Intel trials downloadable CPU upgrades Intel is trialling a processor that can be made to run faster - if you cough up extra cash. The chip is the little known Pentium G6951. It's currently being offered to a "limited number" system builders in the US, Canada, the Netherlands and Spain on a "pilot" basis, but presumably Intel will widen availability if the scheme …
I expect it will not take long for someone to figure out how to unlock the chip without paying intel. And what can intel do about it? If they sold me a chip, it's mine. I can hit it with a hammer or run what ever program I want on it. Is intel going to try and say you don't buy the chip, just a license?
It's not really an upgrade, since the chip is crippled. You are paying to unlock something that you already have.
It would be awesome, if people were all good and trustworthy. In practice you'll have hackers who crack the scheme and sell those cracks, thieves who don't crack the scheme but claim they did, and con men who'll sell codes for CPUs that can't be upgraded. All of which will chip away at Intel's goodwill.
Yep don't think IBM ever really got punished that hard though. Everytime DOJ would request info IBM would pull up with 5 or more semi of documents and bog them down (IBM is the king of redbooks). Also I think IBM usually installed mainframes with extra cpus and memory that were disabled by default and could be turned on for small fortune. I think they could get away with it because you leased the machine from them. Not sure though.
Apparently ICL used to do similar, moving a couple of wires on the back -plane was all that was required to 'upgrade' the CPU speed on some of their machines (think Jurassic computing though)
Of course, you'd have to pay them to 're-qualify' the hardware which involved having a team of cardigan wearing, pipe smoking engineers drinking copious quantities of tea listening to Val Doonican whilst running 'tests' so it definitely wasn't a cheap upgrade in any sense of the word.
Binary Dinosaur icon please
...not quite. If the CPU is capable of doing 3Ghz instead of 2 then it should bloody well do 3Ghz from the start! Selling you an underclocked item only to try and charge you more to make it run at it's capablities later on is a con. Plain and simple. Also hackers would be trying to unlock these things to spit in Intels eye anyhow.
If they sell a CPU underclocked to you, it means that the multiplier and CAS latency is lower than it should be. Changing the settings in the BIOS would allow one to quickly bring the CPU up to stuff fast enough.
Oh, wait, only available to OEMs. I've seen how cripped BIOSes of prefabricated machines are.
I can see this as extortion from Acer and the likes. Want VT-d? Pay up. Want a higher clock speed? Pay up. Want HT? Pay up. The upgrade software basically only changes settings in the CMOS RAM that are not reachable in these restrictive BIOSes (or if it's done in BIOS, cracked versions of the BIOS). Then this will become the hottest kinds of BIOS alongside those customized BIOSes and RPC-1 firmwares for DVD drives.
>>"Selling you an underclocked item only to try and charge you more to make it run at it's capablities later on is a con"
So selling you something you choose to buy, with a performance you know about at the time is a con?
Doesn't that rather assume the purchaser is stupid enough to buy things without knowing the specifications, or that they're just *assuming* that the manufacturer is somehow not-for-profit?
If someone is given the choice of paying $150 for a fast CPU or $100 for a slower one, does it actually matter if the slower one can be optionally 'upgraded earlier?
That doesn't seem morally any different to a CPU manufacturer, (once yields on a new processor permit), making a million dies that could run at speed X, but selling some of them packaged with locked-down multipliers so they run slower, in order to produce a range of products (isn't that what they generally have done).
"They *could* have made them all fast so they *should* have made them fast" doesn't really work very well as an argument.
Even if it wouldn't have cost the manufacturer any more to make them all fast, if they did that and aimed to make the same overall profit by selling them at a single price, that increases the cost of the processor compared to the original lower-priced one.
I could equally argue that rather than selling basic and premium versions of a piece of software, the creators should only be allowed to sell the premium version, since it doesn't cost any more to press a DVD with the premium version on.
However, to make the same total profit from the same number of sales, the price of the single premium version will be higher than the original basic version, and someone who doesn't want the premium version ends up paying more.
That's also assuming that sales figures will be the same - if there is competition at the low end, deciding not to have a cheap low-end version could lose sales numbers overall.
Now, it is *possible* that having a single premium version slightly less expensive than it would otherwise have been might increase sales overall, but that's a commercial call for a manufacturer to make, not some big moral issue.
Speaking for myself (I'm not Dick BTW), I wouldn't say it's wrong from a moral standpoint - I would, however, say it's wrong from a business standpoint.
Regarding the "con" label, it seems to me that this is a question of utility (to put in Economic terms), or more generally one of value. In general people don't expect business to behave as non-profits, but they do have a - possibly irrational - sense of fairness that, IMO, would tend to be offended by this scheme. This stinks of car salesmen up-selling tactics... "oh, you want tires... well that's going to cost you extra." It's not illegal, just shady to us non-marketing, non-used car salesmen types... and one that I expect consumers would not be agreeable to. It's just a damn CPU after all, let me pay $150 for $150's market value worth of CPU and be done with it.
Ultimately this boils down to Intel's incessant need to create and enforce arbitrary tiers in the market. It reminds me somewhat of GM and their 6 brands and 100 models (disclaimer: those are just SWAGs). Maintaining all that mess on the business side incurs significant costs with little, IMO, added value to the consumer. I was on NewEgg the other day looking to price out a build and between AMD and Intel there are 88 different desktop processors (50 and 38 respective) - a scheme like this would take an already confusing catalog and compound the problem by what, 2x, 3x?
Keep it simple :)
>>"It's just a damn CPU after all, let me pay $150 for $150's market value worth of CPU and be done with it."
If you have the option to do that, what's the problem?
With chips, there have been pretty much *always* things either pessimistically marked or permanently internally throttled in order to be sold cheaper than they would otherwise have been sold, and few people seem to ever have been up in arms about how 'unfair' that was.
This idea seems no worse than internal throttling, but it gives people who bought the cheaper slower version of a chip the option of changing their mind later on.
It's not obvious that it would necessarily multiply the numbers of processors available to have a lower-end one being upgradable, since it would seem quite possible that the upgradeable slow one replaces a non-upgradeable slow one.
As for the overwhelming variety, there's a limit to how few processors there could be if a manufacturer is selling some relatively legacy parts, has a few socket types to support and/or mutiple product families including more server-oriented ones and different core counts, and a choice of speeds.
I don't think it's a big conspiracy to confuse the retail buyer, not least because I'm not sure how important the domestic retail market is, compared to the PC-maker market and other industrial/professional markets.
I don't think it's a conspiracy either - hope I didn't make it sound that way. AFAIK, GM never intended to confuse their customers with all their brands and models either, and I don't think they really did. They did, however, incur significant cost overhead by this practice for, IMO, relatively little added value for the consumer... and that's for a significant purpose. For a $150 purchase, speaking for myself at least, I don't see the need for extraneous customization options and would be turned off by the practice.
From a server-side standpoint, with more and more software vendors going to PVU licensing I really don't see this working out well (if they even pursue this for their server chips). It's already frustrating enough to do PVU licensing properly with Intel processors (with Windows OS at least, HT makes a single core show up as two, and you have to confirm that the extra procs aren't real). This would just get worse if you have an 8 core physical processor, where all the cores may or may not be enabled, and HT may or may not be enabled. It would be a mess.
That said, and I think you're spot on with the last point, the real question is how will the PC-makers react to this. I could see them offering "processor X" with some preconfigured set of options - without giving the consumers access to further customization (at least on the initial purchase)... but who knows, maybe they'll buy in with it and resell the processor upgrades. From a consumer standpoint it would make it more frustrating if Processor X from, say, Dell isn't really comparable to Processor X from HP, or Acer... and who's to say that they would even let you know exactly what you're getting.
If the PC-makers really go crazy with this I could see the PC market becoming like home appliances - the mfr's use different model numbers for each retailer so it's pretty much impossible to compare prices and know if you're comparing a 1:1.
All I can really say is that *I* don't like it.
Cheers :)
...but hasn't IBM been doing something similar for years in the midrange/mainframe space? I could have sworn I heard that they sometimes ship machines with, let's say 10 physical processors, of which a company might only use 4 and the remaining 6 are disabled unless you license them.
Anyway, doing this in the consumer market seems like a losing proposition. It still costs Intel X-amount to build the processor and the consumer will cover their costs no matter what. They're not going to sell the crippled version at below cost after all.
It's like a car manufacturer putting AC into every car they make and charging you to "enable" it. I don't see this working out so well. Not only will it tend to offend the average buyer's sense of fairness and utility... but how long do they really think they will be able to withstand the inevitable tsunami of attempts to break their protection mechanism?
On the upside, this would do well to improve competition in the x86/x64 space. "I'll take the uncrippled processor, thank you."
I was under the impression a major cost of each chip is it's design, the materials of nano-tech being - ahem - very small.
Designing fewer models means a greater run per photomask and hence lower costs overall. This *might* even mean the budget/slugged/nerfed version being even cheaper than before which is good for us.
All mainframes used to be like this - the manufacturer only made one model and 'slugged' it to enable it to be sold at varying price points while maintaining economies of scale. The engineers would come in on their regular visits (those were the days) and speed it up to get through their test suite more quickly. the operators, of course, soon found out where the relevant microswitch was 'hidden' - result 8 hours of batch work completed in 2 hours and then down the pub.
Happy days!
There's a fairly decent example of something akin to this in the wild today: Athlon X3 processors. One core was disabled either for performace/stability issues (most likely) or to provide more X3 processors due to demand. We already see a plethora of motherboards with "unlock core" advertising, and many people attempting such. The core was disabled because it may cause your computer to crash, but people still want to unlock it and get an X4 for the price of an X3. Granted, they don't buy an X3 thinking they're going to get an X4, but it's a hope and they'll likely skimp and attempt and use those savings to buy a slightly better component of something else.
However, I only see this scheme working on laptops. If a desktop is involved, in two or three years, a $50 "upgrade" could be a new CPU that performs better than simply enabling HT on a 2-3yr old processor. I'm sure a QX6600 performs better than an E6600 with HT anyway, and they're even socket compatible. I'm sure some discount store still sells the leftovers (lowest is QX8300 or so now). And mentioning socket compatibility, perhaps this is why Intel changes its socket every year? That AM2+ socket is starting to look better and better.
note the intel site only shows it for windows, and windows 7 in particular. so your screwed if you want to run something else.
however, not really that new, certain there was an article on the reg, cant find it now. but AMD selling phenom II x2 that sometimes could be unlocked to giving you a working tri/quad core. and thats done in bios, remember when I was in college reading about how you could unlock athlons with a bit of solder etc.
>note the intel site only shows it for windows, and windows 7 in particular.
Intel already releases firmware patches for processors and has for many years. The tricky thing is preventing malware from "updating" your cpu, so any OS that supports CPU patching requires a properly signed patch.
So what you are looking at for Windows 7 is the signed licence package.
Why does this seem like a bad idea? Because every transistor that is put on a chip costs money. The larger a chip is, the lower the yields are. This means that either these upgradeable chips could be sold at a profit by Intel if they were sold with the maximum capabilities available, but at the price of a chip with the minimum capabilities, or they would risk losing money on making them.
Hence, clearly the only way a scheme like this can work is if Intel is not, in fact, selling chips into a brutally competitive market on razor-thin margins. If that is the case, clearly there is an urgent need for government intervention.
1) When process yields are low, nobody minds that chips with defective cores or cache are sold as lower end models. A Phenom X4 sells for 100% over cost, and Athlon X2 might only sell for 15% over cost.
2) When process yields are high, every chip could be made into a Phenom X4 but there is still demand for Athlon X2 for use in low-end systems. System builders won't pay and don't need so much power, so some chips are made into Athlon X2s, so that AMD can sell more chips and make more overall profit.
3) To prevent remarking and other arbitrage, the disabling mechanism is a combination of shorting some fuses (permanent damage.) In addition, a code on the CPU instructs the BIOS which cores on the CPU are usable.
4) Intel has developed a software-based system that can make a static change to the CPU's configuration. "The upgrade enables changes to the firmware (driven by the Intel® Active Management Technology Management Engine in the chipset) that in turn modify the hardware."
The actual mechanism is not described but I would guess that the chipset is shorting some fuses on the CPU to *enable* the cache and hyperthreading.
This means that it is no longer necessary for the CPU manufacturers to permanently limit a CPU for marketing reasons. It's the same business model, just with more flexibility.
5) This is a positive change. It's better for the environment. If we can upgrade our CPUs there will be fewer CPUs created. If a consumer doesn't want to give Intel the full profit margin upfront, the consumer can pay later for it.
6) If people hate the idea of buying CPUs with locked features, they are not obligated to do so. I recently even paid about $10 extra to buy a Phenom X2 965 with no multiplier locks. But I do have an i7 920 with a locked multiplier, and a Pentium D 905 with locked hyperthreading, and a Celeron D with a locked cache. I got what I paid for. However, I wish this tweak of the business model had arrived earlier so I could unlock my other CPUs.
The comparison between an 'upgraded' processor is not valid.
If you buy a processor, then a year later buy the upgrade, you have a processor that is a year into it's lifespan. If you buy a new processor with the same improved performance you have a part that is at the beginning of it's life.
Yes IBM did and does do this. The difference? The IBM mainframe is leased, the OS is leased, and it's all super-locked down.. the mainframe is not sold as a general-purpose machine the way a PC is. This also helps with reliability, the mainframe already has spare CPUs installed so if a CPU fails a spare can be turned on, instead of being down one CPU until a field tech comes out. Rumor is, this is also one of the reasons IBM got involved so early in the development of Linux on mainframe -- they didn't use threats or anything, but kind of directed Linux development away from running on bare metal (to running under VM -- which is standard on a mainframe, basically every OS runs on a virtual machine) so they would not find out how to unlock CPUs for free.
In the history of computers and technology, there was...
First, there was scareware, make the poor sap think his computer is infected to the high hlit then get him to pay money to get rid of the "infection".
Then came ransonware, Write a script that would encrypt all .doc(x) .lxs(x) .ppt(x) .pdf and so on on the poor saps computer and make him pay fat $$$ for the encryption key.
Now, a new breed of ripoff-ware looms, this one however proves to be the gift that keeps on giving. we all know how intel operates and deep down inside we all know that intel will add a mechanism to disable these upgrades on CPU's that were "illicitly" upgraded, and we can safely bet that intel built in some kind of lock that prevents "downgraded" CPU's from being upgraded again without some special unlock file (to prevent the user from ripping off intel a second time, because we know that protecting the bottom line is more important that actual functionality or usability)... we also know how stellar intel is with their product design and security. With that being said, it wont take long for some asshole in the US, purporting to be from some former soviet state, to write a strain that will cripple any purchased "upgrades" and hold them hostage for large* amounts of money.
This would make a nice little side gig for a less-scrupulous intel engineer (if there was such a thing, arent they all?) or one of his family members or close friends, it would actually be the perfect little scam, the odds of being caught would be slim, especially if they built their web properly.
you know the old saying, a fool and his money... need i say more? one is born every second...
*A figure nicely crafted to maximize profit for the crook, yet be cheaper than replacing the CPU.
>>"we all know how intel operates and deep down inside we all know that intel will add a mechanism to disable these upgrades on CPU's that were "illicitly" upgraded"
To be honest, I don't see any reason why they shouldn't do anything like that, and I'm certainly not speaking as a fan of Intel or some of their *other* business practices, but as someone who tries and succeeds not to be one of their customers (though to be honest, I can't swear I'd still do that if AMD were really uncompetitive).
If someone wanted a full speed chip, they could either buy one at the outset, or upgrade a slower chip legitimately.
If they choose not to do either of those things, they can't blame anyone else if they can't succeed in reliably speeding up the chip they have. They didn't pay for full-speed functionality, so they have no entitlement to it, and no right to moan if they can't get it.
They'd be gambling that they could successfully play the system, and as such, they'd have to act like a grown-up, and take failures as being their own damn fault, just as they'd claim credit for a success.
In the past, I've made informed gambles on slower-marked parts being capable of running stably at higher speeds, in situations where if that didn't work, the part was basically just money wasted.
As things happened, I was lucky, but If I'd been wrong, I wouldn't have had any right to claim the world was unfair because it hadn't let me do what I wanted.
It wouldn't have been a case of either "I save money" or "The manufacturer is a bastard", but either "I save money" or "I lose money".
*My* success if I win, *my* fault if I don't.
Dadz is right: yields will never exactly match market demands so some parts have to be sold as lower-end products (less cache, fewer cores, no HT, lower speeds...). If there's a simple way to get back that latent capacity, that's fine. If you don't want it, don't use it: I'm not sure I would.
And regarding Will's comment on the lifespan of processors, in my experience they're never the first, second or third thing to fail in a PC (my cack-handed attempts at overclocking excluded).
Sounds about as much of a con (though differently implemented) as the 486SX, which was (in most cases) simply a 486DX with the maths co-pro turned off. You could add a 487SX co-pro, but that was actually just a rebadged (and possibly repinned) full blown 486DX - most systems had a co-pro socket for it, but some had just the one socket and you removed the 486SX altogether! Anyone upgrading had basically been conned into buying two fudged 486 processors...
And as others have noted, then there was IBM and their mainframe turbo switch, turning on hardware the customer had paid for but wasn't able to use...
If the hardware is on the chip, then enable it, if it was disabled for stability reason, then disable it, but leave a vector for people to turn it on if they want.
If I buy hardware, I am buying ALL the hardware, and for someone to not allow me to change "my hardware" is a ransom situation. If I buy a CPU, and I want to smash it will a hammer, then I will. If you don't want me doing that, then don't sell hardware. Why do people think that companies can put up a EULA and take that right away? Would you accept this line of thinking in a supermarket, "Sorry sir, I can't sell this hamburger meat to you if you are going to make tacos with it. "?
I will do, whatever the hell I want to do with whatever the hell I buy, the end. If you don't want me hacking my chip to a full speed monster, then don't put the ability to do it in there. I'll do it out of shear principle.
People need to fight this a-la-cart crap. It does not work to help individual people in any way. It completely destroys the concept of equity and value, and will leave us in a society where individuals own nothing.
When companies are allowed to own everything, and you can only "rent" a service, or "lease" a feature, everyone will be poor, and at the mercy of those companies.
>>"If I buy hardware, I am buying ALL the hardware
At the heart of it, you're not buying a specific collection of atoms, you're buying the capabilities the hardware is sold to you as having. It just happens that those effects come instantiated in a physical object.
That's why if the product fails within guarantee, you get a replacement, because you really paid for an effect, not an object.
If you made a guarantee claim and they turned to you and said "Well, sir, you still have all the atoms you bought, so bugger off", you'd probably not be very happy.
I've always bought the cheapest model that has the full compliment of cache, then clocked it past the most expensive model. If this continues I'll be able to buy the bargain basement model, open the extra cache and threads with the inevitable keygen and clock it past the fastest model.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
