Intel to shut down renegade Skylake overclocking with microcode update

http://arstechnica.com/gadgets/2016/02/intel-to-shut-down-renegade-skylake-overclocking-with-microcode-update/

by pavornyoh • 10 years ago

158 127 10 years ago

arstechnica.com

grawlinson• 10 years ago

They're basically disabling functionality in order to make a quick buck.

glogla• 10 years ago

AMD Zen can't come soon enough.

kale2• 10 years ago

My Athlon 740k is getting old. I really hope Zen comes to market soon. I can't bring myself to buy a chip as a stop-gap with Zen around the corner.

I'd upgrade if I could get my hands on an AM4 motherboard. Hopefully we'll see something like the FX-6300 on AM4 early this year, so I can at least get ready to upgrade next year.

venomsnake• 10 years ago

Lets hope Mr. Keller did stellar job at AMD. Once again.

rasz_pl3• 10 years ago

Wouldnt count on that, best AMD marketing predictions promise 40% more IPC, that would still be 1-2 Intel generations behind. AMD marketing always overpromises, so we will end up with something between first-second i5 gen.

roddux3• 10 years ago

What possible reason could they have for this, apart from forcing users to buy their 'unlocked' chips? Is it really worth the nightmarish PR debacle they will undoubtedly face?

Side note (from comments) -- apparently Intel paired this microcode update with a patch to fix CPUs freezing during Prime95. Not cool.

revanx_• 10 years ago

What nigtmarish PR are you talking about? Nobody cares, just like nobody cared when they started shipping cpus with embedded Intel ME.

roddux3• 10 years ago

Nightmarish PR was probably an overstatement, but I would expect a considerable backlash from customers who will be having a free feature removed-- for no reason other than to make Intel money.

Seeing as AMD are getting back into the game with their new Zen architecture, I think this is a very unwise move by Intel.

chinathrow• 10 years ago

"embedded Intel ME."

Do you mean like embedded backdoors?

tspiteri2• 10 years ago

Overclocked CPUs run hotter. For modern technologies that means they can degrade faster, and can fail earlier than their expected lifetime. To be clear, I'm not saying this is the reason Intel did this, but it is one possible reason, and you asked for possible reasons.

scandinavian• 10 years ago

CPU's don't really break in any timespan that matters to the consumer, be that with or without overclocking. How many times have you had a CPU break? I know of no one who has broken their modern CPU due to overclocking or otherwise.

rl32• 10 years ago

Overclocked CPUs will only run hotter if voltage is increased beyond default.

Many CPUs can achieve a significant overclock without raising the voltage.

deckar01• 10 years ago

The irony is that if you are running Prime95 and overclocking, Intel should love you, because you are going to be in the market for a new processor much sooner than the average user.

Prime95 slowly cooked several PCs that I had the pleasure of supporting through my years in college.

I love what Prime95 wants to do, but consumer PCs are not usually designed/maintained with 100% CPU usage in mind.

ajmurmann• 10 years ago

Even more you can probably compare these users to the "sneaker heads" that keep the Nike brand going by waiting in line for underpriced limited edition sneakers.

alexsoft• 10 years ago

AMD adopted same strategy, no overlooking for poor!

dnlrn2• 10 years ago

CPUs are complicated pieces of technology. During the manufacturing process, some parts have a better quality grade than others. The better quality parts allow some overclocking without producing errors and therefore they get put into the overclockable K-processors. The worse parts get put into non-overclockable processors and run fine using the default voltage.

Some of the non-overclockable cpus might work fine after overclocking, some might not. Intel definitely doesn't want the negative press when some kids decides to overclock their non-K CPU and break it during the process. So I understand the decision.

nbzklr• 10 years ago

I've never heard of a CPU "breaking" during non-LN2 Level OC.

girvo• 10 years ago

I have, but that was something like 15 years ago now

onli4• 10 years ago

And that's just not true.

There would be no negative press for intel. Everyone with the shlightest knowledge about overclocking knows that overclocking can damage your parts. And like stated, parts breaking without voltage increase is highly unlikely. But still: Assume I buy an Intel Non-K processor, base-overclock it and it breaks. How on earth would I be able to produce negative press for Intel by publishing that?

It's simply a profit optimization. K-processors cost more, people who want to overclock had the option to buy non-K, that reduced sale numbers of the K line. Also the i7-6700 is clocked way below the i7-6700K, it was a nice option to get the cheaper version and up it to K level, saving 100€ for some time (prices changed).

Behavior like that is why I buy AMD.

Someone• 10 years ago

"Everyone with the slightest knowledge about overclocking knows that overclocking can damage your parts."

It is not necessarily about damaging your parts; that would be the least of their worries. Unreliability of the CPU is the real problem. Your CPU might be 20% faster, but if it incorrectly computes some number in your spreadsheet, corrupts the file, or, worse, corrupts your file system without immediate consequences, or, even worse, makes hardware (drone, self-driving car, nuclear facility) behave incorrectly, they will not just have angry customers, but likely also lawsuits started against them (yes, they might win them, but not necessarily easily; people would complain that they should have closed that hole, given that they knew it was misused)

Also, that 'everybody with the slightest knowledge about overclocking knows' is only relevant as long as overclocking remains a niche thing. If it were mainstream, many of its users would not have 'the slightest knowledge about overclocking'. This microcode update helps keep it that way.

userbinator3• 10 years ago

Have you participated much in the overclocking community? The whole point is that every CPU chip is different and can be overclocked by different amounts, some almost not at all. There is no "negative press", since anything past stock speed is a bonus which is what overclockers are trying to get. If CPUs were not working at stock speeds, that would be a reason for "negative press".

tzs2• 10 years ago

Have you participated much in the overclocking community? The whole point is that every CPU chip is different and can be overclocked by different amounts, some almost not at all.

On the other hand, there is no way that you can actually determine how far a CPU can be overclocked and still maintain full functionality, so it might be best to limit overclocking to systems that will not be used for something of high financial or safety value.

The problem is that fundamentally the hardware is still analog. Digital is an abstraction on top of the underlying analog system. In the digital abstraction, a signal changes instantaneously from 1 to 0 or from 0 to 1. In the underlying analog system, the components carrying the signal have capacitance and resistance. Changing the high voltage that represents 1 to the low voltage that represents 0, or vice versa, involves discharging or charging that capacitance through that resistance, and that takes time.

This sets an upper limit on how quickly that signal at that particular point in the circuit can change digital state.

There are also other ways the analog nature of the underlying circuit leaks into the digital realm. Neighboring components that are in the digital abstraction completely isolated from each other (except through intentional connections) might be coupled by stray capacitances and inductances. This can let signals on one cause noise on the other, or the state of one could change how fast the other can change state.

When a chip is designed the designers can figure out what areas are the most vulnerable to potential analog problems. They can incorporate into their tests checks to make sure that these areas are OK when the chip is operated in spec.

The ideal scenario is that if you clock a chip fast enough to break something, the chip blatantly fails and so you find out right away, and can slow it down a bit.

The frightening scenario is a data dependent glitch, where you end up with something like if the ALU has just completed a division with a negative numerator and an odd denominator and there has just been a branch prediction miss, then the zero flag will be set incorrectly.

oliwarner• 10 years ago

Sorry, the negative press argument is utter nonsense.

If you run any hardware outside specifications, you expect it to fail. People brick phones and ruin engines but there isn't a backlash against people trying to jailbreak their phone or modify their cars. If anything, the people that matter —the enthusiast market for these devices— are demanding that their devices be more customisable. The press and other consumers don't give two hoots about little Jimmy trying to rice 5GHz out of his $100 CPU and turning it into liquid magma. Stupid kid was stupid.

The opposite is true though. If lil Jim manages to get a $5000 part for $100, other consumers are going to factor that into their purchasing decisions.

What is most concerning is that this is a part that has been out and about for a little while. There are dozens of guides recommending certain CPUs for this that Intel are going to patch up now. The articles and their recommendations will remain out there though. It's false advertising by the back door.

daniel-cussen• 10 years ago

Overclocking isn't just a joy ride. It gives you more for your money, reducing the demand for the more premium version.

reflexing• 10 years ago

Good thing is we can load old microcode on Linux

raverbashing• 10 years ago

But then you don't get the Prime95 fix as well

throwaway77673• 10 years ago

Good thing is we can load old microcode on Linux

You cannot load old microcode anywhere. The CPU won't let you.

The OS feeds the CPU a blob, the CPU checks that it's signed by Intel (to prevent modifications), and it additionally checks that the version number is newer than the currently-running code. If it's not, it won't be loaded.

If you have a CPU with the old microcode versions, you can keep it around, but if you update your BIOS you'll find it will bring the new microcode in and you can't downgrade after boot. If you're lucky the BIOS manufacturer wasn't too careful with signing their BIOS and you can replace the microcode blob, but that's a huge hassle.

Laforet2• 10 years ago

Are microcodes written to a writable part of the CPU? I have always assume it resides in the BIOS or equivalent on the motherboard.

dnlrn2• 10 years ago

Microcode is not written to the CPU, it gets loaded on every boot. This can happen during the BIOS POST, during the OS bootloader or even while the OS is booting. Therefore, yes its possible to run older microcode (at least on Linux), since you just have to not write the newer version on boot. If the BIOS contains the new microcode, you can flash the previous version of the BIOS.

snvzz• 10 years ago

I'm glad they're doing this. It'll help plant an image: Proprietary microcode is evil.

astrodust2• 10 years ago

The alternative here is what, exactly? Can you site an example of a better situation?

na85• 10 years ago

It's "cite".

And the better solution is to finally ditch backwards compatibility with the 8086 and implement a sane instruction set that isn't just a virtualized layer à la x86

makomk• 10 years ago

Are Intel still disabling random features in the K versions of their processors for market segmentation reasons? Figuring out what processors had what features locked off was getting truly ridiculous.

onli4• 10 years ago

Yes: http://www.cpu-world.com/Compare_CPUs/Intel_CM8066201919901,... – Trusted Execution in this case.

userbinator3• 10 years ago

In the case of TXT I don't think everyone would consider it a desirable feature since it's part of the trusted computing stuff, but in the past they've also disabled more friendly features like virtualisation.

flyinghamster2• 10 years ago

One thing I've really hated about Intel in recent times has been that you can't really tell what you're getting based on the model number.

I got bitten by this when I built up a system and picked a Core 2 Quad Q8200, only to discover after powering it on that it did NOT have virtualization support.

The problem is only worse now; is that i7-xxxx dual-core or quad-core? Should I get i5-yyyy instead?

lallysingh3• 10 years ago

You have to look them up on ark:

http://ark.intel.com/products/36547/Intel-Core2-Quad-Process...

userbinator3• 10 years ago

Those supporting AMD should remember that they did a similar thing after it was discovered that some processors could have entire cores re-enabled.

On the other hand, if I remember correctly one of the mobo makers soon figured out how to use both the new microcode and the old one to get the best of both worlds; it might've been AsRock too...

Already__Taken• 10 years ago

Those supposedly triple and dual core CPUs that got unlocked to quad cores were apparently a way to sell damaged silicon that the remaining cores passed QA so they just disabled the broken core. To meet demand of these product lines perfectly good quad core CPUs got disabled and sold off as dual/triple editions. It was the luck of the draw if you got one of those and not a genuinely flawed chip.

Now, How much of that is true I don't know but that's what I remember reading at the time. A friend of mine did succesfully unlock his triple and ran for a long time.

masklinn• 10 years ago

Those supposedly triple and dual core CPUs that got unlocked to quad cores were apparently a way to sell damaged silicon that the remaining cores passed QA so they just disabled the broken core.

It's called "binning", and it's not just for defects. CPUs from the same chain can have different tolerances, one will reach 3.5GHz easy and the next one won't be stable beyond 2.8. Those are also put in different bins and sold as different models.

Now, How much of that is true I don't know but that's what I remember reading at the time.

It's completely correct. Selling a model from the exact corresponding bin is ideal, but if you have more demand than the bin provides you get parts from higher bins and gate them. A few years ago that got very very common with Intel parts as they reached tremendously low defect rates, and more or less any CPU you bought would come from the highest bins gated and undermultiplied to whichever model you'd buy.

jason46• 10 years ago

I'm still using all 4 cores in a "triple" core phenom to this day. I believe it is also overclocked roughly 400mhz, my wife sometimes plays games on it.

acveilleux• 10 years ago

This has been going on for a looooong time. Old time intel 486 "SX" chips without math co-processors (floating point unit really) were originally 486 "DX" with either the co-processor disabled or which failed QA. At some point, they made actual SX without FPU built-in.

The "stand alone" math co-processor you could add to an SX based system to gain h/w floating point was actually a full blown 486 dx CPU with a different pin layout (one extra pin) so it couldn't be used as CPU and could therefore be priced differently.

flyinghamster2• 10 years ago

I'm running such a system at this very moment. It has a Phenom II 550 Black Edition (i.e. sold to be overclockable/unlockable), and by toggling a BIOS option, I was able to unlock two more cores. I run it at rated clock speed, and it has been rock-solid since the day I built it up.

I can build a large project in NetBeans as fast on this system as on a hex-core FX-6100 (and I can watch my CPU monitor clearly showing all four cores running during the build).

CPU temperatures stay well within limits with the stock cooler.

sireat2• 10 years ago

The difference is that Intel is forcing this update (via Windows updates and by forcing mobo makers hands). Don't think AMD ever had that kind of power.

I am curious which AMD chips had the unlocking disabled?

I am still rocking AsRock 970 MB with AMD B50 which actually is an Athlon II X3 unlocked into a full blown Phenom II X4 (3 cores with no L3 cache into 4 cores with unlocked L3 cache).

AlexDanger• 10 years ago

So was it just one particular Asrock motherboard that allows this type of overclocking? Whilst my hardcore overclocking days are behind me, I have an i5-6500 and some kind of Asus mobo...I'm curious to take it for a test drive.

onli4• 10 years ago

To my knowledge all of the vendors supports that. You need a Z170 board and search a non-K bios for your model.

jsheard2• 10 years ago

Many Z170 motherboards got this functionality, but keep in mind it's not without caveats.

It breaks integrated graphics, turbo boost, all power saving states, temperature sensing, and cripples the performance of AVX instructions for some reason.

http://overclocking.guide/intel-skylake-non-k-overclocking-b...

alkonaut• 10 years ago

Can't they just disable it for yet unsold cpu:s (e.g. look at serial numbers or something)? Consumers (should) only care if their product is made worse after they purchased it. I can see why someone would be angry if a feature that was advertised is removed after the purchase.

Klathmon2• 10 years ago

IMO this would only be an issue if they advertised that it was overclockable.

If it was advertised with no ability to overclock, you should assume it can't overclock and if you are able, don't assume it will work or will stick around forever.

brillenfux2• 10 years ago

There was a time when I bought something I owned it and could do anything I wanted with it. That time is clearly gone.

rasz_pl3• 10 years ago

nah, they advertised Pentium G3258 as overclockable and later killed OC with microcode (on non Z boards)

qb459• 10 years ago

don't assume it will work or will stick around forever.

I agree with the first part, but no with the second. I don't like manufacturers deliberately breaking products which have already been sold, especially in cases when the sale wouldn't have happened if the product weren't better than specified.

Today it's Intel removing not advertised features, tomorrow it'll be Sony removing advertised ones (or had they already done that?)

lmm• 10 years ago

What if you went by a third-party review? IMO it should be illegal to remove functionality without changing the model number.

wbkang• 10 years ago

Pretty sure the microcode update is applied on every boot both by the mobo and the operating system.

xlayn• 10 years ago

There are two things for this situation

  -On one hand Intel is disabling a "Feature" of their cpus
   as a way of preventing users to get "More expensive 
   performance" without paying for it; you can think of it
   like intel is just covering their back; we can assume the
   difference on their cpus is just related to binning [0]
   and they are just protecting their customers by not 
   allowing less performant cpus perform better therefore 
   increasing reliability.
  -on the other hand you can see how hardware is not
   anymore something you buy and expect to behave in a
   certain way.
   For those paranoid... would this mean they can also alter
   how instructions behave? **cough**security**cough**

[0] https://en.wikipedia.org/wiki/Product_binning

andromeduck4• 10 years ago

Yes microcode can alter the way instructions behave and it's been like this for a decade.

What worries me more is something like Nvidia's Denver architecture because that's actually a full abstraction above machine code.

gioele• 10 years ago

What worries me more is something like Nvidia's Denver architecture because that's actually a full abstraction above machine code.

What relieves me is that RISC-V is demonstrating that opensource hardware can be a reality. This will provide people with a concrete alternative to Intel and ARM chips that does away with (closed) microcode and shadowy marketing/security procedures.

Yes, we are far away from FPGA-ing our CPUs, but I see the years 2010 as the years 1980 for FLOSS code. In the '80s a bunch of people sent around tapes with copies of EMACS, cp and mkdir; 20 years later multibillion dollars infrastructures rely on open source.

Today we use Intel microprocessors fearing that their ME components will spy on us and that secure boot will lock us in. In some years we will just install our Debian-for-HW and be done with it. It is a liberating thought.

awqrre• 10 years ago

All computing in general is somewhat broken if you are looking for absolute security (broken by design)... I wish that I see a solution to this in my lifetime...

m3rc• 10 years ago

There's truly secure computing to be found in projects like libreboot, it just takes half a decade to undo the damage done by proprietary software so your options are limited.

amluto2• 10 years ago

What happens when this microcode patch is applied at runtime on an overclocked CPU?

Remember the Haswell mess when a microcode patch disabled TSX ? If you applied it after user code (glibc) detected TSX support, boom.

qb459• 10 years ago

What happens when this microcode patch is applied at runtime on an overclocked CPU?

Paradox :)

Probably it'll reset conveniently so that you can restore default BIOS settings.

They can't let it run because that wouldn't change anything and I'm not sure if the microcode has a way to reliably switch some off-chip clock generator back to 100MHz.

rasz_pl3• 10 years ago

you get a hang, happened with Win 10 on overclocked G3258. Win10 ships with mcupdate_GenuineIntel.dll containing microcode that ... disables Overclocking

Workaround is renaming/deleting mcupdate_GenuineIntel.dll and not updating BIOS ever again to avoid new microcode, all to keep cheapest 4.4GHz single thread cpu money can get.

dbalan• 10 years ago

What are the ways in which this microcode update is enforced? Is it something that ships with your os? or something that user has to manually install like a BIOS update?

s9ix• 10 years ago

Would probably be a BIOS update distributed from the motherboard manufacturer. As the article states, people could just not update,but finding hardware with the older BIOS will be more and more difficult over time.

bradfa• 10 years ago

Why couldn't the OS just put the microcode update into the EFI bootloader?

I know some recent Intel parts can't even run the EFI firmware without first getting a microcode update, but is Skylake like this too or can the EFI bootloader run without first getting microcode downloaded?

sireat2• 10 years ago

This "software eating the hardware" is worrisome.

It used to be that at least you could depend on the hardware staying the same unless you chose to apply patches yourself.

Is it theoretically possible to change the microcode to actually add more features instead of disabling them?

Let's say Zen actually comes out better than expected and then Intel miraculously releases another update which re-enables OC ability?

quickben• 10 years ago

Microcode giveth, microcode taketh away :)

It is interesting though how Intel waited so far (not even a statement?) , probably just to sell more CPUs overall.

Osiris• 10 years ago

One reason that I've always liked AMD is they have a limited number of products so they don't play games like this. AFAIK, every feature that an architecture supports is available on chips released. The only exception that I'm aware of is "black" edition allows for an unlocked multiplier, meaning you can overclock the CPU without overclocking the BUS.

Intel to shut down renegade Skylake overclocking with microcode update

Related Stories