![]() Praising the growth of a third CPU opcode system sounds to me like wanting to bring back, say, the Scottish historical units of measurement, on the basis that we're "getting bored" with the SI/USCS duopoly. ![]() It's purely a pain point for users, for what appears to be merely a battle of egos among the world's richest megacorporations. The continued survival of x86 and other proprietary instruction sets reeks to me of the days of the "Best Viewed with $" badges. I can type some text in Japanese and email it to a friend, in 2019, and it will just work. Life was worse in every way when every company had their own special encodings for text files, and their own proprietary network protocols. I think it's great that many companies have their own implementation of open standards like Unicode, TCP/IP, and SMTP. All it does is create pointless incompatibilities that users have to deal with. I'll celebrate implementation diversity, but I'm not sure why I'd celebrate interface diversity. That option didn't really exist back in 2004. In case this takes off, there can still be some entrepreneur who cooperates with IBM or NXP or whoever has an efficient Power implementation lying around and shrink it to the smallest process node they can afford, reusing this open hardware design, and send it off to TSMC. The primary goal of a project like this is to prove the concept, and if 28nm NXP PPC chips are the sweet spot (enough performance for it to matter, unlike all those RISC-V micros efficient enough to fit in a laptop without it working double-duty as a grill, unlike Power9 CPUs cheap enough so that people are willing to risk it, unlike the otherwise fabulous TALOS workstation family) for that, that's not too bad. That's different from the PPC -> Intel transition at Apple 15 years ago: the tech is available to anybody who's willing to pay (while back then, Intel was 1-2 generations ahead of everybody else, and there was no indication that this would change). ![]() In short, for a project like this, the goal isn't the very best performance on the market, it's "the more the merrier".ĪRM isn't at 7nm, TSMC is. There are enough ARM laptops out there, this provides another venue for more vendor independence (ARM still exerts tight control over its ISA while PowerPC is a close cousin to Power9 which was opened recently). That's definitely not top-class, but good enough: it's about a factor 3 to the state of the art (compared to factor 2 between G5 and Intel Core in 2006), but in absolute terms it's so much closer. The chip they aim for is built on a 28nm process (which has some impact on voltages used and minimum conversion of power into heat). The G5 chip from 2002 was at 130nm, while the Intel design introduced in 2006 (when 4 year old G5 was the best IBM had to offer) was at 65nm. The move to Intel wasn't because the instruction set is, for some reason, superior (it isn't), it's because IBM couldn't keep up with Intel's fabrication technology. > when the whole reason Apple shifted to Intel 14 years ago was thermals? Wouldn’t ARM be better? Guys over at hackaday probably would love something like this. The NXP T2080 has good GPIO pins and connectivity, so there's a chance that this laptop will be easier to interface with electronically than other devices. This is older-tech for sure, definitely "Chromebook" level of tech, maybe a touch weaker even.ĮDIT2: I think there's something to be said about an "open" design, where the schematics are available for the community to use and extend. 32kB L1 d$ and 32kB L1 I$, 2MB L2$ for the whole chip. e6500 core, which is some form of PowerPC for sure. NXP T2080 chip, 4-core / 8-thread 28nm class chip. modern cloud-based services makes this sort of thing much easier.ĮDIT: Site finally worked for me. Honestly, its probably more important to rack-up a POWER9 Talos II and just SSH into it every once in a while. To serve as a portable development platform, much like Intel Atom can serve as a portable development platform for Intel Xeons.ĭepends on a lot of details however. If binaries were compatible with the Talos II POWER9 HEDT / Server, then there is still a good use of this laptop. Something on the scale of a Chromebook maybe? Maybe something like a NXP-chip, which would still be interesting. ![]() Since POWER9 only exists in the server / HEDT sphere, I doubt this will be a modern POWER9 chip. The link seems to be getting the YCombinator hug-of-death.
0 Comments
Leave a Reply. |