EOMA68 Computing Devices

An Earth-friendly way to easily upgrade and fix your own computer

Jan 16, 2019

2.7.5 Manufacturing, Spectre Analysis

by Luke Kenneth Casson Leighton

In case anyone is wondering why manufacturing has been delayed, Mike’s uncle died in December and his mother is recovering in hospital from surgery. With Chinese New Year coming up, most of the factory staff have already gone home to visit family, which puts us into middle of February before manufacturing is likely to begin. Under the circumstances, there was just absolutely no way I was going to pressure Mike to stick to the previous schedule.

So, I have had a lot of time to think, despite having root canal surgery, which thank goodness is only going to cost around USD $200 here in Taipei. Two treatments so far, and another two quite likely to be needed. No, of course I do not have health insurance. At least now I do not end up so often with utterly exhausting excruciating headaches for 36 to 48 hours.

My thinking is that it may be better to have the first batch of 100 A20 cards sent here to Taipei, and that I do the testing and final assembly. The reason is that the testing is quite involved, and I would rather do it here than risk damaging the cards. There is also simply not enough remaining funds to get me to Shenzhen, where I could oversee the factory staff, who would then also need to be paid, and paying them for testing time is not possible either.

We really, really need those donations, folks.

The last schedule, prior to Mike’s family emergencies, would have had us begin shipping within a few weeks, at which point the second set of orders could be released. Now, that will be delayed by at least four to seven weeks.

In other news: as Mike is on the critical path, there is nothing for me to do except initiate other opportunities, and that means working on the Libre RISC-V Hybrid CPU VPU GPU. The innocent question was asked on the mailing list, "What about Spectre?" and some investigation revealed that it is a total nightmare.

Basically, any out-of-order processor (latest ARM64 cores, and every single Intel and AMD processor of the past 25 years) is vulnerable to timing attacks, due to resource starvation that affects the completion time of past and future instructions, such is the nature of how out-of-order works.

Here’s the interesting fact: the EOMA68-A20, the prototype jz4775 and RK3288 cards, are all completely immune to Spectre, because, by a coincidence, they are all in-order cores.

This got me thinking: I am investigating putting in an EU Grant Application for funding related to privacy and restoring trust in computing devices, for ordinary people. The point being that the in-order cards are immune to Spectre, and if you were wondering if it really is that serious, here is JavaScript Spectre source code.

If successful, this grant would hugely reduce the pressure on the EOMA68 Project, because, at present, it is necessary to both pay for R&D, and living expenses, and plane flights, and pre-production prototyping and manufacturing production runs, all from the same budget. A EUR 50,000 grant would pay for R&D and living expenses, for six to eight months. The laptop housing redesign of the casework and power PCB could be paid for and the RK3288 card completed.

All of this can be done over time: it is a matter of how fast. Funding and especially donations and sponsorship just speeds up the process. So, please, if you would like to help, and would like this long-term project to succeed, do get in touch.


Sign up to receive future updates for EOMA68 Computing Devices.

Subscribe to the Crowd Supply newsletter, highlighting the latest creators and projects