We all know ARM is needed when the lowest power consumption is needed.
But why there exists PowerPC? SPARC? Intel x86 is around here because it is just the most usable architecture, right?
>>54595277
Have all the disadvantages:
>RISC makes compilers hard to implement, CISC makes hardware hard to implement.
>RISC can take more cycles (due to more instructions) to complete a task, CISC will draw more power.
>>54595305
But modern processors aren't CISC, are they?
>>54595277
powerpc is such a beast.
[spoiler]in 2005[/spoiler]
>>54595305
Are you high? The whole point of RISC is to lower the number of clock cycles per instruction and to reduce the size of the Silicon required.
>>54595351
Yes and no. The code they read is CISC but then it gets covered internally into RISC because it is advantageous to do so (even with the conversion process)
>>54595439
Converted*
>>54595277
sparc is open sauce
pity nobody is taking advantage of that.
SPARC is built around high power multithreading. MIPS is good for small embedded systems. PowerPC is good for FPGAs and routers.
>>54595480
Only the SPARC base is open-source. Oracle's involvement and further development is scaring contributors off.
If you want a true open-source ISA, there is one clear modern choice that learned from the mistakes and design wins of all the others: RISC-V. The thesis has a great section analysing other ISAs, actually (basically, why we didn't use them), and is very well worth a look at for an overview.
x86 won through accidents of history, and several very good implementations, but not the quality of the ISA!
Don't forget all the tiny little ones, like ARC, which still lives on, forgotten, inside the Intel ME microcontroller, inside your x86 chip...
>>54595439
No.
The point RISC is to reduce the instruction set to a very small set of really basic operations that complete in one cycle each.
But a complex task, like a division can take 10s of those small instructions.
RISC would have a single division instruction that completes in, say, 2-3 cycles.
So, my point stands.
Src:EE major
>>54595546
>Oracle's involvement and further development is scaring contributors off.
What's the problem with Oracle?
>>54595351
x86 is CISC. However, modern processors have a RISC core and an interpretation layer to convert complex instructions into simpler ones.
ARM is kind of RISC, but it gets increasingly complicated.
>>54595277
x86 is still around because it supports old x86 software.
The ISA isn't very good compared to alternatives.
It's like how women have multiple shoes.
Each one does is good for a specific thing.
>>54595510
>PowerPC is good for FPGAs and routers.
What is some special problem routers have to compute so they can't have a usual MIPS/ARM?
SPARC and power Power Architecture seem to be primarily focused on giving the user a fuckton of cores/threads. They're probably more cost effective and space efficient for a given budget
>>54595277
>We all know ARM is needed when the lowest power consumption is needed.
That's x86 senpai.
>But why there exists PowerPC? SPARC? Intel x86 is around here because it is just the most usable architecture, right?
novelty. x86 is not just the most usable architecture but the most power efficient and software compatible.
See pic related.
>>54595277
>We all know ARM is needed when the lowest power consumption is needed.
nice meme
Not in pic related is the iPhone 6 plus and note 3 (both farther down on the list).
>>54595792
PPC and MIPS(I think) are big-endian so there's less overhead when processing network packets.
I don't know about ARMs endianness so I can't say anything about that.
>>54597478
ARM is all Little Endian since a while
Older ARM processors had both little as big endian versions
>Implying power consumption and and sheer power are the only factors that matter.
>>54595546
>If you want a true open-source ISA, there is one clear modern choice that learned from the mistakes and design wins of all the others: RISC-V.
or openpower with IBM on board.
https://www.raptorengineering.com/TALOS/prerelease.php
Too bad free as in freedom computing isn't free as in beer.
>>54597424
Arm has the lowest power consumption
But that does not apply to meme phone SOCs with such high frequencies and on-die memes like modems or stuff
>>54595790
They're all good for 1 specific thing, shoejobs.
>>54597993
>Arm has the lowest power consumption
No it doesn't. For most cpu work an x86 xeon/atom will use less electricity compared to an A53/A15 ARM CPU. This is why the zenfone 2 has such amazing battery life despite it using a cheap 5.5" 1080p chink panel display.
Phones would have better battery lives if modern x86 core-Ms and airmont/goldmont atoms were used. But they're more expensive than mediakek 8-core turds so phone manufacturers don't use them despite the benefit of better battery life.
>>54598214
Wish to see a x86 CPU for embedded systems or some other micro devices.
>>54595674
Major Sir!
>>54595277
>Intel x86 is around here because it is just the most usable architecture, right?
No, and neither is ARM. They exist because they're cheap, compatible, and useful enough and that's what the market wants.
SPARC is best suited for applications that benefit from extreme parallelism, Sun was putting out native 8-core, 64-thread chips while Intel was still sucking their own dicks marketing two Prescott on a die with fake multithreading as a costly option. The latest chips have 32 cores and 256 threads, at the cost of very lackluster performance on a single core/thread.
POWER chips are optimized for financial and other secure use cases with advanced RAS features and integrated hardware co-processors to speed up things like encryption, it's also a pretty good HPC chip too.
>>54595389
The shitty space heaters that were late-era PowerPC chips =/= modern POWER CPUs, they share an instruction set and a little ancestry, IBM used them both for various purposes, but they really shouldn't be considered equals and compared as such.
Also, obligatory >>>/v/
>>54595439
>The whole point of RISC is to lower the number of clock cycles per instruction
No, not even. Doing that requires complicated instructions that increase complexity, cost, and heat output, RISC chips took shit that a CISC design like the 68000 or 8086 could do is a single cycle and broke it up into simpler instructions, which it made up for by clocking up to bitchin' speeds. At the height of RISC chips that were actually reduced around the early '90s, even the lowly ARM chip in a PDA was clocked almost as high (or even higher) than a high-end x86 offering.
>>54600540
>shitty space heaters
POWER8 is still a 200W+ space heater
>>54597378
Computing isn't about saving five cents on power bills or getting the biggest number in a shitty synthetic benchmark.
I don't think businesses are dumping millions on SPARC/POWER hardware for novelty, either.
>>54600575
But unlike the G5, it's actually worth a shit.
>>54600620
>Computing isn't about saving five cents on power bills or getting the biggest number in a shitty synthetic benchmark.
>I don't think businesses are dumping millions on SPARC/POWER hardware for novelty, either.
On sparc yeah not power though.
>>54600635
>implying the G5 isn't
Even a G4 works as a desktop processor today
>>54600635
Xeon E7 is cheaper and has a similar RAS feature set.
Unless if you're a die hard IBM shop there's no point moving over to POWER
>>54595277
You forgot MIPS.
>>54597378
>>54597424
0.02 shekels have been deposited into your account.
>>54600575
lmao this
>>54600673
MIPS is irrelevant, more so than SPARC.
>>54600695
>MIPS is irrelevant
Sure thing bud
>>54597478
MIPS is bi-endian, although many more MIPS processors are big endian than little, IIRC.
ARM is also bi-endian, although most ARM processors are little endian.
PowerPC is big endian, although POWER8 is bi-endian.
>>54600685
How much is IBM paying you?
>>54600705
It is, it's just a meme cpu architecture. Who the fuck uses MIPS when x86 and ARM exist? Even if you were a hipser nerd you'd use POWER.
>>54600740
>Who the fuck uses MIPS when x86 and ARM exist?
I do, x86 is shit and ARM is a meme
>>54600750
Then enjoy being a meme yourself.
>>54600750
isn't the fastest MIPS in a 20 year old SGI machine? How do you run a web browser on that?
Intel's processors have risc-like microinstructions executed by the "major" instructions anyway, the RISC advantage is that it doesn't need to do the major OP -> minor OP translation and a few more other architecture choices (like being load-store) make it more efficient power-wise. I understand the bad rep it gets from though, especially after benchmarking Creator Ci20 (mips board) I can't be shocked to see people thinking RISC is shit. We just need non-shitty implementations on some dev boards (Cortex-A72 particularly matches Pentium n3540 / core 2 duo e8400, certainly usable for desktop)
>>54600824
http://research.cs.wisc.edu/vertical/papers/2013/hpca13-isa-power-struggles.pdf
>>54600824
x86 is like 90% RISC you nigger, it's more energy efficient than ARM too. Why the fuck would you use a meme architecture that uses more energy and is incompatible with x86 software? Are you legitimately brain damaged?
>>54600655
They're still good and interesting chips, but the G4 and G5 didn't really fair well against their contemporaries.
>>54600658
>"similar"
>but it's cheaper!
Big fucking whoop, Xeons have barely even caught up to Itaniums for mission-critical applications, where hardware costs are the least of your worries when you're going to be paying ten times that in software licensing anyway.
/g/ needs to stop acting like POWER and SPARC are designed and marketed for use in fleet boxes, they're not intended to fill a datacenter, they're intended for specialized jobs where the utmost reliability, availability and consistent performance is more important than saving a paltry hundred or so a month.
>>54600809
Why don't you go ask all of the Nekochanners who are still daily driving Octanes, Tezros and Fuels to this day without giving a single fuck?
There's more to computing than jerking off and mindlessly consuming other people's content.
>>54600879
>Xeons have barely even caught up to Itaniums for mission-critical applications
>Intel's Kirk Skaugen acknowledged the centrality of OS and vendor choices in his aforementioned IDF presentation. Here's the full quote for reference:
>"We used to position Itanium as highest performance, highest reliability... We're still committed to Itanium. It's really now a choice of operating system. Xeon's reliability and performance is now equal [to]—and in some cases better than—Itanium, and they're going to leapfrog [each other] in performance over time."
>"If you like HP-UX, OpenVMS, Nonstop, and [other] mainframe operating systems, we're going to fully support you on Itanium. But now Xeon is in a space where there's no workload on the planet that Xeon can't handle."
>>54600876
>Why the fuck would you use a meme architecture that uses more energy and is incompatible with x86 software?
Well, the biggest problem with x86 is the monopoly/oligopoly created by intel and amd
>>54600902
Link it.
I believe it though, I actually think I read that in an article published five or so years ago.
>>54600924
http://arstechnica.com/business/2011/06/ask-ars-why-itaniumask-ars-with-xeons-improvement-why-bother-with-itanium/
Just search the quotes next time, lazy fucker
>>54600876
>Why the fuck would you use a meme architecture that uses more energy and is incompatible with x86 software? Are you legitimately brain damaged?
I don't know about brain damage but I certainly am mentally mature and apt enough to not lash out at others like a petulant toddler online.
The energy efficiency meme is dumb because in benchmarks they compare the intel's A7 / A9 tier processors (atom) to ARM's Core M tier processors (A72,Krait, Mongoose) and it certainly doesn't help that the morons use 1440p / 4k screens on those devices. Cortex-a53 itself runs of milliwats while the least power consuming processor of intel is using 4.5W
>>54600861
Thanks for the paper, reading it right now.
>>54600923
>so let's change over to an architecture riddled with undocumented proprietary implimentations because I think something bad might happen maybe according to /v/ and other FUD spouters
????
>>54600979
>let's change over to an architecture riddled with undocumented proprietary implimentations
But we're already using x86, senpai.
>>54601001
yeah, and why do you think some shitty chink ARM SoC will somehow be better?
>>54601017
There are actually a few ARM proccesors with totally open firmware and peripherals (chipsets, pci bridges) like the beaglebone, even though the processor design is still classified (for a good reason, ARM doesn't want it stolen)
>>54600809
MIPS CPUs are still in production and can be rather beastly
>>54595688
https://www.youtube.com/watch?v=-zRN7XLCRhc
