Consistently outperforms both Intel and AMD equivalents in single- and multi-threaded workloads

As quoted in , it's not "consistent". But it is very good:

So it's the worst of all worlds.
Intel can do better single core right now and for less money, better multicore (since you can buy more cores for less money)
Not to mention AMD is literally going to murder Intel on the value segment.

POWER is essentially useful for nothing.

>What went wrong
TDP of the chips is so fucking high they often need water cooling just to run. 70% of the fat PS3 console was all heatsink, much like the Power Macs.

>and how do we fix POWER?
Reducing the power consumption is a start, if only to make it marketable to anyone who doesn't have an HVAC installation at home.

>>consistently outperforms both Intel and AMD equivalents in single- and multi-threaded workloads
[citation needed]

>Intel can do better single core right now
Not consistently; Power is competing well even on ST, and the processors themselves aren't all *that* expensive; Raptor sells 4-core CPUs for $375, and 22-core CPUs for $2575 (which is a lot cheaper than 22-core Xeons), and the SMT implementation appears to give much better payouts than both Intel and AMT SMT.

>POWER is essentially useful for nothing.
Also note that, as mentioned, the Raptor boards have totally, 100% free and libre firmware. No IME, no closed boot code, even fully open microcode. If there are vulnerabilities, they can totally be patched (unlike hoping for BIOS/microcode updates from vendors in the x86 world for hardware that wasn't purchased in the last year). That's a very real advantage.

>Power is competing well even on ST
Also, those benchmarks are for Power8, and Power9 is just around the corner.

Both Intel and AMD hate your freedom, though.

sooooo
can i gaym on this

Attached: CjcQBE6.jpg (362x346, 36K)

Open source games and emulators

What's the power consumption like? I'm guessing it's called POWER for a reason (heheheheheheh)

what about running windows in a vm with gpu passthrough, i'd restrict its network access with a whitelist firewall
it would be the ultimate /free/ gayming machine

VMs won't be possible because different architectures. You would have to use an actual emulator, something like QEMU, and I don't know if GPU passthrough would be possible in that case. Probably not. Even if so, the emulated CPU performance would probably be significantly worse than any modern x86 CPU

Emulator, not vm.

Really, if you have that much money you're better off buying a dedicated botnet PC with nothing you care about on it.

i do have a dedicated botnet pc, and use librebooted machines for the rest
even for work, they are getting kind of weak and multi-monitor support is difficult because of gpu blobs
not counting the fact that i have to isolate said botnet devices from the rest of my network and servers, which is also a pain
thanks, sorry for ignorance

Attached: b2dd8fb448695b9bf381abf659662c78.jpg (500x529, 60K)

>even fully open microcode.
you realise that's also a problem if someone magically discovers a 0day and doesn't report it, right?

Security by obscurity is poor security. That issue can happen either way, but here you could audit the code yourself.

Is there anything equivalent for graphics acceleration?

I'm not entirely sure, but I do think that both Intel's and AMD's GPUs are pretty much fully documented, and there are fully free drivers for them already.

AMD GPUs require non-free firmware, as well as Nvidia's

Intel's now too.

checked, satan

i thought you could run amd's "open-gpu" thing without any non-free packages?

Open drivers but there's still non-free firmware on the card itself

install netbsd
you can probably achieve some sort of """high-level""" gpu passthrough if you're dedicated enough to mod qemu enough

>finally out of pre-order status
Fucking finally.

Attached: 1502516332234.gif (300x200, 225K)

gee billy

This.

The OS is basically a CPU driver, so that logic is equivalent to saying you can run Linux therefore intel CPUs are free.

>The OS is basically a CPU driver, so that logic is equivalent to saying you can run Linux therefore intel CPUs are free.
No, but it's not like Power9 is free hardware either.

But everything that runs on it is free - you need zero binary blobs at all, down to the microcode (or at last that's what's advertised). At that point, does it matter whether the hardware is free? The upfront cost and complexity of manufacturing CPUs is so high that I find it highly unlikely any sort of "open hardware" type CPU will ever be feasible.

At least AMD doesn't block out open source drivers with signed firmware

Theoretically, that could be true for GPUs as well. It may not be, but the post you replied to didn't know that.

>The upfront cost and complexity of manufacturing CPUs is so high that I find it highly unlikely any sort of "open hardware" type CPU will ever be feasible.

> ;
>optimised towards asics
But even if it weren't (or if it happens to be easily optimisable to be a good all-purpose CPU instead), is it in production? Does it have any chance of ever actually being produced, and having competitive performance with other modern platforms?

>(or if it happens to be easily optimisable to be a good all-purpose CPU instead)
You're misunderstanding. "Optimized towards ASICs" means that it's optimized for a monolithic, high-performance implementation, rather than for being instanced in FPGAs.

Oh, so in the sense that it's optimized for people to build a specialise CPU for it as opposed to just slapping it on an FPGA? Sorry for brainlet.

No, it's not about using it as a specialized CPU, it's simply about manufacturing it "for real". "ASIC" here simply means "hard implementation", in contrast to soft implementations like FPGAs. Intel and AMD CPUs are also ASICs.

No man should have all that POWER
youtu.be/L53gjP-TtGE

Yeah that's what I meant. I'm not good at this.

Anyway, back to my original question - what chance do you think this has of ever being actually commercially available? Making a design isn't hard. Ok, it is, but relatively speaking if you get a bunch of academics together they'll come up with something (as they did here). The problem is then getting physical production up and running, which is out of reach of pretty much everyone.

Well, right now there isn't much market for RISC-V at all, but even so it's apparently possible for SiFive to sell actual ASIC implementations of it. I wouldn't find it particularly surprising at all if BOOM eventually finds its way into real silicon; there just needs to be enough demand for it.

For that matter, most of the cost in high-performance CPU production is in validation, and arguably that's the perfect kind of thing for an open-source community to take care of.

I mean, if anything, this has already happened. The first practical RISC chips were purely academic projects, and they worked fine.

>Well, right now there isn't much market for RISC-V
Companies don't really care about the ISA. Provide cheap micros with RISC-V or power desktop chips that can compete with intel and people will buy them if the software stack is there.

>if the software stack is there
That's exactly what I meant, because the software stack isn't there. That's probably POWER's single greatest problem as well.

I'm really tired of faggot gamer babies who didn't even know what POWER was before 2016 suddenly shilling the fuck out of it as some shiny new up and coming x86 killer.

IBM doesn't give a shit about the mainstream market, they spent decades trying to get away from it, and even if they didn't, POWER offers fuck all to the mainstream market where PCs are already crashing and burning because they couldn't give less of a fuck about a theoretical 20% performance boost for tasks that already take seconds to complete on anything made in the last decade.

And christ, if you think Intel and AMD are literally the fourth reich because your shitty gaming build cost $50 more than it should have, imagine being fucked in the ass by the very definition of vendor lock-in. IBM is literally the Apple of business.

check my code

leancoding.co/showthread.php?v=3EAFUG

>IBM were literally the Apple of business.
FTFY. In the real world, POWER is governed by a foundation, which is at least as open as ARM is to alternate implementations. That's not saying a whole lot in comparison to RISC-V, but it's certainly more open than x86, which only has more than one implementation strictly because of legacy legal reasons.

It's certainly true that competitors could arise on that basis, but nobody's going to put up with the intermediary ass-fucking and rising costs to get there.

PowerPC is still pretty popular in embedded. Specifically car ECUs and some networking equipment.

So are Z80 and 68000 derivatives, when are those taking over the desktop again?
Being strong in one market doesn't mean it's strong in all of them.

You are a fucking megafaggot

Just mentioning that they're not completely dead in every mainstream market

If anything, System/360 is arguably a nicer architecture than POWER. Just strip out the 60s BCD and string-data instructions and you're left with a surprisingly lean, almost RISC-like instruction set with some extra efficiencies like simple, variable-length encoding.

what would have happened if an asteroid had simultaneously hit intel and amd headquarters back in 1997?
would Windows have gone with PowerPC, MIPS, Alpha, or SPARC? (yes, there was a SPARC version of NT in the works, but it was never released)

Attached: winnt-4.0-ppc-new.install.png (1024x768, 23K)

What gaymes can I play on my Power 720 running System i V7R3?

I can only hope it would have gone with Alpha. Out of all the RISC ISAs, it seems by far the nicest. Arguably even nicer than RISC-V.

>no software
>housefire
and

Attached: 1522545336742.png (2000x1955, 263K)

>POWER is essentially useful for nothing.
/thread

>i want to pay the x86 tax forever

>no software
debian.org/ports/powerpc/

It's not just about the software being strictly available. You can check eg. for the state of optimization for Power.

Sounds like a lot of opportunity to contribute to FOSS

But you're just being pedantic, nobody is working themselves into a frothy lather over fucking routers and ECUs, we're talking about these architectures as alternatives to mainstream personal computers, as all the talk about Intel and AMD (who are only popular in these markets) implies.
PowerPC had the most mainstream developer support out of all of these and stands the best chance by far. NT on MIPS was dead for years in 1997, Alpha and SPARC were awful vendor-locked shitheaps that carried a hefty tax.

But people would probably rather crowdfund Intel and AMD some new executive offices than buy any of them, RISC NT was terrible.

>I want to pay the IBM tax instead
>I want to pay this tax to run the same basic academic shitware nobody cares about on PCs either

>I want to pay the tax and then have to work to actually make it competitive

>then have to work to actually make it competitive
Programming is fun.

Not when it's a job and a waste of time that could be spent solving useful problems.

>The first practical RISC chips were purely academic projects, and they worked fine.
the things people threw five figures at IRISes and SPARCstations for simply weren't doable on affordable commodity PCs, oldschool RISC systems didn't just sell because of the fast processor, they sold because of everything else that came with them; superior system architecture, software and user experience

in 2018 you probably couldn't tell a debadged Talos II apart from a Xeon build in the same case without benchmarks or system information tools, almost every serious alternative to the wintel hegemony is a poorly designed, poorly marketed GNU/Linux box built from the same old commodity parts but with a different CPU in the socket and running the same old GPL stack as a cumstained T420 you bought off of ebay for a tenth of the price

the reason history won't repeat itself is that there's simply nothing tangible or even mildly interesting that these systems have to offer, just nebulous theoretical performance bonuses in niche tasks and talking points for CS undergrads going through their fuck da corporations ethical crusade phase

give me a company that actually tries to do more than just slap a new sticker on the same old taiwanese case and maybe I'll believe there's a future for the likes of RISC-V and POWER on consumer desktops, but until then I'll laugh at every half-assed $6,000 debian vim+firefox machine Jow Forums thinks is going to change everything

Power was the hero we needed but not the hero we deserved

It introduced a genuine competitor to Intel and its clones, then IBM got lazy and everyone who developed for it moved on. the sole exception to this were niche developers and Nintendo.

it'd be sick to see Power come back, even just to introduce some diversity into things. I'm a little surprised it never worked its way into graphics cards in any meaningful way. PPC is optimized for that side of things, but developers are lazy, and having Intel vs. AMD vs. IBM would have made for trouble. It's not hard to understand why, but it's a shame that the marker swings towards excluding outliers and interesting products.

>It introduced a genuine competitor to Intel and its clones, then IBM got lazy and everyone who developed for it moved on.
it was never a genuine competitor in that space, the market profoundly rejected it along with MIPS because the systems were overpriced shit nobody developed for to begin with

wow a $2000 motherboard fucking gay

Click this too while you're at it,

distrowatch.com/dwres.php?waitingdistro=462&resource=links

that's for the old radeon driver
what about the new amdgpu one?

You can get at least 2 x86 systems for the price of one POWER system. They aren't cost-effective.

>we're talking about these architectures as alternatives to mainstream personal computers, as all the talk about Intel and AMD (who are only popular in these markets) implies.
*blocks ur path*

Attached: Windows10-Qualcomm-Snapdragon-1024x576.jpg (1024x576, 41K)

Nobody who can afford something better is going to buy those pieces of shit.

>I want to pay this tax to run the same basic academic shitware nobody cares about on PCs either
>Muh games

AIUI both boom and rocket have had several test silicon already.

distrowatch.com/table.php?distribution=dragonflybsd
The one that matters.

yes we understand that you never used a computer for anything else before you switched to GNU/Linux, but that doesn't mean the rest of us are the same

AmigaOS -> Linux -> Dragonfly.

so what makes you worth taking seriously again when you're obviously a massive contrarian who only knows how to spout memes about things he doesn't understand?

>Amiga
Literally the only reason anyone ever gave a shit about that platform was games.

An open ISA is actually a pretty nice thing for people doing soft-cores in FPGAs or putting microcontrollers in ASICs, because it means they can use off-the-shelf programming tools and still not have to pay licensing costs for the silicon. This is the reason both WDC and nVidia have been engaged in RISC-V.

Give it a couple of years, and it's likely that RISC-V will be walking up from embedded microcontrollers into standard microcontroller packages like what is currently occupied by PIC32s or ARMs, and from there they may just go on to more powerful computers, just like ARM is doing.

>in 2018 you probably couldn't tell a debadged Talos II apart from a Xeon build in the same case
I sure could tell it apart from having replaceable, open firmware, and I'm pretty sure that's the same reason why people like Google is looking at Power. That and CAPI.

How to tell you're from the states.