[RUMOR] Industry source: Lynnfield hard to overclock

[Sampsa]

I suggest you to take following with a grain of salt:

A friend of mine who is working in industry has learned that Intel will be preventing Lynnfield overclocking capabilities. First of all Intel will remove 4 pins from LGA1160 socket and it will be called LGA1156.

Second Lynnfield will only support BCLK (Base clock) 100 and 133 MHz, and no external clock signal fun like with X58 chipset. He is afraid Lynnfield won't be good for overclocking...

Lets hope our own DrWho? will give us some early hints and hopefully he will give me a lecture just like IDF earlier this year



(From The Overclocker issue 2)

source: Muropaketti

[Trouffman]

that's sounds bad, we will need to find other way to oc :/ (hardware modification ?) maybe we should wait more infomrtion to confirm or not this rumor ....

[clonez]

i smell modding "fun"

oh well, guess the midrange overclockers will be sitting on their e8XXX and Q8XXX/9XXX processors for quite some time, if this turns out to be true

[Hamilton]

Isn't this what was rumored about Nehalem as well which turned out to be completely false?

[Cooper]

I remember hipro's 1st overclock was achieved by replacing the quartz clock gen on the board with a faster one. Do i sense deja vu ?

[largon]

I doubt strap-on turboPLLs work with Lynnfield...

[Shintai]

We kinda heard this before. Tho with Bloomfield

Also changing socket sounds odd, even if they remove 4 pins. The 4 extra pins would just be "unused at this stage.

And the external baseclock is the same. Lynnsfield is exactly like Bloomsfield. The QPI is just replaced with PCIe ondie and only 2 memory channels. Everything else functions the same way.

So I put this with "bloomfield" cant OC category

[justthefax]

We kinda heard this before. Tho with Bloomfield

Also changing socket sounds odd, even if they remove 4 pins. The 4 extra pins would just be "unused at this stage.

And the external baseclock is the same. Lynnsfield is exactly like Bloomsfield. The QPI is just replaced with PCIe ondie and only 2 memory channels. Everything else functions the same way.

So I put this with "bloomfield" cant OC category

You seem way more optimistic this can't be so than the actual Intel employee here! Do they keep the viral marketing team out of the loop nowadays?

[Blauhung]

You seem way more optimistic this can't be so than the actual Intel employee here! Do they keep the viral marketing team out of the loop nowadays?

I don't know nothing either, I'm just guessing. But I don't want anyone to feel a false sense of entitlement either. If it turns out locked or just really hard to get them to clock higher, it's not cause Intel is trying to screw you out of your hobby and force you to buy Bloomfield. It is because the platform just works better for mainstream users this way.

I'm still of the opinion that People like DrWho? and others who were helping decide the direction of the Nehalem generation of platforms, saw that the desktop platform was turning more system on a chip. And realizing that this type of platform had limited room or mechanisms for pushing the boundaries, they made a conscious decision to pull the server platform in to fill in that role. Keep in mind that Tylersburg is a server chipset that happens to work very well for a HEDT platform.

[Blauhung]

I'm still thinking this could be true for the Ibex Peak based platforms. There's no apparent connect that carries the clock signal to the CPU. With any of the UP/DP/MP platforms that use Tylersburg, the reference clock must be synchronized between all the sockets and that's why you have clock generation done off die and split/distributed from the Tylersburg chipset through the QPI links connecting everything. With Lynnfield, there's really only 1 component that actually needs that clock signal, the CPU. In the effort of simplicity/cost, power savings, and maybe to some tiny extent performance, just connect the PLL directly to the CPU.

It might just be that Lynnfield is limited to multiplier overclocking only.

[Sampsa]

It might just be that Lynnfield is limited to multiplier overclocking only.

And as we all know, Intel has had multipliers locked upwards in all other models than Extreme Editions..

[Blauhung]

And as we all know, Intel has had multipliers locked upwards in all other models than Extreme Editions..

with the exception of turbo mode

and still there may be some way of tweaking the PLL

[[XC] gomeler]

with the exception of turbo mode

and still there may be some way of tweaking the PLL

I wonder if it'll be programmable. Would be nice if it was adjustable... Intel ya listening?

[Nedjo]

Lets hope our own DrWho? will give us some early hints and hopefully he will give me a lecture just like IDF earlier this year



(From The Overclocker issue 2)

source: Muropaketti

Like he pointed out on Aceshardware forum: he's only giving his personal opinions... and as he demonstrated so well in discussion about intel's "cheating" in Cinebench benchmarks, only giving them if there's something positive or hypeositive to say about his company

In other words his silence will speak also

anyhow, nice finding Sampsa! I guess it's only matter of hours before this hits mainstream "news" sites

[Mats]

I wonder if C2 will be the last one where you actually could overclock even the cheapest model*.
It's like Intel wants to make some room for AMD**: they have no true successor to the C2D*** and simply leaves the <$200 OCing CPU**** market*****.


* Sure, it's all rumors, maybe you can overclock the new ones. Are they called something like Core i5 or what?
** No, AMD may not be able to fill that market even if it will exist, but I think a situation like this will be good for them.
*** I'm not saying that C2 is outdated already, but if these rumors are true then it's hard to see a future upgrade unless you pay in €i7-money.
**** Yeah cheaper €i7 may show up.
***** Yeah I know that very few cares about overclocking in general.

[[XC] gomeler]

Meh, don't think Lynnfield will even show up on my radar. So long as the Core i7 9xx series overclock well then i'll be happy. As for Core 2, I have a feel these 6+ GHz E8600s will stick around for a while

[STaRGaZeR]

Meh, don't think Lynnfield will even show up on my radar. So long as the Core i7 9xx series overclock well then i'll be happy. As for Core 2, I have a feel these 6+ GHz E8600s will stick around for a while

Problem is: after Bloomfield is Westmere, Sandy Bridge, etc. I don't know about Westmere, but Intel showed (in IDF maybe, can't remember) some info suggesting SB will use the same CPU+IMC+PCIe on-die Lynnfield uses, and I must say it's the logical next step. If that turns to be true then the overclocking results and methods of Lynnfield can be very important to determine how the overclocking world is going to be in the future.

[saaya]

hmmm interesting... i heard about this "overclocking lock" too.
ill check if i can find out something at IDF next week here in taipei

I have a feel these 6+ GHz E8600s will stick around for a while

hehehe me2!

and this might just mean that there is a way to tweak it

or not

afaik the pll inside the x58 chipset puts out a beautiful and clean signal at 100 and 133mhz, but if you feed it with 140+mhz it wont work anymore.
on x58 theres some sort of a debug/legacy option for the PLL, which, somehow, allows input clock signals of up to 220. thats how all x58 boards overclock the Bclock (FSB). for Ibex Peak this mode wont exist afaik, so unless the pll is improved somehow we will be stuck with 140Mhz clocks more or less. which is still a nice overclock for a cpu with a Bclock of 100Mhz for default, 40%... but i have a feeling those will be celeron like cpus, so if you want to push a high end cpu, you will need 1366...

im sure there might be a way around this somehow, but it wont be easy... and if the rumors about amd clocking their 45nm quads to 4ghz+ on air with an ipc improvement of 10% then they will be very competitive and steal away market share from intel for sure.

this is goin to be amusing if true
Intel will just shoot themselves in the foot and AMD are moving into same tech as them with similar process will just fly past

surely they learnt from past mistakes, dont mess with overclockers

heheh, yeah i sure hope intel will rethink this

but im not sure intel did this on purpose... maybe they just wanted to save transistors and made the pll very simple. it only HAS to support 100-133, for most systems there wont be any need for more clock options.

then again, look at how many mainstream and high end boards get sold... and whats the main selling feature of those boards? overclocking options!
so i hope intel doesnt underestimate the necessity for overclocking... most people wont actually use it, but its a feature everybody wants, and its a feature that makes a board shine in reviews and in the community which then in return means more people see its a good board and buy it. even if they dont overclock at all...

[Sumanji]

Makes sense.... lock OC'ing on cheaper models and force people to spend the big bucks on X58, Bloomfield and make their memory partners rich with DDR3 sales...

[RejZoR]

Makes sense.... lock OC'ing on cheaper models and force people to spend the big bucks on X58, Bloomfield and make their memory partners rich with DDR3 sales...

They'll strick themself so badly it's gonna hurt them.
Profit comes from massive low and mid range sales, not from high end.
So those who were thinking of buying CPU went with Intel just because of overclockability. Otherwise i'd probably go with AMD in the first place.
At least i know i bought Intel just because of that.
I'll rather buy cheaper LGA775 quad and overclock it and use it till AMD releases something decent or Intel gets over such lame attempts to milk the users...

[Mumak]

The socket change 1160->1156 is true.
BCLK is routed thru the PCH.

[Blauhung]

The socket change 1160->1156 is true.
BCLK is routed thru the PCH.

and this might just mean that there is a way to tweak it

[Mumak]

and this might just mean that there is a way to tweak it

Or might just mean the opposite
But you're better in this HW stuff than I am...

[mikeyakame]

The socket change 1160->1156 is true.
BCLK is routed thru the PCH.

should help to reduce mainboard component cost somewhat. routing BCLK through the PCH and restricting pll oscillation to 133 with dividers for 33mhz pci and 100mhz pci-e makes shielding the input signal for qpi base clock ref almost completely unnecessary since the drive amplitude would most likely be supplied either leaving the PCH or on-die, either way Intel is solving a lot of problems for themselves as a result of third party chipset / board suppliers veering way off course with their applications of the specs, considerably benefitting their long term goals if the on-die signalling is that critical to the chips operation.

[Blauhung]

Problem is: after Bloomfield is Westmere, Sandy Bridge, etc. I don't know about Westmere, but Intel showed (in IDF maybe, can't remember) some info suggesting SB will use the same CPU+IMC+PCIe on-die Lynnfield uses, and I must say it's the logical next step. If that turns to be true then the overclocking results and methods of Lynnfield can be very important to determine how the overclocking world is going to be in the future.

There will always be a need for Tylersburg or Tylersburg-like chipsets with DP and MP configurations, as a QPI "hub" is needed to function as a central hub for PCI and clock synchronization. Bloomfield and the X58 platform is an extension of the DP platform, down into the high end desktop market to satisfy those in this community that enjoy overclocking. Don't worry about Westmere and Sandy Bridge, as there will be HEDT platforms for each that will fit this market. As with nehalem, there will also be the same type of systems as Lynnfield/Havendale with more portions integrated to satisfy the mainstream market.

Makes sense.... lock OC'ing on cheaper models and force people to spend the big bucks on X58, Bloomfield and make their memory partners rich with DDR3 sales...

They'll strick themself so badly it's gonna hurt them.
Profit comes from massive low and mid range sales, not from high end.
So those who were thinking of buying CPU went with Intel just because of overclockability. Otherwise i'd probably go with AMD in the first place.
At least i know i bought Intel just because of that.
I'll rather buy cheaper LGA775 quad and overclock it and use it till AMD releases something decent or Intel gets over such lame attempts to milk the users...

Please again remember that this is not Intel "locking out" overclocking. This is Intel making choices that result in large gains for the items that people who do make massive low and mid rage purchases want. Having a NB on a mainstream platform with Lynnfield's architecture is a complete and total waste of power and money. It's another chip that motherboard manufacturers have to place on the board. It uses extra power, generates more heat, increases cost, and increases complexity of the platform. All things that the people who make massive low and mid range purchases avoid.