Talos II lite

Raptor is going to release a poorfag single socket platform. It's only a mobo, case, and power supply, but it's cheaper than the dual socket motherboard is alone.

phoronix.com/scan.php?page=news_item&px=Raptor-Talos-2-Lite
secure.raptorcs.com/content/TL1BC1/intro.html

Attached: IBM-Power9-Talos-II.jpg (2789x1881, 366K)

Other urls found in this thread:

secure.raptorcs.com/content/TL1BC1/intro.html
hg.mozilla.org/mozilla-central/file/tip/js/src/jit
wiki.mozilla.org/PowerPC#Firefox_on_PowerPC
github.com/gcc-mirror/gcc/blob/master/gcc/config/rs6000/power9.md
lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20160509/354920.html
ibm.com/developerworks/downloads/r/xlcpluslinux/index.html
pgroup.com/products/community.htm
youtube.com/watch?v=wPrUmViN_5c
forge.rust-lang.org/platform-support.html
twitter.com/NSFWRedditGif

HYPE

how long until a desktop PC is available

Brainlet question here can I compile any program (such as Firefox) with gcc so long as I compile it for power9?

Attached: 1511326401906.png (183x275, 4K)

Yes.
You will probably have to because there aren't that many desktop-oriented distros that target ppc64.

Attached: bb66ktp6yix01.jpg (1440x1317, 148K)

POWER9 is ppc64el and all the major distros support it.

>secure.raptorcs.com/content/TL1BC1/intro.html
if it ain't RISC-V don't talk to me

Good.
Can I write programs for it in C?
If i can Im buying one.

The problem with browsers is that they use JIT backends for their javascript engines. It doesn't appear that firefox has a jit backend for Power in their js engine's source tree.

hg.mozilla.org/mozilla-central/file/tip/js/src/jit

>no javascript
truly a botnet-free architecture

Yes

Yes, you won't have the JIT backend enabled, but Firefox has other js engines, so you can still order pizza from the botnet
wiki.mozilla.org/PowerPC#Firefox_on_PowerPC

>You will probably have to because there aren't that many desktop-oriented distros that target ppc64.
Most of them do, including Debian.

>Can I write programs for it in C?
Not only is there explicit support from GCC[2] and LLVM[2], but there's also the option for meme compilers like IBM's XL[3] and PGI's[4] community edition!

[1] : github.com/gcc-mirror/gcc/blob/master/gcc/config/rs6000/power9.md
[2] : lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20160509/354920.html
[3] : ibm.com/developerworks/downloads/r/xlcpluslinux/index.html
[4] : pgroup.com/products/community.htm

excellent

Is this really the fucking state of Jow Forums right now?
No, sorry, Python and JavaScript only.

whats it to you bud?
It was an honest question about something very recent on the market.

I wonder if you could compile TenFourFox with GNUstep and take advantage of AltiVec

Seriously though what does this have to offer over used opterons and xeon blades?

>youtube.com/watch?v=wPrUmViN_5c
0:43

Open hardware

Who releases a processor without a C compiler? It's just an obnoxiously stupid question when you really think about it. You personally don't deserve a ton of shit for it because who the fuck cares, but god damn the POWER hype/shilling in the last two or so years really gets to me sometimes, so many people jacking off to this architecture and these systems who never even heard of it or cared about it until someone pointed out it had an IBM logo on it and made it the next cool Jow Forums fashion accessory/political statement.

In terms of experience it's no different from any other GNU/Linux system, just a different sticker on the front and poorer binary support. Development-wise I don't believe you'd run into any real obstacles.

>poorfag
It's great that it's fully open source down to the firmware but this sure ain't targeting "poorfags". That's a pretty darn expensive computer system. I get that it's competing with Threadripper - but still, it ain't exactly cheap.

Don't you already get that with pi arrays? Why spend 20X more for a power system with basically 0 software support outside of compiling shit yourself?

Attached: images.jpg (253x199, 9K)

How is it competing with threadripper? Last I heard it was struggling to compete with atom ST and MT performance.

Attached: laughingjets.jpg (1080x1086, 100K)

I can finally almost afford a Talos II system. CIA cucks BTFO

>power is significantly better than ARM
k

>an HPC architecture with 8-way SMT is better than a cellphone CPU cluster
Yes.

Not to mention that clusters aren't a drop-in replacement, only certain workloads benefit from them.

>Why spend 20X more for a power system with basically 0 software support outside of compiling shit yourself?
What kind of meme is this? Fedora, Debian, and CentOS all have support with packages. Why waste money on a system that 99% of software can't take advantage of and the other 1% is limited by the Pi's slow IO.

How vulnerable to side channel out of order execution attacks is POWER compared to x86?

IIRC far less than x86 or ARM. The OOB controller on this thing is also open source, which means no Intel ME shenanigans.

half of 10.4fx's reason for existence is its very 10.4-specific implementation of Firefox using nigger rigged native technologies whose GNU/Linux counterparts are already implemented in the mainstream build, AltiVec optimization is a very small part of it and pretty much just a compiler thing with some minor in-code optimizations to squeeze out what little performance you can get from it, AltiVec is really overrated and definitely not a cure-all for a shitty JavaScript engine, take it from someone posting from a 2.3DC G5 that still hits 70C spinning a reCAPTCHA.
>GNUstep
nigger what year is it, you're going to need way more than that to build 10.4fx on GNU/Linux, I don't even think it builds on 10.6+

all in all just build regular Firefox from source with compiler optimizations if you want something more tuned, which includes the usage of AltiVec where possible

I'm unironically considering this as a hobbyist machine

Gentoo with -march=native in make.conf ought to scream on these machines.

would be interesting to see if it's any good
would probably kick even more ass with a proper IBM compiler too if XL still rips GCC a new asshole like it used to (probably does)

The only RISC-V board out there make ARM look like a speed demon and costs $1k

Micro controller vendors here and there. C a shit.

Can I qemu kvm an x86_64 arch and run windblows with a gpu passthrough?

>open hardware
>Raspberry Pi
Pick one

>2
My quad MIPS 3B3000 16-core serverboard can do this, though be it at 1.5ghz, the loongson arch has on-the-chip optimizations for x86_64 emulation via qemu gives about 78% relative performance, and 99% INT perf due to the hardware optimization.

I mean unless you explicitly use platform specific code, as long as it can compile, it can run C programs.

>the chassis alone costs $1,399.99
>minimum $1,920.00 extra for a usable system

No the Chassis comes with a single socket MB, 500W PSU, and a heatsink for the processor.

PPC is nice. Too bad Xeon kicks its ass at almost everything though.

>so many people jacking off to this architecture and these systems who never even heard of it or cared about it until someone pointed out it had an IBM logo on it
so you're mad because you're a hipster that was "doing it before it was cool" or something?

I don't get it. I used to have a ppc mac back in the day, and definitely never liked the idea of an x86 monoculture to begin with, but that "muh IBM" complaint is just pure autism. It wasn't until very recently that desktop caliber non-x86 systems were even available to the general public. And no, I don't consider outrageously large and expensive meme servers as a suitable option for those of us that just wanted non-x86 desktops to work on. The only real competition to this Talos II system until recently has been meme amiga hardware, and some arm-based amd opteron dev kit that's underpowered in comparison. Meanwhile, stuff like the ppc notebook project are still just vaporware, and RPi-like dev boards don't have all the functionality of a full desktop pc.

How OLD can you be ffs?

Old enough to have watched Power rise and then be overtaken by Xeon.

This. I wish instead of being just a meme organization the FSF would contribute to things like this so we could have affordable computers with libre hardware.

This isn't the old time grandpa.

> 3k for an entry level system (16gb of ram, 512gb ssd, 8 cores cpu, no graphics card)
> affordable

Attached: 1526479725832.jpg (161x131, 5K)

>Talos
you have come to seek the word of mighty talos?

Attached: Heimskr.png (600x1000, 1.27M)

It is affordable if you actually need a POWER9 machine.
This is for math dorks not imageboard neckbeards.

Attached: 6zdk2x1jemx01.jpg (2376x1810, 287K)

There's still a reason every one of the top supercomputers uses Xeon and not Power. It was nice but it just can't compete anymore.

Very true, but it looks to have a brighter future than POWER. I.e., there is a decent chance RISC-V will end up in abso-fucking-lutely everything.

>Development-wise I don't believe you'd run into any real obstacles.
I dunno, I used to run into issues caused by code that assumed the world was little-endian on a regular basis.

I just want to know if this will run freebsd as a firewall with PF or opnsense

No, those are x86 only.

Sounds like you need to clean your G5, I've never seen my dp 2GHz model break 45ÂșC even under load

> I don't even think it builds on 10.6+
Literally the reason I suggested GNUstep, it's to get older OS X applications to build. iirc GNUstep specifically targets Tiger compatibility.

>need a POWER9 machine.

Attached: 1526523802683.jpg (378x603, 42K)

In 20 years maybe.

>78% relative performance, and 99% INT perf due to the hardware optimization.
No fucking way.

What's the closest metal way you could execute x86 binaries assuming you wanted to go akin to a wine route instead of passthrough?
Is there anything for emulating as close as this with just binaries or are you stuck with the whole stack of the OS in a VM?

Attached: 1526080339130.jpg (736x736, 95K)

More like 10, I'd say.

That's the Loongson MIPS arch, not POWER9. The Loongson chips actually have x86 decode silicon. TALOS is absolutely not the platform for you if you want proprietary Windows software. Keep a separate PC around for that, and unplug it when not in use.

sure about that?

My idea was to do remote desktop to an x86 system.

pfSense and OPNsense are amd64 only, MAYBE also ARM. You'd need to boot FreeBSD ppc64el and use manual configuration of pf.

Yeah that'd work. Plenty of RDP clients on *nix.

that's just a 2 GHz single-core 970/970FX though, the later DCs are shoving two of those on one chip plus the 2.3 GHz clock which is pretty close to the point of needing to be water cooled at 2.5+
but I still do think it needs to be repasted, the GPU is running way out of acceptable spec in particular, but the point I was trying to make was that AltiVec isn't really isn't the missing piece of the puzzle
maybe you could try to port pieces of 10.4fx over that optimize for the G5 but... we're not working on a G5, this is a far newer and different kind of chip and I really don't think it would be very useful towards it
>GNUstep aims to be compatible with both the OpenStep specification and with Mac OS X. It should be easy to write an application that compiles cleanly under both GNUstep and Cocoa. GNUstep can be expected to support the OSX 10.4 version of the Cocoa APIs
I unironically did think it was further back in the stone age than this, pretty cool, but I still don't think it by itself would help very much, 10.4fx seems to be very much made for 10.4 with all of its limitations in mind

the whole thing just sounds fucking silly considering what 10.4fx is for, you'd probably get much more out of just petitioning Cameron Kaiser to fix the missing parts of Firefox in a way that's actually suited to the system in question since he has a Talos himself

Power isn't open hardware. RISC-V, however, is. This computer that OP linked has free firmware, but the hardware itself is closed.

Unfortunately, "missing parts of Firefox" includes a full Rust toolchain. That needs to get ported to ppc64el (and ideally RISC-V 64 bit) before Firefox Quantum will build.

>forge.rust-lang.org/platform-support.html
>Tier 2
>powerpc64le-unknown-linux-gnu

>guaranteed to build, not guaranteed to work
Ironically I think a dual socket TALOS II build server could help the Rust folks promote that to tier 1.

I'll be ordering one of these single socket machines soon I think. I'll see what it comes pre-loaded with, or try to install RedHat or Debian on it. After that I'll see if there's already a fully working QEMU port with KVM support. If there isn't I'll work on that a bit and see if I can get it working to allow other Anons and POWER users to run x86-64 applications. It probably won't be any better than using a mid range Intel laptop but you'll at least be able to access the massive software libraries of x86 software.

Running VMs will be the best solution until we can get something better in order. Right now I'm working on my own hobbyist OS where I'm porting Debian userland utilities onto MINIX and running stripped down Linux kernels as their own microkernel servers for driver support. On top of that I'm using Genode and trying to work on getting xorg to run so I can some decent window managers and such. I'm thinking about putting a Motif based WM and apps on there. If I like my Talos II system I'm saving for I'll compile and release this new OS for it. Then we'll have a whole new OS built correctly from the start.

>QEMU port with KVM support
KVM will only be useful for Power guests...

Wrong.

>Power.org is still the governing body around the Power Architecture instruction set but specific implementations are now free to use under a liberal license granted by IBM. Processors based on IBM's IP can now be fabricated on any foundry and mixed with other hardware products of the integrator's choice.
?

What a worthless language

Nice. Definitely gonna order a Talos later this year once I start my new job.

Are there benchmarks comparing the 4/8/18/22-core CPUs, and the single-socket vs dual-socket options?

They're actually 970MPs, but the northbridge doesn't support dual-core CPUs so there's only one core active in each processor. Either way you do have twice the cores running.
>the whole thing just sounds fucking silly considering what 10.4fx is for, you'd probably get much more out of just petitioning Cameron Kaiser to fix the missing parts of Firefox in a way that's actually suited to the system in question since he has a Talos himself
I forgot he picked one up, would be neat.

Attached: 1495147543587.png (472x508, 6K)

>math dorks
the fuck are you talking about? what would a mathematician get from POWER9 that they can't from modern x86 chips better and cheaper?

this is for FLOSS purists and security nerds.

You can run qemu x86-64 and do passthrough.
Obviously only games with light cpu usage will work well.

You would need to run in BE mode not LE. TFF author bought Talos so you cn check his blog, I think he mentioned this in one of the last few blogposts

Actually Power has BOTH Spectre and Meltdown, so you need firmware/OS patches for both vulns.
AMD ftw.

>My quad MIPS 3B3000 16-core serverboard
Holy shit where did you get it?

He may be alphabet soup guy since i don't believe anything like that is sold retail by anyone.

>They're actually 970MPs, but the northbridge doesn't support dual-core CPUs so there's only one core active in each processor. Either way you do have twice the cores running.
but they never made the 2.0DP concurrently with any of the dual-core models according to official figures, so that doesn't make sense to me unless they were running out of chips to the point that they were using binned MPs from the next production run in very late systems

whatever the case I think at 2 GHz those things just run way cooler, didn't have any issues with my 2 GHz DCs either

>but they never made the 2.0DP concurrently with any of the dual-core models according to official figures, so that doesn't make sense to me unless they were running out of chips to the point that they were using binned MPs from the next production run in very late systems
I swapped out the 970s for 970MPs

that's an interesting upgrade, did you just want to see if they'd work or was there a processor failure?

I wanted to upgrade to 970FX processors for the die shrink and decided to see if I could get the dual-core 970MPs to work. I may not be able to use all four cores but at least one core from each processor works so it wasn't a total loss and the end result is essentially the same as if I'd gone for 970FXs

>quad MIPS 3B3000 16-core serverboard
What do you run on it?

>Running VMs will be the best solution until we can get something better in order
Please start an affordable POWER-based VPS service, user. The VPS space has way too much Intel (not even x86_64, just Intel), with Scaleway as the sole prominent exception.

For Genode, I'd recommend porting Arcan rather than X directly. Arcan can then run X inside itself.

You don't need to buy their ECC RAM and marked up SSDs. It shouldn't make a difference what you use for that.

Unless I'm retarded.

>ECC RAM and marked up SSDs
They are not that badly marked up, though. The point is that it's still an expensive system compared to AMD64, ARM, or MIPS.

They're marked up for consumer needs, though. If you just want a machine running without binary blobs anywhere, you don't need ECC ram or NVMe SSDs. Just get normal RAM and a SATA SSD. That's what I meant.

The base system is definitely expensive, but not out of reach if you really want it. I'm under 20 years old and still in university right now and I could in theory buy one if I decided that it was worth the opportunity cost.