Lichee Console 4A – RISC-V mini laptop: Review, benchmarks and early issues
3.14.by
3.14.by
A chip that's oddly faster than JH7110 (VisionFive2, Milk-V Mars) in microbenchmarks, but slower in practice (e.g. gcc). Presumably due to smaller cache.
It is also less power efficient, and lacks the upstream support JH7110 enjoys[0]. I would look at Pinetab-V, a tablet-laptop based on that SoC, today.
Better yet, wait for Milk-V Oasis[1] tba this June, as well as other boards based on SG2380, the first announced RISC-V SoC with serious performance: 16x SiFive P670 and 8x X280, all RVA22 compliant, plus vector 1.0 (standard) extension.
0. https://rvspace.org/en/project/JH7110_Upstream_Plan
1. https://community.milkv.io/t/introducing-the-milk-v-oasis-wi...
Do not forget to preorder, to get the unbelievably low price listed in their announcement.
It should be fun to play with it this summer, and further down the line it will become my home server, replacing an old atom board that's considerably weaker.
That's mostly enabled by the builtin I/O (pcie slots, sata ports...).
MilkV Oasis with SG2380 would be the end-game for majority of developers, but they are definitely loosing money if they keep starting price at 120 USD. They don't have it frozen (they changed the SoC specification some weeks ago) thus I wouldn't be surprised to see this slip into 2025. I wouldn't be surprised if this outperforms MilkV Pioneer.
Which is a reasonable assumption to make. SG2380 will surprise many as it launches and gets reviewed and talked about.
I don't know much about the RISC-V ecosystem at the moment, but what do you consider an "unbelievably low price"?
They SoC appears to be £5 to pre-order (EDIT: £5 is for a 20% off coupon), but the motherboard to use it on appears to be ~£1200. Is this considered a good deal for RISC-V right now?
Not sure what you're looking at as you didn't give a link, but perhaps you're confusing Pioneer and Oasis?
The 64 core Pioneer prebuilt with case, power supply, SSD, video card, 128 GB RAM for $2500 is quite reasonably priced against commercially built 16 core x86 machines, if your workload can keep 64 cores busy. Each core is around 1/4 the speed of current x86, but there are four times as many.
Don't forget 64 core x86 is $5000 just for the chip.
This appears to be the (voucher for the) "chip" but I don't see any board to mount it on other than the Pioneer.
Again, I admit I know nothing about the RISC-V ecosystem so apologies, I'm just curious.
The chip is SG2380, and it comes soldered onto the board.
Thus the preorder is for the board, which includes the chip.
This is a Mini-ITX board, and you're meant to provide RAM, storage, power supply, case and whatever else needed.
I presume that the milk-v one might be just the cpu+motherboard, and sipeed included ram and gpu in the price, but thats all speculation.
If it's anywhere in the $300 range I'd be very happy.
Just for scale the specint numbers from the P670 are on Cortex A76 level, so you get what the rp5 has, but you get way more cores.
A78 level, 12 SPECint2006/GHz.
It's P470 (Horse Creek, if it ever appears) and Dubhe-90 that are at A76 8 SPECInt2006/GHz level.
And, for completeness, P870 is 18 SPECInt2006/GHz.
https://rwmj.wordpress.com/2023/07/25/heads-up-lichee-pi-4a-...
powervr is working into getting the drm side merged to the kernel, too.
Edit: Apparently gcc assumes by default, that unaligned loads aren't supported, but with -mtune=size it somehow enables it: https://godbolt.org/z/d9P19aMnn
some discussion around that: https://gcc.gnu.org/bugzilla/show_bug.cgi?format=multiple&id...
I am trying to buy some milk-v duo boards with the CV1800 SOC in EU (without a credit card to use online, and which I can contact with a self-hosted email), anyone can help?
EDIT: the next SGxxxx SOC is actually a full desktop SOC, not an "embedded" one (I was talking about the those). No ARM cores, but still a mpeg and hdmi block, and it seems the GPU hardware programing manual is not public.
The X280 cores are not RVA22 compliant. They are only RV64GCV, so they can't run the same code as the P670 cores. SophGo's datasheet lists these cores under "AI accelerator", paired with a "TPU", so they were never intended to run general purpose code.
You wouldn't be able to run P670s and X280s on and schedule applications between them anyways, because they have different vector lengths.
Not trivial, but also not impossible.
e.g. I understand there's some bits to check whether vector has been used. It would also be abnormal for a process to be in wait (as opposite to ready) within a vector loop.
Having a chip like this one deployed will definitely enable experimentation and research on how to best deal with this situation.
https://www.aliexpress.us/item/3256805658668421.html
Specs: Intel N100 12GB RAM NVME SSD Slot 1280x800 Touchscreen (with pen support) USB-C port (supports charging and DisplayPort) HDMI Ethernet (1 gigabit) WiFi 6 Bluetooth 5.2
It's actually a pretty decent computer for being a no-name Chinese device. It has a backlit keyboard and my Microsoft Surface pen works perfectly on it. The 1280x800 LCD is low-res by todays standards, but at 8" is just fine. The 800 vertical resolution does mean that some modern apps get pushed behind the taskbar (on Windows 10). The cursor touchpad is difficult to be precise with, but you have the touchscreen with pen support if you need to select smaller items.
Around Black Friday it was $245 + tax (Without an SSD, I preferred to install my own name-brand SSD) so keep an eye out for sales.
The UEFI appears to be fully unlocked and has pages and pages of options. There are options to adjust CPU boost and voltage settings but I haven't been able to get the CPU to pull more than 6.5w.
A downside of not buying from a real company: The drivers for the device are emailed to you in a .zip file from the seller. So that is pretty sketchy, but it's the same for any of these Aliexpress devices.
The downsides are the much slower Intel 4200 CPU and less, slower RAM, slightly smaller 7" screen. I would not be surprised if the manufacturer of the Ali laptop is the same as that of the Nanote.
- Have tried a Linux distro on it?
- What is battery life like (especially under Linux)?
- How is the keyboard generally? (Not expecting miracles but is it usable)
The screen is likely sourced from a tablet, so it's default orientation is portrait. This means that the UEFI displays in that orientation, rather than the correct landscape orientation. The included windows drivers correct that.
There is also a sensor that rotates the screen automatically when you fold it into tablet mode. In device manager it is labeled "Bosch Accelerometer".
Wifi is labeled as a "Realtek 8852BE". But the chip looks pretty generic on the mainboard, I'm not sure if it's a real Realtek chip.
The sound card is generically labeled as "High Definition Audio Device".
The keyboard is actually pretty decent. It's only 7.5" wide, and some of the keys are in a strange position. But has a better feel than a MacBook butterfly keyboard.
Battery life I haven't tested. Right now at 77% Windows is estimating "3 hours remaining". Using it lightly I'd estimate around 4 hours of battery life.
Missed having a true pocket computer since the Psion/Windows CE days. I'll probably end up buying one... :-)
This should be fixable in Linux, where you can configure the subpixel order and orientation (horizontal or vertical).
Have fun hosting a CCP exit node
Also, based on reddit posts, GPD's build quality and customer service doesn't seem to be any better than the AliExpress laptop.
I see others claiming GPD has bad build quality. That’s not my experience at all.
I’m currently looking to upgrade to something more powerful I can carry around hopefully just as easy. I’m leaning towards either their GPD Win 4 [0] [1] or the very latest GPD Win Mini.
[0] https://liliputing.com/gpd-win-4-2023-handheld-gaming-pc-wit... [1] https://www.gpd.hk/gpdwin4
[2] https://www.gpd.hk/gpdwinminisdfewr
Their occulink port and USB4 is the real deal and allows eGPUs both for gaming and compute.
Though I’m not a gamer I’m currently leaning towards the Win 4 because it has a dock system and an LTE module, which the Win Mini lacks both. However perhaps the Win Mini keyboard and form factor is more conventional.
Which is a shame, because the Win 2's form-factor and cursor control scheme are basically perfect for what I want out of a handheld PC, and also seemingly unique.
Trade restrictions have given China serious incentives to make RISC-V CPUs domesticly, unlike every other country that can just buy AMD64 and ARM chips. Given the geopolitics of the situation, while RISC-V is gaining marketshare in the microcontroller space, it looks like RISC-V will be the Chinese-bloc CPU ISA while the west sticks with what we already have.
And don't be mistaken, there is no perfect ISA (it is said "average"), it does not exist, only a set of trade-offs and compromises which will not fit all use-cases perfectly.
RISC-V is a US Berkley initiative. Have a look at wikipedia where you'll find most of the answers to your questions.
That said, to be a success, RISC-V will need _extremely_ performant implementations all across the board ("embedded", desktop, server), micro-archs and silicium process. It will have to survive its mistakes: for instance critical bugs in its major micro-archs or design flaws (you have to presume it will happen).
And without access to the best silicium process, it _WON'T_ have performant implementations. Because there is no "enough" in the silicium industry, it wants always more transistors and less power consumption, and each new silicium process brings significant improvements on those metrics.
This is where chinese chip designers are in trouble: Taiwan has the foundries with the best silicium process, and now you get US restrictions on EU EUV tools.
Even though I wish intel and amd to switch to risc-v in the not to far away future, I would give attention at the people over there who would try to torpedo RISC-V. The other one, ARM, well, they will try everything to sabotage it, RISC-V is a death sentence for them... unless they clearly move to RISC-V ultra performant micro-arch design.
I wonder if this is a more common name for silicon in some other part of the world (I'm in the US). I don't think I've ever seen this spelling before today, and I'm not a young person.
The word Silicon 'feels' wrong to a Dutch speaker because it makes it sound like integrated circuits are made out of the same stuff used to assemble aquariums, fake boobs and heat-resistant flexible cookware. That stuff is called 'siliconenrubber' or 'siliconen' (depending on the application and who you ask) in Dutch.
[1] a bit like vi vs emacs for chemistry in other words
The Hitchhiker's Guide to the Galaxy is not just a story, it is the story. The story of our planet, Earth, a planet-sized computer built by the Amaltheans to find the definitive question to life, the Universe and everything for which the answer is '42'. Things went awry when an ark full of telephone cleaners crashed on the planet/computer which made it come up with the wrong question. The book only mentions that they'd have to build a new Earth to run the program again but now it can finally be proven that they did no such thing - they just used the old one and recycled the top layer of the machine's substrate and left us the evidence: sand. Billions of tons of recycled computing equipment cover our beaches, filling up the bottom of the seas and covers parts of the planet.
Think about this the next time you're laying a brick wall, pour some concrete or are annoyed by the gritty stuff between your teeth when you eat your lunch on the beach. You're literally eating computing history and who knows what effects that will have on the final outcome.
And it sounds like you're saying aluminium got corrupted into aluminum, which is not the case.
For a long time I assumed that Silicon is the "unofficial", American word for it, while silicium is the scientific name.
RISC-V likely has a future in mobile devices as it crawls up the performance tree like ARM did. This leaves ARM in a tight squeeze between x86 on the top and RISC-V on the bottom.
And QERV, only slightly bigger, yet dramatically faster.
The US doesn't like it, so they may change the rules again. which becomes a bit of a whackamole.
Sanctions like these gives China a lot more incentive for a reunification with Taiwan. Some say the fabs are rigged to blow if this was to occur. If so, it would have massive negative consequences for the world. Where as new management would have far less of an impact, though some countries might find themselves under sanctions on what type of chips they could buy.
There were rumors like that, but I don't believe it.
The limits nvidia is following are very clear. They cover both total compute power and compute power per square millimeter, in different combinations. nvidia is doing exactly what they were supposed to do. No limits are being pushed, and there is no whack-a-mole on the horizon.
Even if the limits get reduced at some point, it would be because of political winds shifting, not because the current limits aren't doing what they were supposed to do.
It seems strange that the USB hub is using so much power. From a (leaked?) data sheet I don’t see any way to disable it - but it does have firmware which can be updated? There is also a 5V-3.3V LDO built in so I wonder if that may be what is dissipating so much power.
The AMS1153 looks like a similar story…
The case is prototype-quality but fine (the fact that it's metal helps a lot), the screen is nice, the keyboard is even more awkward than it needs to be for the size.
I've been a bit disappointed by the performance in my tests (compiling various projects of mine with GCC), it seems closer to a Raspberry Pi 3 than a 4 or 5.
I imagine it has x64 or arm7 emulator for most builds.
Supposedly TH1520 gets a small uplift when compiled for its proprietary extensions, but I haven't seen a compiler using this and it would have no bearing on application from a standard distribution.
I have both 7110 and TH1520, but I'm looking forward to SiFive P670 powered SG2380 which promises to be an exciting chip.
[1] there is now in the last month or two one (1) board with RVV 1.0 but it's only single core vs the quad core in popular boards, and also has only 512 MB of RAM vs 4 / 8 / 16 GB on the popular boards (and 128 GB on the Pioneer).
SOC is fabbed on a 12nm process and the CPU is under clocked from 2Ghz to 1.5 GHz. Probably the CPU is not intended for laptops and the manufacturer just got a generic board, slapped another 2 boards on top for I/O and built a laptop.
1T Coremark: Pi 3: 3504 Lichee: 6900 Pi 4: 7938
have you run into any toolchain difficulties due to the 0.7.1 nature of the rvv?
I'm using the old gnu toolchain branch from brucehoult and write the benchmarks such that they work in rvv 1.0 and rvv 0.7.1. To help with that I've got a gnu assembler macro file that does 1 to 1 instruction translation when possible and for the other cases I need to #if/#else/#endif.
I'm working on rvv unicode conversions, and plan to release a blog post/article with benchmarks and explanation soon. The problem is that I've uses c intrinsics to write it, so I need to manually translate the assembly code to rvv 0.7.1 compatible code.
That entails
vsetivli -> vsetvli
vle64 -> vle32
vmvNr -> vmv.v.v & 2 vsetvli
vzext.vN -> N/2 vwaddu & N/2+1 vsetvli
Also, the board I've got access to (it might be an eval board, idk) has a bug in vredmax, so I also needed to adjust the code to not use that. (I've yet to investigate that further)
i really hope to have a chance to read your blog post! i haven't tried rvv programming at all
The big problem with these small and cheap systems is that you can't connect NVMe drives to them and you pay a huge performance penalty.
You wouldn't get very much performance anyway, as you only want many PCIe links in order to get bandwidth through them which you're unlikely to do on an anaemic CPU.
to-wit: you wouldn't get good NVMe performance even if you had the pinouts and the CPU had 4x Lanes of PCIe anyway.
... compared to eMMc or sata over usb? You won't get the full performance of the ssd, but i guarantee it will be much better.
Most inexpensive and small SOC just can't justify adding such costs and complexities given that it's very likely their target volume buyer doesn't need such features.
I suspect this RISC-V SOC is priced in the $10-20 range, but I have no true understanding of the cost of this part. In that price range for SOCs almost no one has 4 PCIe lanes on offer, but some do have 1 or 2.
So is it not official Debian? If so, that’s concerning!
RISC-V in a laptop was unthinkable just a few years ago, now it's a reality (at least a prototype) and performance is underwhelming, just like it has been for ARM before.
For a true "freedom machine" I would be perfectly fine with a performance trade-off but it's also not a given that the gap will be so wide forever.
This is an excellent step and to show that it's functioning is huge.
Free how? Just because RISC-V is an open ISA doesn't mean that the silicon is open. None of the manufacturers release the full design of the SoC. Many RISC-V SoCs are still relying on blobs.
Most (all?) current available RISC-V SoCs are designed and produced in China. So I wouldn't be so sure about a SoC designed by Alibaba to be free of blobs, license violations, patent infringement or even hardware backdoors.
I like having an open development process for an ISA, but it's one of the least important factors in total computer openness. Many other boards are probably more free despite using a less free ISA.
The patents on basic x86_64 with SSE2 have all expired anyway.
- RISC-V has miles to go, I worry that with the current pace of ARM it may never catch up. - The assertion you can’t buy UMPCs in 2024 is patently false. GPD and a whole host of manufacturers make em. - A non mobile optimised blog in this day and age? That was a chore to read on the go.
Only assuming development progress happens linearly. In fact, technical progress goes exponentially. It took years to even get to the point to have RISC-V SoCs and now we are seeing the first laptop products. The pace will accelerate, especially with the promise of RISC-V being a cheaper platform to build hardware on than ARM and X86.
Their point was that ARM is also progressing exponentially.
It's more likely to fill in the ESP32 niche then the ARM one
Uhhh ... yes, there is.
Several companies designing very high performance RISC-V were founded in 2021 or so. Most of them have top designers fresh from designing the latest Apple, Intel, AMD chips and they are aiming to match current x86 and Apple performance.
You'll see their products in the market around 2026.
That's a weird assertion. This machine is basically competitive with a Pi 4 (certainly miles ahead of Pi 3, both in CPU power and in having 8 or 16 GB RAM vs 0.5 or 1 GB).
The initial version of the RISC-V spec was ratified in July 2019, and the C910 core in this chip was announced in the same month. The TH1520 SoC hit the market in June 2023 (my board was delivered that month), four years after the CPU core was announced, and also four years after the Pi 4.
Note that the Arm A72 cores in the Pi 4 were announced in February 2015, so it was 4 1/2 years from core to Pi 4 using it, slightly longer than the THead C910 core announcement to Lichee Pi 4A shipping.
The Arm A53 also took about four years from announcement to the Pi 3 and Odroid C2. And the Arm A76 took four years from announcement to the RK3588 and Rock 5 shipping.
This is just the industry.
If the SG2380 machine(s) really come out this year then they will be faster than the A76 RK3588 boards and Pi 5 and around 2 to 2.5 years behind. But their P670 cores are equivalent to Arm A78, which is not available on an SBC -- and there are 16 of them, which is also not available in any cheap Arm-based SBC.
That's halving the gap.
SiFive's fastest core, the P870, is around Arm Cortex-X3 performance, and announced just 16 months later (October 2023 vs June 2022).
The software still isn't really in a working state. But I'd rather pay 30$ to figure that out over 800$.
The Milk-V Mars, with very similar performance to this (see other comments for JH7110 vs TH1520), starts from $39 with 2 GB RAM (and 4 GB or 8 GB options) and of course quad core 1.5 GHz dual-issue CPU. That's around 8x the compute performance of the 1 GB MangoPi MQ-pro, for $10 more. Or, for more bare-bones, the Mars CM is $34.
I still think I'm going to wait a few years before buying another device.
https://github.com/u-boot/u-boot/blob/master/arch/riscv/dts/...
It's been a year for the JH7110 too, so I wouldn't hold my breath.
On Vision Five 2 everything is still softpipe in the official Debian from Star Five!
On the PineTab-V NOTHING involving the 3D part of the GPU works!
YouTube works for consumption, but that's not interesting for production.
To write/film a review of a SoC with GPU without running glgears is backwards.
I think the only hope now is for the upstream to gain a working driver in a year or two.
PowerVR driver is already part of mesa3d mainline[0].
Some definitions for the variant used in JH7110 have been added recently, but it's not yet operational.
Is it so expensive to add 1 lane of pciexpress or at least sata? Not via usb, that doesn't count.
I’m guessing back Linux works but most precompiled stuff not?
If there's an issue with RISC-V it's that it's not one ISA with additional instructions accumulating across generations (which I think describes x86), it's a core ISA with a variety of extension blocks that might or might not be present in your implementation. I think this means that any random RISC-V binary stands a greater chance of not working on a chip you've bought than e.g. any random x86 binary with an x86 chip.
Whereas qemu-user will run e.g. x86 and arm binaries, should you ever need to.
This recent review of VisionFive2[1] has a little more info. There they even run popular games via x86 emulation.
If that's what identifies it
https://docs.u-boot.org/en/latest/board/thead/lpi4a.html https://docs.u-boot.org/en/latest/usage/cmd/bootefi.html
internal rom -> u-boot SPL --> opensbi --> u-boot --> Linux