Redpill me on the PowerPc architecture, was it better than x86...

Redpill me on the PowerPc architecture, was it better than x86? what was the best generation according to their time/competition? worst? was the G5 that bad?

Other urls found in this thread:

yarchive.net/comp/linux/x86.html
youtube.com/watch?v=VLTh4uVJduI
nicovideo.jp/watch/sm18914987
twitter.com/NSFWRedditGif

...

>was it better than x86?
It was until the 2000's

>what was the best generation
G5

They might have been slightly better at FP math which might come up when using photoshop to apply a complex filter to an image or when you're rendering or encoding video. I think x86 was better in general though and x86-64 improved things quite a bit further by expanding the number of named registers and adding new instructions to improve performance.

They are about equal for the longest time. With the 68000 Motorola alot of the Computers with them had alot of custom chips for graphics and sound etc. I am talking about the Amiga which was way ahead of PC for years. Macintosh was also years ahead of PC too......in fact, Macintosh still is in all honesty(Although Apple switched to Intel).

I think it is only Nintendo who still use PowerPC.

No, it shit.

Apple relied on not having to worry about hundred different hardware configuration, and just make it work for a few products.

In essence they installed Gentoo

Depends on what you did, like other "pure" RISC architectures POWER/PPC were strong in floating-point but so-so in integer performance which is most significant for desktop applications. They performed to customer expectations and did a decent job of it, but they weren't the king.

Linus Torvalds also has no love for PPC32:
yarchive.net/comp/linux/x86.html

I hate it when people post this shitty article even if it's just bait to get a rise out of people who actually know what they're talking about, it just shits on marketing for doing what marketing does best and then picks apart all their methodology like they're some kind of special snowflake among vendors. Of course they'll use a prototype system when the actual consumer product isn't even out yet, of course they'll use newer versions of their own compilers and operating systems. Benchmarks are meant to show off hardware at its unrealistic best, and you might as well shit on IBM, SGI, Sun, and pretty much every other non-x86 vendor at that point.

The 68000 line was also generally faster on its own, it was a performance chip after all.

>Apple relied on not having to worry about hundred different hardware configuration
Yeah they did. Did we forget about the clone program already? Or better yet, the massive peripheral/upgrade market the platform maintained well into the 2000s? There were things before the iPhone and dongleBook, you know.

603/750(G3) were used in dickloads of embedded hardware from networking, Sega's Model 3, Mars rovers, and iMacs. The 604 and 620 where very powerful as well. PPC7400(G4) was the pinnacle of 32-bit PPC. PPC970(G5) was very powerful as well. IBM didn't want to put the effort into making a lower power part, so Apple ditched them and subsequently died an early dead. Cell and Xenon were castrated versions. Microsoft wasn't fully retarded and included Altivec/VMX.
I don't think PPC ever had best in class integer performance, but from start to finish it was a floating point monster. They also weren't the first to implement vector units into their CPUs. Nothing even came close to Altivec/VMX until AVX2 came out though. IBM/Motorola truly did a masterful job here.

ps3 cell processor was a significant design innovation coming in during a time period when dual core was just becoming standard on desktops, the cells 8 core design was 10 years ahead.

>8 core
It was like, 1 actual core with 6/7 mini slaves that could do repetitive work really fast or something, iirc.
>tfw ppc never coming back into mainstream market
>stuck with the mobile ARM, PC x86 meme for the rest of time
I blame Sony and Apple for this.

>ps3 cell processor was a significant design innovation

more like significant design retardation

>Macintosh still is in all honesty
How is that possible?! They run on standard PC hardware, you even said that yourself!

x86 is a horrible joke that has been going on for way too long.

>still is

>x86 counter-jerking starting already
Shit's seriously annoying, I wish you niggers would stop and just appreciate this shit for what it is instead of running any potential for interesting and educated discussion into the ground by regurgitating anecdotes and one-liners built on shoddy foundations to push some retarded victim narrative, the "golden age" of exotic, overpriced and lackluster systems with terrible software bases and support to match that you never experienced is never coming back.

It's not good that there's only two manufacturers of desktop CPUs, one which has been having financial problems for the past decade, and both have implemented hardware backdoors.

By the way, it feels really lame when you look through those screenfetch threads and everyone has the same hardware. Would be nice to see stuff like SPARC and POWER for a change.

>I wish you niggers would stop and just appreciate this shit for what it is instead of running any potential for interesting and educated discussion into the ground by regurgitating anecdotes and one-liners built on shoddy foundations
This is the essence of Sup Forums, there is nothing you can do about it.

>I think it is only Nintendo who still use PowerPC.
Certain embedded products still use it from time to time though not as much as MIPS stuff. Hell I think NASA put ppc chips in some of its more recent robots.

>SPARC
in its death throes, thanks oracle. i cry every time i encounter it

Just saying it would be nice to see some variety, didn't mean it would necessarily include SPARC.

>macfag defending shitty performance

Holy shit, Apple literally is like a religion to you people.

I agree but if we see anything in the near future it's because of either ARM finally nipping at intel's heels on the desktop, or the chinese market getting cut off and improving MIPS or some other RISC architecture

>"people"

The entire Ford Sync system is built on ppc32 chips

>was it better than x86?
Being kicked in the head was better than programming x86 in assembly, so yes.

Wrong. G4 was historically known for being the best generation. The G4 had the most performance for the time it was out. The G5 struggled at best during its time, and was far too power hungry.

>The G5 struggled at best during its time
Speak for yourself, my G5 dual was the best space heater I ever bought

RISC-V is beginning to get some serious backing. I don't think it will displace ARM anytime soon, but it's BSD licensed so nobody can kill it

Rampant vendor lock-in and sub-par, expensive, poorly documented hardware isn't good either, it's just the way it is, and it's not the fault of x86 as an architecture.

>and both have implemented hardware backdoors
The actual exploitability of the ME/PSP seems pretty up to debate given the absolute incompetence of the NSA's supposedly impressive surveillance state, and if the shit goes that deep, jumping ship to another chip family isn't going to do much to change that while they're still being manufactured by large American or US-affiliated corporations behind closed doors.

Doesn't mean I can't still shit on it every once and a while.

It's really only microcontrollers at this point though, I doubt there will be desktop CPUs anytime soon.

Where was I defending it? I'm pretty sure my ultimate point was that the G5 was quite lackluster as a pure desktop chip despite its above-average floating point performance.

At certain points, but the true performance kings of the RISC world were MIPS and Alpha.

When HP shut down Alpha development many of the engineers went to AMD and developed k8.

Alpha is not RISC.

It's more CISC than x86 ever got.

That's the first step

Why did Alpha processors have two nubs on them?

source

...are you sure you're not thinking of VAX?

MIPS was always shit though, it peaked at the R10K and went straight into the shitter from there. You bought SGI systems for the operating system, software and the system architecture as a whole, not the shitty CPU.

Other than the Alpha, SPARC and HPPA both were pretty high up.

For mounting the heatsink.

How so?

It has fixed length instruction encoding, a large number of registers, and a small instruction set without any cruft to make assembly easy ... all of the stuff that makes me think RISC

>HPPA
Why do we never hear about this arch?

>For mounting the heatsink.
Man I miss the days when you could get away with no fan (barring some fuck-huge heatsink).

I think the last time i did that was with a pentium 90?

>HPPA
youtube.com/watch?v=VLTh4uVJduI

Seems just nobody ever bought them, the only places I've known of that used them to any capacity were GM and CERN and even here with a huge HP campus down the street most of the corporate/industrial *nix market was dominated by Sun.

Shit, I've only got one and it doesn't even work, most of what I know about them is based on heresay, magazines and forum skimming, I'd like to pick up a 9000/712 some day just to mess with it.

Most of those had fans pointed straight at them one way or another though, even the early ones. The Alpha-based Multias were fucking ovens.

OpenPower seems to be taking off, rumor has it Google is going to use them in their newest datacenter.

>Google's going to use a few custom racks for some obscure bullshit
>taking off
No. Even the fucking i860 and m88k had better runs than that.

The only problem with POWER is that it's expensive as fuck, open or not.

>apple
>massive peripheral/upgrade market
[citation needed]

>OpenPower seems to be taking off
I can only hope. Sucks that Talos didn't get funded.

>Sonnet
>Newertech
>Daystar
>PowerLogix
>OWC
>Radius
>SuperMac
>RasterOps
>and so on
Did you live in a cave? There were a ton of vendors making accelerators, expansion cards, storage peripherals, even Mac clones, and pretty much everything else for the Macintosh ecosystem pretty much up until the G5 era.

>$10,000+ for a pathetic single-socket """workstation""" that was obsolete before production even started and had barely a niche
What were they expecting?

I thought the same thing. I was still disappointed, I guess to really need volume to make up production costs.

Probably why x86 is so persistent.

Yeah it's definitely part of it, the economy of scale helps.

The CPUs themselves are also pretty expensive since they're still high-performance chips that IBM was probably never really intending to sell for desktop/consumer use to any capacity.

>mfw I'll never get my hands on a 1.92GHz PowerBook G4

...

what the fuck

The new Nintendo Switch is running a tegra x1 from Nvidia, which is ARM and Cuda

and it's hilariously bad

>The G5 struggled at best during its time, and was far too power hungry.
>applefag
We are talking about a CPU and not a machine.

PPC would've died sooner or later anyway, because whatever advantage it had at FP operations can now be done significantly faster on GPUs. In fact, CUDA was developed just after PPC was phased out for good in the desktop space, so nothing was really lost. A typical modern PC can do any type of calculation very efficiently.

Too bad it was hampered by insufficient RAM and an under-powered GPU that forced game devs to use the CPU for additional graphics processing, so that great math potential got to see use, but for all the wrong reasons.
If they had doubled up the RAM and preferably kept it unified and with a better GPU in, at the cost of all that useless "home media" crap they stuffed the Phat PS3s with, it would've been a real powerhouse.
This tech made sense back in 2006, but now with very fast and efficient fully-programmable GPUs, the CPU is almost irrelevant for any task relating to gaming except for managing game logic and AI, which hardly requires a monster chip.

This shit again, people don't know the difference between RISC and CISC, just because it has more or complicated instructions than RISC does not make it CISC.

Now they're breaking binary compatibility that goes back to the Gamecube.

>PowerPC is just Apple
Thinkfags have an insane obsesion with that brand

That explain why Macbooks have awesome battery lifes.

Well, the PowerPC 970 was too power hungry for a consumer-friendly Workstation that was the PowerMac in that era. In the server market it was fine.

Daily reminder that virtually, no modern Intel processor is CISC or x86 anymore.
Basically they're RISC processors translating old x86 instructions.

>Basically they're RISC processors translating old x86 instructions.
Sounds extremely wasteful.

Actually is efficient as fuck. There's almost no overhead.

the ps3 cell was 800mhz ahhahahahahhahaha
eight hundred!!!!

>tfw you're watching yet another nip documentary on the semiconductor market, this time about TVs
>toshiba (and every other company) is facing increasing competition from low cost brands like vizio
>and their respone is to build the most expensive TV they can, with a fucking cell cpu as the core that can do 8 PiP at the same time and record it all to a built in 3TB hdd

>tfw none of the other yes-man engineers speak up how godfucking retarded this it

Japanese do not speak up to authority. It bring great shame to famiry

link to documentary?

Was true for most modern processors until recently (even powerpc processors didn't execute much of their ISA without breaking them down first).

Macro-op fusion is pretty much the exact opposite of this, though: Fusing together multiple ISA instructions because they can be executed more efficiently that way - pretty much the most direct anti-risc stance you could take.

BUMP

>falling for the Megahertz myth

So that's why AMD is shit.

ebin

I think it failed because it was limited to ps3 and maybe one tv, wider adoption would have done much good for cell

>tff you'll never have a rackmount Cell server

It's better for specific tasks just like SPARC, MIPS, POWER, and others are. The lead has become so little though that the majority have switched to x86. Much of the lead was eaten away when Intel changed their underlying architecture to closer resemble RISC while maintaining full CISC compatibility.

Na, senpai, the cell was retarded. SPE's were worthless because the bandwidth never made any sense to send anything to SPE. In a real world-non console setup, it make more sense to just send FP/Parallel computations to the GPU where you aren't limited to 512kb of memory or whatever retarded design the SPE's had.

x86 (and x86-64) is still a pants on head architecture, it just turns out if you have 20 billion dollars to spend, you can make any cpu architecture run Crysis on a single core over 10 years.

AMD64 fixed a lot of the deficiencies with x86 in addition to expanding the addressable memory. It may still be an "ugly" architecture but it's practical and functional now. The end user doesn't need to concern themselves with the ugliness it's only the devs that have to directly work with the hardware.

The Cell was designed when GPGPUs were in their infancy.

i would have killed for a 64bit G4. the G4 was a work horse.

>You bought SGI systems for the operating system, software and the system architecture as a whole, not the shitty CPU.
that pretty much summed up prime PPC-apple.

sorry, was busy doing other stuff
>nicovideo.jp/watch/sm18914987