Intel wants to charge $50 to unlock stuff your CPU can already do (2010) | Hacker News Reader

Intel wants to charge $50 to unlock stuff your CPU can already do (2010)

engadget.com

26 pointsvishal012367 comments

I assume this is being posted because something similar has been found on Haswell, or whatever.

Complaining about this is just not knowing what is going on. Developing a processor is a very complicated matter, and it's financially impossible to develop several models for all the possible different performances. So they just design and build the best possible device (with exceptions of course).

Now, I want a not-so-powerful device. Why should I pay the price of all the R&D it took to squeeze the top performance? Maybe I just won't buy it because, hey, it's too expensive.

On the other side, I want the top of the line. And Intel spent lots of money developing it, in practice raising the price for everyone else. Shouldn't I be made to pay more to compensate for that?

At the end of the day is a matter of creating artificial tiers so different needs pay more or less of a share of the R&D and production cost. Something similar happens when paying $100 for an upgrade from a 16GB to 32GB iPod. The NAND Flash chip costs about $5, but I'm paying the premium of the extra investment the company did, the additional supply chains, etc.

Of course this just is invisible to the customers most of the time, and this 'upgrade' cards make it obvious now. Anyone familiar with testing equipment (oscilloscopes, etc...) will know that is a very common marketing technique.

Uh-uh. This is called market segmentation and it's as old as the street. Haven't trains had a first and a second class for centuries?

It's great when I'm not interested in the top tier, I'm really glad the prices for my train tickets are (probably) lower than they would be because I'm being subsidised by people in first class. But I do like having the best CPU and a subconscious part of me feel cheated.

I know how it works, but I understand people feel cheated.

To add to this analogy: Software companies have been doing this for years, develop a pro version and then remove features for an entry level version rather than redevelop a new entry level application.

Example: Cubase historically had 3-4 versions.

Usually a lite/beginner version that severely limited the number of tracks/FX and a few views, a middle range one that included most features, and the Pro version that was unrestricted that featured things that professionals would need (like multiroom mixing so musicians could hear seperate mixes while playing) and then the spin-off Nuendo iteration which largely reused Cubase's codebase but designed for video-post production mixing.

Of those 4 products there's only really two "different" products.

A similar analogy would be cars that are de-tuned for sale to the general public. Modern engine are quite capable of creating far more horsepower and torque than what is available 'from the factory'. You can step up and buy the RS version of the same car with basically the same motor, or you can look for aftermarket workarounds such diesel engine programmers.

In the olden days all chips were created to reach a specific performance level. Those that did not reach that level were simply relabeled as lower performance chips. This is effectively what Intel is doing now. But with improved manufacturing processes fewer and fewer chips 'fail' at testing. So the failure rate is manufactured as well to get the most out of every production run.

Trains are slightly different. First class gives you nicer seats, and potentially other benefits as well, the same scenario would be if first and second class passengers were all together, but first class were allowed off first. Of course, first class could be subsidising second. But it's also possible to structure prices such that it is the other way around.

It would be more like if all the seats on the train had a reversible cushion with a soft side and a hard side - but you had to pay an extra twenty bucks to turn it over.

Why do we need analogies? This situation is not that complicated to understand. Let's skip the inevitable argument over whose useless analogy is better.

More like if you opted out of paying the extra $20, the steward would come by and put stiff boards under your butt to ensure you cannot derive any utility from the cushion that's already there.

If this was framed as charging twenty bucks less for those willing to endure the hard side of the cushion, people would be proclaiming anyone sitting on soft cushions as stupid.

The analogy works, because the primary thing you're paying for on the train isn't comfort, it's being transported. Second class passengers (even the bloody riff-raff in third!) arrives at exactly the same time. Also, first class passengers are let off first, at least on planes.

> the primary thing you're paying for on the train isn't comfort, it's being transported

Depends how you define "primary", and also depends on how tickets are priced.

I don't know enough (or, indeed, anything) about public transport costs/margins/etc. but let's say second class costs $5, first class $10, a nicer seat costs $100 more than a cheap seat, and seats last for 20 journeys before needing to be replaced. In that scenario, the extra cost for first class seat is purely spent on the actual upgrade cost, nothing to do with the fact that you are traveling somewhere.

Of course, in my random price list, the upgrade price (i.e. difference between the two options) covers 100% of the actual cost - in real life, maybe it covers 200% and is therefore subsidising cheaper tickets, or maybe it covers 1% and is being subsidised by cheaper tickets. But wherever it falls there, at least some of the cost is going towards the actual, physical cost.

I define "primary" by observing that practically nobody (save for a very few enthusiasts) ever get on a train or a plane except because they want to arrive at the destination. Riding on a train or a plane to enjoy the journey itself is very rare.

Subsidising:

Though first class represents less than 5 percent of all seats flown on long-haul routes, and business class accounts for 15 percent, those seats combined to generate 40 to 50 percent of airlines’ revenue

http://www.nytimes.com/2011/11/21/business/taking-first-clas...

If I buy a beer for $3 on a train, is the primary purpose of that $3 "transport", because practically nobody gets on a train just to drink beer? What if that $3 is merged into the total cost of your ticket and you get the beer included?

The fact that it seems airlines do underchage for cheap seats and overcharge for expensive ones is irrelevant - I already pointed out that ticket sellers can go in either direction on this. The point I am making is that you are getting a tangible thing that is more expensive (e.g. a nicer seat) - whether you are being ripped off or not is another matter.

Or, in reality, they are NOT subsidizing the cheaper seats... they simply mean that the cheap seats generate 10% profit, and the more expensive seats generate 110% profit. No subsidizing necessary. There's a big difference between subsidies, and product classification. In subsidizing, something is being sold at less than actual cost, in classification, some products make more money.

I see. I think it doesn't cost more to furnish first-class than second class here, they're practically the same, apart from the colour: ugly blue in second class, warm brown in first class. The major difference is that you're not allowed to make noise in first class, which seems like an arbitrary restriction. But it's true the comparison is not perfect.

In the UK the seats are usually totally different (e.g. http://img3.photographersdirect.com/img/13886/wm/pd3295300.j... vs. http://farm4.staticflickr.com/3566/3467484793_42379b7d56.jpg) - with the most noticeable aspect being that you will fit less first class seats into one train carriage, due to size.

The problem with Intel's tiers is that you can't even buy the top of the line device: the concept doesn't exist due to the artificial market segmentation.

For Haswell, you can buy the 3.4Ghz one with TSX (4770), or the 3.5Ghz one without TSX (4770K).

You cannot have the faster speed and TSX. Why? No reason. Well, someone at Intel figured that people that want to have fast multithreaded software must be business users that don't care about clockspeed. Or something. Don't ask me, I think it's the most bonehead retarded thing ever.

Maybe it's just that in that case the silicon won't support it, we don't really know how the architecture is. It's also possible that it heats too much with both things. Who knows.

Indeed it would be stupid not to offer something, just because. Charge $2000 for it and someone will buy it.

Maybe it's just that in that case the silicon won't support it, we don't really know how the architecture is.

Indeed it is conceivable that transactional extensions wouldn't work when the clock multiplier is unlocked, but I can't yet imagine why.

> Why?

My guess is that TSX being a new feature designed for multicore scalability, they're going to pay very close attention to bug reports and crashes on these new systems. They really don't want a bunch of overclockers out there generating mysterious crashdumps. For example, https://blogs.msdn.com/b/oldnewthing/archive/2005/04/12/4075...

(hey gcp!)

Thanks for pointing that out, that sucks. I'm glad I'm not much of an overclocker. Also a shame that they don't seem to have plans for any chips with >4 cores, and they're still disabling hyperthreading on a lot of models.

It probably won't stop me from buying a Haswell, I'm drooling for all those new instructions...

Well, you can, it's called the Xeon 1285v3 (3.6 GHz w/ TSX) and it costs $662 (almost twice as much as the 4770K).

That would explain it. Do those Xeons have the same instruction set as the new i7s?

Yep. They're Haswells.

That actually makes sense, I think. Imagine how simpler and more cost effective it would be to make a single processor tiered at 4 levels via this method versus having the entire fabrication chain for four processors. If the prices are right, then I could buy the lowest tier, and a few years later pay the upgrade and get nearly a whole new processor. As long as the pricing is fair, and some cost savings are passed on to the consumer, this could be win-win, right?

While i agree in part with the idea of spreading R&D cost via locked functionality, i am not sure it's a good idea to legalize it without any regulatory body. And forming these bodies is no trivial matter, So overall am against this idea, simply because i don't want to trust Intel/<other Big Corp(that's close to monopoly)> to pass on meaningful cost savings to the consumer.

Your comment doesn't make sense, the locking functionality is perfectly legal and is actually a net-gain for the consumer (they now have the option to upgrade when otherwise they would have received a hardware locked chip).

Intel can afford to do this simply because they don't face enough competition. Otherwise they would have to give you their very best at competitive prices, or lose clients to the competition. But for several years already this hasn't been the case for the mid-high home range.

It's a worrisome situation.

> Intel can afford to do this simply because they don't face enough competition.

The fact that there will soon be more ARM chips deployed for daily use in the world than Intel ones (if not already) stands in pretty stark contrast to this statement.

They have not competition in the mid-high range home computer sector. I specified as much in the comment, if you read it all.

Intel's Haswell lineup takes market segmentation to a pretty extreme level. There's 22 Haswell cpu models on the lineup. They have clearly targeted some chips for gamers and others for server use.

You can't get an "unlocked" cpu with virtualization, but you can get the same chip locked with virtualization extensions. I guess the purpose is to avoid customers buying unlocked chips and overclock them for a diy server setup and get them to buy the more expensive server chips.

Process binning I can understand, it's a way to try to improve the silicon yield. But artificial crippling, when taken to this kind of extremes feels a bit dishonest to me. But that's just the way the semiconductor industry works, I guess.

>>You can't get an "unlocked" cpu with virtualization, but you can get the same chip locked with virtualization extensions

The 4770K has the core virtualization function VT-x and VT-EPT, just not VT-d. I believe this has often been the case for desktop vs server CPUs; Intel needs to distinguish its desktop vs. server offerings after all. The 4770 is locked and gains VT-d and TSX-NI.

http://ark.intel.com/products/75123/Intel-Core-i7-4770K-Proc...

1. intel likes money.

2. if you want the best (general purpose desktop/server) silicon, you have to buy intel.

3. if you don't have as much money as intel likes, it generously offers you a, um, de-featured version of their silicon at a, erm, discount.

Presumably AMD has all their chip engineers engaged in trying to make their chips more performant.

Yes, they have a legal right to do this.

And we have a legal right to say it's exploitative and immoral.

I own the physical device, but you control how I use it? I'll be looking for a competing product that doesn't do that, thanks.

It's not evil, it's the most effective way to provide a CPU capable of meeting everyone's needs at the cheapest possible prices.

If they had to design and manufacture a chip at every tier it would be far more costly, and they would have to pass those costs on to the consumer.

They did design and manufacture a chip that is completely capable of all of the functionality - the economies of scale are completely in their favor. And now they want to make more money by pretending as though the chip is not capable.

In fact, it took them even MORE effort to design and manufacture in the ability to ACT STUPID, until you pay up.

OK, let's imagine Intel only sold fully enabled chips, so their cheapest processor is now $300. Do you want that? Does your mom want that?

I live in Minneapolis, which is a hub city for Delta airlines. It is quite often MUCH cheaper for me to drive to Chicago, and get a flight that connects through Minneapolis to my final destination, than it is to book a flight from Minneapolis to my final destination. Yes, it makes financial sense for Delta and Intel to do this, from their perspectives. As a consumer, it sucks, and you can't convince me it doesn't suck.

Intel is offering to sell intentionally disabled hardware at a lowered price, because they want the opportunity to up-sell the more advanced features, later.

And I'm perfectly within my rights to say, "Hell, no. And no one else should buy that product, either. If we all reject it, they'll abandon this ploy."

If Intel can manufacture a chip capable of X, Y, and Z, then that's awesome. Don't go out of your way to disable Z, in an attempt to squeeze more money out of me. It doesn't COST THEM ANYTHING to have Z enabled on my chip. It does in fact cost them to add hardware to artificially disable Z.

Intel has figured out how to mass produce Mustang Shelby GT 500's and sell them at the price of a Ford Fiesta... with the ability to "unlock" 600 horsepower.

In order for Intel to be a successful business, they need to sell to every level of the processor market. They need a chip at the 700MHz tier, a chip at the 3.5GHz tier, and dozens of tiers in between. It's expected. It's necessary.

It is absolutely infeasible for them to design and manufacture individual chips at every one of those tiers. They cannot set aside entire factories and supply lines for every level of consumer need. If they were to do so, the price of 700MHz chip would start at $300+, and only go up from there. That's the nature of the economy of CPU design and manufacturing.

Now, instead of doing that, they only design four or five chips, each at the top end of their respective tier, and build into each of those chips a means of "hobbling" them to fill the dozens of tiers in between. This allows them to spread the total cost of the chip across several segments of the market, particularly into those segments that would otherwise be unable to afford to purchase a chip at all.

This does not mean they put additional effort into hobbling those chips, it means they saved a great deal of effort by not explicitly designing and manufacturing a chip that would otherwise fill that specific niche.

What you're asking for is that they only offer chips that include features X, Y, and Z. Those chips cost $600, and there are no other options, because it would not be cost-effective for them to design a chip that only included features X and Y.

Let's really do this conversation from step one. Write down the words you'd like me to say, that would satisfy you. I think you want me to say, "They have a right to sell this product." I've already said that. What more do you want from me???

I declare it's an awful way to relate to ME as their customer to build an amazing chip, and then hobble it. It makes me resent them, and want to seek out their competitors.

Now, explain to me how those emotions they've instilled in me help make them a successful business.

Sure, that's your right as a consumer.

One of the reasons they can get away with it is because there is no competing product.

Functionally, it seems like the logical combination of binning and the market pressure for cheaper machines. (Binning-> http://en.wikipedia.org/wiki/Product_binning)

Say you are Intel and Gateway/Dell/OEM orders X lower-end CPUs... great. However, you don't have enough that binned that far down. So, you have to cripple some ones that are more capable. Do you write it off? No, offer a way to have the end-user pay for the difference! Hey user, you want to unlock the difference between what OEM paid for and what we happened to give them? Here's a painless way to unlock it.

Functionally most of the electronics you buy are capable of more. Take a look at this: http://hackaday.com/2013/03/18/hack-removes-firmware-crippli... Okay, so what happens? Either everything's great and you got a free upgrade or the thing has a bad shader or memory chip and now you have an unreliable card.

I don't agree with the whole up-sale aspect of it, but, it just strikes me as Intel trying to make back some money on what likely was a crap deal for them.

Functionally most of the electronics you buy are capable of more. Take a look at this: http://hackaday.com/2013/03/18/hack-removes-firmware-crippli.... Okay, so what happens? Either everything's great and you got a free upgrade or the thing has a bad shader or memory chip and now you have an unreliable card.

Another thing you might see in a substantially similar hardware product line is that resistors indicate the hardware model, but the lesser models are missing essential components that are present in the higher models. Sure, the firmware will think it's running on the higher model if you swap out the resistors, but it'll promptly crash when it can't find that DAC and EEPROM that are supposed to be there.

This is also something that's often seen in oscilloscopes. For example, many Agilent scopes can be upgraded by buying a key.

The impact for users is twofold. On the one hand users that don't upgrade still have to pay for a few (unused) components. On the other hand, if users decide they want to upgrade later on, they don't have to throw away/sell their 800$ scope or send it back to the factory. You can buy the basic model when you have the money and upgrade later.

It might be a bit awkward to buy something you pretty much own already, but it's still the most rational option.

If only humans were more rational...

When Tektronix aged my 'scope out of support, the last free firmware update unlocked all those extra cost firmware-locked upgrades. A real class act.

I wonder if Intel will do the same a few years down the road?

In raging over this practise, there's an unstated assumption: That the alternative is selling the full, unlocked chip at the consumer price point.

This is very unlikely to be the case. Like first class on a flight, the pricing of the high end is likely subsidising the lower end. Of course, there's still a profit margin on the low end offering, just not enough that it can stand on its own.

So the alternative to this practise is a cheaper high-end chip for everyone, it's more expensive (or same price) high-end chips, which means the low end of the market get's priced out and have to make do with last years model.

Intel is not going to sell you any chip at a loss these days... Intel makes money... they profit... loss-leaders don't give them that.

This is a weird concept and I'm trying to figure out how they intend to market it. It seems to me that people buying cpus look for the best performance at the right price, build the machine, and then 3 or 4 years later start the process over again. They won't unlock the cpu right then and there unless it fits within the price range, and they wont do it years later when they're looking to upgrade their rig because they'll haven't forgotten or there will be something much much better.

IBM has been doing this for Mainframes and pSeries/RISC boxes for a long time. Other mainframes did this as well.

On the pSeries boxes, you buy a base level of CPU cores, and you can buy licenses to unlock additional cores for a period of time. It gives you cloud-ish type features (pay for what you need, when you need it) with on-premise hardware.

It was a really common licensing model when "data processing" was all about batch workloads. Now that we're building virtualized scale-out systems versus client/server systems with fixed resource requirements, it's a model that can make sense again. If you think about it, EC2 is another way of accomplishing the same thing.

The 386SX and 386DX were exactly the same silicon, but the cheaper model had a fuse deliberately blown to disable the maths coprocessor. It all comes down to whether you take the view that you're buying a piece of silicon, or a slice of R&D.

The 386SX was a 16-bit data bus, the 386 was retroactively labelled 386DX to differentiate it.

The 486SX and 486DX were the math co-processor variants of the 486, and they weren't originally "deliberately" blown; they just didn't pass QC and were disabled. Later variants just removed the sections for the FPU from the die.

Note that sometimes this is done for partial failures, for instances if the tests for the maths coprocessor failed, they could just disable that bit and sell it at a lower price point.

Historically that happened a lot more than today however.

No, that was the 486SX/DX. No 386s had built-in FPUs - the 386SX was a DX with a 16-bit data bus and a 24-bit address bus.

Bitrot in my brain... :)

I was certain, but I still looked it up ;-). There's nothing like making a mistake on Hacker News to show one how to eat humble pie.

So did this ever blow over or are these upgradable chips still on the market? Did anyone ever hack the upgrade mechanism?

This is known in the economics literature as product crimping. Perhaps surprisingly, this can make everyone better off (not only Intel but also the purchasers of both the crimped and uncrimped goods). See "Damaged Goods", by Deneckere and McAfee [1].

[1] http://scholar.google.com/scholar_url?hl=en&q=http://www...

The only thing I can see proof of here is that Intel is a monopoly, as they wouldn't be able to get away with this if they were actually competing against each other.

Did you read the paper? They cited a case where crimping occurred in printers even though IBM and HP were competing. Also, ARM and AMD are producing chips, even if they don't target the same subset of the market.

Regardless, the point is that by crimping Intel may be making themselves and their customers better off. Sure, we could posit that with an even more competitive chip market we'd be even better off, but that doesn't negate the fact that we might be worse off than the status quo if crimping were disallowed given the current competitive landscape.

Everyone wants in on DLC, turns out... as much as we all complain about it, we still hungrily buy it.

It's greedy, and douchey, but as long as they don't bring back the Celeron I'm OK with them trying this.

It isn't the case that your average user today can tell the difference between a computer that's brand new, or that's 4 years old. Chips aren't the bottleneck anyway.

Many software vendors do very similar things: by entering a different registration key (that costs more money) you can unlock features of the existing software that were previously unavailable to you.

For example, by entering a registration key, you can upgrade "Windows 7 Professional" to "Windows 7 Ultimate" and get more features. It only takes a few moments to do, so it's clear that little or no additional software is being installed on your machine.

Man AMD had better pull up its socks & provide some competition or there will be much more of this nonsense in future.

I don't really mind this, as all it means is that I have to download a program of some torrent and -viola- better performance.

It's failed model. If the hardware is there people will find a way to unlock it without paying. Haven't license key proven that by now?

Look if you guys are even goign to argue this is a good thign , ive already given up the war on DLC has been lost and it was won by the stupid.

Reply on news.ycombinator.com