8 ms·

Calling this "Software Defined Silicon" instead of what it is - rent-seeking - is audacious...

toast0 5y ago
It's hard to have this discussion until we know what the features are. If it's hardware that implements a (non intel) patent and you have to pay to use it, but not everyone would want the feature, seems reasonable to let those who want it pay for it. Like MPEG-2 support on (early?) raspberry pis. If it's something frivalous, it's frivalous.
- mlyle 5y ago
  A main use is probably core count. AKA sell parts for a lower cost with a low core count, and when people need more performance, let them turn the additional cores on. IBM has had this model on the mainframe for quite some time. Intel gets the benefits of fewer valuable physical SKUs to manage inventory of at distributors... and of very high margin revenue later.
  - fomine3 5y ago
    IIRC Intel did almost same thing at around 2011.
  - toast0 5y ago
    I mean... that feels scummy, but on the other hand, when they've got a mask that has N cores and the yield is good, they fuse off working cores to get the lower core SKUs the market wants. There may be many customers that the IBM model works for. It used to be you could do some fiddling around and get disabled cores undisabled to see if they were disabled for good reasons or not...
thayne 5y ago
Price discrimination would also be an apt description.
kristjansson 5y ago
Intel can make parts to fit a SKU by (a) sorting them based on what elements are working when they come off the proverbial press, (b) physically disabling working elements, or (c) disabling working elements in software with the option to re-enable them. I totally understand the argument that (c) is more objectionable since the chip you buy is prevented from using all the elements it left the factory with. However I think it really shouldn't feel different than (b)? The only difference between getting 8 core part with 8 more cores burned out on its way out the door and an 8 core part with 8 more cores disabled in microcode is that you can double the core count of the latter in place for a price that's presumably proportional to the price difference between the 8c and 16c SKU? Of course there's a slippery slope to a hardware-as-a-service pay-us-x-per-unit-time-or-we-turn-off-your-CPU, but with competition from AMD on x86 and in-house ARM chips from Apple and hyper-scalers, one would think Intel wouldn't want to make their product even _less_ attractive.
- tehjoker 5y ago
  I think the issue is that the economy is supposed to in theory serve people, which most people would think means that if there is no material difference in effort to produce a part, the best most useful technology should be produced and in consumers hands. The theory of market competition says that moves like this should be impossible because it would open Intel up to being undercut by a competitor nearly instantly. However, we can see that monopoly capitalism does not function the way the legitimizing theory does. In fact, it is populated with rent seekers that manipulate the public using market segmentation strategies to profit.
  - vlovich123 5y ago
    Price discrimination happens all the time even when there's no material difference in providing the good/service (e.g. Uber surge pricing).
    - ReactiveJelly 5y ago
      At least surge pricing makes sense - when there are more fares than drivers, the fares have to pay more to be picked up first. With CPUs, it's less intuitive - The people with DRM-hamstrung CPUs are supposed to be getting a cheaper price subsidized by the people buying the product at full price. The cheaper models otherwise would have to be sold at the full price. That's not to say that Intel's prices are necessarily fair at any tier...
milderworkacc 5y ago
As a sibling comment notes, Intel can (and has a history) of doing this physically. Where a manufacturer price discriminates by offering two products, a normal one and an inferior one, it may be the case that the cheapest way to do that is by simply making twice as many “good” ones and intentionally disabling half of them. There is a very famous economics paper by Deneckere and McAfee that outlines exactly this behaviour from Intel in the 386 (…?) days, and how it may even lead to Pareto improvement - and increased consumer welfare. I can’t find it at the moment on mobile, but it should be relatively easily surfaced. Edit: found it, “Damaged Goods” - https://www2.econ.iastate.edu/faculty/langinier/teaching2005/615-fall2005/paper-hom2/damagegoods.pdf https://www2.econ.iastate.edu/faculty/langinier/teaching2005...
- akira2501 5y ago
  I thought the normal and inferior products lines are usually the result of binning, not of intentional production. In which case, it's a way to make good on products that would not normally sell under your brand and would otherwise be disposed of. If your manufacturing efficiency has advanced to the point that you no longer have low quality parts coming out of your binning process, then I think customers are right to wonder why that savings and efficiency aren't being passed on or being used to further advance the product line. In any case, if the purpose of passing the inferior products off under an alternative label is to avoid brand damage, then this seems to be the opposite of that.
  - milderworkacc 5y ago
    In many cases yes, it is the result of binning, but the 486 SX (my first comment incorrectly said 386) was an intentionally disabled 486 DX. From memory, you could then purchase a 487, which was actually just a full-fat 486 DX to slot into your co-processor slot as an 'upgrade'!
  - toast0 5y ago
    My understanding is that binning is typical at the beginning of the product line, but usually process tweaks increase the yield as the product continues. At which point things become complex. If you no longer have N core chips because all of them test good at more cores, but don't want to sell the N+1 core chips at the old N core price, you've raised prices with all that entails.
oceanghost 5y ago
It's actually, "functional pricing."
systemvoltage 5y ago
Wait until you realize that the entire SaaS industry and the subscription model is based on that same concept you're alluding to. I can understand service (say a barber) that needs to have a subscription model. But, most products today such as Adobe Photoshop should just be a one-time fee. Yet, we have come to accept that everything is $/month. You own nothing. The most eggregious of these is browser-based apps. You have zero control over upgrades and whether you want features or not. With a desktop app, atleast you can ignore updates.
- wayoutthere 5y ago
  In practice Adobe Photoshop was effectively a subscription long before the pivot to SaaS. If you used it professionally, the old versions couldn’t open files from the new versions, so if you were part of a creative pipeline that you didn’t fully control, you got pushed into upgrading within a few months every time a new version came out. The switch to SaaS didn’t really end up costing it’s pro users much more since they were spending $600 every 2-3 years on it anyway. But I suspect it did convert at least some of the people who pirated a copy from work into paying customers.
  - systemvoltage 5y ago
    Yeah, SolidWorks also does this. If you get a file from a later version from a customer, you cannot open it on the older system even if it is just a year old. Entire machine shops get sucked into this - they already run on thin margins and it is probably one of the most expensive aspects of running a shop.
  - kristjansson 5y ago
    This is where the JetBrains pricing model is exactly right. If they sold licenses, people would upgrade every few years, after some needling^W marketing. Instead, they sell as a subscription with the option to cancel and keep only the features available as of your last renewal. So someone that would upgrade every release pays about the same average cost and gets a more continuous flow of new features, someone that would buy once can just buy once, and everyone gets the optionality to stop paying and 'keep' what they've already paid for.
gumby 5y ago
IBM was famously past master at this: they’d increase your clock rate temporarily, at a price.
- kristjansson 5y ago
  They still offer stuff like this in the mainframe machines, right? I recall that their z machines have full duplicate processor drawers idle and locked out so that if one dies, IBM can remotely activate the idle one to maintain service?
mschuster91 5y ago
Why? It's the exact same thing as binning, only that now the binning is done in software.
baq 5y ago
free market at work
qeternity 5y ago
How exactly is this rent seeking?
hkt 5y ago
I for one will have no compunction about: a) abandoning CPU vendors that do this, forever if needs be b) pirating features in my CPU
- baq 5y ago
  sorry to break the news for you, but...
magicalhippo 5y ago
Our customers all get the same executable file. Yet if they've paid for it, they get more features as the license will unlock modules. The code is all there for everyone, compiled and ready in case the customer decides to upgrade their license. Is what we're doing "rent-seeking" as well? Of course we could make differentiated builds, just add in a bunch of ifdefs and provide a new executable to the user when they upgrade their license. Would that be any different? The result would be the same, the user would still pay more to get access to the additional functionality.
- microtonal 5y ago
  The cost of copying your executable is near-zero. The cost of producing a CPU die is non-zero. Suppose that ECC support takes a certain die area, then either everyone is already paying for it while buying the CPU or everyone who licenses the feature is paying for the silicon for everyone who doesn't enable it. So, if you are licensing a CPU feature, you may be paying twice for it. Of course, whether you actually do so, depends on the relative cost of producing that area of the die vs. the development cost of that functionality. Another issue is that this probably would not happen in a healthy, competitive market. In fact, the opposite would happen, CPU vendors would try to provide more features at a price point in order to compete (which is probably why AMD does support ECC on many Ryzen SKUs, as an underdog AMD still has to compete). However, if you need a x86_64 machine, you only have two options. Or if you are unlucky and required to purchase from a large vendor like Dell, you may only have one option.
  - magicalhippo 5y ago
    > The cost of producing a CPU die is non-zero. While strictly true, the actual die is a relatively tiny cost when producing a CPU. The majority of the cost is the upfront cost of designing the masks. Once you have those, using this or that mask doesn't make a huge difference. > Suppose that ECC support takes a certain die area, then either everyone is already paying for it while buying the CPU or everyone who licenses the feature is paying for the silicon for everyone who doesn't enable it. As mentioned the majority of the cost of the ECC support is in the R&D, not the physical manufacturing. As such, moving this to a software level can be a win-win for both parties. By selling ECC support to more customers, they can afford to price the bare CPU cheaper. That's exactly what happens with our software. There's no way we could sell to single-person companies at the price we do without having features behind "license walls". > Another issue is that this probably would not happen in a healthy, competitive market. In fact, the opposite would happen [...] One of the reasons people love cloud computing these days is that the cost scales with activity. Get more customers generating more activity? No problem, just add some extra capacity from your cloud provider. Need some additional functionality to service a new market? No need to pay for years for something you don't use until you need it.
wayoutthere 5y ago
I would have much more of a problem with this if Intel was still the only game in town. I get the feeling Intel is effectively ceding parts of the CPU market to ARM and focusing on their high-margin products.

It's hard to have this discussion until we know what the features are. If it's hardware that implements a (non intel) patent and you have to pay to use it, but not everyone would want the feature, seems reasonable to let those who want it pay for it. Like MPEG-2 support on (early?) raspberry pis. If it's something frivalous, it's frivalous.

mlyle 5y ago
A main use is probably core count. AKA sell parts for a lower cost with a low core count, and when people need more performance, let them turn the additional cores on. IBM has had this model on the mainframe for quite some time. Intel gets the benefits of fewer valuable physical SKUs to manage inventory of at distributors... and of very high margin revenue later.
- fomine3 5y ago
  IIRC Intel did almost same thing at around 2011.
- toast0 5y ago
  I mean... that feels scummy, but on the other hand, when they've got a mask that has N cores and the yield is good, they fuse off working cores to get the lower core SKUs the market wants. There may be many customers that the IBM model works for. It used to be you could do some fiddling around and get disabled cores undisabled to see if they were disabled for good reasons or not...

thayne 5y ago

Price discrimination would also be an apt description.

kristjansson 5y ago

Intel can make parts to fit a SKU by (a) sorting them based on what elements are working when they come off the proverbial press, (b) physically disabling working elements, or (c) disabling working elements in software with the option to re-enable them. I totally understand the argument that (c) is more objectionable since the chip you buy is prevented from using all the elements it left the factory with. However I think it really shouldn't feel different than (b)? The only difference between getting 8 core part with 8 more cores burned out on its way out the door and an 8 core part with 8 more cores disabled in microcode is that you can double the core count of the latter in place for a price that's presumably proportional to the price difference between the 8c and 16c SKU? Of course there's a slippery slope to a hardware-as-a-service pay-us-x-per-unit-time-or-we-turn-off-your-CPU, but with competition from AMD on x86 and in-house ARM chips from Apple and hyper-scalers, one would think Intel wouldn't want to make their product even _less_ attractive.

tehjoker 5y ago
I think the issue is that the economy is supposed to in theory serve people, which most people would think means that if there is no material difference in effort to produce a part, the best most useful technology should be produced and in consumers hands. The theory of market competition says that moves like this should be impossible because it would open Intel up to being undercut by a competitor nearly instantly. However, we can see that monopoly capitalism does not function the way the legitimizing theory does. In fact, it is populated with rent seekers that manipulate the public using market segmentation strategies to profit.
- vlovich123 5y ago
  Price discrimination happens all the time even when there's no material difference in providing the good/service (e.g. Uber surge pricing).
  - ReactiveJelly 5y ago
    At least surge pricing makes sense - when there are more fares than drivers, the fares have to pay more to be picked up first. With CPUs, it's less intuitive - The people with DRM-hamstrung CPUs are supposed to be getting a cheaper price subsidized by the people buying the product at full price. The cheaper models otherwise would have to be sold at the full price. That's not to say that Intel's prices are necessarily fair at any tier...

milderworkacc 5y ago

As a sibling comment notes, Intel can (and has a history) of doing this physically. Where a manufacturer price discriminates by offering two products, a normal one and an inferior one, it may be the case that the cheapest way to do that is by simply making twice as many “good” ones and intentionally disabling half of them. There is a very famous economics paper by Deneckere and McAfee that outlines exactly this behaviour from Intel in the 386 (…?) days, and how it may even lead to Pareto improvement - and increased consumer welfare. I can’t find it at the moment on mobile, but it should be relatively easily surfaced. Edit: found it, “Damaged Goods” - https://www2.econ.iastate.edu/faculty/langinier/teaching2005/615-fall2005/paper-hom2/damagegoods.pdf https://www2.econ.iastate.edu/faculty/langinier/teaching2005...

akira2501 5y ago
I thought the normal and inferior products lines are usually the result of binning, not of intentional production. In which case, it's a way to make good on products that would not normally sell under your brand and would otherwise be disposed of. If your manufacturing efficiency has advanced to the point that you no longer have low quality parts coming out of your binning process, then I think customers are right to wonder why that savings and efficiency aren't being passed on or being used to further advance the product line. In any case, if the purpose of passing the inferior products off under an alternative label is to avoid brand damage, then this seems to be the opposite of that.
- milderworkacc 5y ago
  In many cases yes, it is the result of binning, but the 486 SX (my first comment incorrectly said 386) was an intentionally disabled 486 DX. From memory, you could then purchase a 487, which was actually just a full-fat 486 DX to slot into your co-processor slot as an 'upgrade'!
- toast0 5y ago
  My understanding is that binning is typical at the beginning of the product line, but usually process tweaks increase the yield as the product continues. At which point things become complex. If you no longer have N core chips because all of them test good at more cores, but don't want to sell the N+1 core chips at the old N core price, you've raised prices with all that entails.

oceanghost 5y ago

It's actually, "functional pricing."

systemvoltage 5y ago

Wait until you realize that the entire SaaS industry and the subscription model is based on that same concept you're alluding to. I can understand service (say a barber) that needs to have a subscription model. But, most products today such as Adobe Photoshop should just be a one-time fee. Yet, we have come to accept that everything is $/month. You own nothing. The most eggregious of these is browser-based apps. You have zero control over upgrades and whether you want features or not. With a desktop app, atleast you can ignore updates.

wayoutthere 5y ago
In practice Adobe Photoshop was effectively a subscription long before the pivot to SaaS. If you used it professionally, the old versions couldn’t open files from the new versions, so if you were part of a creative pipeline that you didn’t fully control, you got pushed into upgrading within a few months every time a new version came out. The switch to SaaS didn’t really end up costing it’s pro users much more since they were spending $600 every 2-3 years on it anyway. But I suspect it did convert at least some of the people who pirated a copy from work into paying customers.
- systemvoltage 5y ago
  Yeah, SolidWorks also does this. If you get a file from a later version from a customer, you cannot open it on the older system even if it is just a year old. Entire machine shops get sucked into this - they already run on thin margins and it is probably one of the most expensive aspects of running a shop.
- kristjansson 5y ago
  This is where the JetBrains pricing model is exactly right. If they sold licenses, people would upgrade every few years, after some needling^W marketing. Instead, they sell as a subscription with the option to cancel and keep only the features available as of your last renewal. So someone that would upgrade every release pays about the same average cost and gets a more continuous flow of new features, someone that would buy once can just buy once, and everyone gets the optionality to stop paying and 'keep' what they've already paid for.

gumby 5y ago

IBM was famously past master at this: they’d increase your clock rate temporarily, at a price.

kristjansson 5y ago
They still offer stuff like this in the mainframe machines, right? I recall that their z machines have full duplicate processor drawers idle and locked out so that if one dies, IBM can remotely activate the idle one to maintain service?

mschuster91 5y ago

Why? It's the exact same thing as binning, only that now the binning is done in software.

baq 5y ago

free market at work

qeternity 5y ago

How exactly is this rent seeking?

hkt 5y ago

I for one will have no compunction about: a) abandoning CPU vendors that do this, forever if needs be b) pirating features in my CPU

baq 5y ago
sorry to break the news for you, but...

magicalhippo 5y ago

Our customers all get the same executable file. Yet if they've paid for it, they get more features as the license will unlock modules. The code is all there for everyone, compiled and ready in case the customer decides to upgrade their license. Is what we're doing "rent-seeking" as well? Of course we could make differentiated builds, just add in a bunch of ifdefs and provide a new executable to the user when they upgrade their license. Would that be any different? The result would be the same, the user would still pay more to get access to the additional functionality.

microtonal 5y ago
The cost of copying your executable is near-zero. The cost of producing a CPU die is non-zero. Suppose that ECC support takes a certain die area, then either everyone is already paying for it while buying the CPU or everyone who licenses the feature is paying for the silicon for everyone who doesn't enable it. So, if you are licensing a CPU feature, you may be paying twice for it. Of course, whether you actually do so, depends on the relative cost of producing that area of the die vs. the development cost of that functionality. Another issue is that this probably would not happen in a healthy, competitive market. In fact, the opposite would happen, CPU vendors would try to provide more features at a price point in order to compete (which is probably why AMD does support ECC on many Ryzen SKUs, as an underdog AMD still has to compete). However, if you need a x86_64 machine, you only have two options. Or if you are unlucky and required to purchase from a large vendor like Dell, you may only have one option.
- magicalhippo 5y ago
  > The cost of producing a CPU die is non-zero. While strictly true, the actual die is a relatively tiny cost when producing a CPU. The majority of the cost is the upfront cost of designing the masks. Once you have those, using this or that mask doesn't make a huge difference. > Suppose that ECC support takes a certain die area, then either everyone is already paying for it while buying the CPU or everyone who licenses the feature is paying for the silicon for everyone who doesn't enable it. As mentioned the majority of the cost of the ECC support is in the R&D, not the physical manufacturing. As such, moving this to a software level can be a win-win for both parties. By selling ECC support to more customers, they can afford to price the bare CPU cheaper. That's exactly what happens with our software. There's no way we could sell to single-person companies at the price we do without having features behind "license walls". > Another issue is that this probably would not happen in a healthy, competitive market. In fact, the opposite would happen [...] One of the reasons people love cloud computing these days is that the cost scales with activity. Get more customers generating more activity? No problem, just add some extra capacity from your cloud provider. Need some additional functionality to service a new market? No need to pay for years for something you don't use until you need it.

wayoutthere 5y ago

I would have much more of a problem with this if Intel was still the only game in town. I get the feeling Intel is effectively ceding parts of the CPU market to ARM and focusing on their high-margin products.