Intel teases 'software-defined silicon' with Linux kernel contribution – and won't say why Intel has teased a new tech it calls "Software Defined Silicon" (SDSi) but is saying almost nothing about it – and has told The Register it could amount to nothing. SDSi popped up around three weeks ago in a post to the Linux Kernel mailing list, in which an Intel Linux software engineer named David Box described it as "a post …
This will work well if Intel can get great quality from their fabs and purposely SD downgrade capable chips for customers to then pay for enhancements later.
But everyone in the chain wants their money from delivering from end to end. Intel are cutting out a large chunk of that chain, including their own FAB’s by not constantly shipping new.
The way chips were made, I have no idea if it is still the same or not, was that you would aim to have every chip be the best possible but through imperfections during the manufacturing process some parts of the chip may not work as designed. They would test them all to bin the dies by maximum stable clock rate and functioning cores. And traditionally they used to burn silicon fuses to permanently disable parts of the chip that just did not work, and also permanently disable access to perfectly functioning parts that just did not allow the chip to fit into any of the current product ranges sold. Say a 4 core 8 thread silicon die that only had three corectly working cores (6 threads) and one partially work core (1 thread) could never be sold as a 4 core CPU, 3 core CPU is not in any product range, so they would sell it is a dual core (4 thread) CPU and traditionally use blow fuses to permanently remove access to the extra working and partially working cores.
At a guess the silicon fuses have been replaced by the Intel Management Engine, which even reports back what the chip is when queried. And this new mailbox talks to the IME to "re-fuse" disabled fully functioning cores/threads.
At the end of the day they can sell the exact same silicon inside the same package that fits into the exact same socket and then get you to cough up extra cash to enable cores that are there but you did not pay for when you initially bought the chip.
I've got an old G3 iMac I keep pondering on if I should change a couple of the resistors to bump it from a 333mhz chip to 450mhz. No chip swap required.
What I'd really be interested in with this tech is if Intel would unlock all features when the chips aged out and swapped? (sort of like HP and their iLO licenses/older raid arrays so you can find the right licence keys on out-of-support hardware and bump it up in the knowledge that buyer beware.. Hence my jerryrigged streaming server)
GE (fifty years ago) would ship computers with not all of the instructions enabled. Extra money and the field engineer would cut some wire-wraped wires.
The IBM 1401 was similar. There was one instruction, read tape binary, that the field engineers needed to run their diagnostics but it cost extra money to "enable" this already enabled instruction.
There is nothing new under the sun.
Fast forward 2030 when buying a CPU:
You will get one SUB instruction and one JMP on zero instruction for free (i.e. the CPU is Turing complete). All other instructions require additional licenses for specific instruction set groups. There are no perpetual licenses. You can only acquire a license for one month at a time with an option for a 12 month discount license.
The software defined CPU is perfectly normal. It is an innovative solution to cover the development costs of our highly effective processor products. The license system will ensure a more stable revenue stream for better innovation in the CPU space.
Please, don't be alarmed.
On the other hand, you would imagine that the future trend would likely be for an increasing proportion of "serious" processor sales to go to cloud providers where you really need the flexibility to move workloads around between identical systems.
Unless that's actually the target: Amazon or Google or whoever can get cheaper chips for no-frills workloads but run more-demanding workloads on the same hardware when required by temporarily enabling the different features, reducing the different hardware platforms they might otherwise need.
Based on the headline alone, I immediately jumped to FPGAs. Maybe Intel was bringing FPGAs more mainstream, and it could be used in things like hardware video decoders/encoders, so taht you'd have an FPGA rather than a fixed function de/encoder, and on the fly reconfigure it to the appropriate capabilities.
But no. My heart sank the more I read the article.
And…..you aren’t going to like this, but you care why?
The end price to the consumer is determined by market forces and competition. If Intel did this (they won’t, it’s stupid), but if they did then the Total Cost of Ownership will either be lower than you pay now, or they will have no customers.
We’ve got exactly this today with Deezer. Has the amount you spend annually on music gone up, or gone down, since the era of CDs?
Don't worry, I am as healthy a cynic. I'm waiting for some real reviews first.
For the intended use case for me, it's a replacement for a 2015 Macbook Pro; and therefore subject to much the same walled garden use case. That's fine for who it's intended for and what programs will be inflicted upon it.
For real work I have a power hungry ryzen / linux brick.
This is done in the real world already. I know that some manufacturers of mainframe style systems do that. You buy a license for a certain number of cores that are enabled, and can license additional cores should workload increase in the future.
Yes, it is a license to print money. No, I don't really like it. And there are other license models (like paying for the CPU time actually used) by the same vendors. In the end, market (and your company's bean counters) will decide.
This is done in the real world already. I know that some manufacturers of mainframe style systems do that. You buy a license for a certain number of cores that are enabled, and can license additional cores should workload increase in the future.It's usually referred to as "the golden screwdriver" - a technician comes onsite, inserts a 'golden screwdriver' into the hardware, and voila, more CPU/RAM becomes available. Of course, that is what it used to be 60 years ago, but these days it's entering license keys into the devices.
However, this is done usually to reduce actual incurred costs, as otherwise to install actual additional CPUs/RAM/boards would require:
1) inventory - keeping a supply laying around to ship out;
2) shipping costs;
3) site visits by (probably) multiple technicians to install those components (travel costs);
4) time - it could take hours to do that work, so paying the technicians for the hours of work;
5) possible downtime on the systems to install the extra hardware (of course, advanced big-iron could usually do this online).
So, really, the 'golden screwdriver' process did alleviate actual costs to upgradeing such hardware.
But doing this at the CPU-feature-level seems a bit, well, tight.
I'd always understood the variety of SKUs to be how Intel gets a good yield - if function A doesn't work on this chip, it gets packaged up and sold as the SKU that doesn't have that functionality.
If it won't clock up to the max, it's sold as the version with a lower maximum clock speed that it can do.
If Intel start selling licences to use specific pieces of silicon, then they'll also have to make sure all those bits actually work as advertised, even if the customer never buys a licence as a customer who rents a licence only to find it doesn't work is going to be mightily and publicly annoyed.
Label those that has all of the features working as intended as "upgradeable" or whatever the marketing term will be. The question (as @Richard12 wrote) is how many will achieve that level, how many can be sold as the ultra-performance anyway, i.e. if they have surplus stock of those they could maybe shift at a lower price, but with the chance of getting the upgrade-license-fees in the future.
I think it might be interesting for businesses in certain cases in terms of risk and finances etc.
If Intel hadn't spent 81 billion dollars in stock buybacks since 2009, far more than they spent on R&D while they have languished on 14nm, then they wouldn't "need" the new nickel and dime revenue stream. They don't, anyway, because they're getting 50 billion from tax payers to build factories. But at least their former CEO got nice bonuses and shareholders made out.
Does anyone remember the rumours (!) about Cisco/NSA backdoors?
*
Perhaps this is about multi-functional ROM on the chip, with "special software features" enabled remotely?
*
Funnily enough, it must just be a coincidence that Linux and Xeon are way up there in popularity for server environments.
*
Quote (William Burroughs): "The paranoid is a person who knows a little about what is going on."
So long as they restrict this to Xeon they don't have to worry about Apple, since they don't sell servers and have never been interested in the enterprise market.
I could see this backfiring on Intel just as ARM server vendors are trying to push themselves into the enterprise market though. That's who they have to worry about. This might be the impetus needed for ARM servers to expand beyond the cloud market and begin moving into corporate datacenters.
Apple are coin-operated. In a 52-minute infomercial the other day, they introduced very expensive laptops ( albeit juicy ); various colours of fabric for speakers; and a tiered option to have yet another thing in your house that doesn't understand you.
If Apple saw money in server-ish stuff; their would be a flanx of flim-flam men delivering a rigged demo with an IMSR backdrop lickety-spit.
As for intel, the pandemic-induced day drinking doesn't seem to have done them any favours. From what I can see, they are openly admitting that they are out of ideas; are willing to download yours into their chips and charge you for it.
Maybe I could download a RISC-V emulator into a Xeon to ease the transition to being done with both of these dinosaurs.
Intel already sells a lot of "crippled" products. They have only a handful of dies that they use to produce everything, fusing off support for some features to fit their various SKUs. The beancounters decided that rather than selling permanently crippled CPUs, they could sell crippled CPUs that can be uncrippled for a price.
And sorry, this won't benefit RISC-V. It'll benefit AMD and/or ARM. The people who believe RISC-V is going to take over the world because its "open" are the same people who believed phones running generic Linux would take over for the same reason. Instead you got Android, controlled by the world's largest advertising company and getting less open with every version.
Sorry, gotta call you on that one. RISC-V is "free and good enough". The exact attributes that led to Linux's world dominance.
An SOC wannabe company gets barrel-f*ked when it wants to put a dozen or so ARM cores into its chip; while the architectural licensees get to piss them around for giggles.
That is where the gulf lies between < fancy ARM SOC maker > and < rest of the world > lies. RISC-V can unlock that; or at least force ARM out of the per-core licensing model.
Intel are taking the SGI approach of using all their money to buy more smack to shoot up in the alley with, hoping it will bring them back to the glory days. They have maintained a downward trajectory for decades now.
AMD is interesting, since it may surf the intel compat thing for years while intel staggers off a bridge; but it needs to look to the future. That future is also known as distance yourself from x86, bets they are comfortably covering.
> An SOC wannabe company gets barrel-f*ked when it wants to put a dozen or so ARM cores into its chip
You just reminded me of something that made me smile when I saw it, about bumping up the core count.
There is a low performance minimal RISC-V called serv ( https://github.com/olofk/serv ) which is short for SErial-RISC-V at a guess. And the goal of the gateware project is to squeeze as many RISC-V cores into any FPGA as possible. He has been keeping a corescore ( https://github.com/olofk/corescore ) and the currently winning FPGA is able to run 6000 serv RISC-V CPU cores in parallel. He even has a Zephyr OS multithreaded Hello World application!
The developer does some totally insane stuff to reduce the gate count. e.g. When adding two 32-bit numbers they add them one bit at a time using one adder clocked over 32 cycles instead of the normal method where 32 adders are used and add all the bits in parallel in one clock cycle.
Intel tried this before, in 2010, see here: https://en.wikipedia.org/wiki/Intel_Upgrade_Service
This was ... controversial. I'm going to explain what Intel did, why they did it, and why people don't like it. Please note that the downvote arrow is for if I've done a bad job explaining it, please don't downvote just because you don't like what Intel are doing!
Retail shops that sell complete systems can't offer the range of CPUs that you can get online. So Intel cut a deal with certain PC manufacturers, they made a special CPU and artificially disabled some bits of it. When you buy the PC, the shop can try to upsell you a card with a code that you can use to enable them. This way, you can choose to buy the cheaper CPU with less cache, or the faster more expensive CPU with more cache and hyperthreading. And the store doesn't have to keep two different models of expensive PCs in stock. This is good for the store, they can potentially make more profit.
Note that Intel "bin" their CPUs. This means they make a lot of CPUs, test and sort them based on the working features and the speed they work at, then label each chip with the best part number that that particular chip is capable of. "Best" meaning most profit for Intel, obviously. If they are making too many expensive chips and not enough cheap chips, they would normally permanently disable some working parts of the chip, so they can label it with a "worse" part number. Presumably in the case of Intel Upgrade Service, they did that but disabled the working parts of the chip in a special way so it could be re-enabled with a code. They could then sell that for the price of the "worse" chip, but potentially make more money later when they sold the "upgrade".
However, many people didn't like this. They felt that if they'd bought a chip which had certain features inside it, they should be able to use those features without having to pay more to enable them. It brings up a whole host of policy issues. It's pretty much "DRM for hardware", with all the politics around that.
Intel's attempt in 2010 was limited to one processor, and was extended to another three in 2011, but it was discontinued later in 2011. I guess Intel want to try again.
This is something IBM has done with their mainframes -- they are sold over-provisioned (extra CPUs, extra RAM, etc. already installed but inaccessible until the user pays for it).
And more specifically related to CPU features... At some point (around 2000 I think) as they saw interest in running both the existing mainframe workloads and Linux, they found a few CPU instructions they can disable that Linux does not use* but mainframe OSes do, and from then on customers have been able to pay some amount to enable an additional CPU or some lesser amount to enable it for Linux workloads only. In other words, paying extra to enable some additional instructions.
*I suppose gcc or clang don't generate these instructions, so Linux kernel + applications are not going to use them.
Hmm, a modern CPU can have anywhere from 10-35 billion transistors, and take 3-4 months to manufacture starting from a polished wafer. I'm having trouble envisioning any CPU having significant portions of these precious transistors sitting non-functional in the anticipation that a customer *might* decide somewhere down the line to enable additional features. *Maybe* take a gamble with 10% of available transistors. But with global production constrained and customers lined up for products, the vendor (AMD, Intel, etc.) would be much better off using those idle transistors to make additional CPUs.
But the people running these businesses are undoubtedly much smarter than I am, and likely have decades of accumulated data showing which features customers are likely to want.
If it can be used to screw more money out of the customer base overall Intel will have no problem at all with that.
What the customer base thinks about it is another matter, of course.The development work has been done. The production work has been done. If the price to the customer has not paid for all that it looks like the product has been deliberately sold at an initial loss to sucker the buying into paying for an upgrade that isn't an upgrade.
The mainframe industry was notorious for that and I doubt their customers admired them for it. Then the microprocessor came along and started eating away at the corporate computer market.
Who profited from that? Intel. Those who don't learn their history are doomed to repeat it but it seems Intel are going to play the role of the mainframe makers.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
