Intel to capitalize on "unlocked" CPU features?

Discussion in 'PC Industry' started by fellix, Sep 19, 2010.

  1. pcchen

    pcchen Moderator
    Moderator Veteran Subscriber

    Joined:
    Feb 6, 2002
    Messages:
    2,812
    Likes Received:
    217
    Location:
    Taiwan
    I think the only "secure" way to make this work is to make each CPU with its own unique serial number. Then, when you buy an "enable code," the program read your CPU's serial number and send the code with the serial number to Intel's server, and it gives you a new enable code which only works with your CPU. The code you bought is then disabled on the server to make sure it won't be used again (or probably locked to this CPU's serial number).
     
  2. Davros

    Legend

    Joined:
    Jun 7, 2004
    Messages:
    15,858
    Likes Received:
    3,060
    What!!!! Intel would never do such a thing - you crazy man :D
     
  3. rpg.314

    Veteran

    Joined:
    Jul 21, 2008
    Messages:
    4,298
    Likes Received:
    0
    Location:
    /
    If it's a pure sw unlock, then I wonder how long it will be before it is cracked and openly available.
     
  4. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,864
    Likes Received:
    364
    Location:
    35.1415,-90.056
    All the Intel processors (and I assume AMD too) have microcode update support built-in. In Windows device manager, look in your system devices for the "Microcode Update Device." In this particular sense, it's certainly software. But how it will be managed is a different story entirely, and a part that I'm not really sure about.

    What kind of DRM enforcement would they have to adopt to make sure people don't go crazy with the unlocks? Hmm...
     
  5. tongue_of_colicab

    Veteran

    Joined:
    Oct 7, 2004
    Messages:
    3,584
    Likes Received:
    759
    Location:
    Japan
    I'm not sure what Intel wants to accomplish with a ''unlock'' future. The way I see it is people buy a cpu based on a certain performance level because they want that performance and/or think it offers the best bang for the buck. Paying 50 dollars to get some extra mhz, cache or whatever doesn't seem like a good deal to me. Probably the price/performance would be at such a level that you would have been better of buying a faster model to begin with because obviously Intel isn't going to offer upgrades that in total cost mean your cpu will be cheaper and faster than a higher end model.

    I don't really see where the money comes from. Also, wouldn't this mean intel has to sell cpu's as lower end models than they could have sold them? because there needs to be room for the upgrade. I guess its no problem if intel has such good yields that most cpu's could be sold at the higher end but because intel probably sell more in the mid/low range they might get some extra cash of people who after some time decide they want some more power as they don't have the cripple cpu's they otherwise wouldn't have sold as high end models anyway but I really wonder how many people would be interrested in that.

    I kinda see it like overclocking or the unlocking of cores etc. As small group tries to save some money by hunting for models that have a high chance of unlocking or good potential for overclocking but this group probably won't buy upgrades, not to mention the group is pretty small.
     
  6. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,864
    Likes Received:
    364
    Location:
    35.1415,-90.056
    Tongue -- two things:

    First, about what they could accomplish... As a "commoner", you might buy a dual core non-AES-capable processor in your laptop today. Two years from now, you decide to upgrade to Windows 8, and then maybe spend the money for a "cpu upgrade" at the same time. You pick up the hyperthreading option, and then the AES option too. Now your Windows 8 has another ~30% performance to tap into from the extra SMT capacity, plus AES for your disk and network encryption offload. This seems reasonable to me, IMO...

    Second, selling 'high-end' cores as lower devices? They already do this. The 35w i3-330m is the same die as a 35w i7-640m, but with the HT fuses blown, 1Mb of cache fused off, the CPU multiplier changed, and the appropriate clocking fuses for the GPU multipliers blown too. They sell for radically different prices, but the commonality between the two means a radically cheaper production process -- one line for basically the entire Arrandale line (i3, i5 and dual-core i7 all with integrated video.)

    Now obviously there's binning going on, but after a while, your process should mature to the point where binning is pretty much just noise in the background. Now, if you can remove the excess physical work of popping the necessary fuzes and instead just make this a microcode change, your physical process is reduced even further to the point where nearly all processor variations of a single line can be electronically programmable rather than physically enabled.

    Not a bad business idea IMO. People who were already going to buy high will continue to do so, people who buy low will also continue to do so. But now you have a second stream of income for upgrades, an upgrade that costs you absolutely nothing.
     
  7. Blazkowicz

    Legend Veteran

    Joined:
    Dec 24, 2004
    Messages:
    5,607
    Likes Received:
    256
    I see it as a psychological incentive, making the Pentium G look bad in comparison, as an incentive to get an i3 instead.
    the upgrade's value is nuts, as there's less than 20€ between a pentium G and an i3 here (and the i3 is 133MHz faster, though that is background noise at this point)

    or they do it because they can. there were never a technical reason behind de-activating hyperthreading in the first place, probably.
    anyway it's not like we don't have choices, we've reached a point where even the low end computer parts are absurdly powerful.
     
  8. Karoshi

    Newcomer

    Joined:
    Aug 31, 2005
    Messages:
    181
    Likes Received:
    0
    Location:
    Mars
    I didnt like this new option at first, like many others. But as it's been just pointed out by alburquerque, some dies span quite a big range of products, from entry-level to upper-mid-range.

    I've been thinking about it since and it would save me money, maybe. I tend to buy component wise and not to spend too much at once, so i might buy a good mobo but a weak cpu and not too much ram, with my old gpu. Then every 6 months or so i upgrade a component for not too much money, like a new gpu every 2-3 years, or some ram, or maybe a ssd (soon, drool).

    If i could get an unlockable cpu (ideally 4-cores at max unlock) for the price of an entry-level (but above celeron) cpu, i would save money if the price of entry-level-now + mid-range-later is less than entry-level-unlockable-now < mid-range-unlock-later.

    Thinking further about it it seems what could really benefit from unlockables is the notebook market, where component-wise upgrades are not easy. Unlocking an i3 to an i5 for $25-$50 could make sense, specially if the unlock price goes down to reflect current market perf/$ and intel doesnt try to milk the locked-in (because soldered on) upgrade market.
     
  9. Frank

    Frank Certified not a majority
    Veteran

    Joined:
    Sep 21, 2003
    Messages:
    3,187
    Likes Received:
    59
    Location:
    Sittard, the Netherlands
    They do the same for cars: they use the same engine for multiple model ranges, only the motor management chip is programmed differently.
     
  10. doob

    Regular

    Joined:
    May 21, 2005
    Messages:
    392
    Likes Received:
    4
    I have hard doubts that if intel had this back in the core 2 and quads launch i'd be able to buy the exact cpu i currently have for the same price or lower (low end quad 9400) with VT-x enabled.

    One thing is knowing your product has sections fully locked to keep it functional/stable/error free but with key features from its architecture such as VT-x for virtualised environments working.
    The other is having the same product for the same price, with architectural features disabled on purpose to charge extra.(naive to think they'll sell them cheaper compared to past products of the same segment, they may at first, but im pretty convinced they'll slowly raise up, but i dont mind at all if i'll be wrong about it.)

    Maybe in a few years intel will start [strike]selling[/strike] giving away cpu's for free but, tuned to lock after X ammount of cycles and boot into the motherboards bios with internet access with a stylish preppy consumer friendly interface to intel's customer service!
    That would be great! Pre-paid, "rechargable" cpu's, now that's an awe$ome consumer friendly idea!
     
  11. Blazkowicz

    Legend Veteran

    Joined:
    Dec 24, 2004
    Messages:
    5,607
    Likes Received:
    256
    :) why not, after recycling so many ideas from the computer era before I was born and presenting them as new (thin clients, virtualization, "cloud computing" i.e. computer networks, and so on), the next trend would be to lease CPU time by the minute?
     
  12. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,864
    Likes Received:
    364
    Location:
    35.1415,-90.056
    Never worked with a mainframe? ;) Maintenance of your mainframe is billed in total processor cycles between billing intervals. It's been that way for decades, actually. It works quite similarly to cell phone plans -- you figure out about how many cycles you''ll need for a given time span, and buy a plan that fits your expected usage with a bit of wiggle room. If you don't use all the cycles? Well, you still get billed for the base rate. But if you go over? It gets expensive very quickly.
     
  13. doob

    Regular

    Joined:
    May 21, 2005
    Messages:
    392
    Likes Received:
    4
    Applying such model to the (and i was thinking just of) consumer desktop market is plain nonsense/greed in my view. If such would be applied you would no longer own the cpu, we would just be paying rent on them. We may as well apply that to all in-house gadjets, refrigerators, laundry, TV/Media centers etc... just picture the smile on the manufacturors faces! (and web forums :p)
     
  14. pcchen

    pcchen Moderator
    Moderator Veteran Subscriber

    Joined:
    Feb 6, 2002
    Messages:
    2,812
    Likes Received:
    217
    Location:
    Taiwan
    Whether it's nonsense or not does not depend on how you feel, it depends on how well it'd work. Renting looks good for the manufacturers but in reality the overhead of collecting tiny amount of money from a large crowd is very high. That's why you don't see Intel "renting" CPU cycles to consumers because it'd be very expensive to collect money (and recording usage pattern) from hundreds of millions of customers every month, or even every year. On the other hand, it works for mainframe companies because the number of customers is much smaller.
     
  15. doob

    Regular

    Joined:
    May 21, 2005
    Messages:
    392
    Likes Received:
    4
    But that already exists on mainframes, and we shouldn't even bother mentioning them.
    The scenario here is completly different, it's not about some small limited ammount of university/corporate projects that need massive cpu clusters or similar.

    They'r trying to apply the same model to a low-end PC market (desktop/laptops) testing to see if there's a profit to be made, with hopes that shortly after they can profit a bit more out of "enthusiasts" for some e-dick comparison on some benchmarks.

    Even if Intel succeeds, good for them. But from my side, im not interested at all.
     
  16. pcchen

    pcchen Moderator
    Moderator Veteran Subscriber

    Joined:
    Feb 6, 2002
    Messages:
    2,812
    Likes Received:
    217
    Location:
    Taiwan
    First, I don't see why charging for an upgrade won't work. For the consumer, they don't care whether the upgrade is "already there" or not. Upgrading with a key is much simpler and faster than upgrading to a new CPU. Your idea of "rechargable CPU" probably won't work for the reasons I mentioned, but that's not what Intel's doing. So I really don't understand your point.
     
  17. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,864
    Likes Received:
    364
    Location:
    35.1415,-90.056
    Agreed. I only posted what I did simply to point out that the concept of "rental" CPU's already exist, not to suggest that it's anything reasonable or even possible with consumer-grade parts. I agree entirely with pcchen, in that such a "cpu-rental" program would utterly fail in the consumer market.

    Now, as far as selling upgrades in this manner? I really don't have a problem with it. Again, I can already think of several reasons why this would be handy for consumers with limited budget but wanting a machine for longer duration. Buy an i3-330m laptop today, and in a year upgrade to an i7-640m with a little downloadable applet or something.
     
  18. 3dilettante

    Legend Alpha

    Joined:
    Sep 15, 2003
    Messages:
    8,363
    Likes Received:
    3,944
    Location:
    Well within 3d
    The mainframe situation is not quite the same.
    Mainframe processors are not sold in isolation. They are a component of a hardware+software platform coupled with an ongoing service contract.

    Intel is not taking on any additional duties on its part when selling these upgrade codes.
    These upgrade codes themselves are not a problem for consumers unless complications arise.

    1) There are upgrade combinations that do not match an available SKU. For example, if a 8 MiB cache i7 were not sold with hyperthreading, but a 6 MiB cache model can be upgraded to 8 MiB with HT. Since these upgrade codes have prices that are not smaller than the price gap between SKUs, this is little more than gouging.

    2 a) The upgradable chips are not distinguished from those that are not.
    2 b) The product cycle is extended with an intermediate set of core steppings.
    Certain upgrades, like clocks and cache, are influenced by the maturation of the process.
    Others may be influenced by logic bugs in a new design that are fixed in later steppings.
    a) If early stepping cores on a new process with actual defects are not distinguished from those that an be upgraded, then there is a rash of failed updates and angry customers.
    b) Separating out the non-upgradable chips as a different SKU may allow Intel to sneak an extra beat (or two...) into its tick-tock strategy.
    Logic bugs affect new designs. Process maturation affects the old design on a new process.
    On an individual scale, this is not a source of fraud, but it does mean that the overall acheivable improvement in product generations may potentially go down, since the temptation exists to milk a given design or node through a phantom SKU.

    3) Security of the upgrade process is breached. There have been some concerns about the security through obscurity that CPU microcode updates tend to rely on. A software update of CPUs to higher value feature sets adds a much stronger motive to crack what has traditionally not been verified as being safe.

    4) The upgrades are sold to consumers at large with unverifiable setups, potentially conflicting with the OS, motherboard, or platform specifications.

    The most conservative implementation is that these upgrades (and what seems to be similar to this trial run) only be sold through OEMs, for chips that are either OEM-only SKUs or specially verified.
    There would be revenue sharing with the OEM, who should have something closer to a service contract with the consumer than a part supplier like Intel.
     
  19. doob

    Regular

    Joined:
    May 21, 2005
    Messages:
    392
    Likes Received:
    4
    Buying a cpu today to upgrade a year or 2 later for something like hyperthreading and/or doubled cache is likely not worth the money spent (unless the price they charge is something very small, as in $20€) on the upgrade vs the performance increase, specially if the ram is holding back the upgrade the customer thinks will get.
    And for laptops that will just be good to help drain the battery life faster(also would the laptop be able to stably handle the extra TDP say during summer under heavy use? There would be a good chance to increase problems), unless its always plugged on.

    It is rare ammong people i know that upgrade their pc, specially a component such as the cpu a year or even 2 after they purchase it(have any of you? or someone you know? likely not many, specially in the low-end), and in my view doesn't make much sense to upgrade it after that, say 5 or 6 years as time passes further in it's a better choice to just save that upgrade money to invest in a complete new motherboard, cpu and ram.

    In the end i would spend more than just the supposed cpu upgrade to unlock some extra performance but in terms of performance return vs money spent it's a better investment, since the old system is likely being hold back by the ram's speed or chipset performance.
     
    #39 doob, Oct 21, 2010
    Last edited by a moderator: Oct 21, 2010
  20. Albuquerque

    Albuquerque Red-headed step child
    Veteran

    Joined:
    Jun 17, 2004
    Messages:
    3,864
    Likes Received:
    364
    Location:
    35.1415,-90.056
    Guess it depends. My laptop has 8Gb of ram. an OCZ Vertex 2 SSD and Win7 64; my laptop hits a CPU bottleneck just turning it on. If I had an i3-330m, I'd be waiting probably another fifteen seconds just to get my desktop. Since the initial price difference between an i3-330 and an i7-640 would be several hundred dollars, I don't expect it to necessarily be cheaper to upgrade it after it leaves the factory. But I also don't expect it to cost more than the "bare" i7 processor would, either. And maybe I only 'need' an i5-520m's worth of performance out of it, or at least maybe the upgrade cost to that level of performance makes more sense... Dunno.

    Well, the i3-330m, the i5-520m, and the i7-640m are all 35W TDP processors. In fact, they're all the exact same die with a bit of binning perhaps and some parts fused off and settings hard-wired. The i3-330m has no turboboost, no AES, 3mb of cache and runs at 2.13ghz. The i5-520m has 2/4 bins for turboboost, does have AES, 3mb of cache, and runs at 2.4ghz. The i7-640m has 3/5 bins for turboboost, does have AES, 4mb of cache, and runs at 2.8Ghz. But they're all the same die, and they all are 35W TDP parts that feature onboard video.

    As for the rest? Who are you to say what is more interesting to the entire customer base of Intel's wide range of processors? I can imagine LOTS of sales for this sort of upgrade, which is what they're REALLY after.

    It's not going to stop my personal upgrade path; I buy what I want and typically upgrade the entire machine when it's "slow". But my usage (and likely, your usage) is not the typical usage of the grand majority of PC users out there.
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...