. Right now the reasons to switch just aren’t there.
Ditto for Google with their hardware too btw. RISC-V is a science project at the moment, and until somebody builds something significant with it (or at the very least, until its full instruction set is sorted out and standardised - privilege levels weren’t even sorted out until mid 2019) and there’s a significant developer community with significant OS support for it, it’s a non-starter really for someone building real multi-tasking OS products at scale like Apple or Google are, when ARM is available and well proven.