Hacker News Reader

Comment by jdietrich | Hacker News Reader

This is how the industry has always worked. The 486SX was a 486DX with the FPU burned off with a laser. Initially this was to make use of parts with a defective FPU, but parts with perfectly good FPUs were often disabled to fulfil demand.

CPUs are produced to a single design within a processor series. The ones with no faults are clocked the fastest and have all their features enabled. Parts with lots of flaws are sold off as crippled budget processors. AMD's triple-core processors seem a bit strange until you realise that they're a quad-core part with one faulty core.

Overclockers have known about this for years, identifying countless 'good batches' of processors and GPUs that were perfectly capable of being clocked faster, but were sold as cheap parts to fulfil demand.

If this seems like a con, then the whole IC fabrication industry is a con - artificially crippling ICs has been standard practice for decades.

View on HN

This is different. In order to be upgraded via software, the chip has to be able to function at the upgraded specs. So Intel can't take a broken batch and sell them in this type of machine anymore; they have to put the good one in no matter what.

That's what makes it seem arbitrary and annoying. When you buy a low-spec chip, you get it cheap because it could be defective. But when you buy this, it's perfectly functional, just artificially limited. You aren't helping Intel increase their yield and getting some savings for helping them -- you are just being fucked because they feel like fucking you.

I hated Intel for a long time, but then they started making good products and contributing drivers to Linux... but if they keep this sort of thing up, that goodwill is all going to erode and AMD will have to opportunity to get the enthusiast market again. (Their "we'll sue you for using the HDCP crack" is similarly goodwill-eroding, especially because there is nobody they can even sue.)

I'm genuinely curious - why do you find this to be a problem? Isn't the decision to upgrade (or not) completely up to you? How is this any different than enabling a feature set on a cisco router with a license key, or adding users to an exchange server. In both of those cases the systems are already capable of performing the function. Are you opposed to all of these classes of 'pay to enable' functionality?

Because it implies that I'm "licensing" the blob of engraved silicon sitting in my computer. How long until we're just "licensing" everything?

No, you're buying in installments.

I am not renting the privilege of using my computer from Intel; I purchased a piece of hardware they happen to produce. I own it now, and I shouldn't have to maintain a relationship with Intel any more than I should have to continue paying fees to Honda for the car I own outright.

Corporations are renting the privilege of using Exchange from Microsoft because they want a corporation to stand behind their email system. If they didn't want to maintain a relationship with a major corporation, they could just as easily have done it with open-source software.

Cisco is entitled to sell me more sophisticated software, but not to prevent me outright from using the hardware I bought the way I want to with my own software. And they don't. My company wants features that would be too expensive with a unified Cisco solution, so we run Cisco phones flashed with custom open-source firmware connected to a Trixbox.

>> But when you buy this, it's perfectly functional, just artificially limited

Sadly this line reminds me of certain phones and media players

But this is no DRM-ful app store. The upgraded computer will run the same instruction set, just faster. The mechanism to disable the bigger cache until a code is entered inhibits nothing people would be doing with their Intel CPUs today. Perhaps the homebrew microcode scene suffers, but that's a niche.

This is just simple price discrimination. They could just add $50 to the price of each chip and then people who didn't want improved performance pay it anyway. This way those who want it pay for it, those who don't don't.

Nothing wrong with AMD at the moment, they have a 12 core chip out!

From a technical perspective, you're absolutely correct. This is just the logical next step in what's been going on for 20 years.

But from a consumer's perspective, this is a big shift. It's taking what was once an open secret in the industry and shoving it right in the faces of consumers. And I wouldn't underestimate the backlash that can come from something like that.

Its still nothing new. IBM was doing this with mainframes in the '70s. The mainframes they installed typically had more horsepower than your were licensed for. When you needed more power, the IBM rep came over with an impressive suitcase of equipment, locked himself inside your server room and simply flipped a switch and then ran some diagnostics for an hour or so to further the illusion.

IBM didn't even pretend that that they "sold" you the mainframes, though; they were explicitly rented.

So you're saying that IBM got their representitives to do more than necessary to cover up the fact that the upgrade as just a flip of a switch? That they were conscious of what their customers would think - business customers that are probably far more understanding of this type of arrangement than the average consumer?

The fact that we know it's been going on a while means nothing to the average consumer. Crippling of CPUs used to be magic that happened at the chip fabs that only people connected to the industry knew about. Now consumers are being openly told that their chips have been crippled in software, and if they pay more, that magical switch will be flipped.

I don't know if there'll be a big backlash or not, but this is not business as usual. Intel's small step forward has suddenly made the pratice visible.

As recently as five years ago, IBM would do a mainframe deal sort of like this: You get X MIPS available, but only YY% of those MIPS will be turned on with the option of turning on additional computing power during the period of the deal. There was no pretense that the hardware wasn't already present.

AFAIK mainframe business still works exactly like this.

And I see no problem with it.

It is good on two fronts.

1. If you're savvy you can hack the chip thus getting more for less (Premium on your ability). 2. If you're not savvy then you can buy exactly what you're after at price point that is acceptable to you (thus ensuring a proper price/performance).

People feel cheated since they didn't get more than what they paid for? I always assumed that trade happens when two parties reach a level where both "agree" that they get what they expected out of the deal.

There is no need to go all ape on Intel -> IC manufacturing is a cutthroat business with insane risks and low margins. If Intel can profit from different people willing to pay more for the "same" thing (and yes for anyone that knows and understands Amdahl's law its perfectly clear that within a microchip generation difference between high and low end is mostly superficial - regarding to system performance).

I recommend reading Spolsky's Camels and Rubber Duckies: http://www.joelonsoftware.com/articles/CamelsandRubberDuckie...

But is this really a problem? I can buy a quad-core CPU restricted to 1.33GHz on the cheap and 6 months down the line I can put down a little more cash to get it unlocked to a 2.00GHz.

Why is this a bad change? Currently if I buy a 1.33GHz and want to upgrade later to a more powerful one, I'm buying the second processor and dropping a lot more money than I'd be putting down to unlock more processing power of my existing CPU.

This would be like selling all Kindles with 3G built in. You can save say $10 off the activation of 3G if you buy it already activated, or you can wait and pay a little more in the long run, but less up front.

I don't see this as a bad thing, but I understand why people would misunderstand as to why it is.

If you put a restricted CPU in a budget computer, it's the same as putting a flawed under-clocked CPU in a budget computer save that you can upgrade it later for less money than replacing your CPU to the equivalent. (Edit: note that it also puts less strain on CPU manufacturers as they're not producing 2 CPU's whenever someone buys a computer to upgrade, which in the long run would probably mean more quality control could be implemented in the production process meaning less flawed CPU's)

> I can buy a quad-core CPU restricted to 1.33GHz on the cheap and 6 months down the line I can put down a little more cash to get it unlocked to a 2.00GHz.

Actually, you're buying a 2.00Ghz part on the super cheap, you just aren't allowed to use all of what you bought under this kind of payment scheme. This brings up two downsides to this:

1) Consumer's know that the thing they purchased is not really under control...many people want to be able to do whatever they want with what they purchased. This is basically the same as a pay-per-use license like with software today...which most people absolutely hate.

2) It tells the market that the 2.00Ghz part is way overpriced. It's a powerful signal of how bad of a value the company is offering since you actually can but the 2.00Ghz part at the lower price, it's just artificially broken. How long till crackers figure out their way past such a lockout scheme?

This isn't how pricing works. Prices track value (in the abstract) and the market (in reality); unless the good we're discussing is sorghum, prices don't track costs.

If pricing did work this way, Photoshop Elements would have cost the same as Photoshop.

If I may ask, why is sorghum your standard example? Just because the word "sorghum" sounds so sweet (and I agree), or did you have a transformative sorghum experience as a child?

It's the least valuable commodity I can think of.

> Prices track value (in the abstract) and the market (in reality)

I think that's what I said. In the abstract, consumers will feel like they are getting gypped by paying for the fully unlocked part knowing that the exact same thing is available at a cheaper price -- part of the consumer's mindset will be "we know chip manufacturer A has to make a profit, but they're clearly making a profit on the cheaper part or they wouldn't sell it, the premium part then is just a con-job and manufacturer A is just getting greedy.

This reduces the desirability of the premium part. It'd be like selling a 6-cylinder car and a 4-cylinder car, only the 4-cylinder is actually a 6-cylinder with two of the pistons turned off (and can be enabled by paying some premium and having the other two pistons turned back on -- don't think something like this hasn't already been thought of in the car industry). People will just buy the 4-banger and take it to their local ricer shop and have the other two turned on and a type-R sticker put on the back.

In reality, it'll reduce the price of the premium part as demand shifts to the cheaper (but exactly the same) alternative OR, if the price of the premium part drops so low as to be extremely close or at parity with the slower part, people will stop buying the lower priced part. Either way, in reality, they should have just sold one part and not wasted everybody's time.

I don't work in the industry, but I'd imagine some of the cheaper budget chips don't make profit on their own--they require the more expensive non-defective chips to break even.

It's even possible for the lower-priced chips to make a loss, as long as their volume is sufficient to pay for a large share of the development costs that make the high-margin chips possible.

With the exception that Photoshop Elements can't be "unlocked" to become full blown Photoshop, your comparison here doesn't work.

I don't think the average consumer has a clear enough understanding of how a computer works that they would automatically view this in a negative light.

Depends whether by "consumers" you mean people who buy a computer as a unit, or people who just buy the CPU.

Reminds me of a 100MHz AMD DX4 I had, circa 1996. I was handed a signal 11 whenever I tried to compile a certain C program.

After going over every inch of the source code and finding nothing wrong, I'd almost given up when I discovered the Sig 11 FAQ (http://www.bitwizard.nl/sig11/) Clocking the CPU down to 90MHz, no more crashes. Turned out my gray market dealer had sold me an overclocked 90.

Is the fact that Intel is testing the market for this a sign of fabrication processes having improved to the point where yields have outgrown the binning process?

Then again, here's a thought: This test marketing is being done with ancient-architecture Pentium chips. Perhaps this is a consequence of fabricating these chips with less-dense architectures on equipment built for the tolerances required by the latest generation. I don't know enough about the process to say for sure, but I could imagine this resulting in a significant improvement in yields compared to the Pentium's heyday (or to Core i7 chips today).

These chips are not "ancient architecture" Pentiums. It's because Intel branding is fucking insane. For example, the Core i3, i5, and i7 names are meaningless, except to vaguely suggest increasing performance: some low-power-draw i5 chips are put into laptops and given an i7 label.

As for this CPU, recently Intel revived the Pentium name and positioned it as a low-performance chip positioned about midway between the Celeron line and the Core i3 line.

Now Pentium can be a low-end i3.

Intel doesn't want people to really compare which chip they are buying. They just want people to feel dirty for buying low-end chips.

Absolutely - the binning process is 100% commercially driven now. Yields are so high there's no need for this from a yield-management perspective.

Well, the simple answer is that is is a con. Disabling hardware capabilities that have already been confirmed by QA is nothing less than an attempt to subvert the laws of supply and demand. The chipmakers are artificially restricting supply of high-end parts in order to maintain the pricing structure they want. They can only get away with it because they are a duopoly, and Moore's Law means that the market forces that could eliminate this practice can't ever catch up.

Cracking encryption is the new overclocking

Reply on news.ycombinator.com