Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10 • Hacker News Board

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#181

Post by danilocampos » Tue, Mar 19, 2013, 6:23 PM UTC

GhotiFish wrote: post #179

danilocampos wrote: post #175

Earlier quoted context omitted.

You seem really excited about being angry about this, and I'd hate to deprive of your rage-joy. So, sure.

I appreciate hacker news preventing anyone under 500 from downvoting comments. It makes sense really. People use downvotes as an analog for disagreement, but those kinds of actions obfuscate proper discussion. Really what they should be used for is dissuading bad thinking, trolling, or poor rhetoric. What do you think?

I think whatever you're getting at has sailed right over my head. But I'm glad you're having fun! That's really what it's all about.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#182

Post by cookiecaper » Tue, Mar 19, 2013, 6:34 PM UTC

jsmeaton wrote: post #125

cookiecaper wrote: post #94

Earlier quoted context omitted.

In the case of software as a service, different pricing tiers are at least semi-justifiable because you can argue that additional featuresets require more system resources to be added to the hardware infrastructure undergirding the site. If you want the more demanding features, you must compensate the provider for the extra infrastructure hit they'll be taking. Therefore, it doesn't feel as lame as intentionally perf…

As far as more demanding resources go, that's not always the case but in a lot of them it would be. Think then about the various versions of Visual Studio or even Windows (Pro/Home) etc. The argument is still the same. And with this nvidia case, the hardware hack only lights up new features in the drivers. It's exactly the same principle, and one others have shown is acceptable market practice.

Visual Studio and Windows aren't software as a service, and I suppose the argument in traditional client apps is that "the advanced features cost a lot more money to develop, so we need to recoup that cost". Neither case is as egregious as starting everything at a high-level baseline and then intentionally decreasing/damaging some parts to make them into "lower-end" models. Software is usually developed the opposite way -- you start at the baseline, get something together, and then say, "Oh, what about feature X? That'll take a while to develop..." and some guy says "Well, we can make it a premium feature". It just doesn't have the same feel as nvidia's hardware hack.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#183

Post by cma » Tue, Mar 19, 2013, 6:51 PM UTC

Peaker wrote: post #144

sliverstorm wrote: post #25

Earlier quoted context omitted.

Exactly, binning is good for the business AND for the consumer.

It's good in the sense that it's not clear there's a better alternative, but there is a painful point with a significant part of the market using a crippled design and losing value -- just to make the market segmentation possible. If there was some other way to do the market segmentation: those who derive $X of value from the chip pay $X, and those who derive $Y pay $Y, without the crippling, then more economic value…

That's actually how we primarily fund the education system these companies rely on for their R&D hiring. Which itself utilizes a whole other layer of price discrimination for need based tuition cost adjustments.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#184

Post by cma » Tue, Mar 19, 2013, 6:55 PM UTC

danilocampos wrote: post #29

sargun wrote: post #5

I don't entirely know how I feel about this hack. I feel like this is really similar to unlocking the "Professional" version of a piece of software with a serial key generator. Although, I understand hardware is dramatically different than software, it's using the same principal -- the same reusable bits to assemble many different products, and then just using tools to mask features unless a user pays for them. Then…

It's not about principles, it's about economics. This is a matter of price discrimination. Nvidia wants to be sure they have a chance to sell their gear to every single person who has money to spend on graphics cards at a range of price points. Consumers have smaller wallets than professionals, but both have similar needs. So the company performs a "hack" that allows them to extract maximum revenues from each segment…

Sennheiser may have very well manufactured that whole incident in order to sell more of their overpriced-at-any-level headphones. Marketing people love to make people think they are getting a steal.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#185

Post by nooneelse » Tue, Mar 19, 2013, 6:57 PM UTC

wtallis wrote: post #31

sliverstorm wrote: post #25

Earlier quoted context omitted.

Exactly, binning is good for the business AND for the consumer.

When "binning" refers to the practice of putting premium prices on parts that test as being capable of operating at higher clock speeds or lower voltages, that's good for the consumer. When it refers to die harvesting - selling the chip with a defective section disabled - that's also good for the consumer. When it refers to crippling a chip that has already passed QA, it's a symptom of insufficient competitive pressu…

> ...it's a symptom of insufficient competitive pressure.

The extra value a producer is able to capture from price discrimination due to a relative lack of competition can be thought of as one of the ways in which the deal between producers and society called "intellectual property" functions to shift rewards for innovation toward to the producer.

Something thought "insufficient" for one end, might be in proper indeed for some other end.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#186

Post by cma » Tue, Mar 19, 2013, 7:01 PM UTC

chii wrote: post #130

latimer wrote: post #12

Earlier quoted context omitted.

This is a good point. When you buy hardware, you are now the owner of it so they shouldn't be able to stop you from doing it, just like once you buy a phone you should be able to unlock it or install whatever software you want on it.

> just like once you buy a phone you should be able to unlock it or install whatever software you want on it. unless, of course you signed an agreement to not do that, in exchange for a cheaper price on the phone and some period of contract with a particular carrier...

Which is handled by contract law; it isn't clear why companies that want to do that should get a subsidy via criminal law DMCA penalties. Just like it isn't clear why white-only diners should have received free services from government thugs to throw out African American "trespassers".

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#187

Post by snikolic » Tue, Mar 19, 2013, 7:10 PM UTC

oh_sigh wrote: post #178

joelthelion wrote: post #123

Earlier quoted context omitted.

It's just a situation where capitalism doesn't lead to the most economically efficient solution. The company and the consumers are rational, yet for no additional cost, everyone could be benefiting from better hardware. Unfortunately I don't have a better system to propose :)

Don't you feel a little stupid for bringing up the term "economically efficient" when you don't know what you're talking about? "Economically efficient" is not equivalent to "the cheapest possible goods"

Could you elaborate in a less rude and more informative manner?

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#188

Post by comrade_ogilvy » Tue, Mar 19, 2013, 7:16 PM UTC

dhimes wrote: post #141

alpine01 wrote: post #138

Earlier quoted context omitted.

It's also worth noting that the electrical performance of your cheaper card's chip may not be as good as the higher grade chip, so may overheat more or have a failure earlier in it's life. On chip lines like these, there's usually a process called "Speed Binning" where basically when you test the performance of each chip, and you put the higher performing ones in a different "Bin" than the slightly less efficient one…

I had no idea that the manufacturing variation among chips is large enough to create performance tiers. Pass/fail I get, but this sounds like more extreme. Like, the difference between C and fail may be such that some non-negligible percentage of chips that test into the C bin should really be in the fail bin.

Keep in mind that chips can be tested at different frequencies with different amounts of input and expected output. So that C bin could have a perfect test score under certain reasonable conditions.

Clever little hardware hacks gives you oodles of geek street cred, but it may mean putting up with occasional bizarre behavior so that cred is well-earned. When silicon fails, the apparent effects can defy all logic. We can be talking about a logical AND that does something else entirely <0.001% of the time.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#189

Post by Peaker » Tue, Mar 19, 2013, 7:16 PM UTC

Sumaso wrote: post #147

Peaker wrote: post #143

Earlier quoted context omitted.

Say you have 2 audiences: Gamers and professionals. For gamers, a low performance value yields a value of $10, and a high performance value yields a value of $12. For professionals, a low performance value yields a value of $5, and a high performance value yields a value of $20. If nVidia sold a single design to all of them, with a high performance value, then gamers would be willing to pay less than $12, and profess…

However in your latter scenario, the net loss is only 2$ versus your original scenario where the net loss is 8$. (lose 8$ per card to the perfessional, versus gamer losing 2$ of performance) So its socially optimum to continue to do what Nvidia is doing ;).

In the original scenario, there is no economic value loss, there is only monetary loss for nVidia. i.e: the surplass 8$ value are captured by the professionals rather than nVidia.

Re: Resistor hack turns a Nvidia GTX690 into a Quadro K5000 or Tesla K10

#190

Post by sergiosgc » Tue, Mar 19, 2013, 7:43 PM UTC

chii wrote: post #124

sergiosgc wrote: post #111

Earlier quoted context omitted.

It isn't about just being profitable, it's about extracting maximum value from every customer. Ideally, the company would haggle with each and every customer, and get each customer to pay the highest price he in particular accepts. Since haggling is not feasible in mass markets, customers get grouped in segments, where the price is set to the maximum the segment will accept. Clearly, if you misclassify a hacker into…

Yep. in my view, if the product is already profitable at a lower segment, and they manage to sell it to the pros for a higher price, then the company is making undeserved money, rightly so, bears the anger of their customers when found out.

I can't break it to you any more easily than: You are flat wrong. Segmentation is micro-economics 101 and is an expected company behaviour.

Perhaps I can explain briefly. The "deserved"(sic) price is a value where both buyer and seller are comfortable doing business. In most cases, this is not a single value, but a range of values. At the minimum threshold the seller turns comfortable, and then there's a range where the transaction is possible and at the maximum the buyer turns uncomfortable.

Now, while for a given product and fixed seller, the minimum value of the agreement range is fixed, the maximum is dependent on the buyer. Yes, it is possible to sell all products at the minimum, but then you are giving all shared value to the buyer. Even if you are aiming for fairness, the transaction should occur at the middle of this agreement zone (the buyer buys at less than his maximum price, and the seller sells above his minimum price).

You could tackle this difference between consumers on a case by case through individual negotiation. Obviously this is not practical, so the next best thing is segmentation.

You may think this as unfair, but this is a result of seeing a half-empty glass. Segmentation allows for a company to subsidize "cheaper" products using "premium" products. If a cheaper product covers variable costs (but not fixed costs), and does not cannibalize the premium products, segmentation allows for prices below what would be possible if the burden of fixed costs had to be assigned to the cheaper segment. In industries where most of the cost is fixed (as in semiconductors), segmentation is key to achieving large volumes without compromising the ability to profit.