An update at Power11 hits the market this month prompts Quote:

Quote:

I once heard testing a CPU is like building software, but you are only allowed to compile once every 3 months and it costs a $1 million each time. That was a LONG time ago and costs have gone up exponentially since then.

Testing a new CPU now is horrifically expensive.

You need to build a model first to test how well it's going to work. An emulator basically, probably just testing the logic at first then moving to a cycle accurate model.

Then you need to design it, and make sure it matches the model. You'll need to simulate it. By that I mean you have to have a cycle accurate model of all the gates, busses clocks etc.
That requires an immense level of computing power, it typically requires a supercomputer. Even then it's simulating at thousands to millions of times slower than the real thing.

Once you know it (mostly) works you can turn it into a bitfile and put it into FPGAs. You need top end FPGAs and lots of them in a system designed for this. These things are probably in the millions these days but they only run hundreds of time slower than the real thing.

Ultimately, the real test is to build a test chip. On the latest processes that's $50s million.

Update:

>> S1 by a chip design company called Solid Silicon

> S1 no longer mentioned on the website. What gives?

Solid Silicon's OpenPOWER membership "cancelled: 2024-08-31" (with "2024" probably being a typo).


Edit: Website no longer reachable.

[ Edited by Andreas_Wolf 21.07.2026 - 09:55 ]

Addendum:

>> Oh my... just the kind of news one would expect and is
>> accustomed to in the Amiga scene. :-D

> It doesn't get better:
> https://www.phoronix.com/news/Libre-SoC-NLNet-Funding-Lost

https://news.tuxmachines.org/n/2025/10/25/Libre_Hardware_Founder_s_Project_Sabotaged.shtml
https://ninazeng.substack.com/p/libre-hardware-founders-project-sabotaged (full article with objection as comment)

"MIROSHNIKOV was tasked with liasing with me in sep 23 to make carefully-crafted modifications to the Financial Audit Records of the libre-soc project. instead within ninety minutes he sabotaged the financial records, did not inform me, and i was forced to call an emergency freeze and audit, a day later on discovering the sabotage. in the meantime, BHARVANI had been caught embezzling EUR 35,000 from the NGI POINTER Programme. he had not done the work, had not purchased the asset required, and was fighting tooth and nail to claim otherwise. when hard evidence of his fraudulent activity was presented he became vengeful and aggressive, and when i called an emergency audit and requested that everyone help with the audit he instead illegally called a meeting attempting to seize control of the libre-soc project, its resources, its finances and its operation. [...] the illegal meeting progressed, [...] attended by LEWIS CALDERWOOD MIROSHNIKOV and MARGARITIS. they "voted" not realising that the project operates on unanimous decision making. [...] MARGARITIS in the meantime had received a contract as an employee of ARM and was doing sub-standard work for NGI Search as a result. [...] during a meeting [...] we had it confirmed that MARGARITIS and BHARVANI were attempting to seize control of the NGI Search Grant, not realising that it was a contract not a Memorandum of Understanding. it was very clear that they intended to "divvy up" the EUR 150,000 grant between themselves, and a very pissed-off Mirko had to bluntly disabuse them of this additional fraudulent attempt at embezzlement of EU Funds. in addition [...]: MARGARITIS was attempting to steal patentable material jointly owned by himself, myself, and JACOB LIFSHAY. he was not in any happy that i stood firmly in the way of this effort at blatant technology theft. [...] CALDERWOOD wilfully and intentionally mis-represented the facts [...] in order to bolster his case of fraudulently stealing both RED Semiconductor Ltd and [...] Simple-V, through aliasing to VISC. that [...] is Weapons-Grade Military Supercomputing Technology, intended to be hidden in plain sight as COTS products. it can be used to make Hypersonic missiles more reliable, aid criminals and foreign powers to enhance AI fraud - all of which constitutes a direct threat to the security of the United Kingdom, the EU and its allies given that CALDERWOOD and LEWIS are attempting to sell my stolen technology to the highest bidder. [...] CALDERWOOD [...] will stop at nothing to get what he wants: control of my Weapons-Grade technology and to sell it to the highest bidder."
https://lists.libre-soc.org/pipermail/libre-soc-dev/2025-November/007022.html

"i was deeply concerned that the UK, [...] the DSIT and the ChipStart Programme would be brought into disrepute by [...] theft of some extremely powerful Supercomputing tecchnology, which CALDERWOOD and I spent six months [...] to evaluate, concluding in 2021 that it was reasonable to achieve an underestimated revenue from hardware sales of GBP 1.6 billion by year 6-7. [...] my strategy to contact the Ministry of Defense with a view to quietly receiving approval for COTS hardware was blatantly disregarded by LEWIS despite clear and unequivocable instructions from the Board that he was to seek out sponsorship and a grant through the MOD's Research Programme, obtaining approval by CESG as a dual-use COTS hardware product. [...] CALDERWOOD [...] has claimed publicly that RED Semiconductor Ltd is a software-only company which applied for the hardware-only ChipStart Programme [...]. it is very easy to verify this statement as false: RED Semiconductor Ltd was the recipient of 2020 NGI POINTER funds on the development of an EU-funded hardware 180nm proof-of-concept ASIC implementing an early prototype of my Trademarked Technology. it was also the recipient of additional funds under the EU NGI Assure Programme. RED Semiconductor Ltd was also the recipient of NLnet Grant funding under the 2019.02.029.Coriolis2 hardware Grant, and the NLnet 2019.10.043.Wishbone hardware development Grant [...]
RED Semiconductor Ltd positioned itself as a hardware-only company. the missing information that allowed CALDERWOOD to claim that RED Semiconductor Ltd was a software-only company is that they falsely and fraudulently claimed to have ownership of [...] Simple-V, under the alias "VISC", where i was, personally, as the sole exclusive owner [...], funded by the EU to develop the Simple-V Service Mark and associated Standards and Certification Programmes, with a view to licensing them to hardware-only companies such as RED Semiconductor Ltd. no license was issued and they have not requested one, and neither has RED Semiconductor International Ltd."
https://lists.libre-soc.org/pipermail/libre-soc-dev/2025-November/007026.html

"i in my capacity as a RED Semiconductor Ltd Director completed the documentation associated with the hardware acceleration achieved by my Weapons-Grade Military Supercomputing capable [...] technology, Simple-V [...]
that money, issued by the European Union, funded RED Semiconductor Ltd in their fraudulent efforts to claim ownership of my technology, presenting themselves as the Licensee of my technology that in the field of cryptography could speed up algorithms by a factor of 100. [...] this is why my Trademarked Technology, Simple-V, is so powerful. the power consumption reduction is immense, and the simplicity of the resultant algorithms means that they are more easily verifiable as secure. the benefits to the United Kingdom's defense should be blindingly obvious, and that that security is threatened by CALDERWOOD and LEWIS's theft just as equally obvious. this was all supposed to be "hidden in plain sight" as affordable mass-volume COTS products so that the capability was easily available to the United Kingdom and its Allies without its enemies suspecting anything or being able to get their hands on comparative technology. that capability is now not only jeapordised, but available on the open market to the highest bidder, thanks to CALDERWOOD and LEWIS's blatant and criminal fraud and theft."
https://lists.libre-soc.org/pipermail/libre-soc-dev/2025-November/007024.html

"the value […] is almost incalculable: ten-digit to twelve sums are not unreasonable. worse: it is Weapons-grade Military tech that was supposed to be "vetted" by GCHQ as dual-use. LEWIS blatantly disregarded the Board's explicit instruction to seek out MOD Grants."
https://lists.libre-soc.org/pipermail/libre-soc-dev/2025-November/007037.html


Edit: added more quotes

[ Edited by Andreas_Wolf 20.11.2025 - 14:38 ]

Addendum:

> https://www.phoronix.com/news/Libre-SoC-NLNet-Funding-Lost

Just stumbled upon this:

https://www.phoronix.com/news/Libre-Chip-NLnet-Grant (08/2025)
https://libre-chip.org

The Libre-Chip project has been run for 2 years by 3 former Libre-SOC project members and attempts to initially create a, you guessed it, Power ISA CPU, namely "a high-performance superscalar CPU with speculative execution [...] that [...] doesn't suffer from any speculative-execution data leaks", i.e. that "doesn't have any spectre-style bugs even though it still is OoO superscalar with speculative execution".
https://libre-chip.org/first_arch/
https://libre-chip.org/grants/nlnet-first.html

Eventually, this is supposed to lead to a CPU "that will support a programmable decoder [...] so the CPU can be programmed to decode and run at full speed just about any ISA you select, by handling the most common instructions entirely in hardware, with software fallback for decoding some of the less common instructions which can still easily be executed in hardware [...], and full software emulation for the remaining instructions."
https://libre-chip.org/grants/cpu_with_programmable_decoder.html

Update:

> Solid Silicon's OpenPOWER membership "cancelled: 2024-08-31"
> (with "2024" probably being a typo).

Raptor Computing System's OpenPOWER membership "cancelled: 2026-09-01".

Hah, not that long ago I checked Raptor's website and you could still order Talos, now all is out of stock. Any idea what happened?

> Hah, not that long ago I checked Raptor's website and
> you could still order Talos, now all is out of stock.

Yes, this happened between 25th July and 2nd September. Blackbird went out of stock between 17th June and 3rd July.

> Any idea what happened?

I guess producing premium-priced boards/systems with 9-year-old POWER9 isn't really worthwhile any longer in Power11 era.

Update:

>>>> the v2.06/2.07 A2I and A2O cores cannot be used commercially under the
>>>> license by 3rd parties unless those cores be upgraded to ISA v3.0C or
>>>> v3.1 level, which includes, among other things, replacing the embedded
>>>> supervisor instructions (from Book III-E of Power ISA v2) by non-embedded
>>>> ("server") supervisor instructions (from Book III-S of Power ISA v2, or from
>>>> Book III of Power ISA v3 or pre-v2.03 PowerPC ISA). This is quite an effort.

>>> From the president of the OpenPOWER Foundation:
>>> "Call for Interns: Help Bring the A2O Core to Power ISA 3.X Compliance! [...]"
>>> https://www.linkedin.com/pulse/call-interns-help-bring-a2o-core-power-isa-3x-ganesan-narayanasamy-jgq1e

>> they have to be aware of the terrible optics of "chips designed by interns".

> Absolutely. Besides, I doubt that anything tangible will come to fruition
> using this approach.

Follow-up approach started in July 2025: Chips designed using Agentic AI

Project website: https://powercommons.org
NLnet project page: https://nlnet.nl/project/PowerCommons/ (funding announced in April)
Repository: https://codeberg.org/PowerCommons/a2o/src/branch/main
Introduction video: https://www.youtube.com/watch?v=pmZmBZlC4UQ
Repositories for new MMU: https://github.com/ARX-0/Microwatt_MMU, https://github.com/ARX-0/a2o_MMU (to be ported from MicroWatt)
Patreon: https://www.patreon.com/cw/PowerCommons ("we build the world's first openPower-based open-source computing stack. We start by reviving the A2O core and then design the world's first openPower-based laptop.")

Looking at the presented roadmap, this project seems already way behind. Additionally, there seem to be some misconceptions on https://powercommons.org/projects/a2o/ and other pages:

A2O...
- was never productized. BlueGene/Q used A2I, not A2O.
- implements Power ISA v2.07, not v2.06 (which is implemented by A2I).
- is 2-way SMT, not 4-way SMT (which is A2I).
- uses AXU-connected standard FPU, not AXU-connected QPU (QPX SIMD/FP unit). AXU-connected QPU was only ever contained in A2I-based Power BQC in BlueGene/Q (as opposed to A2I-based PowerEN, which was supposed to contain AXU-connected standard FPU just like used by A2O) and has not been open-sourced.