What went wrong?

Intel BTFO them in laptops

>>59103929
RISCucks can't compete

File: silly_chart.png (34KB, 774x606px) Image search: [Google]

34KB, 774x606px

>>59103929
The economies of scale.

>>59104363
x86 killed Power even before, it was just that Apple took its time to dump it.

>>59103929
PowerPC is shit compared to Intel x86. Even RISC-V shits all over it.

What is unique about PowerPC from an architectural standpoint? I don't just mean whether is has good software support

>>59103929
Simply too power hungry. Even videogames gave up on it, PS3 and X360 were housefires.

>>59104597

Apple is and always has been an utterly insignificant % of the market to the point of being meaningless. If this were phones you'd have to consider them, but nobody ever put PowerPC in a phone.

>>59105183
RISCs are easy to program to and fast as fuck, but PowerPC was limited to specific stuff, videogames when were used only for videogames could get lots of juice out of it, but nowadays when it is needed to be multitasking diferent kinds of applications powerPC isnt that good.

>>59105266
>when it is needed to be multitasking diferent kinds of applications powerPC isnt that good.
why not? it supports openmp and shit

>>59105266
>RISCs are easy to program
Only if you are a compiler. CISC are usually more friendly to humans.

>>59104822
RISCV is paper
>>59105213
You're basing this off G5 which wasn't even bad unless you had a dual CPU dual core PowerMac. The Nvidia and AMD GPUs most certainly used more power in those consoles. PPC was extremely power efficient compared to x86 as a whole.

>>59105374
I'd also like to add, AltiVec shat all over Intel's SSE. It wasn't until AVX2 that Intel had something truly better.

File: g5-spec2000.png (33KB, 611x638px) Image search: [Google]

33KB, 611x638px

>>59103929
Lackluster as a desktop part. Workstations were shit.
>>59104597
Not really, just on the desktop.

Developing for 800 threads is way fucking harder than developing for two

>>59103929
>What went wrong?
Apple wanted fewer than 100 million PPC CPUs a year so IBM didn't think it was worth their while to burn PPCs at the same process/speed as their POWER CPUs.
They didn't improve the heat envelope either.
So teh Steve went Intel.

>>59103929
>What went wrong?

It was too risky.

>>59104363
>ARM
>killing anything
my money is we keep x86 forever and ever, until we get some quantum architecture
and even that will run on some extended quantum x86 instruction set

>>59106216
>So teh Steve went Intel.
He did, but not due to IBM not improving their CPU's for laptops
PA Semi was doing it, incredibly well, with their sole consumer being Apple
Then jobs went to Intel just when PA Semi was releasing their power efficient G5, literally bankrupting the company and buying them
This way they got some of the most talented people for their SoC's, that still dominate the market

>>59106324
ARM killed everything else on the smartphone market, including x86

File: P3lYSlG.jpg (92KB, 640x480px) Image search: [Google]

92KB, 640x480px

>>59105247
>nobody ever put PPC in a phone

>>59106579
Smartphones aren't desktops, and ARM is the x86 of the smartphone and small mobile device realm. Neither will kill the other in their target markets because there is no reason for either to ditch their established markets. Architectural migration are nothing but a waste of time in 2017, the shit show of the '90s and 2000s showed us that.

>>59105316
Well, lets assume this is true for the sake of the argument. maybe some years ago could be relevant, most programmers today dodnt know shit about compilers, let alone assembly languague or the RISC vs CISC differences

RISC is more power efficient than CISC because you need less space for the CPU because all those "friendly" CISC instructions need extra hardware to decode and execute them. RISC instruction set is "limited" but also leaner.

I'm not familiar with modern CPU architectures but ARM won the mobile race because, as far as I know their architecture doesnt have as much cruft as x86 so they are way less power hungry.

Maybe if compaq didnt kill the Alpha, or if the MIPS didnt followed the way of the dodo and Apple didnt ditch the PPC they could be rolling in dough of mobile CPU market.

RISC is not bad, part of its "demise" could be attributed to Intel's monopoly and business practices.

>>59106880
>MIPS didnt followed the way of the dodo
Friendly reminder that the fastest supercomputer in the world is running MIPS thanks to the chinks
Along with a plethora of trashy routers and other embedded crap

>>59105566
That was against P4. Athlon 64 was what made G5 really obsolete.

>>59106216
Mainly G5s couldn't go into laptops and G4s while more power efficient were obsolete.

>>59106565
That is some alternative history. PWRficient from PA was not ready by the time Apple switched, and it would not be performance-competetive with Core 2 Duo at all, basically it would bring dual-core to macbooks but that was it. It was also not a laptop CPU, it was embedded chip, for network intrasftructure and such cases. There was never even any possibility of them being used in Apple, who needed large volumes and also a proven supplier that could be trusted for delivering - which a startup that has yet to prove its first chip was clearly not.

>>59106324
ARM supports x86 emulation now. Microsoft demonstrated full Windows Photoshop running on Windows 10 on ARM because Windows 10 on ARM now supports win32 emulation. It could be easier to ditch x86 than you think because ARM tablets/laptops/maybe even desktops can still run win32 applications.

My bet is that eventually the only independent computing device most people will carry around will be their phone, which will be powered by ARM. People will be able to dock their phone into a laptop or a desktop dock or they'll be able to stream content to their TV and stuff.

>>59106880
>doesnt have as much cruft as x86 so they are way less power hungry
It has enough of cruft, it's like three ISAs in once by now, with some having alternative ways to do the same shit with slightly different behavior and some instructions being rather weird. ARM isn't very RISCy actually.

The reason it has caught on is becaue the ISA and cores are licensable as IP. Hence many vendors can use it. Also the business/licensing/development model seems to allow for quick development of SoCs and there are other advantages that don't boil down to the ISA directly.

File: BYTE-1993-01_0152.jpg (421KB, 1156x1626px) Image search: [Google]

421KB, 1156x1626px

>>59106880
>RISC is not bad, part of its "demise" could be attributed to Intel's monopoly and business practices.
Intel themselves maintained their own pure RISC chips for a time, the reality lies in multiple factors. RISC servers were brutalized by the dotcom bubble bursting that sent Sun and its competitors who rode it straight to the top right back into the shitter, while the inevitability of commodity hardware becoming powerful enough to do 95% of tasks that once required a workstation blew RISC off the desktop too.

Not to mention the mismanagement, between Itanium cock sucking and shitty, outdated marketing strategies, it's a wonder it wasn't even more of a bloodbath than it already was. And in the end, the hardware really wasn't all that great. You got a 20% faster chip at most at 200% of the price with 5% of the software base of a decent NT workstation made by a shitty vendor that will try its damnedest to make sure you're stuck with them as long as possible. More often it wasn't even this clear-cut, most mainstay RISC platforms towards the end were severely lackluster, and not in a way that can be waved away with the nebulous point of "R&D costs" when even the might of companies like IBM and Sun could barely produce a competitive chip in the desktop space. 2000s era Sun was especially a pathetic husk of its former self on this front.

In short, quit it with simplistic the 'it was just another victim of Intel' shit.

>>59106903
SunWay is home-grown and has absolutely nothing to do with MIPS, it's closer to the Alpha if anything, even then it's merely "influenced".

>>59107431
I don't understand how this is relevant at all, a newer A64 or equivalent 64-bit P4 would just make it even more apparent.