[–]▶ No.915534>>915568 >>915634 >>915648 >>917090 >>922294 [Watch Thread][Show All Posts]
> For those that have been interested in the Talos II POWER-based system that is fully open-source down to the firmware but have been put off by its cost, Raptor Computer Systems today announced the Talos II Lite that is a slightly cut-down version of the Talos II Workstation.
> The Talos II Lite is still a very competent beast of a system and features a single POWER9 CPU socket, EATX chassis, 500W ATX power supply, and is sold as a barebones package. The Talos II Lite motherboard supports up to the 22-core POWER9 CPU, eight DDR4 ECC RAM slots, one PCI Express 4.0 x16 slot, one PCI Express 4.0 x8 slot, dual Gigabit Ethernet, four USB 3.0 ports, and one USB 2.0 port.
> The Talos II Lite is expected to begin shipping in July. For their pre-ordering, the bare chassis (motherboard + PSU + case + heatsink) is $1,399 USD while a 4-core POWER9 CPU is an additional $375, the eight-core CPU $595 USD, 18-core CPU $1375, and the 22-core model is $2575 USD. Besides the base chassis and POWER9 CPU expendature, there is also the DDR4 ECC Registered memory and any other peripherals desires for your libre system.
> While it's still a significant chunk of change, the $1,399 USD bare chassis is quite the markdown compared to the $4,925.00 price-tag on the Talos II Secure Workstation (that price does include two 4-core POWER9 CPUs). More details on the Talos II Lite system at RaptorCS.com. We should be getting remote access soon to the Talos II Lite for some fresh POWER9 comparison benchmarking.
https://www.phoronix.com/scan.php?page=news_item&px=Raptor-Talos-2-Lite
▶ No.915541>>915636 >>915904
There's already an active thread about this meme architecture my totally legitimate poster with absolutely no financial agency to this project
▶ No.915550>>915564 >>915754
reminder it's not libre if the relevant docs showing there's no botnet are behind a $100,000/yr paywall.
▶ No.915564
>>915550
Go back to your int* spinoff
▶ No.915568
>>915534 (OP)
>$1,399
>no CPU
>no RAM
>no GPU
>no storage
Great deal, I'll take 6!
▶ No.915632
So, it's essentially the mainboard, with one socket. It's cheaper, relatively speaking. But still expensive nonetheless.
▶ No.915634>>915918
>>915534 (OP)
>The Talos II Lite is expected to begin shipping in July. For their pre-ordering, the bare chassis (motherboard + PSU + case + heatsink) is $1,399 USD
>$50 psu/case & $10 heatsink
$1340 motherboard.
▶ No.915636
>>915541
If he was CIA, he would be trashing POWER and promoting Intel. Why can't people appreciate technology?
▶ No.915648>>915747
>>915534 (OP)
>post about more accessible version of more free and powerful computing platform
>shills immediately begin trashing it with ravenous, frosting posts
I want the worthless sub-humans to leave.
▶ No.915689>>915726 >>915754 >>915919
Honestly, when seeing the benchmarks with x264, I was too disappointing to think about it.
▶ No.915726
>>915689
You're talking about the set of benchmarks where the only time it lost to a much more expensive system was in x264 encoding and won in all others. I don't think you could be any more obvious and should consider suicide, since you suck at even this job.
▶ No.915747>>915749 >>915754
>>915648
I'm still waiting for the power10 boards, im hoping this turns into at least a sub-community in computers ala raspberry pis and sbc's. Hopefully they can order larger batches and lower the price next time. It's just too much right now.
This is still great, though. The devs of the board are listening. This board, however, is a half-measure to satisfy us/sell something while they work on the dedicated single-cpu board. It's just the talos II with half the components missing ffs.
I've been waiting this long, i can wait a year longer.
▶ No.915749>>915754 >>916186
>>915747
I wouldn't hold your breath for a cheaper POWER board unless you're looking to buy one of these used on ebay.
▶ No.915754>>916038 >>919138
>>915747
I'm excited as well and willing to wait a little while longer for the prices to come down. POWER10 should be here in a few years and hopefully they'll have more stuff aimed at the consumer level. I'm not going Intel or AMD for my next build so the cost isn't that bad when you consider what you're getting for the money.
>>915749
The price will come down with time as it does for everything else. These provide enough benefit that people will buy them for servers. In a few years time you'll be able to grab POWER9 hardware off ebay for a good price but even the new stuff shouldn't be as high as it is at the moment.
>>915550
>implying these docs won't leak to the internet as soon as some autistic anon gets ahold of them.
>>915689
It's still plenty fast and will only get faster once we can start optimizing software for the chip. I find it odd that benchmarks suddenly matter when we're discussing POWER9 but when we're talking about botnet-CPUs armies of shills show up to say the benchmarks aren't relevant. It's fun to see them flip-flop on basic issues like that depending on the day of the week.
Anything is better than x86/64 even if it's slower initially.
▶ No.915865>>915877 >>915892 >>915906 >>916107
So I've been following the Talos II & POWER9 for some time, but I've only just realized today that POWER is RISC. I was under the impression it was CISC. Will this prohibit widespread desktop/workstation adoption, considering CISC (x86) is the current desktop/workstation standard? And what advantages does POWER have over RISC-V, the latter being the ISA that big names (incl Google) are putting their weight behind?
▶ No.915877>>915879
>>915865
Modern x86 is RISC anyway, the CISC instructions are microcoded.
▶ No.915879>>915884 >>915898 >>915921
>>915877
when your processors have a minix subsystem built into them, they are anything but reduced
▶ No.915884
>>915879
After they put it on the die, sure.
▶ No.915892>>915900
>>915865
>And what advantages does POWER have over RISC-V, the latter being the ISA that big names (incl Google) are putting their weight behind?
POWER skillfully convinced all the libre fans that's it's libre hardware while it clearly is not.
▶ No.915894
power also has it's own version of ME, the firmware is just open source, but without the documents on it, which are behind a 100k/year paywall, there's no telling if it's really doing what it says it's doing.
there's literally no difference between this and nuking an intel processor with me cleaner, you get the same level of assurance.
▶ No.915898>>915962
>>915879
Not any different than all ARM systems having a LittleKernel based OS running in them
▶ No.915900>>915906
>>915892
Care to explain how?
▶ No.915904
>>915541
Intel shills like you have been on suicide watch since January. I love it.
▶ No.915906>>923708
>>915865
>the latter being the ISA that big names (incl Google) are putting their weight behind?
Internally Google uses POWER a lot, POWER and RISC-V are two solutions for two problems.
POWER is for high performance where you need your system to be as fast as possible, RISC-V is for applications where you need low power consumption such as phones or IoT devices. Google is pushing RISC-V maturation in order to further its mobile and IoT goals.
>>915900
>Care to explain how?
Hes salty because when he heard 'free' he thought he was going to get all the RTL and such gratis so he could make his own Power9 CPUs. In practice POWER is actually slightly more open than RISC-V since RTL is available if you join the OpenPOWER foundation (the lowest tier is something like $1000 for individuals if I remember correctly) while RTL for production RISC-V chips is proprietary (only stripped down 'trial' RTL is available from companies like SiFive).
▶ No.915918
>>915634
Design and sell your own, faggot.
▶ No.915919
>>915689
>multimedia benchmark with lots of hand-crafted x86 SSE/AVX assembly routines
>representative of anything
▶ No.915944>>916001
>>915921
What's your beef though, why's this make you so mad? You're sageing and reeing but it doesn't make sense unless you're literally an Intel shill.
Explain yourself.
▶ No.915962
>>915898
No shit. Why the fuck are we choosing POWER over ARM and x86 then?
▶ No.916001
>>915944
i hate tripfags and so should you
polite sage for offtopic
▶ No.916038
>>915754
>>implying these docs won't leak to the internet as soon as some autistic anon gets ahold of them.
>claim your shit is libre
>it isn't
>it's fine, someone will leak it
According to your shilling logic, Windows NT is open source as well
▶ No.916050>>916054 >>916056 >>916067 >>916070 >>916072 >>922192
can it run windows? Could I buy this computer and run my everyday applications?
▶ No.916054
>>916050
When did Windows users start browsing /tech/? Don't post here if you aren't using free software.
▶ No.916056
>>916050
You have bigger problems than closed source hardware if you use Windows for your "everyday" computing.
>>916050
Summer is coming.
▶ No.916067
>>916050
>putting botnet software on your non-botnet hardware.
peak degeneracy, why even have such a thought?
▶ No.916070
▶ No.916072>>916156
>>916050
Windows users OUT!!! REEEEE!!
▶ No.916094
DISREGARD THAT I SUCK COCKS
▶ No.916115>>916145
>>916107
I posted retard pepe deliberately
▶ No.916145
>>916115
RISC and CISC have no bearing on a platform's success, which is determined by what (kind of) applications it can run and by its strong points
▶ No.916156>>916181
>>916072
tripfags first REEEEEEEEEEEEEE!!!
▶ No.916180
Getting a Talos Lite soon anons, I have to sell off some Stinkpads first though.
▶ No.916181
>>916156
A tripfag didn't make the thread. You're just an ass bothered Intel shill or a glowdark mad about Talos.
▶ No.916186
>>915749
This, it's a bespoke item for people who know what they want.
▶ No.916213>>916271
>Blatant shilling is only okay if it complies with the /tech/ hive mind
If this was a series of threads about some Shitel workstation being created in succession like this before the last one is even fucking dead they would've been deleted by now
▶ No.916253
Look at how mad the Intel Sage Brigade is here.
▶ No.916271>>916309
>>916213
>why are people hating on stuff they think is bad and praising stuff they think is good???
You're too autistic for /tech/, that's something.
▶ No.916280>>916296
>>916226
Quad core with 4 threads per core.
▶ No.916296>>916366 >>917089
>>916280
Real threads too, not "hyper" jewtel threads.
▶ No.916309>>916316
>>916271
>Why are people selectively hating on stuff that doesn't fit their autistic narrative no you're the autistic one!
▶ No.916316>>917092
>>916309
You should stop sage bombing good threads. Plenty of shitty bait here to sperge out on.
Or are you an Intel shill trying to SHUT IT DOWN?
Either way, kill yourself.
▶ No.916366>>917089 >>922217
>>916296
Same thing, retard.
▶ No.917072
▶ No.917089>>917170
>>916296
>>916366
this, threads are nothing but shilling and provide no performance benefit over the core by itself the moment you put a load on the core.
▶ No.917090
>>915534 (OP)
>Talos II Lite Secure Workstation
>firmware: Embedded GNU/Windows Freedesktop
Secure! Freedoms!
PLEASE BUY
▶ No.917092>>917094
>>916316
>He thinks sage is a downvote
>Retarded newfags still triggered by the word 'sage' in the email field
▶ No.917094>>917102
>>917092
When you sage one thread and then bump a bunch of shitty threads it is in effect a downvote.
▶ No.917095>>917102
Back on topic, this is relatively inexpensive for a high performance machine. A typical high-end Dull workstation costs as much.
▶ No.917102>>917332
>>917094
This thread is clearly the pinnacle of quality for this board? Get fucked shill. we already have a thread about this that I have not saged once.
>>917095
You're right brb buying 50 of them
▶ No.917170>>919669 >>919671
>>917089
This is not true.
▶ No.917332
>>917102
Where's the Talos II Lite thread?
▶ No.917429>>917490
ONE, NOTHING WRONG WITH ME
TWO, NOTHING WRONG WITH ME
▶ No.917490>>917544
>>917429
how many is it now? i lost count.
I wish i had enough money for either of these boards. I mostly want one because i believe this board will not be obsolete(for desktop/code neet use case) for at least 7 years, maybe longer if the community keeps writing firmware updates like they do for the asus d16.
▶ No.917544>>917609 >>917754
>>917490
At least two dozen fundamental architectural vulnerabilities revealed in the last six months, the patches for which either are unavailable, can't be made at all, won't for national security reasons, and which aren't even being widely discussed among IT professionals who ostensible are concerned with security.
It's a sorry state and a state or non-state actor is free to exploit stock exchanges, banks, health care, federal, and military systems in the mean while. It's not obvious yet but this could be the end of the Internet.
▶ No.917609
>>917544
It's probably better that way. Internet went full retard over 10 years ago.
▶ No.917754
>>917544
>It's not obvious yet but this could be the end of the Internet.
I think a lot of these vulns require people to rollback Intel optimizations until we see the initial fear of a 50% performance hit become a reality.
It won't be the end of the internet, most backbones aren't x86(switching gear), people like Google and IBM who have invested heavily in POWER will be okay.
Worst case is the majority of the internet will experience a 50% slowdown.
▶ No.917871>>919245 >>922293
Too bad all modern chips won't be reduced to 0.01% of their speed. That would actually be fun to watch.
▶ No.919138
>>915754
I'm stoked for those docs once they leak, though I'm still waiting on the IBM manual on the WiiU cpu.
▶ No.919245
>>917871
Wait for Spectre variant 4562212. By that point, all the mitigations will pile up catastrophically.
▶ No.919669
>>917170
this is true faggot show some benchmarks where (((hyper))) threads perform better than a single core when pinned at 100%.
▶ No.919671>>922211
>>917170
a recent benchmark showing hyperthreading reduces performance.
https://www.techpowerup.com/forums/threads/gaming-benchmarks-core-i7-6700k-hyperthreading-test.219417/
>muh gaymes
there's a suspicious lack of benchmarks publicly available on this, I wonder (((why)))
▶ No.922192>>922246
>>916050
Out of curiosity, how would it perform with an emulated x86 single threaded program?
▶ No.922211
>>919671
>A few games with poor HT support perform slightly worse
>Hur Dur threads must be bad
This is why game benchmarks are bullshit. They speak more so for the game rather then the CPU
▶ No.922217
>>916366
They're not the same thing. A hardware thread has all the units necessary for any operation a thread may perform (add, xor, etc). An hyperthread shares those units, so if the two hyperthreads sharing them use different operations they can practically double performance. If they use the exact same you'll lose performance unless the effect on memory accesses is worth it. In my experience they more often than not worsen it, but in some cases you can see some gain from ht.
▶ No.922246
>>922192
Probably not very well. It's still layers of emulation that you're having to wade your way through.
▶ No.922288
Will the guy who has the 36 core Talos II, please run this so we have more than the 16 core Raptor system to go on?
https://github.com/phoronix-test-suite/phoronix-test-suite/
▶ No.922293
>>917871
They'd just ignore the vulns at that point and tell the good goym that everything is okay until the world finally burns to the ground.The last remnant of our dying society would be their hands, clutched around a dollar bill, held high to preserve it from the flames.
▶ No.922294>>922309
>>915534 (OP)
How do the graphics drivers work? Can I just throw my R9-290x in and let it go or do I need their special drivers?
▶ No.922309
>>922294
>All AMD GPUs currently have DMA issues (limited to 32-bit, which can cause crashes) with the current Talos II firmware. This is expected to be fixed in future firmware updates.
https://wiki.raptorcs.com/wiki/Talos_II/Hardware_Compatibility_List
▶ No.923708>>924059
>>915906
>POWER is for high performance where you need your system to be as fast as possible, RISC-V is for applications where you need low power consumption such as phones or IoT devices.
Is there a chance that RISC-V could replace POWER in the future? It would be kinda lame if two incompatible closed architectures (x86-64 and ARM64) were replaced by two incompatible open architectures (PPC 64-bit and RISC-V 64-bit) instead of one.
▶ No.924059
>>923708
>Is there a chance that RISC-V could replace POWER in the future?
Maybe in 15 years IF people try, every RISC-V partner are aiming for the arm space, so the microarch is all being designed for embedded chips.
There is nothing wrong with the ISA, it in theory could but no one is trying to do that right now.
>were replaced by two incompatible open architectures (PPC 64-bit and RISC-V 64-bit) instead of one.
Still better than two closed, maybe even better, that's a lot of power for RISC-V to hold.
▶ No.924060>>924065 >>924144
>bare chassis 1400$
fucking why
is there any good non-x86 non-arm compooter for less than 300?
▶ No.924065>>924078
>>924060
>is there any good non-x86 non-arm compooter for less than 300?
There isn't any good x86 or arm computer for less than 300 bucks either. And where I live, I'll be looking at at least five times that much for a good rig, ten times that by the time I finish with it.
So yeah, you should get off NEETbux and get a job. The upside is that leaving your parent's basement feels quite Libre.
▶ No.924078>>924181
>>924065
You can buy an old 8-bit computer for less than $300, so it can be done on NEET budget. Modern computers can't be trusted, no matter what. They might "say" it's open and doesn't have cianigger backdoors and "oops we bugged again!" holes all over the place, but that's a bald faced lie. Some are just a little better than others, that's all.
▶ No.924144
>>924060
If ARM-tier performance is okay, there's a few modern things using other ISAs out there, like Longsoon's MIPS chinkshit.
▶ No.924181
>>924078
How much work do you get done on your Z80 homebrew Anon? Are you shitposting from it right now? Serious question btw.
▶ No.929415