I am asking this because, I noticed the thread about amd possible in mnt’s future. Also, I know on some coreboot sellers websites such as novacustom, that it is possible to coreboot newer intel gens if the OEM allows it. Intel me would need to be disabled of course though and probably open source EC firmware would be good…
I have also looked and seen some processors that are intel, are very ultra lightweight in gen 10+
Even if this was for, say, one of the less functional more low voltage processors, it might still interest some out there.
I had the category as uncategorized because I wondered about this for both mnt reform devices. Pocket and regular.
A processor like this for example:
Intel® Celeron® Processor 7300
Uses 9 watts unless its turbo. Which would probably be good to disable anyhow.
Just 6 watts. Personally, I wouldn’t care for X86 too much as it doesn’t offer any major advantage over something like the RK3588.
The Intel N100 can definitely outperform in a few workloads and lose in others making the two roughly on the same level, but the Intel also has issues with Intel ME and encumbered firmware as you noted. For the level of effort required with X86, I would say that MNT’s resources would be better spent on the RK3588.
I haven’t bought an MNT Reform yet, but it is the machine I want. My current issue is that only the i.MX8MQ can currently support all of the features I need, but it lacks the performance I need. The N100 or the RK3588 would be able to do everything, but… between the two the RK3588 would likely be far less work.
I do suppose that it is worth considering X86 for things like Zoom, Slack, Adobe, and so on. The thing is, it may be possible to run those on ARM with a combination of WINE and box86 and box64.
X86 on the reform is possible if there are System on Module available with the processor and all the required connectivity to make it work with the Reform mainboard. If that exists, then anybody is free to create a CPU adapter board which physically and electronically fits into the Reform mainboard. Then those interested in X86 do not need to wait for MNT to do the work. That said, I’m not aware of modern X86 SoMs that meet these requirements.
About performance: I used the reform with the imx8mq for over a year and it was fine as long as I wasn’t using websites with heavy javascript like youtube or twitch and friends. A bit more than six weeks ago I upgraded from imx8mq to a311d and I can now confidently say, that for my personal workload, the a311d provides what I need. Example: the testsuite of my main project which took more than 4.5 hours on my old intel laptop now runs in only a bit more than 4 hours so in that sense it’s even an upgrade in performance for me. 1080p@60 youtube and twitch work flawlessly with software decoding. Clearly, the RK3588 will be another massive performance boost but not one that I require but just a “nice to have”. As far as RAM goes, the only times that the 4 GB ram became a problem for me was when trying to compile C++ projects on all 6 cores. Doing so, even kills my 8 GB swap. So I’m just careful to compile suspicious projects with -j4 instead.
About proprietary x86 applications: I’m in the fortunate position that I do not need the usual x86-only suspects – except games. For those, I had a lot of success with box64. For example, according to the ingame-timer, I already played 10:46 hours Stardew Valley in multiplayer networked co-op on my reform. Gameplay does not differ in any perceivable way from playing the game on x86. Zero crashes.
The performance boost is one reason I want this, but the other is, compatibility with operating systems I love. Unixlike ones specifically. 6W would be cool.
Only a few things would be needed though, make write protection easily disabled + intel me disabled and of course coreboot potential as previously stated.
But yeah, if this happened, I might consider getting another mnt reform device. I ordered a pocket mnt reform, but having two devices from mnt, one purely for speed and one for low level stuff, would be pretty cool.
I gather intel is betting heavily on RISC-V as something they can dominate. When they make a tasty CPU with this ISA then yes I think that we might welcome Intel to Reform.
As for x86, please no. Let’s not perpetuate an architecture that, in a rational world, should have been replaced by PowerPC (and others) back in the 1990s.
My objection is precisely that x86 is the safe, compatible, middle of the road choice. It is one half of the Wintel establishment that we are trying to rebel against, no?
I do have some sympathy to the compatibility argument because I like Haiku which does not yet run on ARM. But I can choose from millions of used x86 machines for much less than the Reform - machines which may be obsolescent but have more than enough muscle to run Haiku well.
I recently did some investigating into the feasibility of specifically the 6W Intel N100 on a Reform-compatible SoM with at least 8 GB RAM. Many mini PCs on the market are able to hit this target for under $200 with a (dubious) Win11 license included.
A few weeks ago I asked a manufacturer if they could respin their N100-based design into a module. The design work would cost around ~$20K USD and the minimum order quantity would be ~200 modules.
Back of the napkin math: let’s say parts & mfg for each module are comparable to the existing Intel mini PCs at around $200, and splitting the design cost across 200 modules would make it roughly $300 per module for native x86, less if more modules can be made and sold. We’d still need someone to help with bring-up, but the drivers are already there thankfully.
There’s a chance we could crowdfund it but the Reform community is small and I don’t know that a lot of users would opt for this option with compelling ARM alternatives on the horizon. I’m not of the mind that having Intel as an option is inherently bad, as I think that choice is key to Reform adoption, and that there’s a spectrum of open to less-open it already offers.
Unfortunately the AMD chip options out there are not very good, mostly 2 cores at around 15W and on older process nodes, and I bet they cost a lot more than the Intel stuff because Intel is subsidizing adoption of their platform somewhat.
Wintel? Nah… if someone did that and created a lintel, as in linux + intel maybe… but its no big issue. ARM64, Risc-V and other architectures are moving closer to getting somewhere as I may have mentioned elsewhere on this forum.
I was mulling this as well, specifically because I can not fully leave behind proprietary software for many work and personal usecases yet and having an x86_64 MNT board would bridge the gap and provide me with all the MNT benefits and longevity while not forcing a clean break from a 30 year x86 computing baggage overnight.
The i3-N300 is particularly compelling from a performance/watt standpoint, being marginally faster than my desktop i7-6700k for 1/13th the TDP, making it a viable desktop replacement for me at least.
I don’t think we will see amd64 based CPU module soon, MNT Research relies on available SoC(most likely been some form of compute module like raspberry pi cm4) to make the CPU module.
Manufacture Intel or AMD board from grounds up would be too much work for MNT Research to do.
I can understand what is stopping x86 development for MNT, but I really feel that moving to a x86 SoC would help move MNT out of obscurity, and get MNT much more revenue which is greatly deserved.
That said, and as I have written elsewhere, these ARM CPUs are very capable and are not as limiting as one might initially suspect.
Any x86 will be so low powered due to design constraints such as power and passive cooling that it would be comparable to a much older “typical” laptop. Just buy reconditioned Thinkpad if you need compatibility - it’ll be much cheaper. Those interesting OS we run on our Reforms (Sculpt, 9Front) also run on Thinkpad models, and if you can make do with boring old linux it’ll run on anything you can find.
It is not like anyone is proposing to carry around a quiver of compute modules of different architectures and change them in the field according to need, which might have been a valid argument for x86 modules over just having a secondary computer.
Production seemingly already struggles to keep pace with demand, so that is not the problem.
