 | RISC-V | 581 Members |
| RISC-V is a free and open ISA. Part of #riscv:matrix.org. Bridged to Telegram: https://t.me/riscv. Unofficial. | 46 Servers |
| RISC-V | 581 Members |
| RISC-V is a free and open ISA. Part of #riscv:matrix.org. Bridged to Telegram: https://t.me/riscv. Unofficial. | 46 Servers |
| Sender | Message | Time |
|---|---|---|
| 25 Jul 2024 | ||
 Bruce Hoult | In reply to @finagolfin:matrix.org I suspect we're looking at another year or two before the first Android RISC-V smartphoneyes of course. No one has ever said differently. 2026-2027 is the date. | 13:16:25 |
 Daniel aka CyReVolt 🐢 | There is going to be another RISC-V Tech Session in 30 mins: TL;DR security, overflows, fun! | 13:30:59 |
 finagolfin | 3-4 years before the first device is a ridiculous lead time in my opinion. That tells me nobody is really investing it much. | 13:31:13 |
| finagolfin | * 3-4 years before the first device is a ridiculous lead time in my opinion. That tells me nobody is really investing in it much. | 13:32:03 |
| finagolfin | Not surprising though, eg the Android NDK team doesn't seem to have many resources for one, as they're always saying they're not staffed for this or that on their github issue tracker. That is ridiculous considering how many billions Android brings in, and games are the biggest money-maker for mobile apps, which disproportionately use the NDK. Many signs that google has descended into bureacratic chaos: https://ln.hixie.ch/?start=1700627373&count=1 | 13:35:49 |
| Daniel aka CyReVolt 🐢 | Many Xooglers write such stuff, I know a handful personally. It is a giant company after all. | 13:38:58 |
| Bruce Hoult | https://x.com/seblu84/status/1815369952695660710 | 13:27:00 |
| Bruce Hoult | but to help keep things moving along, see https://www.scaleway.com/en/docs/bare-metal/elastic-metal/reference-content/elastic-metal-rv1-guidelines/#android | 13:26:28 |
| finagolfin | Glad to see it, though not sure who it's aimed at. As they say, "Despite @Google’s recent conservative stance on supporting RISC-V on @Android," ie I'm not the only one saying that. 😉 | 13:51:06 |
 H | ulinux RISCV rv32ima ASIC fresh from https://x.com/splinedrive/status/1815477715018911785 | 15:07:52 |
 edolnx | In reply to @finagolfin:matrix.orgIt typically takes 2-3 years just to get Silicon made at this point, especially at production volumes. Lots of work happens behind the scenes to make it look like a new product comes out every year. The DVT teams at "The Fruit Company" are working on M6 or later at this point, and M4 hasn't even hit volume production. This stuff takes time. There is a lot of movement and interest in RISC-V but for legacy industry reasons, it's all happening behind closed doors and bound by NDA. That's why there is no "support" | 15:36:44 |
| finagolfin | Sure, I followed the Qualcomm-Nuvia odyssey, which took 3.5 years from acquisition to the big launch last month: https://arstechnica.com/gadgets/2021/03/qualcomm-closes-nuvia-deal-promises-in-house-laptop-cpus-by-2h-2022/ | 15:43:15 |
| edolnx | And five years of development before that, all of it behind closed doors | 15:46:24 |
| finagolfin | In reply to @edolnx:matrix.orgThe article says the company was founded in 2019, so only two years. | 15:48:23 |
| edolnx | Right, two years in public view. Three before that behind closed doors. | 15:54:50 |
| edolnx | They didn't pick a company name until 2019 | 15:55:04 |
| finagolfin | The linked article says all the founders left their previous jobs and started the company in 2019: https://techcrunch.com/2019/11/15/three-of-apple-and-googles-former-star-chip-designers-launch-nuvia-with-53m-in-series-a-funding/ | 15:57:12 |
| edolnx | Correct, started the company in 2019. Were working on designs and plans for years before that. They tried to launch within Apple first, and when Apple was not interested they left with the relationships in-tact to form a company. Many years of development and R&D before they "launched" nuvia | 15:58:30 |
| edolnx | That's why they were acquired in "only" two years. As opposed to Rivos, Aekena, Tenstorrent, and Ventana that are many years in and yet to "launch" a product. These companies were starting with a blank piece of paper. Nuvia started with a complete design in-hand. | 15:59:52 |
| edolnx | My point is that these things take a lot of time and effort before anything "real" happens. So lack of a "release" in less than 3 years after announcement doesn't mean that there is "no interest" or that it's vaporware. This is all incredibly hard, and in the space of RISC-V all incredibly new. Almost every product has shipped late because of unforseen integration issues in the verification stage, or on test silicon. These things take a lot of time. | 16:02:16 |
| finagolfin | In reply to @edolnx:matrix.orgGulati and Bruno had been working at Google from 2017-2019, without Williams. Anyway, none of what you claim is in these articles: what's your source that they had three years of work prior to founding? | 16:06:42 |
| edolnx | Discussions with executives and VPs in 2019\ | 16:07:40 |
| finagolfin | Alright, will take your word for it. Academic really, as the point I was going to make was that their launch this summer was much bigger than the Android RISC-V smartphone will finally be. | 16:09:27 |
| edolnx | Correct. Qualcomm was launching a halo product with the Snapdragon X Elite and has a massive stake in getting this product out the door before Arm sues them into oblivion. The Android RISC-V launch will be at the low end phones probably launching in areas outside of the US and EU in a year or two, and like every other non-halo product Android Hardware Launch will be completely ignored because it does not have a multi-million dollar marketing budget. | 16:12:36 |
| finagolfin | In reply to @edolnx:matrix.org Sure, can't disagree with your basic point, as I've said that before too. I think our only disagreement might be on whether reading the tea leaves, google and the hardware vendors are investing much into this effort. Scaleway apparently agrees with me that google's interest has waned. | 16:15:29 |
| edolnx | I'm part of the Android SIG within RISC-V - there is a lot of interest. But it's at the low end of the market and entirely under NDA. Everything that is publicly visible in AOSP is years behind what development deals are happening behind closed doors under NDA. This comes up at every meeting: There is a lot happening, but nothing that can be commented on publicly. The press is reading the tea leaves based on AOSP commit messages. Android development is not done in the open, it's done behind closed doors and then tossed over the "open source fence" several weeks after a launch. It has always happened this way. It is correct that the AOSP teams cannot get CI resources, they are limited to what is publicly available and right now that is limited to QEMU. Android cannot even publicly state what the hardware requirements are for Android on RISC-V yet, that is still under NDA. There are products that exist which hint that they are getting close, and much like the Arm SBC ecosystem, most of the RISC-V SBCs are going to be SoCs designed for Android repurposed to run "generic Linux" directly because it makes no economic sense to do anything but that. | 16:21:23 |
| finagolfin | Heh, well, if you're on the Android SIG and know what's happening behind closed doors, thanks for the non-specific info (not a critique, though it may read like it, I know specific info may be embargoed) that deals are happening: that is reassuring. 😊 I'm particularly happy to hear about more RISC-V SBCs, as I feel that has been under-invested in so far and is important. | 16:31:44 |
| 26 Jul 2024 | ||
| Daniel aka CyReVolt 🐢 | making baby steps again porting the DRAM code to Rust for the SpacemiT K1x | 13:54:46 |
| edolnx | Awesome, I have been swamped with other things and haven't made and progress on the C part yet | 14:44:22 |
 John Carr joined the room. | 14:57:45 | |