/tech/ - Technology★

I only hope the radiator can accomodates standard fans, this time.

Do we have open source consumer-tier PSU units yet? Not counting Facebook's consortium for cheap and well integrated server hardware, I haven't heard any news on power supplies last time I checked.

It's important because Israelis use Jewish Desert Lightning Magick to infiltrate operating systems, when TEMPEST practices are not followed.

Also, FOSH mice. Plebbitors have been all over FOSH keyboards because of

>muh cherries

but it would be a neat thing to find FOSH mice.

I've never dealt with M.2 form factor SSDs so I'm not familiar with them. Let's say that this new board has 2 of those slots, since it's linked directly to the PCIe lanes I'm assuming it doesn't require a storage controller to use them?

>>982529

ok anon how tf can you post that without sauce this is torture

If Talos did a new custom board just for a single socket system they must be doing quite well.

Or are we looking at something different here? On board cpu maybe? Sub 1000 dollar?

tfw might actually get one now, this is sexy.

>two ram slots tho

how much is it going to cost?

>>982474 (OP)

>probably over $500 motherboard

>comes with an embedded CoC linux & pajeet code

gnoo

>>982553

>ELINAR

>The AI Miner enables any organization to detect privacy data (for example data under EU GDPR) and then you can take action on it.

>Easy-to-use solution for GDPR or other data discovery.

>Train artificial intelligence using your own data so that AI is able to detect privacy data unique for each customers’ way of doing business.

>Makes unstructured data simple to access and utilize.

>The GDPR solution has been recognized and proven by IBM -- We were selected as a Wildcard Champion of the global IBM Watson Build challenge, and were one of eight companies at the challenge finals.

>https://www.elinar.com/artificial-intelligence/gdpr-solution-ai/

BOTNET

>>982523

How would you be certifying a system to TEMPEST specs anyway without access to classified documents and an EM test chamber? Simply adding a GPU or swapping a motherboard in an existing SDIP-27 Level A system is enough to compromize its shielding. A DIY solution might be OK if it's completely air-tight though, otherwise you need to know exactly what wavelength any emanations might be so you can get your ventilation mesh the right size.

>>982554

They're open to forks and are working on replacing the existing chain of boot-loaders with Coreboot.

What is the best GFX card to pair with POWER9 CPUs if someone is wanting to build a system for day to day desktop use and encoding video?

>>982578

Vega 64 if you want FOSS drivers.

Are their any suitable cards if the only demands are:

>performance comparable with Intel HD graphics

>Open enough to roll your own drivers if needed.

>Cheap and passively cooled

>>982579

>Vega 64

Are they ever going to come down in price? Why are these cards still so expensive now that it isn't possible to mine bitcoin with them? Are the jews really still taking advantage of all the faggots that think they'll become rich by buying $800 GPUs?

>>982523

I'm pretty sure the psu is the only part of a modern computer you can actually build.

>>982574

I do know that some of them online (pretty sure cryptome), as I needed access to them for what I was doing in Iraq. So ofc I had to find a leaked document or 2 in order to actually do my job properly. That's Army logistics for you.

>>982589

Right now getting GPUs to work requires a lot a of tinkering.

https://wiki.raptorcs.com/wiki/Talos_II/Hardware_Compatibility_List#AMD

For cheap and passively cooled the R5 230 is suitable.

>>982593

The type of memory it uses is still in very high demand so they won't come down to a reasonable price until at least 2020. The price has nothing to do with crypto currency, DDR4 is also still very high and is will stay that way for awhile.

So where's the specs?

>>982554

>>comes with an embedded CoC linux & pajeet code

Until someone comes up with a way to fund some boards to donate to Open/NetBSD you will be forever CoCked with just linux.

MicroATX is pretty much an absolute waste of time compared to standard

If it's not MiniITX you might as well just go standard because there's just no point to it

>>982589

The on-board AST2500 on the Talos II pretty much fits that, but I'm not sure if

a) this new board has it

b) if you can get them as PCIe cards

>>982843

>It's the cheapest form factor they could put a socket on

maybe i'm getting confused here but we're talking about the size of the motherboard ?

MicroATX is basically just annoying, there aren't many cases you can put them in. I just don't see the point when you can just release a standard sized motherboard, with some more expansion options. If you don't take the step of going to mini-itx you might as well stay standard is all i'm saying but if they seriously save money somehow by sticking with microATX, I mean, ok I guess ? But how much would they be saving ? $20 a board they markup by how much ?

>>982610

I don't buy these RAM pricing shenanigans for a second. There's price fixing going on.

>>982877

Of course, China is suing RAM manufacturers for price fixing.

>>982549

>>982536

>>982553

Just use a PCIe extender from the PCIe 8x slot and plug it into a PCIe 16x to M.2 card, I got one for a dozen doubloons. It might come with chinese botnet though.

>>982863

>But how much would they be saving ? $20 a board they markup by how much ?

For the quantities they are looking at they are probably saving at least $100 per PCB compared to a full sized ATX mobo.

>>982978

It's going to take a decade AT LEAST before RISC-V has the microarch to compete with Intel and AMD.

I will be happy to support RISC-V but you're going to be waiting a long time with no freedom in the mean time(if ever).

>>982988

Don't buy those chips. Someone will make a kickstarter or whatever to fund a fully open chip.

>>982997

Also, if you can into FPGA, you can probably outright make yourself a RISC-V today. Someone may already have, in fact.

>>982978

I'm OP and as far as I know this is the first time a modern POWER type CPU has been made affordable to the general public. There hasn't been a computer encumbered by proprietary code and powerful enough for basic desktop use for almost a decade so this is big news.

Ever since UEFI was introduced with locked bootloaders I've been sick of the x86 platform and am a bit hyped for any suitable replacement. I'm glad RISC-V is around but it'll take years before anything useful comes out of it for the average user, until then POWER and some older intel and AMD chips are all we have.

>>983243

If you don't know the difference or that there is a difference between UEFI and EFI then don't waste people's time by posting here.

>>982554

I bet it's probably $1k for motherboard and CPU considering they still haven't mentioned the price. If I weren't NEET, I'd get one yesterday.

>>983231

If it is secure boot you dislike, you can turn it off or put it under your control by replacing the Platform Key. I'm not a big fan of them replacing my good old BIOS, but UEFI is not nearly as bad as secure boot on ARM, where they use it to lock you out of your smartphone. I figure UEFI could be used to do this, but so far I haven't heard of it.

>>983262

Why wouldn't you buy instead some FPGA gear and learn to make your own hardware? That's the only way we're going to get out of this mess. With actually open hardware, not by trusting yet another vendor.

>>983231

holy fuck thats a sexy radeon pro

and damn that slim

>>983304

That's what you get for buying a smartphone though. With ARM dev boards, you're in control and don't need to use secureboot. Also pick one that works with U-boot, not that RPi type stuff that needs Broadcom firmware blob to start the ARM cores.

>>983304

I've only had to deal with UEFI when installing older versions of Windows for other people and it turns what was once an easy simple process into a nightmare. Thanks to UEFI telling someone who's somewhat interested in GNU/Linux to try it out on their own computer has become impossible, instead of hitting a key at the boot screen to get to a menu of boot options they need to follow a complicated tutorial just to run a live distro.

>>983306

I want a real OS and programs to run on it not an expensive hobby computer that can barely do anything useful. And a vendor that has 'fully open computer' as a selling point while trying to get RYF certification from the FSF is anything but 'another vendor'.

And as an ex PPC user that only uses free software know that 99% of the programs I use will work fine on this platform. If a new architecture comes around that's better or POWER starts being locked down it'll be easy to switch.

power9 also infected with spectre and meltdown.

>>982474 (OP)

>Hopefully the SATA ports in the corner will be optional for those that don't want any proprietary code in their computer.

What do you mean by this? Do the SATA ports require proprietary firmware?

>>983340

My issue with ARM is the non free video drivers most SoCs use. That board uses a Mali type GPU.

https://redmine.replicant.us/projects/replicant/wiki/GraphicsResearch#ARM-Mali-and-Lima

>>983346

Lima project is still active, here's the site:

https://gitlab.freedesktop.org/lima/web/wikis/home

The Freescale i.MX6 SoC already have open source GPU driver.

>>982523

>open source

>switched-mode power supply

What kind of ricer chinesium have you been fucking buying?

>>982549

>broadcom wifi chip

into the trash it goes

>>983310

The only smartphones I have are peoples old discarded ones, which I have unlocked. They are just lying there waiting for a project.

>>983328

True. It is a pain in the ass and legacy BIOS mode is planned to be removed by 2020 *sigh*.

>>982474 (OP)

>Hopefully the SATA ports in the corner will be optional for those that don't want any proprietary code in their computer.

>>983342

>What do you mean by this? Do the SATA ports require proprietary firmware?

The Talos II had (optional) Microsemi SAS controllers on them which required non-free firmware, >>982549 clearly has a Marvel SATA controller. The Marvell SATA chips don't require firmware.

>>983003

>just buy an FPGA and implement a CPU on it for your desktop!!!

Unless your standard for acceptable desktop CPU performance is a low end Pentium III you are a fucking idiot.

>>983262

I would be willing to bet its under $1000 for the mobo and a 4 core CPU.

>>982997

>>983003

How fucking stupid are you? Is this some high level Intel shilling?

High performance FPGAs have proprietary ARM cores and decoder cores, if you run a RISC-V CPU off an FPGA you're basically running a modified ARM CPU unless you want the slowest CPU in the west.

>Kickstarter!

Oh yeah bro, billion dollar goal here we come!

>>983306

Do you know how to wire PCIE cards and SATA/SAS hard drives to FPGA? I don't.

>>983340

ARM and current RISC-V are too slow. Even the Caviums which cost $2k are slower in almost every benchmark compared to similarly priced Talos II.

Looks really nice. We're being spoiled with this, honestly. I fully support these guys.

>>982978

You're retarded. The only thing free about RISC-V is the ISA. The chips SiFive make are 100% proprietary. OpenPower is more free than RISC-V.

>>983252

There's only a difference of you want to try really hard to be a contrarian autist

>>983306

Do you realize just how much fucking work it would be to design your own processor architecture, write compilers for it, port thousands of programs to your architecture, and of course there's no telling whether you need to make custom peripherals too because obviously any peripherals and computer parts you can buy weren't designed for your custom architecture so there's potentially years' worth of engineering required on that part, oh and also you're the sole developer and maintainer for all of this so 100% of the time and money being poured into this project is yours. There's a reason why nobody actually does this.

reminder that power is not open

<oy vey yes it is

link the schematics

>>983487

>there's no telling whether you need to make custom peripherals too because obviously any peripherals and computer parts you can buy weren't designed for your custom architecture

Most of the time peripherals are just memory mapped so as long as you use a common interconnect such as AXI (which was developed by ARM and made free to use a while back) then existing peripheral designs can be added to the chip and just work.

You are right about everything else though, even tech-litterate people give little thought to the countless man-hours it takes to design and produce even just the CPU in their computer.

>>983465

>Do you know how to wire PCIE cards and SATA/SAS hard drives to FPGA?

I am not going to say its easy since I literally do that kind of stuff as a job but its not as hard as you probably think. Its just a case of if its worth it protip: it isnt.

>>983469

>The chips SiFive make are 100% proprietary.

They aren't 100% proprietary, SiFive put the Chisel code for their CPU up on github but thats only a small part of the design. Critical parts of their design, including the fucking DDR4 interface, is proprietary which has led to them literally telling the community to reverse engineer the binary init code if they want to make a libre boot option.

Chisel has made designing a CPU a much easier task than it was, but the hard part has always been the peripherals and drivers. If the SiFive guys weren't just a bunch of virtue signaling hipsters they would have created all the other peripherals as well (its not like the standards aren't open), but they didn't because thats a fuckton of work and outside the realm of their CompSci degrees.

>>983453

>High performance FPGAs have proprietary ARM cores and decoder cores

It depends on what type of "FPGA", some are an ARM SoC with an FPGA attached via the AXI bus (such as the Zynq series) while most are just a standalone FPGA.

>>983470

>same BMC chip as the dual socket board and sound which seems like a waste.

In the grand scheme of things those aren't that expensive and add some nice functionality that quite a few people would want.

>>983490

I'm pretty sure this issue has been brought up before here and that someone who inquired about said you need to pay a few thousand dollars to become a member to their org to get the schematics. And that the only reason why they aren't public is because nobody has bothered to post them on the internet not because of NDAs.

I could be wrong but really the schematics not being made public isn't that big an issue for me. If I had to guess it's probably done this way because of vaguely worded laws on the export of technology that has encryption built into it.

What's the point of buying one when Gentoo ppc64le doesn't exist, right now?

>>983557

>Gentoo

Nice meme dude kill yourself, also Power9 CPUs can run in both big and little endian modes.

>>983566

Just because an OS could be made so minimal that it runs amazingly fast on a tiny SBC like the PiZero doesn't mean that it would be useful. Have you ever tried to compile and test kernel modules on such a device? its not a pleasant experience.

You might not need anything more than a single 1GHz ARM core and 512MB of RAM but some of us actually do more than just code hello world. I have written code that takes hours to run its test suite even with 64 threads at 4GHz and will happily eat up tens of GB of RAM, there is no fucking way I am switching to a PiZero even if my system sits mostly idle 99% of the time.

>>983500

The thing is that even if you get the Verilog/VHDL for the Power9 cores its not as if you can verify that is whats actually in the CPU you buy. This is another thing the monkeys masturbating over SiFive seem to forget, just because a bunch of Chisel for a RISC-V CPU core is up on github doesn't mean that's whats in the silicon. The most useful thing you can have is the documentation for the hardware registers and such, because at least then you can write all the software that runs on it.

>>983594

You could just set up a dedicated compilation machine and write a script that sends over files to be compiled with a makefile and you get the executable back once it's done.

I actually prefer to do my actual coding on an SBC, it limits distractions and is absolutely silent.

>>983688

I'd get it at that price, as long as it'd run NetBSD.

>>983690

Talos doesn't seem to be supported by NetBSD so...

>>983566

You're like the kid that tells other kids real ninjas can kill people with a blade of grass. You don't know what you're talking about and you aren't going to use a $5 Pi as your main computer.

>>983566

The blight of software bloat is real but only in consumer products where the upgrade treadmill keeps speeding up to sell newer products. For hardware like POWER that is to be used in super computers and data centers the operating systems and programs that are designed to run on it are often as lean and optimized as possible.

The most expensive thing about running a super computer today isn't the hardware but the energy cost so there's an economic incentive to squeeze out as much computational power as possible per watt. That's the reason why all the biggest super computers have been using GPUs instead of throwing more cores at it.

https://www.livescience.com/18072-rise-titans-exascale-supercomputers-leap-power-hurdle.html

https://cacm.acm.org/news/192296-supercomputings-super-energy-needs-and-what-to-do-about-them/fulltext

>>983594

It may run in big and little endian modes but there's only the Adélie distro that uses BE. And so does FreeBSD which is in the process of being ported to POWER9. Once that gets ported NetBSD is sure to follow.

>>983688

$1k with an 8 core CPU and complete system seems a reasonable price.

>>983594

RISCOS runs on the Pi Zero, and it's useful. Nobody's going to do the kind of stuff you're talking about on it, but that's not what it's for.

>>983725

Well I got news for you. Over the years my main computer has consisted of: Z80-based stuff, Amiga 500, 486DX/33, and then later just faster x86 shits. They were all perfectly useful, but what stands out is how utterly insane the bloatage has become in the past decade or so. Browsers like Firefox & Co. are a prime example. The whole thing sucks, and it's not work to keep going down that trend, just because everyone else is doing in. If you want to jump off a cliff with the rest of them, that's your problem.

>>982521

It can accommodate any fan with enough duct tape.

>>982978

>don't do anything, just wait for Hitler goy

>>>/auschwitz/

>>>/gaschamber/

>>>/oven/

>>983749

>Browsers like Firefox & Co. are a prime example.

It's not the browsers. It's the webappssites they run. Web browsers have been relegated to a platform for DOM/JavaScript applications called for retarded historical reasons "websites", and most of those sites are written by a bunch of niggers and pajeets on crack or worse, by hipsters. ANY browser capable of running that shite is gonna gobble RAM like OP semen.

>>983747

>That's the reason why all the biggest super computers have been using GPUs instead of throwing more cores at it.

Thats not entirely true, in terms of compute per watt manycore CPU based systems outperform GPU based systems and until recently the most powerful supercomputer was a manycore CPU based one. The reason why the US supercomputers use GPUs is because the manycore based ones are all Chinese and Japanese designs and Nvidia is US based.

Not that the trend is going to last since Xilinx just released their new FPGA compute platform which BTFOs even the best Nvidia GPUs by a significant margin.

>It may run in big and little endian modes but there's only the Adélie distro that uses BE.

I am pretty sure the default PPC Gentoo is BE since it was made for the Power7 CPUs which only supported BE. Typically PPC64 refers to BE while PPC64LE specifies LE.

>$1k with an 8 core CPU and complete system seems a reasonable price.

When you consider that's 32 threads while the cheapest 32 thread x86 CPU is $700 alone its a pretty good deal.

>>983690

>remote compilation/execution

I have used it in the past and always found it cumbersome, even with plugins to IDEs which automate the process.

>I actually prefer to do my actual coding on an SBC, it limits distractions and is absolutely silent.

Well if that is what works for you then stick with it, everyone is different.

Nothing new to report asides for an entry on their wiki, only things on there that hasn't been mentioned here are

>Three Gigabit Ethernet ports provided by a Broadcom BCM5719 chip.

Hopefully it'll support Ethernet bonding

>CPUs with more than eight cores are not supported due to power delivery limitations.

>CPUs with eight cores will have a slightly lower all-core-turbo frequency. Nominal frequency is unaffected.

https://wiki.raptorcs.com/wiki/Blackbird

>>983780

>When the /pol/bot malfunctions

IBM has an R&D Center in Israel

>>984098

It's both though. Firefox is a huge codebase, bigger than the OpenBSD kernel. Starting it up on my computer, it sucks up ~ 160 MB, without even going anywhere (browser homepage is set to my home directory). On top of that it does shit in the background, or tries to anyway (since I have their site blocked):

1539000327965 addons.productaddons WARN Failed downloading XML, status: 0, reason: error

In comparison, I have a Links browser process that's about half that size, even though it's got a lot of html and big images cached in memory. Xorg itself is also about half that size. And everything else running on my machine uses much less memory.

I just started Netsurf also to compare. This browser only uses 30 MB, and starts up instantly compared to Firefox who's doing (or trying to do) who knows what in the background.

I wanna run OpenBSD on one but there isn't a PPC64 port reeeeee

>>984294

Run it in qemu on a minimal netbsd install.

>>984258

No, it's the websites. Both of the browsers you compared Firefox to have zero support for Javascript

>>984405

>it comes out four fucking years from now

You are confusing the Talos I with the Talos II, the first one was Power8 and never reached its funding goal while the second was Power9 and shipped on time for most orders.

>nerfed form

I have no fucking idea what you are referring to.

>>984431

Netsurf doesn't have "zero support for javascript".

FPGAs can run OSs. https://www.youtube.com/watch?v=zYL42WQfgGI

>>984891

Hmm, passive cooling?

Neat, I might just make a HTPC with this.

>>985783

Why? You can get Ryzen ITX boards, and there are some incredibly tiny cases in that form factor now. I personally would want a Talos as a workstation, not a HTPC. Here is a video detailing what I mean:

https://youtube.com/watch?v=oa9xbnIdrFw

Shovel it down then.

>>983500

>Associate/Academic membership $0

Troll harder son

>>986477

>x86 is half-way secure

Who the fuck made this figure?

>>986588

Too be fair it is far more audited than other archs

>>986588

To be fair the figure was made before meltdown, spector and foreshadow.

>>986616

2 of those effect other archs as well, it's not necesarilly the architecture at fault, just a really poor implementation of one of its feature-sets. There are probably millions of ARM chips out there with silicon bugs that we just don't bother testing for.

>>986477

Zaius/BarreleyeG2 machines also use LaGrange, so there is some alternative to IBM

A Google+Rackspace machine would be more open than an IBM anyway, though I would rather trust Raptor

>>986655

>A Google+Rackspace machine would be more open than an IBM anyway,

Anything Google is going to be equally as open as x86-64 and IBM PPC, that is to say, psudo-open (ie. "our drivers are all open source guys!") but still relying on firmware blobs

This'll sound retarded but I'm pretty excited about this board. 8 cores, 32GB ram, easy way to permanently switch to *nix as there is no other option. I have a nice silverstone matx case from a few years ago that has somewhat of a 1970s stereo aesthetic. Assuming I can figure out a decent video editing solution to mess around with and something equivalent to photoshop, i'll say goodbye to windows 7 and windows in general forever.

>>989480

If it'd come down in price some more I'd make the switch too. Going to wait and hope it sees adoption with richfags. You're only really losing games and some speed early-on for intensive things like video encoding on the CPU. If I had the money I would have just jumped on the dual-cpu board.

>>982997

If the kickstarter for Talos I failed, mind you that was just to make a mainboard, in what parallel universe are you imagining that < 150 nm process chip will actually get funded?

I'd be that by the time you actually gathered enough money, you'd have to readjust for inflation and keep going.

>>984294

Yeah, `macppc` is probably the best starting point, since the discontinued `socppc` port may have embedded-specific (Book III-E) stuff that's not even in the Power ISA anymore.

Appendix:

macppc = PowerPC Big Endian 32 bit, New World Macs (OpenFirmware)

I'm not positive, but Power ISA 3 either removed entirely, or merged with Book III-S, Book III-E.

>>992825

Honestly, the only thing is fear, is lack of Altivec support for some software. x264 (and proably rav1e) is okay, but vapoursynth filters and some emulators like rpcs3 aren't.

Still waiting for it; maybe waiting for power10 would be a more prudent move.

>>989562

The only games you're losing are newer ones that aren't worth your time. All the classics from the late 1990's to mid 2000's are getting FLOSS engine replacements as most of the modding communities have given up on trying to expand old engines or keep them running on newer hardware/OSes. Those engine reimplementations should work on POWER if ported correctly and older games will work fine through software emulation.

The real issue for gaming will be the poor state of GPU drivers, IBM is stuck with a partnership with Nvidia and AMD is making their own x86 chips so won't be supporting an alternative CPU architecture.

>>992825

I haven't been on /g/ in years, are their threads on POWER worth visiting through the archive?

>>992851

>>992897

Looking through:

https://wiki.raptorcs.com/wiki/Troubleshooting/GPU

We have the following config necessary:

If you have a discrete GPU:

* are you sure you want to add GPU blob to PNOR?

* is your kernel updated?

* Xorg is fucked; either 1) disable 2D GPU 2) ply with xorg.conf

* make sure

* may need param `amdgpu.dc=0`

If only 2D GPU, disable 3D api calls:

* disable GLAMOR

* disable OpenGL on KDE

Pretty mild issues for suddenly going non-x86 with hardware that's mainly used on x86.

Especially surprising to see nothing specific to running Big Endian or Little Endian. Not so on Chromium or Firefox:

https://wiki.raptorcs.com/wiki/Porting/Chromium/BE

https://wiki.raptorcs.com/wiki/Porting/Firefox

>>992894

>>992851

Found the archive of it:

https://archived.moe/g/thread/68035607

https://www.raptorcs.com/

>Basic Talos™ II Bundle (Single CPU)

>TL2B01

>Basic Talos II Bundle (Single CPU)

>Order online for $2,499.00

Wish I could afford :(

Guys. If I set up a gofundme do you think you could buy me a Talos™ II Entry-Level Developer System for $2,149.99?

https://www.raptorcs.com/content/TLSDS3/intro.html

>>992911

GPU issues in Linux are already bad enough but when you add a niche platform to the mix there's only a handful of people filling bug reports or bothering to fix it. It also increases the price of the system quite a bit since you've got to use a newer card that's popular with other users instead of going with an old video card you already have.

>>992913

That wasn't as terrible as I was expecting. With as many replies in single day that this thread has in a month the signal to noise ratio is pretty poor.

>>993015

I see they've improved their entry level developer system, it used to come with a slightly defective CPU that couldn't do virtualization and was missing some Spectre or Meltdown protection.

They've already announced that the blackbird will be shipping in Q1 2019 so the price for preorders will be coming shortly.

>>993020

You're thinking of the Special Developer Systems:

https://secure.raptorcs.com/content/TLSDS1/intro.html

https://secure.raptorcs.com/content/TLSDS2/intro.html

4 core DD2.1 stepping - $1600

16 core DD2.1 stepping - $2000

That's $550 less in exchange for a slightly defective CPU, seems fair to me

>>982474 (OP)

Video of the Blackbird announcement is online:

https://www.youtube.com/watch?v=Tvey2Nsc3oY

let's get those view numbers up!

>>993100

He's right about the GPU being a blackbox that shouldn't be trusted but there isn't anything that can change this in the foreseeable future.

>Interestingly, it transpires that there may be contractual obstacles to this. According to posts by an AMD employee on the Phoronix forums (see linked post and all following posts by the same user), AMD is contractually obligated not to reveal details about the internals of their GPUs, because those GPUs implement DRM and security by obscurity and the tiresomeness of reverse engineering is apparently part of what Hollywood's counting on. In other words, AMD may be contractually precluded from revealing the internals of a DRM-free GPU (to enable open source firmware to be developed, for example), because that DRM-free GPU would be so closely related, in terms of internal architecture, to a DRM-implementing GPU that it would provide useful information to people trying to break that DRM.

>It's basically stated that because of the above contractual obligations, it would be necessary for a DRM-free but owner-controllable design to differ significantly in its internals from a DRM'd design simply for the sake of being different, so that it doesn't reveal information about the DRM'd design.

>Thus, for AMD to make CPUs or GPUs which are DRM-free but owner controllable would be likely to require substantial engineering R&D, for no other reason but than to satisfy these gratutitous contractual requirements. Intel/AMD are unlikely to do this unless they perceive sufficient market demand.

https://www.devever.net/~hl/intelme

There were a few attempts at creating an open GPU over the last decade but they've all fizzled out. This is huge problem in the world of free and open computing with no real solution at the moment.

>>993114

>https://www.devever.net/~hl/intelme

Fun fact: that's the same guy who is reverse engineering the NIC:

https://wiki.raptorcs.com/wiki/Project_Ortega

>>992851

>The only games you're losing are newer ones that aren't worth your time.

sour_grapes.png

Also it's hilarious how quickly freetards turn their back to an incredibly popular use case that moves billions each year in both hardware and software.

>>993297

popular != worth your time

>>993297

> freetards turn their back

* Unreal 4 is already proven to work on POWER9

* The next thing to do to encourage development would be to increase the userbase on Power

* In the meantime their are open source games you can play

Most of us are not saying the grapes are sour, we're saying that we can't reach them yet and there's no point starving and staring at grapes you can't reach, when you could be eating other fruit that is accessible.

If you don't like modern AAA games, like >>992851 , it just makes the switch to Power all the easier

>>993324

>popular != worth your time

popular == worth many people's time

>>993330

You make reasonable points, 992851 did not.

Saying "modern vidya support is almost non-existent, but at least UE4 should work" is an ok assessment, saying "you don't need modern gaming support" is laughable.

I see way too much of the latter in FOSS.

>>993671

That's a terrible argument. There's a tons of unpopular things that are worth spending time on. And there are tons of people with misplaced priorities.

>>993687

>There's a tons of unpopular things that are worth spending time on.

Spending YOUR time on them, maybe.

But because they are unpopular, you won't find many willing to help you (or the money to pay someone to do it).

And with less manpower it's harder to get anything done, see the thousands of one-maintainer forks that die off.

>And there are tons of people with misplaced priorities.

Everyone but you, I suppose.

>>993671

>saying "you don't need modern gaming support" is laughable.

Well it's the truth. For one thing no modern major publisher is supporting Linux and those smaller ones that do aren't going to support a niche CPU architecture, best case scenario they give the code to the public and let them deal with porting it. If you're interested in a POWER9 workstation or desktop you clearly don't need or want modern proprietary games. For classic gaming you have thousands of titles available with a few newer FLOSS games being developed for it.

The argument that the free software movement hasn't adequately supported this market doesn't belong in this thread. I agree with it, especially when it comes to getting younger programmers involved.

>https://github.com/openbmc

Is there a way to not use this fucking shitware with network access right in? Because this is a deal breaker for me.

>>994154

Also,

>https://github.com/openbmc/docs

>Technical Steering Committee

>Brad Bishop (chair), IBM

>Nancy Yuen, Google

>Sai Dasari, Facebook

>James Mihm, Intel

>Ali Larijani, Microsoft

Trust us, go.. guys!

>>994154

>>994156

I don't see any blobs in there, do you? If you're going to not use something based on what company has been working on it, you should probably stop using computers altogether.

>>993330

>* Unreal 4 is already proven to work on POWER9

And your point is ?

I acknowledge that the UE is visually nice but the "open" core license is completely proprietary, they can even legally force you to quit the quit that you are working one.

>>994161

An always connected systemd distribution running some shitware made by pajeets under the yoke of PRISM frontends like Microsoft is okay? Pretty high standards you have.

>>12361106

I'm pretty sure the always connected thing is only an issue if you're using the ethernet jack that's connected to the BMC chip. Running a standard Linux OS that's created and overseen by the Linux Foundation under the main OS which is also probably going to be Linux isn't going to be an issue for most people.

At least with this platform an alternative can be created if there's enough demand or maybe by porting U-boot/coreboot. Also the guy who runs Raptor Engineering tried to make a free firmware for these machines a few years ago and wasn't able to

https://www.phoronix.com/scan.php?page=news_item&px=Librecore-Formation

>>994167

D-BUS is not systemd. It sounds more like you have an issue with Linux in general, in which case you should definitely stay away from this machine and stick with something more BSD friendly like Intel with IME.

>>994276

People buying Talos workstations aren't "most people", though.

>I'm pretty sure the always connected thing is only an issue if you're using the ethernet jack that's connected to the BMC chip.

A source for this would be nice, because their wiki lets me think this isn't the case.

>>994282

>The OpenBMC stack uses technologies such as Yocto, OpenEmbedded, systemd, and D-Bus to allow easy customization for your server platform.

It sounds like you should learn to read.

Honestly, I'm sure I'm not the only one for who this is a sine qua non condition to buy it, because this is a lot more pozz than using something like stock Ubuntu.

The best reason to get this board is because the motherboard firmware has no blobs.

The CPU architecture is still not OSH or FOSH, and the OS still requires a few proprietary drivers, but the fact that the mobo firmware has no blobs is great.

The next step is to remove the NIC firmware blobs from the Linux kernel, which is the basis for the broadcom reverse engineering competition.

>>994154

Disable BMC access via the motherboard firmware that is not a BIOS but is similar.

>>994387

If you're sure this is possible, then it's okay; not perfect, but bearable.

>>994387

>OS still requires a few proprietary drivers

Please stop trying to FUD

It has always been made clear that all the drivers are open source.

The "blobs" we talk about for the NIC, or optionally the SAS controller or AMD GPU, are blobs that are executed on that device, NOT IN THE KERNEL, and NOT ON THE CPU.

Since POWER9 does have IOMMU, the NIC can't DMA sensitive memory even if it's firmware was compromised

New news from Phrononix now that the guy who runs the site has a Talos 2 machine with dual 22 core CPUs. On their forums the CEO of Raptor Computers said a LaGrange based computer is in the works, the ones they offer now use smaller Sforza CPUs.

The only real benefit I can see would be twice as many memory channels per CPU for 2tb max which is what AMD's Epyc already supports.

>>994941

Oh and Intel just announced new CPUs with 12 memory channels, right now their max is 1.5TB.

https://arstechnica.com/gadgets/2018/11/intel-announces-cascade-lake-xeons-48-cores-and-12-channel-memory-per-socket/

>>983409

>>983262

$875 for the Blackbird motherboard alone. I hate it when I'm so right that I'm wrong.

https://www.strawpoll.me/16840668

>>997546

including a 8-core cpu, taxes and shipping it would cost $1822 or €1610.

more than i could reasonably afford, too bad.

Wait for POWER10 for:

* More vuln fixes; especially needed seeing how they're coming in numbers, right now.

* Better performances to try to alleviate a little the fact that a lot of software doesn't have the same Altivec code they have for SSE/AVX.

Question: Is it true that the Talos chips are all fabbed in America? Including the motherboard assembly? I want to avoid Chinese trash as much as possible

>>997644

I could have sworn I saw made in the USA being tossed around, but I can't find anything to back that up. I would email the company.

>>997645

IIRC the CEO of Raptor said it in a tweet that the boards and CPU are all American made after the whole "Chinese Spy chip" hoax fiasco

>>997661

>>997645

Found the tweet

http://archive.is/JL5GV'

Wouldn't hurt to email anyways to verify. I really want to save up for one now though

There's been an update posted to the PowerPC Notebook project but it's pretty lackluster.

>>997546

Browsing through the tweets where that potential price was mentioned it looks like Wendell from Level1Tech is going to be getting one for testing during the upcoming Thanksgiving holidays in America.

https://twitter.com/RaptorCompSys/status/1062551451220787200

>>997626

You'll have to wait until mid or late 2021 best case scenario and it's going to be using DDR5 so expect to spend a small fortune on memory again. When Raptor releases their POWER10 boards it'll be the perfect time to grab an older generation model from people that have the money to upgrade like the guy who owns this beast >>993101 .

>>986079

white men use the terminal

>>992838

>buying $2000+ hardware to run the redhat botnet

>>997596

Are you in EU? If so, express your interest to vikings.net. They are willing to become a reseller for Raptor in EU, but they need volume to make it tenable.

>>994402

It is an almost miraculous improvement at this point, I want to buy one meself

>>994406

>Please stop trying to FUD

https://www.raptorcs.com/TALOSII/nic_fw_contest.php

So you were right, it's not the drivers that are proprietary, but the firmware that interfaces with the driver.

Apologies anonymous.

>>997837

The only non-systemD distros at the moment are Gentoo and FreeBSD. It's not easy to avoid corporate bloat from IBM products.

>>992894

>tfw an anon posts a pic of your computer

Weird feeling, made me realize its nearly been 6 months already.

>>994387

>Disable BMC access via the motherboard firmware that is not a BIOS but is similar.

The BMC only has access to the top RJ45 connector, you can just use the bottom one if you don't want to allow the BMC access to the network.

You can't really disable the BMC itself anyway since its what manages the power, in fact when you first connect the PSU to the mains and turn it on you have to wait a few minutes for the BMC to boot before you can power on the system.

>>982978

SiFive claimed to be open source, yet kept everyone in the dark on their proprietary pre-reset-vector code.

https://www.bunniestudios.com/blog/?p=5127

RISC-V has no truly FOSS option unless someone makes a physical CPU, and that requires significant capital to do, just to be able to do a bare minimum wafer run. Ultimately there will come one but it will be significantly more expensive.

Power8/9 is FOSS and is a great move away from the horrible x86 cancer.

Is this possible to use different ventirads, or at least different fans?

>>999351

>bitching about a server grade mobo having to boot its BMC before you can do anything

Welcome to the enterprise world.

There's a new post on Talospace, comparing x86 emulation performance between QEMU and HQEMU (which does some JIT compilation via LLVM)

https://www.talospace.com/2018/11/sometimes-its-necessary-running-x8664.html

>>983490

I mean if you cannot use kicad to diy your own psu .... jsut kys fag

>>983490

>muh schematics

open hardware != open software

>>982474 (OP)

If I'm leaving x86 for good its going to be for RISC-V

Fuck IBM, they are part of the problem

>>997596

its fun how everyone is suddenly a poorfag in these threads but in other threads they say that they have high paying jobs.

>>999813

Even when I had a job that paid more than the US median income, I didn't spend anymore money than absolutely necessary. Computers get outdated very quickly, and are not a good investment unless you're buying something that's needed to make money.

>>1000007

>computers get outdated

Normalcattle thinking. Why would it get outdated? Because you use a retarded OS that can't run unless it has 64GB of RAM? A computer gets outdated when you're limited by it (as opposed to feeling limited because you can't run your AAA gaymes).

We should be getting a review from Level1Techs and Anandtech by the looks of it:

https://twitter.com/tekwendell/status/1062566741048549376

https://twitter.com/IanCutress/status/1065006007162675203

>>999813

It’s almost as if there were multiple distinct posters on this board.

Anyway, even if you earn plenty of money, it might be tied up if you’re married and have children and a mortgage to pay.

>>1000011

>as opposed to feeling limited because you can't run your AAA gaymes

"your use case is wrong", lmao loonix fags

>>1000074

>I'm compelled by a mysterious force to always buy the latest goyware

Retarded hedonist.

>>1000011

They always get outdated at some point. I don't play modern games, but I need to use the bloated botnet-compliant web browser in order to do stuff like online banking, manage investments, or even just buy stuff online. My old Thinkpad T30 with 512 MB RAM wasn't cutting it anymore for that. Now I have a cheap ARM board that does everything I need. I didn't want another x86 anyway, and the Talos isn't worth the cost to me.

>>1000098

Fair enough, but having the option for an OpenPOWER system, I would never trust an ARM box as my main system.

In terms of price and form factor though, there are may situations where OpenPOWER isn't an option, so then I'd say, go with the most owner-controlled system you can get.

>>1000098

They don't really, though. Especially now that video, image and audio codec performances have basically peaked (I'm talking about transparent bitrates, of course) with av1, avif/heif and opus/flac (maybe tak when it's open sourced).

>>983490

ISA: https://ibm.ent.box.com/s/1hzcwkwf8rbju5h9iyf44wm94amnlcrv

processor datasheets(for the sforza types that talos 2 uses): https://ibm.ent.box.com/s/g8cscr0du396455j8vk1o3yuyl6d5nmy

Although they have 10x more info on how to use their processor than either intel or amd, you're right about the hardware schematics, so i won't be buying one.

>>1000597

In one of the OpenPOWER formware talks on YouTube they talk about how the chip bootup firmware (hostboot) is designed to handle multiple core failure, and to boot only the cores it has twsted to be functional. Haven't heard of x86 chips doing that.

For uarch though, you'd probably need to ask someone who writes assembly for it, maybe the TenFourFox devs? Don't think you'll find them here though.

>>1000781

Aren't tri core CPUs just quad cores with a failed core? So there'd have to be something telling it to only handle those cores

>>1000597

Replacing the old spinning rust disc with SSD is gona bring benefits in seek times and random reads regardless of the version of SATA being used. And it would work better as swap then the mechanical disc.

But that wont help you that much anyway.

I doubt the pcb has the required address lines for adding larger DRAM chips. Did some folks upgrade the flash in the original eeePC and in some macbooks?

Even if you could do it successfully.

Interesting idea tho.

I have some mini routers I should upgrade the SDRAM and flash in.

>>1000597

>consoles

I read somewhere that console chips are modified as to disable every security feature in favour of speed, since they are closed ecosystems and malware is not a concern

Preorders are online.

Looks like it's $799 for the board through cyber monday

https://secure.raptorcs.com/content/BK1MB1/intro.html

>>1000941

The bundle with 4 core CPU is $999.99

https://secure.raptorcs.com/content/BK1B01/intro.html

>>1000943

Damn that is mighty tempting.

>>1000996

>Anyone know what the white connector inside the blank space next to the ATX power is for?

Its a 4 pin CPU power connector

Ignore what I said here >>1000998, I misread the question.

Its probably a Flexver connector, there is one on the TalosII motherboards as well.

https://www.crowdsupply.com/raptor-computing-systems/talos-secure-workstation/updates/talos-fpga-functions-and-responsibilities-part-2

>>1001000

One of them could be, judging by the checkout options it could also be an OCuLink port which would enable me to get an OCuLink to U.2 Cable for $50 bucks, meaning I can keep an nvme ssd while still having the x8 pcie free.

>>1000798

I mean it does this check on every full boot when hostboot runs.

If inbetween reboots you popped out your four core and managed to scrape the die to damage a particular core, on next boot it will detect that, disable the core, and tell Linux about it via OPAL.

Welcome to enterprise grade my friend.

>>1000222

That POWER stuff doesn't buy me anything more than my cheap ARM Cortex-A7 does. I already have open boot firmware via u-boot, and I'm not forced to use any blobs, neither in firmware nor drivers, in order to operate the board. The only way I could justify spending more money is if I needed more performance, but this little A20 dual core does everything I need for now. I think the Talos is more a niche market for people who need servers or high performance workstation. I'm not interested in any of that stuff.

>>1001160

Uboot doesn't need any blobs for the A7? Only a few days ago I was reading about Librem5's crazy loophole to get RYF certification despite needing a blob for DRAM PHY initialization

>>1001171

Every ARM board is different. I don't know about the one you're talking about, but the RPi is well-known to need a closed firmware blob in the GPU in order to even start the ARM cores. However, there are much more open ones that need no blobs at all to run. Here's a starting point:

https://www.fsf.org/resources/hw/single-board-computers

And here for sunxi (Allwinnner SoC's) specifically:

http://linux-sunxi.org/Buying_guide

>>1000943

>>982474 (OP)

8-core bundle now available, this one includes heatsink/fan assembly.

$1330

https://raptorcs.com/content/BK1B02/intro.html

Apparently they unbundled the heatsink for the four core bundle because it can work with the 2U fanless heatsink if there is enough airflow in the case

https://www.twitter.com/RaptorCompSys/status/1066138653724946432

>>1001507

Margins on that would be pretty low, and Raptor doesn't (or at least claims not to) have enough profit margin of their own to be able to give deep discounts purely for _resellers_,

https://twitter.com/RaptorCompSys/status/1066405150024232960

They have always been open to volume discounts, however.

>>982877

kill yourself retard faggot

>>1001507

>>1001539

And I doubt Sys76 would actually agree to that.

Purism, would actually care, but RCS was shitting on them about their RYF proprietary code loophole for DRAM PHY initialization, so there are probably still hard feelings there.

In my mind:

Raptor > Purism > Sys76 > Chromebooks (coreboot) > other "Linux hardware" brands

Then there are the RYF companies that barely do anything like:

Vikings, Technoethical, Minifree

Although Minifree did pay Raptor to port coreboot to the ASUS D16

…and then asked Raptor to port the D8 and refused to pay after the job was done.

>I paid 100% of the KGPE-D16 contract with was 75k USD I did not pay the KCMA-D8 contract which was 15k.

https://www.phoronix.com/scan.php?page=news_item&px=Librecore-Formation

So I don't blame Raptor for being really cautious about a potential business deal with another RYF company like Vikings; they've been burned in the past to the tune of $15,000 - although Minifree claims this was just 2 weeks work.

>>1001507

Despite Sytem76's "open" Gigabyte motherboards with proprietary BIOS being borderline fraudulent, I wish Raptor would just ignore them. Slinging mud at the competition is low brow.

>>1001555

>doesn't like banter

>on imageboards

What are you doing, nigger.

>>1001558

This is a business, not anon talking about Chinese cartoons. Let other people rip into System76 as they have already done.

>>982529

>>982539

Kore wa nani?

Just had a funny thought. Are we going to see IBM "blue-ify" Red Hat? Running on this platform, IBM would be back in the workstation biz (even if inadvertently). Not saying that's a good thing or not.

I've been thinking of getting into this platform but before jumping in I'd like to know what software packages I currently have that aren't supported. Is there a way to get a list for powerpc and ppc64 in the debian repository that I could easily compare to my 'dpkg --get-selections > installed-software' list?

>>1001856

The last workstation IBM made was the Power Systems S814 in 2014 so it's unlikely that they're going back into that business.

>>1001872

S814 was still a server, it just had a tower form factor for smaller customers without server racks

IIRC:

S - server

8 - POWER8

1 - 1 socket

4 - 4U

No idea what would automate searching your packages for those without PPC builds, maybe `dose` would help:

https://wiki.debian.org/Arm64Port#Using_dose_to_see_what_is_multiarch-buildable

Well the Blackbird black/cyber friday/monday/whatever sale is over so prices are now at:

900 USD - board

1175 USD - board + 4-core (bundle)

1500 USD - board + 8-core + HSF (bundle)

Sale pricing had been

800 USD - board

1000 USD - board + 4core

1330 USD - board + 8core + HSF

>>1001507

>made in the US

Dropped.

>>1003280

Does the 4-core bundle make any sense without a heat sink? Can I just use a cheap one? Which one?

>>1004166

Just get the one that the 8 core comes with, the passive 2U one is only suitable if you have a reasonably high airflow case. The socket has a somewhat unique mounting arrangement and the heatsink is what provides the clamping force for the CPU similar to the way the LGA3647 sockets work.

>>1003579

this

>made in the US

enjoy ///interdiction///

>>992838

>red hat

the hat knows

>anon@mememachine

>built in telemetry RHEL testing area

>>1003579

>not made in china

Bought. Gas yourself, chink homo.

>>993242

>>994387

>>994406

>>998217

Current progress on Ortega; looking good

https://wiki.raptorcs.com/w/index.php?title=Project_Ortega&diff=1831

>MIPS side

> Reversing: done

>APE side

> Reversing: good progress

Keep in mind that the process is:

<HLandau

1. Study/understand code (Reverse Engineering)

2. Reimplement in C (RTG (Reversed Tigon))

3. Create natural language spec.

<Community

4. Create FLOSS implementation of natural language spec., therefore not derived from any proprietary code

timeline of Phoronix benchmarking:

Nov 27

https://www.phoronix.com/scan.php?page=article&item=power9-x86-servers

Nov 8

https://www.phoronix.com/scan.php?page=article&item=power9-threadripper-core9

June 25

https://www.phoronix.com/scan.php?page=article&item=power9-talos-2

April 4

https://www.phoronix.com/scan.php?page=article&item=power9-epyc-xeon

Fedora 29 is apparently working well apart from GNOME's web-app and browser extension based shell extension installer system.

https://www.talospace.com/2018/12/fedora-29-mini-review-on-talos-ii.html

>>1008365

Who cares about shit distros, though? I want a Gentoo stage3 already.

>>1008385

http://distfiles.gentoo.org/releases/ppc/autobuilds/20181214T111003Z/stage3-ppc64-20181214T111003Z.tar.xz

>>1008680

Well, that's big endian, though. But yeah, it's possible. My only problem is the OpenBMC shit.

>>997837

>buying the entire redhat botnet to make it run on your $2000+ hardware well

t. IBM

>>999339

Fans likely yes, but you're stuck with stock rads for now since nobody makes mounting kits for the Sforza socket.

>>1000838

>I read somewhere that console chips are modified as to disable every security feature in favour of speed, since they are closed ecosystems and malware is not a concern

False. Malware (DRM) is not only a concern, it's a major required feature.

sauce on monitor on op image?

>>1010119

No idea, but the source tweet looks to be

https://mobile.twitter.com/stewartsmith/status/1047444730450251776

and it sounds like it was posted before the official Blackbird announcement, so it’s probably a desktop background set by Raptor, which means you’re likely looking for something with a reuse license. I can’t see a business risking some/any kind of infringement over a desktop background.

It's still scheduled for something like 2020, but POWER10 will be manufactured at Samsung, since Globalfoundries gave up on 7nm

https://fuse.wikichip.org/news/1953/ibm-chooses-samsung-7nm-euv-for-next-gen-power-and-z-microprocessors/

Hopefully we'll be seeing a TalosIII then.

OBS is working on Talos II

One of the wiki guys did some livestreams of misc hacking:

https://www.youtube.com/playlist?list=PLR1sAeBSEVccA1OklGgItEmJSwUPKrO9H

>>982877

There is some price fixing going on but I don't think HBM is affected as much by that as it is the fact that it's expensive to make and thus low in supply.

>>982863

>MicroATX is basically just annoying, there aren't many cases you can put them in

Any case that fits full ATX can fit mATX.

>>1017413

>mATX

heh. I look at that and I can only see:

milliATX

Did Raptor ever release new firmware to fix the DMA issues with graphics cards for the original board?

https://wiki.raptorcs.com/wiki/Talos_II/Hardware_Compatibility_List

>>1022746

This was actually fixed in the amdgpu driver - I believe starting with kernel 4.20 (I'm still running patched 4.17).

>>982988

>We'll all be waiting until we're fucking dead for a fully open source processor.

It's not as bad as you make out. Several fully open source IP cores have been available for a while. That's not the same as a functioning processor, of course, but the LowRISC project is coming along nicely to deliver a fully open SOC. There's a good chance we'll have Raspberry-pi class products within a few years. It will probably be much longer before we have anything in the server/workstation class like POWER9, but we'll get there eventually, it's inevitable.

>>1023141

There already was OpenSPARC

That, um, really took off as you can tell.

My bet is that we'll see plenty of open silicon for Pi-tier machines, but all the high end RISCV designs will be locked down way more than IBM's POWER chips are now.

>>1024844

Although I guess China did make their FeiTang飞腾 chips, so OpenSPARC was used for something; I'm sure you'll get a great response if you ask them about their GPL compliance.