This looks like a new hacking project waiting to happen.

[ThunderSqueak]

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

http://www.engadget.com/2010/09/18/intel-wants-to-charge-50-to-unlock-stuff-your-cpu-can-already-d/

[slburris]

Back in the day (well, about 25 years ago), I worked in the computer
science department of a local university.  They had a minicomputer
(I think it was a VAX 11/780).  The manufacturer (DEC) had a speed
upgrade which consisted of removing a single wire from the backplane
of the computer.

This upgrade was never purchased.  Instead the wire was usually removed
and put back into place whenever the system needed service :-)

At least you had to physically change something -- it wasn't just a software
unlock like the Intel CPU, or the Rigol DS1052E.

Scott

[rossmoffett]

At least you had to physically change something -- it wasn't just a software
unlock like the Intel CPU, or the Rigol DS1052E.

Well, to be fair, for a little while the only published way to mod the Rigol DS1052E for 100 MHz bandwidth was to desolder a capacitor, that's how I did it before the hacker masses showed up to find the software method.  

[alm]

This is just stupid. I already dislike software with locked features that can be enabled by paying extra, but at least you can argue that 99% of the price of the software is in development, and distributing different versions would actually increase the price. I can't imagine developing the extra cache costs more than the die area, and it's not an issue of yield either (they are already certified to be stable with the extra cache/hyperthreading, since they don't any testing before activating it). You're just paying for market differentiation. At least for the Rigol scope, they may have more stringent specs and do calibration all the way to 100MHz.

I wonder where they store that information. As far as I know, CPU's don't have any persistent storage (unless they included some EEPROM for the first time). Modifications like this are usually done by physically cutting traces on the die before packaging. Patching the BIOS would work until the processor gets separated from the motherboard (probably not that likely given the market for low-end CPUs), assuming the BIOS updater understands to leave this area alone (like it does for SLIC certificates). Patching CMOS memory would be destroyed when it's reset or the battery runs down. Installing a service that's run at startup would preclude it from working with any other OS and require reinstallation after reloading Windows (or replacing a hard drive).

I'm not sure how much interest there will be in hacking this, since the overclocking community typically isn't that interested in low-end CPUs. It's probably not too hard if you analyze what the upgrade application does for someone with time and skill.

[NiHaoMike]

That kind of "softmodding" has been done for years already. I have had several wireless routers where upgrading the firmware "unlocked" two extra channels. And of course, the once famous Radeon 9500 to 9700 mods, not to mention overclocking...

[alm]

That kind of "softmodding" has been done for years already. I have had several wireless routers where upgrading the firmware "unlocked" two extra channels. And of course, the once famous Radeon 9500 to 9700 mods, not to mention overclocking...

The difference is that this is softmodding by the manufacturer, that's fairly new as far as I know. In cases like the Radeon 9500 or Phenom X3, it was about hardware that was disabled by the manufacturer (either because it failed a test or because they ran out of failed onces and relabeled a good part), and there was no supported way to enable this. Just like regular overclocking: If you're lucky and are comfortable with less margin than the original design, you may be able to exceed factory specs. In this case, enabling the features is supported (so it is working), as long as you pay Intel.

[Bored@Work]

ARM does this for some time, too. If you have an ARM with the Jazelle extension you are supposed to pay ARM extra to be allowed to use the BXJ machine code instruction to activate it.

Big cell phone manufacturers with Java on their phone indeed payed ARM so they could use Jazelle to speed up their Java on cell phones.

[hans]

It's funny that the news article says they are currently experimenting this on low-end chips.

Well, low-end chips are cheaper, so $50 is quite a big lot. Why not buy a better chip in the first place? This is probably an easy catch for Intel if people think they got a low-end PC and want to make it faster.

[Zero999]

I'm sure it won't be very long before there's a crack and there'll probably some non-mainstream Linux distros which instal it by default.