Intel plans to deliberately limit Sandy Bridge overclocking

[prava]

IMO what saaya is trying to get (not that is easy to understand with such amount of quotes ) is just a little explanation about why overclocking has been down-graded from other arquitectures.

I don't think he never meant to challenge the team behind SB, but he is just trying to understand the why. In a way he is like me, as we are not ok when we are told no unless they tell us why, because that way we understand the reason fully and are able to use it in another level and comprehend the reasons behind it.

Just my 2 cents.

PS: if Francois can't talk about that "why", he should just say so, aka "can't talk about why we did things the way we did but can assure you we had our reasons".

[Manicdan]

IMO what saaya is trying to get (not that is easy to understand with such amount of quotes ) is just a little explanation about why overclocking has been down-graded from other arquitectures.

I don't think he never meant to challenge the team behind SB,

i think he did a little, and i think he is right
if the team is so awesome, as that has seem to have been pointed out constantly, why did they get stuck, or was this really a decision from managers who wanted more $ per mm2

i cant recall a single cpu that ever used different bus speeds, everything on stock parts was always controlled with multipliers, so why did this one have to change?

[Chumbucket843]

With all Due respect, you are trying to teach driving to the world champion of F1 ... the team working on SandyB is probably one of the best in the industry.

It is always easy to say "They should add this, or that" , or "adding a simple divider or multiplier" ... If it was that simple, it would have been done ...

It is always easy from your desk to say what to do, just try to even do one cell of those processors, and you ll realize that the complexity of the subsystem you try to do is so much larger than what your brain can comprehend without years of training ...

your company has been designing microprocessors for 40 years. they should be able to understand what's going on. any good hardware designer should have an grasp on how modules work in parallel. imo it's not that hard. the only things that are hard to understand are the simplest as they can not be explained in simpler terms.

to put what i am saying as simply as possible, i dont care if you say it's hard. i have very high standards for a company with 40 billion dollars of revenue annually and you have made a sub par cpu. keep in mind this is coming from a guy who owns 2 i7's. yes, i cheated on the 920. she just doesnt appreciate my overclocking abilities anymore. i even OCed here till she reached Tjmax.

The architects and designers of SandyBridge are very hard working people, and thinking that we did not try to put a clock divider or multiplier is kind of insulting.

they should work smarter, not harder. you can say what you want about your design teams but the chip consumers buy is going to reveal more. imo intel has not shafted me with sandy bridge, they just made something i dont care about. maybe some one else will deliver.

[[XC] gomeler]

After playing with a 655k I must say I hope there is at least some wiggle room for the bclk even if it requires going below 100MHz(if that's indeed the baseclock stock MHz). With 100MHz increments possible the cores can easily get within well.. 100MHz of what they will max out at. I however worry about the other fun things like memory and uncore. I'd hate to have to choose between 2400C9, 2133C7, or 1866C6. Where's my 2050C6

[Hornet331]

Theres no uncore anymore in SB, it runs the same freq. as the core clock, due to the ringbus.

[Movieman]

Theres no uncore anymore in SB, it runs the same freq. as the core clock, due to the ringbus.

And thats the whole issue in a nutshell.
It's the lack of being able to tweak that has people upset and I understand those feelings.
It is that tweaking that shows these people's talents and ultimately shows a platforms max capability.
Ialso heard that some of the motherboard manufacturer's are trying to find a way around this WITH Intels help.
Let's see how this plays out before we tar and feather them.

[Hornet331]

People have a short memory and are exaggerating. There was no uncore part in any intel cpu till nehalem, even though there where cpus that had L3 cache.

Actually im glad they removed the sparate clock domain for the L3 cache and now it runs at core frequency. You don't have to worry anymore about at which speed you have to run your uncore part so that you can run a certain mem frequency. Good old stuff from pre nehalem returns.

[Movieman]

People have a short memory and are exaggerating. There was no uncore part in any intel cpu till nehalem, even though there where cpus that had L3 cache.

Actually im glad they removed the sparate clock domain for the L3 cache and now it runs at core frequency. You don't have to worry anymore about at which speed you have to run your uncore part so that you can run a certain mem frequency. Good old stuff from pre nehalem returns.

No, your right, you don't have to worry, but then again look at what we can do with those Nehalems that we wouldn't if they were designed like SB..

[Hornet331]

No, your right, you don't have to worry, but then again look at what we can do with those Nehalems that we wouldn't if they were designed like SB..

Which is no problem since most will use S2011 anyway as it replaces S1366, nothing changed there.

[Movieman]

Which is no problem since most will use S2011 anyway as it replaces S1366, nothing changed there.

But,But I don't want them to replace socket 1366.
It's been SOOOOOOOO good to us!

[kl0012]

But,But I don't want them to replace socket 1366.
It's been SOOOOOOOO good to us!

Wow! That is a bit strange argument. S775 was good also... as well as socket 7. And I'm sure S2011 will be good also.

[Movieman]

Wow! That is a bit strange argument. S775 was good also... as well as socket 7. And I'm sure S2011 will be good also.

Comment made only 1/2 seriously.
Yes, agree, moving on is smart.

[Chumbucket843]

Theres no uncore anymore in SB, it runs the same freq. as the core clock, due to the ringbus.

they changed the name of the uncore to the system agent. yes, it no longer has the L3 but it's still the "uncore". it still has important stuff like PCIe, mem controllers, power control unit, DMI, etc.

[Hornet331]

they changed the name of the uncore to the system agent. yes, it no longer has the L3 but it's still the "uncore". it still has important stuff like PCIe, mem controllers, power control unit, DMI, etc.

Yes but its far more seperated then in nehalem, where memcontroller and L3 cache where the main parts of the uncore.

[geo]

And thats the whole issue in a nutshell.
It's the lack of being able to tweak that has people upset and I understand those feelings.
It is that tweaking that shows these people's talents and ultimately shows a platforms max capability.
Ialso heard that some of the motherboard manufacturer's are trying to find a way around this WITH Intels help.
Let's see how this plays out before we tar and feather them.

[Chumbucket843]

Yes but its far more seperated then in nehalem, where memcontroller and L3 cache where the main parts of the uncore.

define "separated" and separated from what? what the uncore does has changed in SB but it's goal is still the same (easier on die integration of system components).

[Hornet331]

define "separated" and separated from what? what the uncore does has changed in SB but it's goal is still the same (easier on die integration of system components).

Seperated as in not anymore bound to the mem controller. In nahelem the L3 is is on a different voltage and frequency plane then the core. In SB the L3 is on the same voltage and freq. plane as the cores.

Now everything down to the L3 counts as core, and everything else as SA (beside the IGP and the en-/decoder logic).

[iboomalot]

Intel so worried about money and people OCing a 300.00 chip over a 1200.00 chip they need to narrow the price gap.

price the CPUs at 250 , 350 ,450 and 980X type chips at 750 not $1,200

at $750 they make good profit and you aren't paying double or tripple the price for a simple multi unlock.

I mean a i7-950 is $279.00 for 3.06ghz and a i7-960 is 579.00 for 3.2ghz and a i7-980X is $949.00 for 3.33ghz (per newegg)

anyone else see the greedy BS going on???

[PatRaceTin]

Anybody know about Z68 Chipset?