Originally posted by: ectx
Originally posted by: Xcobra
my logic: you buy it, you can do whatever the hell you want with it.
Agree - however, if you mess up anything through your oc, don't ask for a RMA. Ay least when it it clear that the early demise is due to the oc.
I guess overclocking my Black Editon processor means my soul is as black as the box it came in.
This is my (basic) understanding of how processor clock speed and voltages are decided:
If Intel has 1000 chips produced and they need only 50 QX9770s(random numbers picked for this explanation) at whatever given voltage, they will test 50 chips for those specifications. If those 50 chips pass, then they won't test any more chips for QX9770 specifications. They will then test for the next model down to get enough chips for that processor model. That means that some chips might have been able to run at the QX9770 specifications but were never tested for it.
Let's assume the QX9770 runs at something like like 1.2 volts at 3.2Ghz. If Intel tests 120 Q9550s for 2.83Ghz at 1.2 volts and all they needed is 120 Q9550s, that's all they'll do. They don't test the the chips beyond the limits of what they are trying to get them to run at at the time. A Q9550 that passes testing for running at 2.83Ghz at 1.2 volts will be guaranteed to run at that clock speed but not guaranteed even a tiny bit higher. However, that doesn't mean that it isn't in fact the same silicon that can reach 3.2Ghz or maybe even higher on the same 1.2 volts. The consumer gets to figure that out if they so desire. This is all I'm saying. A Q9550 or a Q9650 could in fact be the same grade silicon as the QX9770. Of course something like a Q9550S is tested to even higher standards than the Q9550, but even that doesn't mean it's necessarily better, just that the guaranteed minimum is better.
To summarize, a 3.2Ghz 1.2 volt processor is not necessarily better than a 2.83Ghz 1.2 volt processor that was never tested for a higher clock speed. That 2.83Ghz processor just might clock as high at 1.2 volts as the 3.2Ghz one can. A 2.83Ghz 1.2 volt processor is better than a processor that couldn't reach 2.83Ghz without 1.25 volts though, I understand that.
I do understand that there are differing qualities in processors, and I suppose I shouldn't have made it sound like they are all the same. The poster I was quoting was implying that all $1000 processors are higher quality than all lower priced processors silicon wise. I was countering by saying(or attempting to) that it's in fact much more likely for a lower priced processor to be as good as the $1000 processor than not.(At least for the Core 2 line)
Binning is not really like that, at least it isn't in the rest of the world (Intel has their own fabs so I guess they could be different but it'd be surprising). The first thing to note is that when you own a fab you want it to be running 24/7/365 every minute not running is essentially money lost. So up until end of life they are likely churning these things out more or less constantly.
The second thing is that binning isn't just pass/fail and it doesn't just happen after packaging. Packaging parts is expensive so you don't want to package parts that don't work so you test and bin at the wafer. So you'd run say 50 different tests and based on the results of those a part would fall into various bins. Some bins are pass/fail others are things like speed or power measurements. Based on that you burn some values into the cpu (like say the multiplier lock) and then pass the part on to be packaged. This is all done automatically.
So lets say, for example, that the power limit is 100W at operating frequency. If your 9550 can run at the same speed as a 9770 but runs at 100.1W it will be binned lower. Odds are, technically speaking, your 9550 isn't as good but it is entirely possible that it is close.
I'm simplifying things here and of course I don't know Intel's process as I'm sure its not publicly released but I do know in general how these things work.
You both remind me of the blind men and an elephant parable...not realizing you both are describing different aspects of the total testing/binning/packaging life-cycle of an IC regardless the IDM involved.
http://en.wikipedia.org/wiki/Blind_men_and_an_elephant
Dug is right, and you are too, the key here is to realize that neither of you are speaking to the process in its entirety and that neither of you are discussing aspects that are actually inviolate and mutually exclusive of the other.
For beginners I would not be surprised if the two of you are using the term binning differently, no two IDM's use it to mean the same explicit steps in their cycle...the same goes for terms like test and validation...they kinda mean the same thing but when you get into the nitty-gritty minutia of the process flow you find out that one man's "test phase" is another man's "pre-bin validate phase", etc.
Unless you've managed a project that involved you becoming intimately knowledgeable of the test/bin/package loop, versus say being exposed to it simply thru workshops and second-hand accounts, you are asking a lot of yourself to know exactly what that elephant looks like and you are assuming a lot when you conclude that others who describe the process slightly differently than how you perceive it to exist are the ones who are in error.
So lets say, for example, that the power limit is 100W at operating frequency. If your 9550 can run at the same speed as a 9770 but runs at 100.1W it will be binned lower. Odds are, technically speaking, your 9550 isn't as good but it is entirely possible that it is close.
.
Great, now that you let that cat out of the bag, we're going to see motherboards that display a "CPU EULA", when switching processors or booting up for the first time.
Heck, that might be an interesting feature, especially when the motherboard declares, "CPU overclocked -- WARRANTY VOID".
Of course, to make that stick, you would need a small bit of flash memory on the CPU, or maybe just a fuse, that the mobo would blow once you overclocked, so then Intel or whomever would know to not take back the CPU.
Funnily enough I am involved in the test/bin/package loop but as I've only ever been exposed to it at one company I can only speak from my experience there.
Funny, OP almost had a point... have you been under a rock and not realized AMD is releasing chips with unlocked multi's? What the hell do you think the purpose of an unlocked multi is??? What do you think AMD's Overdrive software does? (Rolls eyes)
Truly a village idiot!
What about people who underclock their processors?
There are more expensive low power chips rated at the same power. You're stealing those.
I bought a 965 and disabled a core, then underclocked it. I'm not taking *ANY* chances.