The Intel “Horse Creek” developer board is compact computer board featuring 8GB of DDR5 memory, a PCIe 5.0 slot and SD card reader for storage, and support for Linux-based software.
But while this is an Intel product, the heart of the system is a SiFive Performance P550 quad-core processor based on RISC-V architecture rather than an x86 chip from Intel. The companies announced they were working together on the Horse Creek platform last year, and Intel recently showed off a Horse Creek prototype during its Intel Innovation 2022 event.
According to a report from WikiChip, the Horse Creek board features a SiFive P550 64-bit processor with four cores clocked at 2.2 GHz, a 13 stage, 3-issue out-of-order pipelines, support for DDR5-5600 memory and 8 lanes of PCIe 5.0.
It’s manufactured using an Intel 4 process, because while Intel is probably best known for designing and building its own chips, the company is also positioning itself as a foundry that can build chips for clients like SiFive… which is part of what makes the Horse Creek platform interesting.
The chip doesn’t have a built-in GPU, but Intel has demonstrated the Horse Creek developer board playing games and videos, as well as other applications, suggesting that the CPU is powerful enough on its own for basic computing tasks.
There’s no word on when Horse Creek developer boards will be available for purchase or how much they might cost.
You can find more details at WikiChip.
At the RISC-V summit at the end of last year, SiFive announced that their HiFive Pro P550 Development System (previously code-named Horse Creek) would be commercially available this Summer.
I never understood the excitement around RISC-V from consumers.
I understand chip makers being excited since they’re not restricted by/paying ARM. However, that seems like it’d result in an even bigger mess than the ARM ecosystem has been.
It’s going to be a mess like ARM was and is. Maybe more.
I know a few companies I’m working with are implementing non-standard/proprietary features into RISC-V co-processors in future products and definitely don’t plan on open sourcing code nor providing binary blobs long-term.
I wish Intel the best, but putting four P550 cores in their latest DDR and high speed IO is like putting a Volkswagen engine into a Ferrari. It shows the potential of the engine, but no one wants to pay Ferrari prices… unless you are getting a Ferrari engine too. Intel is a one trick pony. They make great SOCs for anything plugged into the wall.
Perhaps Intel are seeking to enter other sectors of the market – and take the fight back to Arm on their terms? Or at least, make some money off companies who are intent on doing so.
This processor’s got the performance of a top-flight ARM chip from 2017, I believe : not quite head-to head with Arm yet, but already sufficient for a lot of processing needs.
More like 2016 (Kirin 950-Kirin 970), as I think these are close to the performance of the ARM Cortex-A73. Even though they claim it’s superior/equal to the ARM Cortex-A75 I take that with a HUGE grain of salt, and rather wait for in-depth testing and benchmarks from trustworthy sources that are independent.
And remember, there’s very little software optimisations on there, so potentially it could be worse than the ARM architecture.
I was impressed with RISCV, but not any more. We’ve seen more theoreticals than actual prototypes, developer units, and commerical products. Kudos to Intel for driving this a little. Since it’s off to a rocky start, where RISCV may never hit critical mass. And if it does, there’s a high likelihood that it will be a proprietary device and have very little to do with the open-source community.
The whole industry has seen huge disruption, probably since 2016, what with
– the trade dispute (which continues to escalate)
– Arm’s precarious ownership and
– Tim Cook’s retiring of X86 from Apple products. (Another huge sector diminishing for Intel.)
China’s backing of RISC-V alone assures it of a significant position in the marketplace. But other countries are concerned by the US’s weaponisation of technologies too, and realise the imperative for digital sovereignty in an age when humanity relies on technology.
Manufacturers (since 2016) have recognized RISC-V to have the same ball-park efficiency as ARM, and yet are relieved not to have the authoritarian onlooker telling them what they can and can’t do.
Significant OEMs are already using, or are planning to use RISC-V for co-processor roles – and that’s only likely to increase, as more powerful SoCs become available. Industry’s membership of the RISC-V Foundation has grown enormously, and includes many of the big players.
Consumer electronics, with its myriad of apps, is obviously the hardest hurdle, but if RISC-V is the direction of travel, then the apps will follow.
Warning long rant:
But it won’t. RISC-V getting massive marketshare in Mainland China has little ramifications for India, Russia, USA, Europe, and other locales. Keep in mind, we won’t be sure of what type of backdoors or security compromises those systems will have. This is open-source in theory, but in practice it will not be so.
The most probable scenario is adoption of RISC-V as a co-processor. Which means it is not a CPU competitor (ARM or x86), so this conversation is utterly meaningless.
If your assertion was true, that RISC-V has better efficiency than ARM (or heck, even close to it). We should’ve seen a buttload of products shipped already from the entry level to the high-end. But as it is, they’re struggling to even get SBC out the door and into hands. So either its untrue, or they have catastrophic mis-management. Either case prints a pessimistic outlook for it.
With that all said, I am very impressed with the 2015-2017 technology; 16nm lithography, 8x Cortex-A73, and even Android v6. They still make for a good smartphone. It was a substantial leap from the previous 1-2 years; Android v4.4, 32nm node, and 4x Cortex A17. We saw all of those improvements come together in 2016, and finally initiate a competitive alternative to the established system (Intel 6th-gen Core-M x86_64, +14nm node, Windows 10). If RISC-V can come and replicate that 2016 level, it would be a major starting point. That’s the starting point. Because from that point forward, we haven’t actually progressed very far in 2022: 4nm node, 8x Cortex-A78, and AndroidOS (v12). The next leap will be in 2024 with true-3nm, 12x Cortex-A730, and hopefully more optimisations to Linux and Android to take advantage of it.
Response to Kangal’s “rant”
“RISC-V getting massive marketshare in Mainland China has little ramifications for India, Russia, USA, Europe, and other locales.”
India are committed to RISC-V. Russia too, for obvious reasons. The EU are talking about digital sovereignty, and have some RISC-V projects ongoing as well (one member state already has a satellite launched, with RISC-V on an FPGA). As for the US – DARPA, NASA, Google, Intel and the military, are all invested in RISC-V. The bigger point though, is that Chinais a huge producer of electronics, and their products are exported globally.
“The most probable scenario is adoption of RISC-V as a co-processor.”
Arm have already had to amend their business practices because of RISC-V’s progress in this sector. Co-processors are already in Nvidia and Samsung devices, and Apple are said to be looking at it too.
“If your assertion was true,”
It’s not my assertion : it’s Dave Ditzel’s – “We assumed that RISC-V would probably lose 30% to 40% in compiler efficiency [versus Arm or MIPS or SPARC] because it’s so simple,” says Ditzel. “But our compiler guys benchmarked it, and darned if it wasn’t within 1%.” (from an article in 2017). Nvidia said that the reason they’d chosen to develop some RISC-V processors is that it afforded them the freedom to produce more efficient designs. I heard that Arm are having to relax their grip for some design companies. Seems not for Qualcomm/Nuvia, though.
“that RISC-V has better efficiency than ARM (or heck, even close to it). We should’ve seen a buttload of products shipped already from the entry level to the high-end. ”
There are! Calista Redmond suggested the figure of 10Billion a few weeks ago, and I think that’s likely to be a very conservative estimate, given that some processors won’t be marked as RISC-V.
“Which means it is not a CPU competitor (ARM or x86), so this conversation is utterly meaningless.”
Obviously, like the ARM ISA, RISC-V started with micro-controllers, but more powerful SoCs, like the SiFive one in this article are coming along all the time, including Esperanto’s first effort. With out-of-order SoC and the vector and matrix extensions now ratified, much more powerful processors are set to arrive within the next 6 months or so.
“But as it is, they’re struggling to even get SBC out the door and into hands.”
There are any number of boards (I have a list, but won’t make the comment longer), including the SiFIve & Sipeed SBCs and StarFive & Pine64 are about be delivered. The only “struggle” was from Beagle, which was sad. There’s also the recent news of the ROMA laptop, and promise of at least 2 phones coming soon.
“If RISC-V can come and replicate that 2016 level, it would be a major starting point.”
Agreed. If I were Intel, having just lost 10% of their business (from Apple), I’d be looking to build a M2-beater, wouldn’t you?
“Keep in mind, we won’t be sure of what type of backdoors or security compromises those systems will have. This is open-source in theory, but in practice it will not be so.”
This is just scare-mongering. The reality is that Google and Microsoft do as much snooping on users of tech as the Agencies do – which is equally odious.