All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
risc-v hosting?
not sure if right category so feel free to move my post etc etc
someone sent me this earlier: https://www.phoronix.com/news/Milk-V-Oasis
12 2.5ghz cores + 4 1.6ghz e-cores, up to 64gb lpddr5, m.2 nvme slot, 4 sata ports, dual 2.5g ethernet, all of this starting at just 150 usd (120 if preordered) (optimistic). shipping in q3 2024 (also optimistic)
seems like risc-v is moving very fast and seems way more promising to me than arm ever did
it's crazy to me how cheap hardware can get when you don't have to pay shit like royalties and licenses
i think if this actually ships for the advertised price risc-v will actually become a feasible architecture for desktop computing, but how far away do you think we are from datacenters filled with risc-v boards providing super cheap hosting? or do you think this is just vaporware?
Comments
I have a few risc-v boards and unfortunately their performance is too low compared to standard/consumer x86/arm devices. So no, they are not for hosting, at least today.
I hope it is matter of a few years, would like to see more ARM/RISC-V server grade hardware soon
ARM is already in deployment in many datacenters.
Also:
https://www.reuters.com/technology/nvidia-make-arm-based-pc-chips-major-new-challenge-intel-2023-10-23/
That's what I meant under "more hardware"
My dream is to see a low power chip capable of emulating both ARM and x86 at close to native speeds to make the transition until everyone compiles for the new architecture.
ATM my bet is on ARM but I am not sure, there might be new kids on the block soon, multi-billion investments are being mounted in this battle. The Wintel days are truly over.
Unknown company, in affiliation with unknown company, on a non-mature & pioneer like platform offering a preorder with unknown delivery date? Sign Us Up!
LOL! Humor aside, our new MD platform could host these, but i'm worried about the man hours this would take.
We could take the risC maybe, at least if some others chip in on the, i bet countless, manhours it will require to test this stuff out and asking the developers to fix firmware issues ... maybe.
Performance wise these would suck. Performance to Dollar wise, these would suck.
99% of the cost will be on the operating costs and man power. But we could risk it, given sufficient interest.
I see a future for risc, but it's a looong road ahead. Maybe even 2 decades before IPC and volume is even remotely close enough to be price competitive with x86 CISC CPUs.
But if you guys can arrange something like 32-64 node preorder in a manner that we would cover atleast the initial investment + some of the labor we might be persuaded.
The preorder 5$ is to get a "coupon to buy it at 120$" ... so, that's kinda expensive. BIOS/Firmware will suck arse, none of pre-existing installers will work, even if PXE booting works ... So yeah we would probs need 10k €+ in preorders.
In comparison; That's about the price for Intel N100 mITX board we can trust to work reliably ... so these are Guaranteed to cost more, and perform worse
I see it more like 2 years. We will see.
Indeed, we shall see. It is interesting regardless, lots of potential for the low end market.
Right now RISC-V is a long way from being mass-market.
Until maybe 6 months ago, the fastest chips you could get were a couple of hundred MHz. There's recently been a dual-core 1GHz RISC-V chip released, but as far as I know, everything that's used it so far has had the second core disabled. Maybe this is chip yield, maybe it's hardware bugs, maybe it's just waiting for kernel support. Who knows.
I think it was always inevitable there'd be interest from cheap manufacturers initially, but ARM is so entrenched in mobiles (and specifically although Android was designed to be processor-agnostic, nowadays most software relies heavily on native compiled code) that I didn't think it'd be mass market.
But I think the US really messed up when they started restricting Arm from dealing with Chinese chip designers. That market is massive and if you restrict the biggest players in their biggest domestic market, they're obviously going to look at other options. And there RISC-V is - a potentially more capable ISA, patent-free and much cheaper to implement, but still a few years behind ARM and previously no real driver to push investment. But now there is a driver, and it falls neatly into China's plan of wanting to reduce its reliance on foreign tech and improve domestic chip fabrication. They've practically been handed RISC-V on a plate and it ticks nearly all the boxes, but most significantly, the US has no way to control how they use this tech going forwards.
If we fast forward a couple of years, I think all the big Chinese mobile brands will have shifted to RISC-V for their flagship models, simply because trying to stay with Arm will cause their development to stagnate and fall behind non-Chinese brands. Perhaps they'll start off with a few slower lower-end phones first so that developers will start shipping apps built with RISC-V native as well as Arm-native code so that the ecosystem is ready when they do launch the high end phones. When that happens, it's inevitable that all the Chinese market will switch in a relatively short time, because without licensing fees and import restrictions they'll cheaper to manufacture.
RISC-V isn't ready to take over the market yet, but it has at least reached the critical mass of interest and there are loads of companies with deep pockets who now want it to succeed. And when it does, we'll then find that China has a lot of cheap powerful chips available and a mature Android environment based on it, able to produce RISC-V phones at a lower price point than Arm phones and then the rest of the manufacturers in other countries will be at a disadvantage.
Maybe it'll just cause a fracturing of the mobile space between RISC-V in China and Arm in the rest of the world, but as per OP's original idea, I think it's inevitable that hosting will increasingly have a space RISC-V. After all, the last couple of years with ARM becoming available, it's shown lots of people that the transition away from Intel isn't all that hard but can offer performance and cost benefits.
One final note is that Arm now has a fairly well defined standard way of detecting what functionality is available on each chip, and so features can be auto-detected and enabled by the kernel. For RISC-V, that's still quite embryonic and there's no standardisation between vendors beyond the ISA, so you need to build specific kernels targeting each vendor's features and peripheral memory maps rather than a single kernel that can support a wide range of different hardware. I'm sure that too will change as popularity increases.