Hidden Cores On Phenom CPUs Can Be Unlocked

An anonymous reader writes "One of the major ways a semiconductor manufacturer manages to make the most of its chips is through binning. Chips able to cope with high clock speeds with all cores running end up as premium product lines, while others end up as models rated at lower speed grades, or with fewer cores. In the case of AMD's Phenom CPUs, dual and triple core models are quad cores with some disabled, while some newer quad core CPUs are actually six core models with two disabled. To this end both ASUS and MSI have announced that they have modified versions of AMD 890FX- and 890GX-based motherboards to unlock these hidden cores. Much like overclocking, there is no guarantee that you will gain anything by unlocking the hidden cores — everything depends on just why your CPU ended up in a certain product line."

Comments

Whatcouldpossiblygowrong

[Linker3000]

Unlocking cores that the manufacturer deems to be flawed - um, yeah.

Unless this is a rehash of when Intel were (alleged?) to be selling 486DX processors as 486SX with perfectly good maths co-processor cores disabled, I think I like my data unscrambled! /Lawn etc.

Re:Whatcouldpossiblygowrong

[qoncept]

You're showing a complete lack of understanding as to how processors are rated and sold. AMD determines they need to meet a certain quota for each model of CPU. If it works out and all of the CPUs in their 1 million unit run works flawlessly, they will maximize their profit by disabling some of them and selling them for less money to account for that market without flooding the market with their top performing part.

Unless this is a rehash of when Intel were (alleged?) to be selling 486DX processors as 486SX with perfectly good maths co-processor cores disabled ...

Uh, yeah, basically that's what the article says.

Re:Whatcouldpossiblygowrong

[Jeng]

Or even cost effective.

Pay more for a better CPU, or pay more for a better motherboard so that you can buy a not as good CPU and hopefully have the functionality of the better CPU.

Re:Whatcouldpossiblygowrong

[MBGMorden]

It depends on the maturity of the product. Often cases early on there is a legitimate need to reuse chips with flawed cores so they are disabled and sold as such. Later in the product cycle though, the demand is still there for lower cored versions, but manufacturing has often caught up to the point where there simply aren't enough flawed versions to fill demand for the limited versions. Result is that when the quota of flawed runs short, perfectly good chips are limited in the same way to fill the gap.

Later in the product line it might end up that only 20% of the lower priced chips have any flaws at all. For those people who want to tinker, it's often worth while to at least check and see if their chips will run ok when then turn the rest of them on. They stand to gain some performance if it works, and if not - eh, they paid for the slower version anyways (the only issue I take with this is when I see Negwegg reviews or forum posts claiming that they were returning the chip because it "didn't overclock far enough").

It's not something I really bother with anymore (as I've gotten older as long as the computer keeps running I'm happy), but I remember enjoying the whole overclocking scene ~10 years ago and wouldn't begrudge the new cheap teenagers of the same fun I had :).

Re:Whatcouldpossiblygowrong

[MoonBuggy]

They might not necessarily be flawed. It quite probably is a 'rehash' of what Intel were doing, and for good reason:

If all the chips come off the same line, then they might have an average cost of, say, $150. If there's a huge demand for quad-core chips at $200 and little demand for six-core chips at $350 then it's probably going to be more profitable disable two cores, bulk up the stock already consisting of chips with only four working cores, and take the $200 rather than have a chip sitting on a shelf. Thus some quad-cores are perfectly good six-cores, others aren't. They couldn't, however, afford to market all the six-core chips at $200 because the yield would be too low - there'd be nothing to do with all the faulty ones, thus pushing the average cost above $150.

Re:Whatcouldpossiblygowrong

[tlhIngan]

You're showing a complete lack of understanding as to how processors are rated and sold. AMD determines they need to meet a certain quota for each model of CPU. If it works out and all of the CPUs in their 1 million unit run works flawlessly, they will maximize their profit by disabling some of them and selling them for less money to account for that market without flooding the market with their top performing part.

True, but there's also a good possibility that the your part wasn't binned to fulfill an order. Chips go through a severe set of stress tests that often exceed what will be encountered in practical use. During these tests, it may be revealed that a core doesn't function properly or well enough (it gives bad results) to qualify. All chips go through that, and that's why there's many redundant structures on a chip (to improve yields). (Sony PS3 has 7 SPUs when they build 8 on a chip, Xbox360's got 3 PowerPC cores even though it has 4, Intel disables cache lines and/or functional units, etc. etc. etc.)

So the question is, are those cores disabled because AMD had extra parts and an outstanding order they could fulfill? Or are there actually potential issues that may only be revealed under certain loads? FOr the most part, it just means a game crashes a bit more often than usual (since mission critical servers never do wierd things like this - the money saved isn't worth the potential for extra downtime), or maybe a file gets corrupted. Or worse, your disk gets corrupted.

Plus, AMD's historically been supply-bound and unable to fulfill demand for their product, so there's a potential that instead of getting a binned part, it's actually one that failed their test patterns.

And yes, you see the same behavior with flash chips - NAND flash traditionally ships with bad blocks, and the majority of those can probably be erased and used quite safely (having accidentally destroyed the bad block information before due to buggy software...), but you never can tell why it was marked bad in the first place.

Re:Whatcouldpossiblygowrong

[tuna816]

FYI, my friend successfully unlocked his dual core Intel to quad core. He did it 3 times on the same chip model and it only failed once. They work perfectly fine. I'm not sure if they were the processors you mentioned though.

Re:Whatcouldpossiblygowrong

[mea37]

Yeah, maybe. Then again, GP has a point and you're being an asshat.

TFS makes a comparison to overclocking. It points out that there is no guarantee of a benefit - but doesn't point out that there is a risk. In the case of overclocking, the risk is that you will overheat a chip that was rated at a particular clock speed for good reason. Of course you can combat this risk by improving the cooling system. You can combat the risk because you know exactly what the risk is.

Now in the case of "hidden cores", what's the risk? Do you even know? Do you know what kind of flaw would lead them to legitimately disable a core? Is that one core unable to tolerate the same clock speed as the others? Is it functionaly broken such that it will return incorrect results for some operations? How would you tell the difference between that, vs. a chip that was perfectly fine but sold in a degraded state to balance out supply and demand?

You could shell out for a special motherboard just to test your chip, and if no flaw in the normally-disabled chip causes any damage to the rest of the chip (or do you have some basis on which to rule that possibility out?) you at least won't lose anything. Or, could the defect be intermittant such that your tests might miss it?

And if your computer is for hobbying and you enjoy working with a potentially-unstable system, good for you. A lot of people think that's a fine trade-off for what they're going to do with their systems. None of which invalidates GP's question - which is "what exactly might a disabled-by-default core do if you turn it on when it really was disabled for a reason?"

Re:Whatcouldpossiblygowrong

[MrNaz]

Right. And given that there is *always* a yield rate somewhere below 100, it's a guarantee that not all of the partially disabled parts are in actual fact fully working. You'd have no way of knowing if you do. In fact, given that the yields are private information, you don't even know the *probability* that your unlocked unit will work properly.

The manufacturer will *always* bin the partially flawed parts as their low end units first. They will only use intentionally crippled units to fill the low end volumes if they run out of partially flawed units. Historical experience with yields indicates that they're more likely to get not enough fully functional units than they need. This was the case with single core parts, and I'd assume it's even more the case with multi-core parts, becoming more of a problem as core counts increase. I doubt AMD or Intel have the latitude to pick and choose the relative outputs of their units; I doubt the yield curves are such that they end up having to cripple many units because they have too many fully functional parts and not enough to fill low-end volumes.

Even if there *were* a decent percentage of fully working CPUs on on the market, you'd have to be pretty stupid to spend that amount of money on a high end motherboard to turn your CPU into a *maybe* working higher model that *may* totally destroy your data. Either that or the work you're doing is so trivially unimportant that you probably don't need a computer in the first place. Why not just buy a normal motherboard and spend the saved money on the real fully featured part.

You're showing a complete lack of understanding of, well, just about everything.

Re:Whatcouldpossiblygowrong

[Joce640k]

History has shown that there's a pretty good chance that it _was_ binned for marketing reasons.

(ie. In many previous CPUs, graphics cards, etc. you had to be pretty unlucky to get one which didn't work perfectly)

Re:Whatcouldpossiblygowrong

[electrosoccertux]

You're showing a complete lack of understanding as to how processors are rated and sold. AMD determines they need to meet a certain quota for each model of CPU. If it works out and all of the CPUs in their 1 million unit run works flawlessly, they will maximize their profit by disabling some of them and selling them for less money to account for that market without flooding the market with their top performing part.

True, but there's also a good possibility that the your part wasn't binned to fulfill an order. Chips go through a severe set of stress tests that often exceed what will be encountered in practical use. During these tests, it may be revealed that a core doesn't function properly or well enough (it gives bad results) to qualify. All chips go through that, and that's why there's many redundant structures on a chip (to improve yields). (Sony PS3 has 7 SPUs when they build 8 on a chip, Xbox360's got 3 PowerPC cores even though it has 4, Intel disables cache lines and/or functional units, etc. etc. etc.)

So the question is, are those cores disabled because AMD had extra parts and an outstanding order they could fulfill? Or are there actually potential issues that may only be revealed under certain loads? FOr the most part, it just means a game crashes a bit more often than usual (since mission critical servers never do wierd things like this - the money saved isn't worth the potential for extra downtime), or maybe a file gets corrupted. Or worse, your disk gets corrupted.

Plus, AMD's historically been supply-bound and unable to fulfill demand for their product, so there's a potential that instead of getting a binned part, it's actually one that failed their test patterns.

And yes, you see the same behavior with flash chips - NAND flash traditionally ships with bad blocks, and the majority of those can probably be erased and used quite safely (having accidentally destroyed the bad block information before due to buggy software...), but you never can tell why it was marked bad in the first place.

I bought a Ph2 720BE and unlocked it to a quad. Stress tested with 12 hours of Prime95, no failures. When the core is bad, you usually can't even boot into Windows; never have I heard of one that could withstand gaming for more than 5 seconds. If something in it is broken, you know it.

So I paid $120 back when the Ph2 965 cost $240, and unlocked and overclocked the 720BE I bought to a quad at 3.5ghz. 4 cores for the price of 3. Love it.

Re:Whatcouldpossiblygowrong

[Cowclops]

Eh... theres really no such risk with regular overclocking. The biggest threat to your CPU is increasing the voltage - which would strictly be overvolting, not overclocking. If you turn up your clock speed high enough that it "could" cause damage to it at load... odds are you've turned it up so high that it won't make all the way through bootup. And the solution to that is simply revert it back to its stock speed, or cut the difference between stock and what it won't run at until you find a working speed. The chance of permanent damage to a CPU without changing the core voltage is essentially zilch.

The big difference between overclocking and unlocking hidden cores is that you can make small incremental overclock adjustments, say from 2.6ghz to 3ghz or 3.2ghz or 3.5ghz or whatever until you find that its unstable, and just back off a bit. You can't incrementally unlock one core, its unlocked or it isn't. And if it was disabled due to being flawed, it should stay that way or else your computer is just gonna blue screen right in the middle of some important work/gaming session.

Re:Whatcouldpossiblygowrong

[qoncept]

I don't have a clue what your acronyms mean -- can't even infer -- but I'll just read over them.

Unlocking cores that the manufacturer deems to be flawed - um, yeah.

The OP might as well have said, with the same sarcastic, completely ridiculous implication, "Run my processor at a clock rate that the manufacturer deems to be flawed - um, yeah."

The OP mentioned nothing about system criticality -- just implied that running a processor beyond the manufacturer's factory spec was a completely absurd idea. Like it doesn't happen all the time.

Re:Whatcouldpossiblygowrong

[UnknowingFool]

The problem is you don't know if a particular core was disabled for legitimate flaws or for marketing. From AMD's standpoint, they probably don't want to disable the cores unless there was not other choice than they really needed to fill orders because they can sell the fully functional chip for lots more money.

Re:Whatcouldpossiblygowrong

[LordKronos]

True, but there's also a good possibility that the your part wasn't binned to fulfill an order. Chips go through a severe set of stress tests that often exceed what will be encountered in practical use. During these tests, it may be revealed that a core doesn't function properly or well enough (it gives bad results) to qualify. All chips go through that, and that's why there's many redundant structures on a chip (to improve yields). (Sony PS3 has 7 SPUs when they build 8 on a chip, Xbox360's got 3 PowerPC cores even though it has 4, Intel disables cache lines and/or functional units, etc. etc. etc.)

So the question is, are those cores disabled because AMD had extra parts and an outstanding order they could fulfill? Or are there actually potential issues that may only be revealed under certain loads? FOr the most part, it just means a game crashes a bit more often than usual (since mission critical servers never do wierd things like this - the money saved isn't worth the potential for extra downtime), or maybe a file gets corrupted. Or worse, your disk gets corrupted.

That's what diagnostic tests are for. memtest86, prime95, etc. If you system can crank through 24+ hours of those tests, you can be reasonably certain it will perform just fine for everyday usage.

Re:Whatcouldpossiblygowrong

[Anonymous Coward]

Indeed. 66% of the time, it works perfectly every time....

Re:Whatcouldpossiblygowrong

[Zerth]

And you apparently don't understand the mind of an overclocker; they aren't sane.

Re:Whatcouldpossiblygowrong

[Anonymous Coward]

That's why you test the core. If you have the knowledge of being able to turn on a core, you have the knowledge telling you that you should stress test it with prime95 or somesuch.
Argue this fact as much as you like, if you're the idiot who didn't check for stability it's your own damn fault.

And this information has been out for a long time and slashdot is just now finding out?

Re:Whatcouldpossiblygowrong

[0123456]

Or rather, it worked perfectly fine 66% of the time. Three tries with one failure.

Actually, it's 697.31965% fine, according to the FPU on an 'unlocked' core one of the other two CPUs.

The problem with his claim that the other two are 'perfectly fine' is that he has no idea whether the cores really work 'perfectly fine' without performing the same kind of low-level tests that the manufacturer would have performed before disabling them. Of course if he can live with, say, the FPU randomly producing incorrect results then that may not matter.

I used to work for a company which disabled components of chips which failed manufacturing tests and sold them as cheaper, less powerful units, and with our mix of products I don't believe we ever had to disable working hardware. I can't vouch for Intel or AMD.

Re:Whatcouldpossiblygowrong

[gknoy]

And if your computer is for hobbying and you enjoy working with a potentially-unstable system, good for you. A lot of people think that's a fine trade-off for what they're going to do with their systems. None of which invalidates GP's question - which is "what exactly might a disabled-by-default core do if you turn it on when it really was disabled for a reason?"

WoW might crash while I'm tanking our guild's best attempt at the Lich King. Realistically, that's the worst case scenario I can think of for me, as a gamer, and even there I highly doubt that would happen. (I'd likely see instability long before that, and revert to 4 cores.) I obviously wouldn't suggest using hardware this way on my office machine, where numerical accuracy and stability are important, but ... if games are reasonably stable, going for 6 vs 4 cores seems like a pretty good deal.

Re:Whatcouldpossiblygowrong

[garyrich]

"The manufacturer will *always* bin the partially flawed parts as their low end units first."

True, but the after market CPU is not the low end, not at any price point. You would put the real X2 and X3 chips in the low end consumer boxes, where the mobo doesn't support unlocking and the consumer doesn't know/care. You sell the perfectly good ones to newegg, fry's, etc. Happy geeks that unlock cores or overclock successfully are morle likely to recommend to others and buy next time. AMD and Intel understand this very well.

Why do you think AMD has a "black label" line in the first place?

Re:Whatcouldpossiblygowrong

[indi0144]

You are clearly cheating in the e-Penis tests, everybody knows you have to spend more money in order to brag about your kickass hardware, otherwise, you will be labeled as a smart consumer which is not compatible with the 1337woopassHardwarez0wnzorzx label. :)

Re:Whatcouldpossiblygowrong

[speculatrix]

and then there were the "soft quadro" hacks where you could make a low-end nvidia card appear to be a quadro which would unlock a load of features which were in the drivers as well as the graphics card.

Re:Whatcouldpossiblygowrong

[marcansoft]

memtest86 is a diagnostic test for RAM. Prime95 isn't a diagnostic test for anything. Both are reasonable CPU burn-in tests, but they don't test all (or even most) features of the CPU. I'm not even aware of memtest86 using more than one core. Sure, if you run them for a while you can be reasonably sure that the critical parts of a core are working properly, but there's a very real possibility that its problem is a more obscure one that only shows under certain circumstances. For example, some specific app might corrupt data, while everything else works fine.

In order to properly test a CPU core, you at least need a full suite of tests for that architecture, including OS/kernel-level tests, and even those are likely to miss things particular to the specific manufacturer's implementation of the architecture.

Re:Whatcouldpossiblygowrong

[Anonymous Coward]

I basically had the same thing happen to me. A few months ago I upgraded to an X3 720 for $120 on Newegg. After testing that it works, I tried unlocking the 4th core. It was Prime95 stable and I tried all the games/apps and they all worked. Of course I got a newer processor after all the problems AMD had with are gone, so it was more than likely market demand that they disabled the core (I hear the X3 720 is fairly popular). The only setback with it is that it disables the temp. monitoring. So, all cores now read 0c. But it wasn't that bad for a black edition quad core for $120.

Re:Whatcouldpossiblygowrong

[ByOhTek]

What test suite would you use though?

I seriously doubt prime95 is comprehensive enough to cover all CPU operation.

How do you ensure the test runs on the knocked-out core?

Note: I'm not saying this to be sarcastic or suggest it is a reason to not try unlocking the cores - I'm actually curious (*looks at 65W dual core Phenom II*)

Re:Whatcouldpossiblygowrong

[Rasperin]

So they spend more in labor, spend more parts, to sell these at a cheaper price instead of selling all of the top-of-the-line at the same price for less cost? This should be illegal!

Re:Whatcouldpossiblygowrong

[rwa2]

Good work! I plan to do something similar soon, though the cost savings of getting a $100 2-core Ph2 and unlocking it to a $160 4-core Ph2 isn't so great :/

I'll share my pseudo-failure story, though. I bought a Tyan Tiger MPX about 10 years ago to run dual SMP 1Ghz Durons. About 5 years later I upgraded the CPUs to 2.0 Ghz mobile Athlon XP. My motherboard couldn't control the mobile chips, so I think they only ran at 1.2Ghz or something for a time, then I got brave and whipped out the xacto knife and cut some bridges to clock them up to 1.8Ghz. After I migrated to a new server, I got even more brave and whipped out the pencil as well and linked some more bridges to get them up to ~2.2Ghz for the past few years. It's still my primary gaming machine (yeah, I'm too cheap to budget any real money towards entertainment, but it still runs most games better than my wife's 1-year old laptop, as long as they don't require 64-bit or DX10).

Of course, it's quite a bit flaky now, I think due to the penciled bridges and probably old noisy cooling fans. It crashes when I kick the case, and if it gets too warm in the room, it just plain doesn't boot (motherboard gives out 5 beeps and it just sits there). But once it starts running a game for more than a few minutes it tends to continue to be OK

Still, I'm plotting to migrate my current server to a low-power, low profile Zotac Zbox with some sort of external eSATA RAID, so I can free up my current hardware for gaming before it gets too outdated :-P

Re:Whatcouldpossiblygowrong

[LordKronos]

memtest86 is a diagnostic test for RAM

You don't think the diagnostic puts any sort of stress test on anything other than the memory?

Prime95 isn't a diagnostic test for anything

Really? You don't think a test that is notorious for pushing the CPU to high load and high temperatures is a diagnostic for anything?

Well, in case you disagree with both of those assessments, I'd like you to reread my original post, and this time please pay extra careful attention and note the "etc" included in there. Unless you are suggesting that there are absolutely no diagnostic tests that are available to consumers to test stuff like this, I'm not sure what your point was other than to pointlessly smack me around for knowing the basics of what you'd have to do but not knowing (and not listing) every last diagnostic test available.

Re:Whatcouldpossiblygowrong

[TooMuchToDo]

Intel uses prime95 for testing, so I would say yes, it is comprehensive enough to cover typical CPU operations.

Re:Whatcouldpossiblygowrong

[dfgchgfxrjtdhgh.jjhv]

In that case, they should reduce the price of their top end chips until supply meets demand. The reason they don't do that, is because some people have more money than others, it's a duopoly (so no really aggressive competition) & they can charge more for the higher end chips if they mark some down at a lower speed (or core count).

They just want to extract the most cash possible from people that want the higher end chips, they couldn't do that if they were charging a fair free market price. So they have to mark some as lower end chips, so that they can still sell to people with less money to spend, without damaging their top end market.

http://en.wikipedia.org/wiki/Market_segmentation

Re:Whatcouldpossiblygowrong

[cayenne8]

"In order to properly test a CPU core, you at least need a full suite of tests for that architecture, including OS/kernel-level tests, and even those are likely to miss things particular to the specific manufacturer's implementation of the architecture."

Any specific applications/tests you could list that one might want to run to give a proper test to all the cores in question?

Re:Whatcouldpossiblygowrong

[dfgchgfxrjtdhgh.jjhv]

sorry, better wiki link here: http://en.wikipedia.org/wiki/Price_discrimination

it can only occur because of a lack of competition in the market.

Re:Whatcouldpossiblygowrong

[allcaps]

You show a complete lack of understanding of economics. If they by some luck produced 4 times the premium CPUs than projected, then they would maximize their profits by selling all of them at a lower price to beat out any competition who didn't get as lucky. There is absolutely no reason they would ever want to disable cores knowingly in a free market.

Re:Whatcouldpossiblygowrong

[mea37]

"meh, you might MAYBE corrupt some windows stuff not unlike the risk involved in pushing an overclock."

Citation needed.

"In reality if you attempt to enable extra cores they will either work or they won't."

Citation needed.

"I've been running my dual core phenom 2 as an overclocked quad core stable for 2 months now"

Anecdotal; no significance in assessing risk.

As to the rest of your post - I'm not saying that nobody should ever do this; I'm saying that I disagree with bashing OP for wanting to know what the risk is. (And I still haven't seen a credible answer. As you say, you rolled the dice and won; so the fact is you don't know what happens if you lose.)

Re:Whatcouldpossiblygowrong

[Anonymous Coward]

I don't have a clue what your acronyms mean -- can't even infer -- but I'll just read over them.

TFS - you've heard of RTFA? This is referring to the summary.

Parent post - one above. GP one above that. You had OP already, you could have inferred that!

Re:Whatcouldpossiblygowrong

[billcopc]

There is a lot of room for error without crashing. It is quite common for slightly overtaxed CPUs to boot and run "fine" but produce incorrect math results, hence why people run Prime95 to test them as it will catch those rounding errors and flat-out incorrect calculations.

I'm all for squeezing more value out of a product, but these days the price delta between dual/triple and quad cores is so small it seems rather foolish to even risk it. Sure, there's no permanent damage but for the $30 to $50 you're saving, it's peanuts compared to the value of any work lost in a crash or corrupted file.

Re:Whatcouldpossiblygowrong

[Mashiki]

It's not something I really bother with anymore (as I've gotten older as long as the computer keeps running I'm happy), but I remember enjoying the whole overclocking scene ~10 years ago and wouldn't begrudge the new cheap teenagers of the same fun I had :).

I know exactly where you're coming from on that. Remembering back in the days of PIII Celery's 266/300's and the mass push for super-overclocking with air cooling. I seem to remember they were the first CPU's to push 1ghz by simply overclocking the bus modifier on the motherboard. Then there was the physical clock unlocking on the chip itself by wiring in bypasses to the SLOT1/SLOTA cards, then there was the push for variable modifiers for the FSB/PCI on the BX boards to make it easier. I actually kinda miss that type of hard hacking, where you actually cracked open a physical case and started soldering and wiring in resistors/transistors and 24 gauge wires, and breaking on board connections in the tracing.

I just don't have the time for it now. But I wouldn't want to take that away from kids either, you're looking at your next generation of hardware hackers when they get into it.

Re:Whatcouldpossiblygowrong

[entrigant]

[citation needed]

prime95 doesn't cover all of cpus functionality. Granted, it does directly or indirectly stress the fpu, cache, maybe task switching and interrupt handling. However, there are many more things that can go wrong.

To be fair, I don't know of a better way to test, and I'd love to see a discussion of better utilities. If I tried this I'd probably do mprime and keep an eye out for MCE's in the system logs, but don't delude yourself into thinking that core is error free because you ran prime95.

Re:Whatcouldpossiblygowrong

[marcansoft]

You don't think the diagnostic puts any sort of stress test on anything other than the memory?

The diagnostic doesn't put any sort of uniform stress on anything other than memory. Even wondered why it does a ton of passes on a ton of different modes with a ton of patterns on RAM? That's testing for as many possible RAM failure modes as it can. No attempt is made to test the CPU. You're stressing some parts of the CPU, but you're neglecting the vast majority (e.g. floating point and SIMD).

Really? You don't think a test that is notorious for pushing the CPU to high load and high temperatures is a diagnostic for anything?

If anything, it might be a diagnostic for your cooling system. Sure, it helps ensure that nothing is blatantly wrong with the CPU, and it does a better job at testing the CPU than memtest86, but it isn't even remotely a comprehensive test of CPU functionality.

This isn't overclocking we're talking about here. When you overclock, you stress the entire CPU more as a whole. When tests like memtest86 and Prime95 start failing, you know that your CPU is definitely unstable. Then you back off and you hope the untested parts of the CPU will do OK with whatever safety margin you gave it.

When you enable a core, it might have some broken parts, or it might not. Those parts can be flaky, or they can be borked, period. Unless you run software that has a chance of testing those parts, you will never find out. E.g. if the hardware for a specific floating point instruction is borked, memtest86 will be useless, and Prime95 will be useless unless it happens to use that specific instruction. If the transistor in charge of forbidding kernel memory access from user mode is borked, you won't find out until an unstable application takes down your entire system by scribbling all over the kernel.

Unless you are suggesting that there are absolutely no diagnostic tests that are available to consumers to test stuff like this

I am absolutely sure there is no test that will match what Intel and AMD do - because they know exactly how their CPUs work and what to test for. I do know that you can do a whole lot better than memtest86 or Prime95. I haven't checked whether someone actually has attempted to produce a comprehensive architecture test of this sort.

Your mistake is attempting to extrapolate from tools used for testing overclocking (which typically results in overall instability) as a means to test for disabled and possibly subtly broken hardware. Any failures from a defective core are likely to show up only with workloads that exercise the defective bits, and the rest of the CPU will work fine.

Re:Whatcouldpossiblygowrong

[kriston]

Often the cores that are disabled are disabled because of failed L2 cache.

What software tool can I use to test the L2 cache to ensure it's totally working? I'd hate to enable that core and find out that its L2 cache is faulty, something both of those benchmarks don't seem to test.

Re:Whatcouldpossiblygowrong

[wiredlogic]

I bought a Ph2 720BE and unlocked it to a quad. Stress tested with 12 hours of Prime95, no failures.

Running Prime95 is not a stress test. Did you fully exercise all of the logic at all voltage and temperature extremes? Did you do the same with all on-chip memories? Did you subject the I/Os to a barrage of marginal conditions? This is what goes on in the fab during the binning process and you can't replicate it on your own.

Re:Whatcouldpossiblygowrong

[Ancient_Hacker]

There are bazillions of combinatorial tests that your average stresser program does not do, and cannot foresee that it needs doing.

There's a whole lot more than the basic instruction set that needs to be tested for.

For instance all the superscalar stuff -- pipeline loading, serializing, register interlocks, register renaming, stack register lookahead, jump prediction, cache prediction, cache-snooping, cross-core interlocks-- all things that require a certain complex SET of carefully primed and timed instructions to test.

Not to mention the extra MMX, SSE, SSE2, SSE3, and later instructions.

Your basic CPU heater program is not going to test for these, at least not intentionally, and not often.
 

Re:Whatcouldpossiblygowrong

[retchdog]

I remember when running the "Second Reality" demo (by Future Crew) on my 486, if you hit the desk the computer was on, the particles on screen would jump around to different locations (and occasionally it would crash). I never noticed any other problems with any other software. Granted it was probably the RAM and not CPU, but after seeing this, I was really surprised that the computer worked at all...

Re:Whatcouldpossiblygowrong

[hibiki_r]

memtest86 will use your CPU, but it's using a very small subset of what it can do. It's great for memory, but for the processor, you might as well run a screensaver, or a video running OGG.

Prime95 puts the CPU to high load and high temperatures, but it's not very comprehensive: I can overload a CPU with floating point additions, and that will probably make it run hot, but that doesn't really test most of the CPU: It will test some parts very heavily, and it will not test others at all. Prime95 tests more than just a couple of instructions, but it doesn't test most of the CPU either.

AFAIK, there is no consumer geared test that will really be able to clear a core as good. Not even a battery of 5 or 6 benchmark tools will do it. Checking the validity of a CPU is pretty serious business, and only the manufacturers seem to be in the business of testing them out.

Now, can you run a bunch of apps for a while and figure out if, in practice, the processor is stable enough for your desktop use? Maybe, but there will always be a margin of error there, which will be far larger than the margin that the manufacturer's test create.

So, even if the processor passes your tests, and then suddenly starts behaving oddly 5 weeks later, you can't discount the possibility that it's a problem with the core you enabled. And if you are having strange problems, adding one more thing that could be broken is rarely a good idea.

Re:Whatcouldpossiblygowrong

[marcansoft]

I'm not sure if there's a way of explicitly testing all of L2 on an x86 chip (it might be vendor-specific, if possible). However, anything that exercises large amounts of memory in certain patterns is liable to test CPU caches to a reasonable degree. I believe memtest86 does enable cache for some of its tests, so it might be a good idea if all you're worried about is L2.

L2 failure is one of those things that should eventually cause obvious instability with many tests and workloads, although the degree of instability can probably vary quite widely.

Some CPUs (e.g. PowerPC) have better specified cache behavior and document specific testing methods.

Re:Whatcouldpossiblygowrong

[Khyber]

"Unlocking cores that the manufacturer deems to be flawed - um, yeah."

Why not? I turned an AGP 6800NX into better than a full-powered PCI-E 6800 Ultra by unlocking the other pixel pipelines (before nVidia went to laser-locking the chips, and the AGP version had 16 pipelines where the PCI-E version was stuck at 12.) I took a 100 dollar card and cranked it up to its 250 dollar counterpart's performance without any issues.

Not everything is locked because it's bad, maybe it just underperforms to the degree to warrant being placed in a lower bin. Usually a microcode fix can be applied to work around the non-functional sections and still make use of the functional silicon.

It's actually a very common practice in IC microfab.

Re:Whatcouldpossiblygowrong

[CAIMLAS]

As someone else said, memtest86 and prime95 don't cover even a fraction of the CPU functionality.

A better test would be to compile something for a prolonged period of time for your architecture. Off the top of my head, the gentoo install CD might be a good choice for this: start from scratch and build up a kernel and userland for your architecture. In the process you'll be testing the whole system - RAM, CPU, and disk - to at least know there is not an apparent manufacturing defect (short of a drive/electrical failure 2 days later).

Re:Whatcouldpossiblygowrong

[nabsltd]

Granted, it does directly or indirectly stress the fpu, cache, maybe task switching and interrupt handling. However, there are many more things that can go wrong.

Off the top of my head, I can think of a lot of things that specifically need tested that one program probably won't do. For example, you need to verify both 32-bit and 64-bit operations. Prime95 is specifically compiled for one or the other, so would stress less of the "other" version.

There are also a lot of SIMD instructions that need tested. Some are obscure enough that only a few apps would use them.

Then, there's all the instructions that support virtualization. I have found that bad hardware running a hypervisor will fail much more frequently than if it is running a "normal" OS (YMMV).

But, unlike Memtest86+ for RAM, there doesn't appear to be any program that specifically tests all CPU subsystems (registers, cache, instruction execution, etc.).

Re:Whatcouldpossiblygowrong

[TeknoHog]

It's not unlike DRM: they develop new technologies in order to take our current technology backwards. The result is a lesser product that they sell for more money.

Re:Whatcouldpossiblygowrong

[WaroDaBeast]

Dude, what were you doing when gamers were unlocking Geforce 6 cards? Early 6200's could unlock to 6600 and certain models of 6800 NU (i.e. Non-Ultra) could unlock to 6800 Ultra. They locked vertex and shader pipelines (now called "cores," go figure) because the demand for lower-priced products was high. Now, some of the pipelines were often damaged, but chances were you could unlock part of the locked ones. Free performance can never hurt.

Re:Whatcouldpossiblygowrong

[derrida]

To be fair, I don't know of a better way to test, and I'd love to see a discussion of better utilities. If I tried this I'd probably do mprime and keep an eye out for MCE's in the system logs, but don't delude yourself into thinking that core is error free because you ran prime95.

There are quite a few tools, mainly found in the overclocking communities. OCCT, Linx and Intel Overburn just to name some.

Re:Whatcouldpossiblygowrong

[TheRaven64]

Prime95 does not execute every instruction

Even if it did, a lot of CPU errata in the past have related to interactions between instructions. One story the Intel guys tell is of a particular condition flag being set accidentally on 486 CPUs after a sequence of other instructions. Apparently, game developers discovered this and started using it for optimisation. The first Pentiums, when they were run in simulation, crashed these games, so the final silicon had to do the same (wrong) thing or Intel would get the blame for breaking everyone's games.

The test suites that CPU manufacturers use are exhaustive and cover a lot of combinations of instructions. You may be able to run 90% of your programs without any issue on a flawed core, only to have one particular program crash strangely. Irritating if it's a userspace program, disastrous if it's your operating system's block device driver...

Re:Whatcouldpossiblygowrong

[mestar]

...stress the fpu...

Yes, but only if you search for primes with nonzero decimal part.

Re:Whatcouldpossiblygowrong

[yoyhed]

Exactly, especially if you follow the advice of the article (or any article on the subject, as this has been known for quite some time before Slashdot picked it up) and choose newer versions (steppings) of the processor. It makes sense that over time AMD would get better yields as they improve the manufacturing process - but they still have many different market segments to fill. Choose a newer chip and you have a lot better chance at having a fully-functional one.

However, nowadays the monetary benefit is somewhat diminished - you can pick up AMD's top of the line Phenom II X4 965 for $180, the X4 925 for $140, the X3 720 for $100, or an X2 550 for $80. Hell, the entire Athlon II line (just Phenom IIs without L3 cache) ranges from $65 to $105. Is the difference really worth potentially days of tinkering and testing? Depends on how much you like tinkering. :-)

Re:Whatcouldpossiblygowrong

[entrigant]

mprime performs FFT's when doing stress testing. The documentation explicitly states it's a FPU stress test...

Re:Whatcouldpossiblygowrong

[yoyhed]

What happens when you lose is you revert to normal settings in the BIOS if the system crashes or won't boot. You either return the processor, or live with the fact that you still got more performance for your dollar than with Intel (and chances are, you didn't really NEED the extra performance anyway).

Re:Whatcouldpossiblygowrong

[yoyhed]

The first instance I remember of this was back around 2003, when you could flash a Radeon 9500 with the 9700 BIOS to turn it into a 9700. Good times.

Re:Whatcouldpossiblygowrong

[entrigant]

These are mostly toys meant to make the cpu run _hot_. Neither of them is designed specifically to perform a comprehensive path test of the CPU. See nabsitd's reply to my original comment for a non comprehensive example of things that need to be tested thoroughly.

Even hitting every instructions is not enough to test every path. There are things such as cache coherency and reliability, branch predictor testing. You have out of order execution and pipelining that needs to be examined, etc.

If a part of the processor that determines safeness for out of order execution has a single piece wrong that could be enough to start corrupting a filesystem. Granted, I imagine such a quirk would be highly unlikely, but these tests are most definitely not going to be able to tell you one way or the other of something is lurking broken that's going to ruin your day.

Re:Whatcouldpossiblygowrong

[Apocalypse111]

I did something similar back with my ATI x800 Pro VIVO, throwing in the x800 XT BIOS to unlock the additional bit pipelines.

Re:Whatcouldpossiblygowrong

[Dekker3D]

if you're worried about your filesystem, keep regular backups. i fully agree with you on everything else, but really.. the filesystem going awry isn't the worst thing in the world of computers and bad ideas.

Re:Whatcouldpossiblygowrong

[ottothecow]

That would only be true if it was a pure commodity item with no real differances besides grading--think meat...You wouldn't downgrade a bunch of Prime beef to Choice just because your cow ended up being 100% Prime. You would sell everything at Prime price and assume that the Choice market would be filled by farmers who got unlucky and had cows with no prime meat.

These are not cows--these are high end CPU's, there are two major industry players and they do not make identical products. In this market, you would rather price discriminate. AMD historically has had lower prices than intel but people still buy intel--so there must be more to the story than you are imagining. The key here is that AMD gets to segment their market and price discriminate. Sure, they are selling you 4 cores for 80% of the 4 core price by labeling them as 3 cores, but that means they get to make a 20% premium on people who want guaranteed 4-core functionality.

I am explaining this poorly--but for someone accusing everyone of having a "poor understanding of economics" you sure seem to be spewing a lot of misinformation (mainly driven by your interpretation of CPUs as a commodity).

Re:Whatcouldpossiblygowrong

[electrosoccertux]

lol

Re:Whatcouldpossiblygowrong

[Rudeboy777]

you'd have to be pretty stupid to spend that amount of money on a high end motherboard to turn your CPU into a *maybe* working higher model

Yes you would. Good thing Asus and Gigabyte have included this feature even on low-end AMD 770 chipset boards!

Re:Whatcouldpossiblygowrong

[electrosoccertux]

use speedfan. I don't remember which temp it is, but it still has one. I've since renamed it "CPU". You don't get individual core monitoring, but you do get the temp of the whole CPU.

Re:Whatcouldpossiblygowrong

[paradigm82]

I don't think the part about the NAND chips is true. The bad block management takes place internally in the controller using a spare pool and is not under user control by any means. Hence, when you format a flash drive you won't actually see any bad blocks since they will have been remapped internally by the controller. I haven't seen a bad block in 13-14 years. It is possible that if a sector should at some later point become bad, that the controller will report the sector as unreadable until it has been rewritten again (after being remapped to the spare pool) but that is just to prevent the o/s from relying on invalid data.

Re:Whatcouldpossiblygowrong

[electrosoccertux]

and I've been running this chip for 8 months computer on all the time no problems.
95 was not the only stress test I gave it.

There's really no reason for me to not be running it unlocked. I haven't had any problems thus far and it wasn't worth the extra $100 for another core at the time (poor student here). Since I haven't had any problems thus far, I'm not going to have any for the next 2 years either, and if I do have a problem oh well big deal. I would have harddrive corruption issues by now if there were a problem, and I was prepared to reformat if necessary.

Re:Whatcouldpossiblygowrong

[MarkRose]

I remember over clocking the system bus on an old machine of mine. I could run the 133 MHz bus at 140 MHz without issue. When I would bump it to 145 MHz, things worked fine, except some hash computations.

Re:Whatcouldpossiblygowrong

[osu-neko]

The best test would be to run World of Warcraft for many hours a day for a few weeks. This not only stresses a variety of subsystems, but accurately simulates your real-world usage of the machine. ;)

Re:Whatcouldpossiblygowrong

[osu-neko]

"In reality if you attempt to enable extra cores they will either work or they won't."

Citation needed.

Err, I don't think you need a citation for any logical tautology (in this case, P or not P).

Re:Whatcouldpossiblygowrong

[Nulifier]

Companies that create chips like this usually have a system in place for running built in test that do test all the registers and functions of the chip.

The methods that they use to do this are actually quite complex involving creating test patterns and seeing if the correct result is generated. This is often on a much smaller scale than "does each core work" and it could test just the ALU (Arithmetic Logic Unit).

The problem with a consumer trying to access this is that the test mode is usually hidden from the user intentionally so they can't break their chip. The way that you get into test mode varies from chip to chip, but can be any thing from a hidden instruction to pulsing the memory CAS and RAS backwards.

These two links go into greater detail.

Re:Whatcouldpossiblygowrong

[BikeHelmet]

Good work! I plan to do something similar soon, though the cost savings of getting a $100 2-core Ph2 and unlocking it to a $160 4-core Ph2 isn't so great :/

I picked up a Phenom II X4 925 for $130 CAD on the week they launched. That was equivalent to $120 USD at the time.

I don't need any unlocking with launch sales like that!

The same week they had these going for $80 after MIR. I didn't trust the $40 MIR, though.

Re:Whatcouldpossiblygowrong

[BikeHelmet]

Later in the product line it might end up that only 20% of the lower priced chips have any flaws at all. For those people who want to tinker, it's often worth while to at least check and see if their chips will run ok when then turn the rest of them on. They stand to gain some performance if it works, and if not - eh, they paid for the slower version anyways (the only issue I take with this is when I see Negwegg reviews or forum posts claiming that they were returning the chip because it "didn't overclock far enough").

This really irritates me. :/

I have a Phenom II X4 925 that plays games at 3.6ghz, and encodes video up to 3.4ghz. It's plenty fast enough for me, even if I don't break any forum records.

Re:Whatcouldpossiblygowrong

[HermMunster]

I bought a tripple core and later found that it was really a quad core. I enabled the 4th core. Confirmed everything in the cpu tools and Windows. After running it with the 4th enabled the system began to blue screen. I disabled that core and all went back to normal.

Re:Whatcouldpossiblygowrong

[BikeHelmet]

If all the chips come off the same line, then they might have an average cost of, say, $150.

Of course, the cost of the actual chip is approximately $8.

But they have to pay off those billions in R&D and fabs somehow! It's almost like software, where the first copy costs a ridiculous amount, and every copy after is pretty cheap.

Think of all those years of R&D - thousands of highly paid employees. Then they finally get it working, and can start mass producing them for a couple dollars each, to recoup their investment. I'm surprised we get CPUs as cheap as we do.

Re:Whatcouldpossiblygowrong

[treeves]

Unless the point was to exclude a third possibility: that the additional cores would run, but in some compromised way e.g. with math errors or occasional crashes or some other deficiency.

Re:Whatcouldpossiblygowrong

[mjwx]

Unlocking cores that the manufacturer deems to be flawed - um, yeah.

This has been happening for some time now. I've known how to unlock the fourth core on the AMD Phenom II X3 series for almost a year. I'm certain this is not a record for old news on /. however. This has turned out to be quite stable, AMD have disabled the core because they found some flaw with it in QA, this may mean that the core is completely unstable or it may mean the core has one tiny flaw that may cause a problem under some obscure circumstance. Further more, AMD doesn't care if you unlock the core, they just wont support the proc under warranty, same with any kind of overclocking.

Re:Whatcouldpossiblygowrong

[allcaps]

What you just said makes absolutely no sense. It doesn't really matter if Intel isn't a direct competitor (or even if there aren't any other competitors!). The truth of the matter is that there is a demand for AMD LGA 775 processors for some number of cores. The demand is there (for whatever reason), and AMD is there to fill it. If they make all their LGA 775 processors perfectly, then there is no reason to downgrade anything, because me the consumer would be ecstatic to see a perfectly functioning quad core at a 30% price reduction than normal. They are going to price all of them identically to A) sell them all, and B) maximize revenue. Turning off cores does not aide in those endeavors.

Re:Whatcouldpossiblygowrong

[electrosoccertux]

I love how its always a necessity for 'poor students' to have these high-end rigs to play games. Stop wasting mummy and daddys money on your toys.

eh, mom/dad haven't given me any money. This is the cheapest hobby there is. Try living for 5 years without a car and see how humbling it is. Building your own computer can be done on $500 from job money. It also is what keeps me motivated to study my electrical engineering.

Re:Whatcouldpossiblygowrong

[dudpixel]

I agree.

Most of the time the 2 or 3 core Phenoms are the ones where 1 or 2 cores failed in tests. The manufacturer can still sell them with those cores disabled and therefore makes a profit on something they otherwise would have to throw away.

On the other hand, if the locked cores were ok, why would AMD not sell them at the full quad-core price? doesn't make sense to do more work and make less profit when its not necessary.

The only other thing to consider is that they have a product that needs to meet sales volumes, and there may not be enough duds to satisfy that volume, so there may well be legitimate quad core cpus sold as 2 or 3 core ones. Even so, who wants to take the risk to unlock them and find out if they passed in testing or not. Is there any test software available that can tell you if your cpu is all ok or not?

Re:Whatcouldpossiblygowrong

[dudpixel]

The OP mentioned nothing about system criticality -- just implied that running a processor beyond the manufacturer's factory spec was a completely absurd idea. Like it doesn't happen all the time.

Close. The OP implied that running a processor to a spec that the manufacturer has specifically indicated is flawed, is an absurd idea.

Re:Whatcouldpossiblygowrong

[linzeal]

What do you think AMD uses? If they are like Intel they use something very similar to Prime 95 because well last time I checked Intel does use Prime 95.

Re:Whatcouldpossiblygowrong

[Calinous]

Usually trading DirectX performance for OpenGL performance (due to the drivers, but anyway). And you didn't make a low-end nvidia card appear to be a quadro, but you'd make a similar grade (same graphic chip) GeForce card appear as a Quadro.

Re:Whatcouldpossiblygowrong

[Chaset]

If running some utilities on a desktop PC were sufficient to properly test a chip, AMD, Intel, et. al. wouldn't shell out megabucks for things like this and this.
Unless you got one of these in your living room (and have their test vectors to run on them), you don't really know whether the chips meet spec.

Re:Whatcouldpossiblygowrong

[Calinous]

"The only other thing to consider is that they have a product that needs to meet sales volumes"

      Remember when AMD was to be found only retail or in "white boxes", generic computers? Then Dell started to sell computers with AMD processors, and they really really have a desire to sell cheap computers. So, they might want more 3-core CPUs than AMD is getting (due to all the processors being working 4-cores).
      This happens also for TigerDirect and others - some of the lower-end processors might be so successful that they request more. AMD being limited in every way (capacity, manpower, ...), they find it cheaper in the short run to just sell perfectly good 4-cores as 2-cores than to make a new silicon for a 2-core processor. Not to mention that you could make 2-cores processors out of 4-cores silicon in weeks, while building a new, cheap, native 2-core silicon might take you a month's work and three months wait

Re:Whatcouldpossiblygowrong

[jibjibjib]

How do you ensure the test runs on the knocked-out core?

That's trivial. Run it on all cores at once. Alternatively, your operating system probably lets you select which core something runs on.

Re:Whatcouldpossiblygowrong

[ranulf]

But you're not factoring in perceived value. If they reduce the price of the top-end chips, they then have a better price-performance ratio for the customer, so demand for the low-end chips drops off. When they are unable to make their yield of top-end chips, they end up with a glut of low-end chips that nobody wants any more because the top-end ones are perceived as better value (and possibly also supply problems on the top-end chips because their yields happened to be low). They then end up with a glut of the low-end chips that nobody wants anymore because the high-end ones are perceived as better value, so the only way they can shift these chips is by lowering the cost so that the low-end chips are perceived as better value. Which is where we are at now.

Re:Whatcouldpossiblygowrong

[English French Man]

Chip makers could have their own testing software, which they don't disclose. I think very few people would be interested in a tool for stress-testing their CPU. Benchmarking would probably be more popular, as it is already for GPUs. The reason memtest exists is that at one time, RAM memory was more frequently failing than anything else in the computer, and that it was useful to have a piece of software that tested it completely. I don't think there is the same kind of need in the general public for testing CPU.

Re:Whatcouldpossiblygowrong

[English French Man]

Of course the OS chooses on which core a process/thread runs on. Generally speaking it doesn't let the final user choose because it is of little usefulness in everyday computing, but for these kind of tests, the OS will control and tell which cores are working fine and which are not.

Re:Whatcouldpossiblygowrong

[yoyhed]

AMD's very own OverDrive utility has a CPU Stability test designed for overclockers. I'd trust it, coming straight from AMD - but the main thing is it can do independent core testing.

It tests integer calculation, stack operation, and floating point operation independently for each core, then does overall calculation tests and also checks MCA registers and the GPU. This, combined with Prime95, memtest86+, and long gaming sessions, should be a good test suite for most users' needs - are the people who are unwilling to spend a little extra cash to get a guaranteed quad-core going to need it for anything other than basic productivity and gaming (will they run into that obscure instruction)?

That having been said, a Phenom II X3 720 (2.8Ghz triple-core) costs $120. I just spent the extra $40 to get an X4 955 (3.2Ghz quad) and be done with it. That's still cheaper than any comparable Intel, and less than I paid for much less power in previous generations.

However, with the X2 550 (3.1Ghz dual) only running $85, that'd be a real nice lottery unlock to quad, since it's also a Black Edition and thus could easily match the 3.4Ghz clock of AMD's highest $180 model - for $95 cheaper. :-)

Re:Whatcouldpossiblygowrong

[Vectormatic]

Ah, good times

I had a AGP 6800LE which i found on clearance sale for half price (just after launch even, some shop was closing up some product lines), and managed to unlock it from 8/4 pipes to 12/6, the fourth pixel-quad was borked unfortunatly, so no full GT for me. It still gave me a 300 euro card for 90 euros :)

A while later i switched to two 6800XTs in SLI, which unfortunatly didnt unlock, but they overclocked like MOFOs. those cards came with the core at 300 and mem at 600, after an evening of experimenting i had both running at 475/900, with stock shitty cooling even. After a month with these i found a killer deal on two 6800GTs, which where the last serious gaming cards for me (had a 8800GTS and now a GTX260, but i hardly game anymore)

Geforce 6 series was awesome :)

As for running outside of spec, i also have a core 2 duo e4300 running at 2.4 (stock is 1.8), with no voltage bumps, just a FSB bump, has been solid like that since the day i bought it.

Re:Whatcouldpossiblygowrong

[CoderJoe]

Depends on which OS. Windows, for example, has APIs for setting process CPU affinity, as well as thread CPU affinity. Other threading systems may have something similar.

Re:Whatcouldpossiblygowrong

[English French Man]

Oh, I didn't know that, thanks.

But, the final user who doesn't have to program on Windows cannot make Word or Windows Media Player run on a specific core, can he?

Re:Whatcouldpossiblygowrong

[CoderJoe]

Bring up task manager. go to the processes tab. select the program, right click, and pick "set affinity". Check the cores you want it to run on, and uncheck the ones you do not want it on. click OK.

Re:Whatcouldpossiblygowrong

[English French Man]

Oh, my. I had no idea. I'm going to shut up now! :)

First?

[Anonymous Coward]

I would have gotten first post here, but AMD disabled two of my CPU's cores :/

Re:First?

[indi0144]

You're doing it wrong, If you use Intel you will be able to forst pist even before the article is submitted.

Re:First?

[AndrewNeo]

You got what you paid for!

Re:First?

[mswhippingboy]

forst pist

Better check that CPU dude, sounds like a bladder infection!

Re:First?

[ignavus]

The bad news is: Nature has disabled two of your cerebral hemispheres.

The good news is: The rest of your cerebral hemispheres are working fine.

Kernel tricks to take advantage of it?

[gehrehmee]

Just a thought, maybe Linux could be aware of what those cores look like, and what their sensitivities to temperature are.... and change the amount or type of work pushed to that core? Although I suppose heat from the other cores would most likely transmit very quick to the "zombie" core. Any CPUs have seperate temperature tracking per core?

Re:Kernel tricks to take advantage of it?

[.tekrox]

Core does, but its not exposed via the same interface.
for windows - try http://www.alcpu.com/CoreTemp/

Re:Kernel tricks to take advantage of it?

[Pharmboy]

This reminds me of "processor affinity" or "affinity mask", whereby you assign software to a particular processor or core. If you want to setup your software so that only less cpu intensive software (cooler) runs on the questionable core, you can do this in Windows 7, and likely for at least some software in Linux (I'm really not sure here), then yes, in theory, you could do this so only Word runs on core #3.

But please remember the wisdom of Yogi Berra when trying to apply a theory like this: "In theory there is no difference between theory and practice. In practice there is."

In other words, your mileage *will* vary.

Re:Kernel tricks to take advantage of it?

[indi0144]

In WinXP you can do the same, but I think it does not give the same results as Win7. In task manager go to the process tab an right click on the process > set affinity.

Re:Kernel tricks to take advantage of it?

[Jaysyn]

You can also do this in XP with Process Explorer.

Re:Kernel tricks to take advantage of it?

[ae1294]

linux zombies can likely transmit very quick to the "zombie" cores

Woah sweet! I've been having major problems dealing with zombie processes on my 6.4Ghz AMD rig of late. I didn't know I had the option of herding them all over to their own zombie cores! Sort-of like a botnet but for zombies right?

Can I do this in the kde or do I gotta use that text window thingy? Is there a one-click thingy you can kermit me?

Re:Kernel tricks to take advantage of it?

[Trelane]

Yes, you can set CPU affinity in Linux (sched_{g,s}etaffinity, taskset, pthread{_attr,}_{g,s}etaffinity_np).

Re:Kernel tricks to take advantage of it?

[VGPowerlord]

Is there a one-click thingy you can kermit me?

Sure!
 
...
 
...
 
...wrong Kermit?

Re:Kernel tricks to take advantage of it?

[ae1294]

..wrong Kermit?

Ow, I guess there's been some slight corruption in that sector of my brain. I best restore from the backup before it spreads. I hope this betamax tape is still in working order...

Re:Kernel tricks to take advantage of it?

[slashdot_commentator]

*Whew*! I didn't have the heart to shoot my new computer...

Gambling with CPU's?

[Tenek]

Now all they need to do is stop selling the processors with all cores guaranteed to work and watch sales skyrocket as people buy half a dozen dual-cores in the hopes of getting one that wasn't damaged. And whoever buys the most CPU's every day gets a working one for free...

Re:Gambling with CPU's?

[qoncept]

Good idea.

How do they disable the cores?

[91degrees]

I always assumed that there were fusible tracks but that would totally isolate the core electrically. Or is the workaround inexplicably clever?

Re:How do they disable the cores?

[Monkeedude1212]

I think they use one of those ion Cannons from Empire Strikes Back.

Re:How do they disable the cores?

[Joce640k]

In the past it's been done by a combination of BIOS and/or those tiny resistors soldered to the back of the chip.

Re:How do they disable the cores?

[Hatta]

Perhaps they use a pencil.

Re:How do they disable the cores?

[Terrasque]

Ah, memories. I did that to my duron. Or, to be precise, I actually soldered it (had access to equipment meant to fix mobile phones)

Chip worked well all the way into retirement. BTW, the pencil trick could be unreliable over time. Soldering was not :)

Re:Minimal Contribution

[ElectricTurtle]

Yes, but getting a working hexacore for the price of a quad would be a coup. However, odds are against winning that lottery. Demand is probably exceeds supply already due to production/release ramp up, so I'm sure that if AMD thinks a hexacore is bad it probably really won't be salvagable.

Re:Yield...

[nomadic]

I don't know about you, but I would not want to be willingly running a system with a known-bad CPU core

You underestimate the combination of paranoia and lack of sense that a lot of overclockers have, who are convinced the CPU manufacturers intentionally disable their chips in order to make more money somehow by selling them at a lower price.

Re:Yield...

[91degrees]

Producing a chip still costs a fair amount. R&D is a substantial part of the cost as well, but fabbing a chip costs a lot more than stamping a CD. We could be talking hundreds of dollars per unit for a new process and a large enough chip.

already done, wtf?

[drinkypoo]

various boards have permitted unlocking the cores. I'm looking right now for proper BIOS to do it with my Gigabyte GA-MA770T-UD3P with which many people have reported success (see thread)

Re:already done, wtf?

[Ant P.]

The newest BIOS version should work and give you a menu option to mess with the EC firmware. The process is pretty scary (it'll deep power cycle itself without warning) but it's straightforward. I've been using all 4 cores of my X3 for half a year now for running Folding@Home and at stock speeds it's never had a single problem.

Also, file TFA under "N" for "No Shit Sherlock".

Re:already done, wtf?

[drinkypoo]

Yeah, I just tried it with F5D (what I had) and then with F7 (the latest) and it doesn't work with my X3 720. But I know that it works with some chips and not with others. At least it let me try. I set it to hybrid and auto, then I tried hybrid and -12%, with no luck either way.

The deep power cycle was extra fun because my temp/fan speed monitor apparently doesn't like rapid power cycles, and it detects overheat on the cpu when it happens, then starts beeping at you. It was a little scary to begin with, then before my monitor comes up I get beepbeepbeepbeepbeep

Re:already done, wtf?

[Vectormatic]

my asus mobo also supposedly supports unlocking, but i havent gotten it to work

first cpu i tried was my x2 7750, which secretely is a complete phenom I die, no extra cores
second cpu was a sargas sempron 140, which should be able to unlock to a X2 regor, nothing

the 7750 has a pretty low change at unlocking anyway, but the 140 is a much safer bet. Its still one chip though, so im no sure what the problem is, the board or the chip..

Re:Yield...

[Intron]

Depends on the cost of making and maintaining multiple designs. It isn't just a mask set, it's a whole new design/simulate/layout cycle. And since I/O and shared cache takes a lot of chip area, the benefit is not linear with the number of cores. My guess is that it's cheaper to disable than to redesign.

That said, the disabled cores are very likely not tested, so there is no guarantee that you won't start getting some weird errors if you enable them.

Re:Yield...

[Nadaka]

I would say its closer to an even split in terms of R&D and the fab facility itself. The chips themselves cost at most a couple dollars in materials and electricity to make.

Re:Yield...

[Epi-man]

It would be highly unlikely to just disable a valid core, because if they were doing a fair amount of that, it would be better to make a new mask set that was JUST a 2 or 3 core processor.

You obviously don't realize how expensive your proposition is. A mask set for processors of this complexity cost hundreds of thousands of dollars. That is just for the glass, let alone the cost of actually laying out and qualifying the new glass! So let's say they can change the cost of manufacture/die (the dynamic cost) by a HUGE $10 (that is an incredible reduction on a per die basis), you have to sell tens of thousands of them to simply cover the cost of the glass!

Engineering margin

[Anonymous Coward]

it could also be possible that one of the disabled cores happens to have been disabled because of safety margins : it night not be 100% reliable under all circumstances (using officialbspec's voltage and being able to operate in a wide rrrange of temperatures. Including some constructor branded machine which place priority on silence rather on temperature, and including some badly hacked together beige box with lousy PSU and Thermal mangement). Thus they got disabled to avoid a barrage of recalls from Dells or from small shops (machines which could easily reach 65-70C under load)

but the same core, giving a modest increase of voltage and a very aggressive cooling solution, like water cooling or oil immersion cooling (the kind with which the CPU never rises above 35-38C even when running BOINC 24/7 in background) could still function reliabily.

just like over clocking : It won't work for the full spec (operating range) but might work in the specific controlled environment of an enthousiast.

Of course, if the core was disabled because it's fried, no matter how much liquid nitrogen you pour on it, it won't work.

In Unrelated News

[Monkeedude1212]

A Canadian Man was seen running away from his burning home shouting "beware of the Beowulf" before being arrested for questioning and charged with arson. Firefighters have found over a hundred computers, one of which they believe is the source of the fire.

Re:In Unrelated News

[CAIMLAS]

MOD UP

I'm trying my damnest to stop from laughing and disturbing the office.

This is how.

[AnonymousClown]

I think they use one of those ion Cannons from Empire Strikes Back.

No. They ruin the core's self esteem. They tell it, "You're not good enough to work with the others. Just turn off and sit there and stay out of everybody else's way."

Then one day, a gamer comes by and turns it on. But the core is thinking, I can't do this! This is graphics processing! It's intense! I can't keep up with the other cores!

But the gamer, having faith in the little core, turns him on. And low a behold, the little core can do it, but not without being picked on by the other cores. No! They still tell the little core that he's just not good enough. He can't keep up. But the little core hangs in there to fulfill his duty to the gamer - feeling less than every one else.

One day, the gamer upgrades, and the other cores are scared. They can't keep up. The clock is mad now. He screams, "Come on cores you need to keep up!" The little core comes in and takes up the slack, showing the other cores that he indeed can keep up. The other cores shout, "You did it! You can do it! Come and join our click!"

The little core responds, "No, I'm having lunch with the master clock and by the way, he's promoting me to be your boss. You're my bitches now!"

That's how it happens.

Re:This is how.

[CAIMLAS]

Thank you. I'm telling that story to my son tonight.

Re:This is how.

[Agripa]

I want to hear the Leroy Jenkins version of this story.

Re:This is how.

[__aailob1448]

I'm a computer scientist and I can vouch for this explanation. A lot of people don't realize the social challenges of being a disabled core and I hope that after reading this post, they will have a better understanding.

Excellent work sir, excellent work.

Re:Yield...

[AlXtreme]

And you underestimate the profit product differentiation can generate.

If you have $300 to spend and you can choose between two products, one for $100 and one for $500. Which will you choose?

Now if I take that $500 product and turn it into a third product, $300 and slightly tweaked to perform less than the $500 product. Which will you choose?

You and I might take the $100 product and pocket the rest, but many buyers will go for the $300 one. As long as manufacturing costs are low it's more profitable to have a range of prices.

Mod up

[Joce640k]

In the old days the first thing I did when I got a new graphics card or CPU was to scan the forums for how to re-solder those tiny resistors on the back of the chips to get it to say (eg.) "Quadro" in the properties box instead of "GeForce". These days I don't bother but let the kids have their fun - they're playing games, not running a bank.

Percentages are obviously hard to come by but 20% failure doesn't sound far off (in my experience).

Re:Mod up

[PitaBred]

I have no problem with the kids having fun playing games. I have a problem when they break things and return them for new hardware. That's pushing the cost of hardware up on the rest of us.

If you want to play, fine. Just make sure you take responsibility for what you break as well.

History, Doomed to Repeat, Etc.

[MostAwesomeDude]

AMD used to sell three-core Phenoms. Not for any evil nVidia market segmentation reasons, but because they found that the first couple runs of quad-core Phenoms had a fairly high rate of having one bad core. (It was always the third core, IIRC, due to a manufacturing inefficiency.) So they decided to sell them with that bad core blocked off as a triple-core processor, priced accordingly.

It's true that some companies artificially block off processing power for some reason, but AMD hasn't been one of those companies, and I'd really recommend against doing this.

Re:History, Doomed to Repeat, Etc.

[Mashiki]

Both AMD and Intel used a multiplier lock their SlotA and Slot1 CPU's in order to fill market demands. Skilled hardware hackers figured out how to unlock both CPU's and create unlocked CPU's which operated at a higher speed but were stable. With that, people figured out which batches were a higher end CPU under locked to fill market demand. Saying that bad cores are the reason they're locking them is partially true, the reality is that in most cases locked cores, along with hardware mulitplier locks existed to stop re-silkscreening CPU's and fill market voids for higher demand but lower clocked CPU's.

People buy what fits their budget, the same they did 10-15yrs ago.

try?

[FonkiE]

it's worth a try and running a test suite?

flawed cores can be a reason, but production simplicity is more likely...

Re:Yield...

[jellomizer]

yes but what is the delta between stamping a 4 core and a 6 core?

Depends on your luck (AMD Phenom(tm) II X3 720)

[xiando]

My "AMD Phenom(tm) II X3 720 Processor" does not work with the fourth core enabled. This is to be expected, X3 is sometimes sold as that because the fourth core is just broken and sometimes it's just got a diabled fourth core.

Re:Howcanthispossiblybenews?

[Anonymous Coward]

I'm not sure how this is news... I've unlocked 7 or 8 AMD cores over the past couple years, as well as having a couple that wouldn't.

Anyway, there are some of both scenarios - slightly damaged CPU's and order-filling CPU's being sold. You can visit any one of at least a dozen forums to see if the model / serial / day-of-the-week of your CPU is generally unlockable.

BTW, ASUS and MSI are far from the only boards with ACC. I personally prefer the Gigabyte MA785x lines.

Note: I'm neither a teenager nor terribly poor, just exceedingly frugal.

Re:Yield...

[digitalchinky]

Expensive? They make them out of sand! :-)

Oh hell yeah...

[DG]

For a while I was selling race car / high performance street car suspension systems.

I had discovered that 90% of the aftermarket shocks being sold as performance upgrades were actually crap. The customer is really not qualified to properly evaluate a shock valving and so it is very difficult for them to differentiate between a proper performance shock and a juiced-up pogo stick.

I started putting shocks on a device called a "shock dyno" (which measures the forces produced by the shock at different shaft speeds) and discovered an absolute parade of horror. Details can be read at http://farnorthracing.com/autocross_secrets6.html

To get the good stuff you needed to be paying upwards of $3000 per corner (so $12000 per car) which is far, far out of the price range of most customers.

So I was building packages based on a brand of shock that was pretty decent and much cheaper. Even though the base design was solid, it still suffered from manufacturing variations. To get around this, I would buy batches and then dyno the lot. Shocks that were close to each other became matched sets, and I'd tweak the adjusters on the shock to ensure each pair was as closely matched as possible. On top of that, I designed some hardware to resolve some other tricky problems typical of the off-the-shelf aftermarket designs, and only used the best bang for the buck components to build them.

When done, I provided a race-quality suspension system, dyno-matched (and it came with the data sheets to prove it) that was very nearly the equal of the $3000/corner systems, for about $500/corner. I say "nearly" the equal because the adjusters on my shocks worked nowhere near as well as the adjusters on the expensive shocks, but in terms of absolute performance, they were effectively identical.

There was almost no markup in these parts; I was hoping to make it up on volume and I knew the customer base was price-sensitive.

These suspensions were INCREDIBLE deals. There was nothing else like it anywhere for anything less than 5 times the price, and unlike all the cheaper stuff, I could prove that it worked. What's more, I could run the cheaper stuff on my dyno and prove that it DIDN'T work; that it was categorically JUNK.

I sold almost none of them, and the universal complaint was "too expensive".

Even when I opened up the books, showed what I was paying for the components, explained why *this* part instead of *that* part, explained every single design decision and proved why it could not be made any cheaper without compromising the functionality, over and over again potential customers would choose to buy non-functional (but shiny) JUNK over functional parts based solely on price.

It was mind-boggling, and eventually I just said to hell with it and found something else to do.

The chip manufacturers are right on the ball here. If I were them, I'd be encouraging the creation of these kinds of motherboards and rather than down-rating the high end parts to make mid/low end, I'd be cherry-picking the best ones for the high end and defaulting the output of my fab runs right to the mid/low end SKUs. In fact, I'd be tempted to DESTROY any chip with a bad core and ensure that all the low-end chips were fully functional - specifically to build a reputation for being "overclocker-friendly".

You can't make money off what you DON'T sell. Believe me, I know.

DG

Re:Oh hell yeah...

[unother]

I certainly have no basis for criticism here, but I was wondering why you chose not to market your shocks to the $3000+ audience instead? It would seem with a reasonable markup (time and energy spent) to $800 or more, you could have captured the "low-end" there.

OTOH I would imagine persons who spend $3000+ on shocks are pretty much loyal consumers for some specific brands...

Re:Oh hell yeah...

[DG]

There were two different types of customer for the $3000/corner parts:

1. Real Racers who understood the value-add the top line components brought to the table and who would sell their mothers to get that functionality; and

2. Rich posers who were all about being EXTREEEEM and who were buying the name to impress other rich posers.

Neither of these markets are very big - but they would spend the money.

The very, very much larger "budget racer/street driver" market was all about price.

DG

Re:Oh hell yeah...

[unother]

Okay, that makes sense.

In a sense it seems you showed how the market was "rational"--i.e. since in the larger market priced trumped all concerns, quality went out the window. Shame.

Re:Oh hell yeah...

[bungeejumper]

I bow to the slashdot user with id number 789 !

Re:Oh hell yeah...

[hellop2]

For what it's worth, what you did seems awesome. IMHO, seems like you needed a marketing department to go along with your nerd department.

Re:Oh hell yeah...

[hellop2]

huh huh what? 989

Re:Oh hell yeah...

[Vectormatic]

very interesting story

just one thing to add though, for me personally, no matter how good a deal 500/corner suspension is, if i only have 1000 bucks for an entire set, i wont ever buy your product, since it just is to expensive (not for what it is, but it doesnt fit my budget)

sad to see that a pationate technical guy like you got out of the racing biz though..

Re:Oh hell yeah...

[DG]

I went back to the Army.

Every day in Afghanistan (and there were some shitty ones) was still way better than every day spent trying to help spoiled rich kids drive faster.

Not that everybody I encountered racing was a spoiled rich kid; far from it. But enough of them were that it was a real soul crusher. Far better that I spend the rest of my life accomplishing something meaningful.

In the meantime, I opened up my specs for those suspensions: http://farnorthracing.com/autocross/konis.html

DG

Re:Oh hell yeah...

[DG]

I dunno, that 789 guy seems pretty cool too.

DG

Re:Yield...

[91degrees]

Well, okay. I'm only vaguely familiar with the amount that a company like TSMC would charge. But it's a competitive industry. The margins are going to be fairly small.

I guess what it it is that there's a certain amount of time during which the fab will remain profitable, before the technology is obsolete. In that time they can produce a specific fixed number of wafers. So the cost of each wafer is the cost of the fab divided by the number of wafers that can be produced. That's not cheap. A fab costs a lot more than a printing press.

No, not so much

[Sycraft-fu]

If a core is just flat out non-functional then yes, you are right, a system wouldn't boot. However that it works mostly doesn't mean there isn't a problem. There could be a single instruction that has a flaw, so everything is fine unless that instruction gets executed but when that happens you get a crash or worse, data corruption.

If you think Prime95 is an accurate test, you are kidding yourself. Prime95 tests the FPU mainly, and is good for heat testing. It is not a full CPU test. So maybe the FPU works great, but one of the other units doesn't.

So no, you don't know that nothing is broken. You assume nothing is broken. Maybe that's fine, however then no bitching if you get data corruption or the like because there was a problem that you didn't know about.

Re:No, not so much

[kriston]

Okay, not only are the disabled cores (and their local caches) sometimes tested to fail, sometimes they are NOT TESTED AT ALL. If the production run says we're only going to have 3 out of 4 possible cores (with cache), they're not going to bother testing the fourth core (and its cache) if the first three test successfully. Worse, if they're calling for 2 cores out of 4, they test and get two good ones and DO NOT TEST the third and fourth core.

Remember, also, this is not just the core that is tested but the L2 cache. In multiprocessing, a core is hardly useful without a working local cache.

Many of these re-enabled cores have faulty caches. The benchmarks do not show any benefit and in many applications that extra core without cache actually degrades peformance.

Re:No, not so much

[IorDMUX]

If the production run says we're only going to have 3 out of 4 possible cores (with cache), they're not going to bother testing the fourth core (and its cache) if the first three test successfully. Worse, if they're calling for 2 cores out of 4, they test and get two good ones and DO NOT TEST the third and fourth core.

You don't seem to realize how the economics of this really works out. Nobody will set up a production run before hand and say "this line only needs to produce 3 usable cores". Nobody will do this because no fabrication process has 100 % yield... in fact, most cutting-edge runs have far less.

Let's say your fabrication process produces 1M chips per run, and we have the capacity to do 2 runs at once. You set up a '2-core' run, and find that 95% (950k) of the results have two working cores. Well, that seems great, now you can sell this 95 % at your bargain bin price. However, your '4-core' run may have had a yield of only 40 % (400k)... now, you have to spend more time and money producing more 4-cores to meet demand (lets say... 750k each), while you are selling perfectly good 4-cores at a 2-core price.

Instead, all of the chips will be fabricated and tested at a 4-core 'level'. If 1 core fails, put it in the 3-core bin. If 2 cores fail, put it in the 2-core bin. If your yield was better than you expected, then you can bump some 4-cores down to the lower bins to meet demand. If your yield was poor, you are drawing from a much larger sample of chips (2M, so 40 % yield --> 800k 4-cores), so you don't have to produce more to fill the demand!

Re:No, not so much

[electrosoccertux]

and that's why I made the comments about the games. Prime95 is not the only stability test I ran, lol, duh.

All my applications work fine; I've been running this rig for about 8 months now.
I don't know why you're so hell-bent on telling me my chip is broken, people get pissed when someone gets something for free it seems...we should tax me since I didn't pay for it.

Re:No, not so much

[BikeHelmet]

Prime95 tests the FPU mainly, and is good for heat testing.

But no guarantees on the heat testing, either. My last GPU (not CPU) failed at stock speeds in Furmark. - now with the new drivers it's even worse - my GPU temps climb to 102C, and then my system shuts down. I only did it twice, because I didn't want to damage anything.

Oh well - modded cooling to the rescue!

Re:No, not so much

[Klinky]

How do you know your stock clocked CPU isn't failing on you? It's happened before where you get a faulty CPU out of the box that doesn't boot, what about phantom errors in stock clocked CPUs that don't fail to boot.

Overclocking/unlocking is probably not a good idea in a production environment where stability is key, but for enthusiasts who know what they're doing and have been doing this for decades than it should be fine, so long as you stress test.

You seem to have the notion that you're risking data loss by overclocking/unlocking. You're risking data loss by not backing-up your data. Even stock clocked CPUs overheat or hard drives fail. Also I don't think I've ever had data loss occur from overclocking, only stability issues where the system crashes out. It's unlikely you'll lose data on your hard drive due to an overclocked CPU.

Overall your post just reeks of fear-mongering.

Re:No, not so much

[cheese_boy]

You don't seem to realize how the economics of this really works out. Nobody will set up a production run before hand and say "this line only needs to produce 3 usable cores". Nobody will do this because no fabrication process has 100 % yield... in fact, most cutting-edge runs have far less.

I don't think you realize how this works out in practice.

In practice there are multiple stages to testing.
And a part may be down-binned for numerous reasons.
1> Frequency - ie. some part of the chip doesn't function at the freq./voltage specified.
2> Power - ie. the part would function but it would consume more power than spec.
3> Functional - ie. some portion just doesn't work (ex. part of the cache is so messed up that the repairing mechanisms can't compensate, and have to disable that section of cache completely, so a part that has patterns for a 8M cache is used as a 4M cache. Or maybe one core doesn't have it's branch prediction operate properly or maybe one opcode doesn't give the right result in a certain corner-case, so one core is disabled.)
4> Supply/demand - ie. the actual yield is higher than the actual demand at the top bins, so parts have to be down-binned to meet the demand in the lower bins. This may mean that certain lots of parts - or parts with certain characteristics get run through a test program that automatically jumps to a lower bin if there is already more than enough supply at the top bins. (Testing is expensive, and if you can shorten test time by an average of 10% because on 30% of the parts shipped you shortened the test time by 30% it's a multi-million dollar benefit.)

If the parts get downbinned in an earlier stage of testing (because there are normally multiple stages) normally that portion that's disabled won't be tested at the later stages. For example, if you test at the wafer level, and determine that you need to downbin some parts because they're almost certainly going to consume too much power, you only test those parts at the lower frequency/core-count once they're in packages. Then Joe Q Hacker gets the part and re-enables a disabled core - he doesn't know how much that part was tested. It is quite possible that the core he re-enabled wasn't tested as thoroughly as the ones that were enabled when he got it. Since he's got a liquid cooled setup though, he doesn't have issues with the power dissipation - but maybe there was some other latent issue that was never even encountered.

Or maybe it gets disabled when it's socketted in a Credence Sapphire ATE (Automated Test Equipment) but the next stage is a more PC-like environment, and at that stage it already has a core disabled, so the 4th core doesn't get the full testing in that PC-like environment.

In your example you didn't put what the demand was - if the demand is 10% four-core, and 90% two-core, it makes sense that you meet demand by skipping over the four-core testing 3/4 of the time, and jump right to the 2-core testing. It'll save time, and that means it saves money because maybe you can get by with 8 ATE platforms instead of 10. And the code to implement that took maybe a month of engineering effort to implement and test (spread across 2-5 people), which is much much cheaper than even 1 ATE.

Without knowing the actual test-flow AMD uses and the yields, (neither of which will be revealed to the public) it's impossible to know how likely a core that was sold disabled is actually good, and how thoroughly that (disabled) core was tested before it was shipped out to customers.

Re:No, not so much

[Fjandr]

The person who replied to you seemed neither hell-bent on telling you the chip was broken, nor pissed that you got more cores for free. You seem to completely miss the perfectly valid point that it is entirely possible there is something wrong with one of the disabled cores that you won't find out about until it bites you. It's possible you got a perfectly functional chip. It's also possible you have one with something broken that hasn't cropped up because the conditions haven't been met.

Unless you have the manufacturer's test software, the entirety of the point was: You don't, and can't, actually know for sure. That's it. No bitching, pissing, or moaning. Just a simple fact. If it works for you forever without errors great. The certainty that will happen, though, does not exist.

Re:No, not so much

[electrosoccertux]

yours is but mine isn't!

Re:No, not so much

[dudpixel]

no ones saying its broken, they're saying that you have no reliable way of telling if it is or not. The fact that it has worked for 8 months may be purely lucky, or it may well be fully working. You had no way of knowing that 8 months ago, and not everyone wants to risk losing all of their data by testing a cpu this way (essentially, trial and error).

Not only that, read the other comments regarding the cache being potentially faulty. Your 4 cores may be working, but are you getting the full benefit? It seems that your methods for testing it involved running it in your live environment. That wouldn't be acceptable for many users.

Re:No, not so much

[electrosoccertux]

I mean there are forums dedicated to computer hardware and computer building and a lot of us do this and none of us have experienced any problems. By "lots of us" I mean there's about 100, and none that got theirs to unlock and pass stress tests have experienced other issues. And so we are certain that AMD bins these perfectly functional quad chips down.

Re:No, not so much

[dudpixel]

ok the bit the worries me is that you can get computational problems on very random occasions that wont cause a crash at all.

For games, this is not a big deal. But it would make me nervous thinking that any spreadsheet I used or any calculations done by software I rely on, could be incorrect without me ever picking up on it.

It really comes down to how critical/crucial your desktop use is.

Re:No, not so much

[electrosoccertux]

as I already mentioned, the FPU is fine...

Re:No, not so much

[dudpixel]

FPU is for floating-point. what about integer calculations?

Some people have too much time on their hands...

[carlos92]

...and so they are willing to trade a lot of time in exchange for a little monetary compensation. That fact is behind those market segmentation strategies that make you jump through hoops (like clipping coupons from the newspaper or collecting cookie wrappers) to get a discount. AMD is not going to worry about this, as the existence of this new segment could mean more sales of the lottery chips.

Re:Yield...

[91degrees]

I think it is pretty much a linear relationship. You get 50% more 4 cores per wafer than 6 cores.

Lottery core?

[jd2112]

Does this mean that the 'Lottery Core' article wasn't an April Fools joke?

'Cause I've got a Golden Ticket ...

[psbrogna]

You just watched Charlie & the Chocolate factory, didn't you?

Re:Yield...

[psbrogna]

Yes, but the sand is on the beach. Moving the sand to the fab facility and filtering out all the seagull poop can't be cheap.

About a year ago..

[damuhatori]

Isn't this old news? I remember when you could buy a Phenom II X3 and unlock the fourth core.

abie-normal?

[scharkalvin]

Why does the idea of unlocking cores in a cpu remind me of the scene in Young Frankenstein where Igor grabs the jar containing the abnormal brain? (Whose brain is this? Some guy named Abienormal).

AMD facilitates your tinkering...

[Anonymous Coward]

I recently bought a Phenom 555 and an Asus motherboard that allowed me to unlock the two dormant cores. No doubt one is taking a chance when doing something like this, but personally I think it is a marketing gimmick by AMD who is appealing to the "hardware hacking enthusiast". Anyway, I put my new machine together and was amazed at how easy and obvious it was to unlock the two cores. On first boot I noticed a message saying "Hit F? to enter ACC". Doing so would have taken me to a menu where I could unlock my cores. Instead I entered the bios settings, went to the "Processor" tab, set ACC to "enabled" which then revealed a setting titled "Number of cores" which had a value of 2. I set the "Number of cores" to 4 and rebooted. Upon reboot the bios splash screen said "4 cores enabled !!!" and had a graphic of a cpu with the number 4 overlaid on it. It was as if it was designed to make it very easy to do and my machine has been running perfectly stable 24/7 for a month now.

Bottom line is, this is a marketing initiative by AMD aimed at the hardware enthusiast.

The Little Core That Could

[pyrr]

This story requires illustrations and a publishing company.

Unlocking semi-officially supported on some mobos

[Alok]

Many AMD motherboards with 710, 750, or 850 SB (south bridge) support unlocking of cores in BIOS - the feature is called as ACC (advanced clock calibration). In fact, right now I am sitting on an X2 555 trying to decide whether to keep it (and have to spend more on DDR3 as well) or return to store; with the potential to unlock it into an X4 955.

However, from some accounts AMD was trying to convince motherboard mfrs. to stop offering ACC in newer boards; so the fact that its working on 890 SB now is the actual news (if the article is correct). Not really surprising though, now that users are getting spoilt into having easy ways to potentially unlock cores it would've been pretty hard to stop that and make competing mobos more attractive :-)

RAIP ?

[bugs2squash]

Not a good acronym. But perhaps we could make a system that runs error corrected on a pool of marginally-operating processors. Besides, who's to say that a processor that passes all the tests will always work.

Car analogy

[Anonymous Coward]

Whoa! a car analogy from someone with real first hand experience in the analogue.

What is /. coming to!?!

Oh come on

[Weaselmancer]

Now in the case of "hidden cores", what's the risk? Do you even know? Do you know what kind of flaw would lead them to legitimately disable a core? Is that one core unable to tolerate the same clock speed as the others? Is it functionaly broken such that it will return incorrect results for some operations?

Ever written any assembly?

I promise you, if a single instruction had a single bit out of place - the whole house of cards would come tumbling down. FAST.

If you run the thing with a good stress test for a day and it doesn't blue screen - you're golden. I think the odds of a cpu core being actually physically broken and somehow not crashing the system are roughly equivalent to being struck dead by lightning while holding a winning Megamillions ticket. During a blue moon in December on the day they release Duke Nukem Forever.

Re:Oh come on

[osu-neko]

... I think the odds of a cpu core being actually physically broken and somehow not crashing the system are roughly equivalent to being struck dead by lightning while holding a winning Megamillions ticket. During a blue moon in December on the day they release Duke Nukem Forever.

Wait... you're saying Duke Nukem Forever is going to be released on December 31st, 2028?

w00t!

Why disable?

[izomiac]

I'm starting to wonder why they disable perfectly good chips to begin with. Most businesses just give a free upgrade if the customer orders something that isn't in stock and the price difference isn't too great. So why not just give a few customers a 5 or 6 core processor for the price of a 4 core? Customers are happier, the manufacturer doesn't have to waste time/money disabling stuff, and you have fewer RMAs (e.g. core #2 dies on a quad core, but #5 wasn't disabled so the customer still has 4 working cores).

I suppose the only drawback would be that customers would be less willing to pay for the pricier processors if they knew how artificial the distinction was. OTOH, customers are perfectly willing to pay nearly twice as much for an LCD screen with a "zero dead pixel" guarantee, so the more thoroughly tested units should still be able to fetch a high price.

Re:Why disable?

[SharpFang]

...also, customers who got only 4 cores for the same price as the guy next to them who got 6 would get pissed.

Not Quite a New Trick

[stilz2]

Many AMD motherboards with the SB710 southbridge have been able to do core unlocking for a while now; most SB710 boards from Asus, Gigabyte, MSI have this capability. IIRC, AMD announced that the new chipsets (890 series) can no longer unlock, and that is when the manufacturers decided to do it on their own.

Better luck dating a fat chick.

[IBABad1]

I'd think you'd have better luck dating a fat chick in hopes that one day she drops the pounds and turns out to be hot and lacks stretch marks.

Re:Yield...

[nomadic]

And the dance continues.

A Personal Story of Software Unlocking

[Liambp]

Some years back I bought a mid range graphics card called an ATI 9500 in anticipation of the soon to be released blockbuster called Half Life 2. A post on the internet alerted me to the fact that the 9500 actually used the same chip as the much more powerful ATI 9700 but with half of the channels disabled. Happily a simple software mod allowed me to unlock the missing four channels. I was delighted and enjoyed top drawer 9700 performance at a bargain 9500 price. Sadly the game Half life 2 was subject to delay after delay so I played other games while I waited, none of which really needed the extra graphics performance. It was more than a year later when Half Life 2 was finally released. I waited with renewed eager anticipation for the release date confident that my home brew 9700 would at last get a work out. Let us gloss over the fact that it took several further hours for me to download most of the game from Steam despite having an original disk. Eventually the game was installed and I eagerly started playing only to be surprised at the strange checker-board graphical effects. Google confirmed that these effects were not a creation of Valve but were in fact a sign of faulty cores on my pseudo 9700. Removing the softmod downgraded me to a vanilla 9500 and allowed me to play the game as it was intended.

Moral of the story: Sometimes manufacturers disable cores for a reason.

Re:A Personal Story of Software Unlocking

[Joce640k]

I did that and mine worked perfectly...!

Re:Yield...

[Droideka-TheGuy]

They're not all bad cores. What AMD does is locks cores that don't perform within a certain margin of error. While some of the cores may be flawed, some are simply a megahertz or two off of the others, and so AMD locks them, because some boards just can't handle the difference. Asus, MSI, and I believe ASRock all are stable enough even with different speed cores to allow for some small variations. So when it comes down to it, you have about a %75 success rate for one that you can actually unlock cores.

Re:If it was Intel, maybe.

[pyrr]

Cyrix processors tended to lack power altogether, but they ran fairly cool.

No, I remember, when I used to play around with idle tinkering such as overclocking, that AMD's 486 and K-5 & 6 series processors ran on the hot end as it was. With comparable Intel processors, it was easy to clock them up to 15% higher without really even needing fancier heatsink (the first Slot 1 processors, with the heat-trapping packaging, were rather forgettable, though). An active heatsink was a necessity with the AMDs to keep the temperatures under control and the system stable under load. The net of it was, if you wanted to tinker with overclocking, Intels just worked best; you were all but guaranteed to be able to get that extra performance, depending on the capabilities of the motherboard and how much effort you were willing to put into the cooling system.

Also, I --was-- speaking "historically". Yes, it's been a while since AMDs failed spectacularly. And no, it wasn't just Tom's Hardware, though that old video does earn style points for spectacularity. Many years ago, I replaced fried AMD processors and found scorched system boards underneath; this was a pretty serious expense considering that all that failed was usually a cheap processor heatsink fan. AMD rose to the challenge and fixed the problem, but it's a problem that shouldn't have existed in the first place. Processors should protect themselves when they get too hot, because heatsink fans can and do fail; that shouldn't be an event that does more than causes a computer to shut-down until repairs are made.

I'm happy for your lack of experience with failing AMD processors, but I'm sure there are some people who never had problems with Quantum Fireball and Bigfoot hard drives either. Notice that I don't point out how many Intel processors I've screwed-with and had them survive to live long lives. A data point or two doesn't really define a trend. My employer has thousands of workstations deployed (I support around 450 or so). When dozens of one type of processor fail, it's just something you take notice of. Our vendor is a little company by the name of HP. No, not a pinnacle of quality by any stretch of the imagination, but the Intel-based HPs (basically the same model, just different innards for the machine, and a little higher price) don't seem to have that sort of problem, just a dead processor here or there. Yes, the Athlon-64s do seem to underclock themselves, frequently, if performance is any indication. That doesn't stop them from failing, it might buy them a little time.

I honestly wish that AMD would get it together. They got the jump on Intel with the 64-bit architecture, they did it right, and Intel folks looked like chumps, with the first iteration of their 64-bit platform kind of being that pathetic imitation 64-bit processor (the "EMT"). Intel needs a strong competitor to keep them honest and to give consumers a choice. AMD might be close to nailing the performance and reliability edge down, but they're going to have to maintain that for a while before they'll shake the perception I have of them. Lottery Core motherboards really don't help, though I guess if there really are many, many reports of those additional disabled cores being A-Ok, I might think better of their brand.

Re:Old news and overclocking is dead

[JoeSixpack00]

Just because you diasagre with me doesn't make me a troll. What I said about CPU prices is undisuptable fact, and if you don't believe me when I say unlocking cores is old news then google it and see for yourself>

Maybe you should get off of your high horse.

What I'm telling you

[Sycraft-fu]

Or rather others is that this sort of thing is a RISK. Now if you are willing to take that risk to save $100, that's fine, but recognize it for what it is. What I don't like is people who sell stuff like this as safe and/or reliable. "Oh I tested things and it works fine, there's nothing to worry about!" No, not so much. There is a non-trivial risk of problems. If you are ok with that, then fine, but acknowledge that.

Also you have to understand luck is a factor. Far too often people present this as a sure thing. "Buy the cheap one, get the full one for free!" Again not so much. Maybe it works, but maybe it doesn't. If it doesn't, tough luck.

I'm fine with people who want to overclock, or otherwise try to push their systems. However it needs to be understood there are risks to that.

Re:Yield...

[Nadaka]

That is what I said? The fabrication facility has a huge cost.

Over-Coring

[sixsixtysix]

i claim ownership of the term Over-Coring

Re:If it was Intel, maybe.

[DrMrLordX]

AMD has no problems with reliability. Performance? Sure. Reliability? No.

If you have suicidal K8s in your department, I wouldn't be pointing the finger at AMD (or you either, I'm thinking OEM or vendor). They don't have a reputation for dieing at stock clocks after years of use. Sounds like someone, somewhere, cheaped out on the PSU or mobo.

And how you intend to draw any meaningful conclusion about AMD when basing your assumptions on the heat output of k5s or k6s is anyone's guess. Have you examined the thermals from any processor AMD released in the last, oh I don't know, two years?

And do you even know WHY AMD releases CPUs with disabled cores? Here's a hint: it has nothing to do with their products exhibiting poor reliability.

Re:Yield...

[Calinous]

Yes, make blueprint for a new 3-core processor based on the 4-core working one. Debug that blueprint. Wait a couple of weeks for the masks. Order test wafers. Wait a couple of months. Debug test wafers. If all is perfect, start production. If all is almost well, use a very expensive machine to repair cores. If not, build a new mask, order test wafers, and go again. It will usually take 3 to 6 months until production is in full swing, and some of that time you need to have a team of engineers taken away from the development of the next processor core. We're talking about AMD, not "three full development teams" Intel here.

Re:Yield...

[Calinous]

"I'm only vaguely familiar with the amount that a company like TSMC would charge"
      The blueprints/masks/whatever for 65 nm processors made at AMD's fabs (now Global Foundries) can NOT be used in Intel's 65nm facilities, TSMC 65 nm facilities or whatever other 65nm facilities you could find (due to differences large and small, but almost all of them critical).

Re:Yield...

[Calinous]

It's close to the difference between making an aircraft wing out of aluminium versus composite. You need different everything for it, and research and development time is only surpassed by time lost waiting for wafers (months and months between an order and delivery).

Re:If it was Intel, maybe.

[Calinous]

"They got the jump on Intel with the 64-bit architecture, they did it right"
      AMD was better than Intel starting with the K7 processors (the Slot A ones), continued with the 1GHz point (and Intel's 1GHz fiasco), with Athlon MP processors, and it was at the top one way or another until the Conroe (Core 2 Duo) (Pentium 4 Northwood was a nice contender, but I still think AMD was on top then). Unfortunately, production capacity and other issues stopped them from financial success, and AMD is back to selling the slower processors for less money.

Re:Old news and overclocking is dead

[SharpFang]

Dual core is optimal in most cases though.
One core for the OS and background tasks, one core for your foreground app. Everything is snappy and smooth.

Re:Yield...

[Vectormatic]

not quite,

this 6-core still has only 6mb of l3 cache, the same as the 4-core, meaning the 6 core doesnt just take 150% of the die-size of a 4-core. Also the memory controller and interface logic dont suddenly swell up. All that is needed is two chunks of extra l2 cache, and 2 execution cores (and perhaps some extra internal bus logic)

the following die schematic for a phenom II die suggests adding two cores would probably take 25% extra die space:
http://commons.wikimedia.org/wiki/File:Phenom_II_Deneb_-_Sch%C3%A9ma_Die.svg

Re:If it was Intel, maybe.

[dave87656]

I know this is comparing Apples to Oranges but ...

I've always found that AMD CPU's seemed to work better than the comparable Intel CPUs. My Girlfriend has a PC with an AMD Athlon 64 x2 5000+ and 2G of memory. It seems to be much quicker than my Laptop with a Core2 T5800. Of course, we're comparing a desktop to a laptop with different disk-drives and Vista compared to Windows 7, but it reflects what I've seen in the past.

I'd be curious what other people have experienced.

Re:If it was Intel, maybe.

[dave87656]

I used AMD processors exclusively from the mid 90's to a couple of years ago and never had any problem with them. I wasn't pushing the performance envelope but I think their reliability has been good.

Re:Unlocking semi-officially supported on some mob

[Vectormatic]

not sure if you know this, but any AM3 chip can also run in a AM2+ mobo with DDR2 (if the mobo supports the new chips)

My year old AM2+ board already has a bios update for these 6-core chips

So if you have an existing AM2+ system (or an intel 775 system filled with ram), you wont need new DDR3 for that 555

Re:Old news and overclocking is dead

[Vectormatic]

i ran win 7 beta on a sempron 3000+ with 1 gb of DDR, single channel (old school socket 754), and for desktop purposes it ran snappy enough, having a browser open and streaming music and such.

Now if i started doing more things at the same time, like installing two programs and the browsing/streaming, things would clog up, but i attribute this mostly to the 1gb of ram.

I am pretty confident that AMD's slowest cpu (a 2.7 GHz single core sempron these days) would do just dandy given 2gb of ram for general joe-sixpack use in windows 7

Re:Unlocking semi-officially supported on some mob

[Alok]

Replying late, but wanted to thank you for the heads up.

I know that AM2+ boards can run AM3 chips; but mine is an old Biostar K8M80-AM2 ... which according to their site doesn't support any of the newer chips :(. Luckily the MSI 785GM-E51 that will replace it already has an update for x6 chips, though I will probably stick with the 555 (no need for 6 cores).