Arch Linux

zerophase

Member

Registered: 2015-09-03

Posts: 239

Know if Intel daughterboards will work with AMD?

Trying to figure out how difficult getting these Ghost Canyon Intel boards to function properly on an AMD system. This is a machine for compilation, and on Threadripper I could get an extra 80 threads, slotting these boards in for relatively cheap. The cpu should be overclockable too. Never tried hooking a second computer up as a daughter board.

Should Arch just recognize these exist? Should there be a means of me playing around with the bios from my main board?

Last edited by zerophase (2021-06-22 11:26:27)

Slithery

Administrator

From: Norfolk, UK

Registered: 2013-12-01

Posts: 5,776

Re: Know if Intel daughterboards will work with AMD?

I'm pretty sure that won't work. Those boards are PCIe host devices, not client devices. They're not designed to plug into other computers.

---

No, it didn't "fix" anything. It just shifted the brokeness one space to the right. - jasonwryan
Closing -- for deletion; Banning -- for muppetry. - jasonwryan

aur - dotfiles

progandy

Member

Registered: 2012-05-17

Posts: 5,317

Re: Know if Intel daughterboards will work with AMD?

You could use these "compute modules" over ethernet as a distcc node, but you'd need a case / backplane for them and it would function just like any other standalone computer. Alternatively, you could configure the boards as some sort of kubernetes / docker / nomad / ... cluster and distribute your workload that way.
Edit: You can design and use a very fast interconnect just like ExpressFabric, but that is not plug-and-play and not cheap.
https://www.kernel.org/doc/html/latest/ … i/ntb.html
https://www.dolphinics.com/products/eXp … Linux.html
(distcc + thunderbolt networking would be easier)

Last edited by progandy (2021-06-22 13:16:50)

---

| alias CUTF='LANG=en_XX.UTF-8@POSIX ' | alias ENGLISH='LANG=C.UTF-8 ' |

zerophase

Member

Registered: 2015-09-03

Posts: 239

Re: Know if Intel daughterboards will work with AMD?

Would I be able to see the devices if plugged in through pcie at least?

I just saw a video from Gamers Nexus mentioning they could be used as compute units. I'm sure that would require distcc. But, at the end of the day would I be able to see the cpus through Arch? Would I need arch installed on each machine?

Are there actual compute boards intentionally designed for this use case? I'm thinking those might be cheaper or more cost effective. Basically, if this ends up being a cluster computer would I potentially see any performance gains over just connecting through the network? Not necessarily just compilation, but for long running tasks, which require multiple cores. So, I'd be starting the tasks up on each cpu present individually from the mainboard.

Last edited by zerophase (2021-06-22 14:33:48)

Lone_Wolf

Administrator

From: Netherlands, Europe

Registered: 2005-10-04

Posts: 15,086

Re: Know if Intel daughterboards will work with AMD?

There are cards that can speedup certain tasks by using dedicated hardware.

AMD has the CDNA cards to accelerate gpgpu tasks, intel has programmable fpga cards.

You mention compilation, a big bottleneck for compilation is often IO , not number of available cores.
Have you considered very fast storage like intel Optane ?

---

Disliking systemd intensely, but not satisfied with alternatives so focusing on taming systemd.

clean chroot building not flexible enough ?
Try clean chroot manager by graysky

zerophase

Member

Registered: 2015-09-03

Posts: 239

Re: Know if Intel daughterboards will work with AMD?

I've tried fast storage that definitely helps. The bottleneck seems to be running compile jobs in parallel so far. But, that was on a 16 thread cpu. I have a 128 thread cpu on my new machine, and just trying to figure out if I can squeeze out extra threads, or compile dependencies in parallel for the main project.  I'm compiling from RAM, and not sure if current Optane provides an advantage. Those NUCS can be configured to come with m.2 nvme drives. Just trying to figure out if there's a use for them. It's going to be a machine basically optimized for compiling Unreal as fast as possible. Been getting stuck with 45 minute compile jobs for the main project. The dependencies take between 30 seconds and ten minutes, and there's like ten of them.

I believe the Intel fpgas are too specialized for my particular use case, while these would work. I'm just trying to figure out why it has the ability to slot into a pcie slot if there's no advantage to using it that way.

EDIT: @Lone_Wolf Your signature just reminded me I need to pick up a CMOS battery, or risk losing my overclock in a few days. Been getting weird bugs with the MOBO taking me into the bios update screen if I power cycle. Computer is about five years old too.

Last edited by zerophase (2021-06-22 14:58:36)