How Intel Will Lock Overclocking on LGA1160 (Mainstream Nehalem) - Fud
Quote:
No easy work around
As we told you some time ago, Intel has put in what can only be called an overclocking lock in the upcoming LGA1160 processors which are currently going under the codenames of Lynnfield and Havendale.
Until now we didn't know how this had been implemented, but we've learned some more about it and it looks like there is no easy work around. As these processors have the memory controller and a few more bits integrated into the CPU itself, Intel suddenly has a lot more control than it has had with past designs that utilized a separate chipset which contains the memory controller.
As Intel seems to want to push the much more expensive Bloomfield platform to overclockers,
the company implemented a lock that prevents these new processors from being overclocked by adding two PLL clock generators, one inside the CPU itself and one in the PCH.
This might not sound like it's a big problem in itself, but what Intel has done is that these two will clock generators will reference eachother and this means that just changing the bus speed won't have any effect if you're trying to overclock the CPU as it will dissregard the information from the PCH if it's not a correct value.
There might still be options for overclocking these CPU's, but our understanding is that this isn't easily done and Intel doesn't want it to be and as such Intel is unlikely to give out this information to third party motherboard manufacturers.
The sum of all this is that Intel might loose out a few enthusiast users that don't have deep enough pockets to get a Bloomfield system to AMD, which might not be a bad thing for AMD in the end.
Source:
http://www.fudzilla.com/index.php?op...7255&Itemid=35
Old thread:
http://www.xtremesystems.org/forums/...d.php?t=185374
Does it really makes sense? and would they seriously do that? I really hope it ain't so.
dynamic single core frequency change?
Quote:
Originally Posted by
Blauhung
Not really, Bloomfield is single socket so its more of the replacement for the X38/X48 platform. I have no clue if there's a dual socket enthusiast board in the works, but with the loss of the FBDIMMM requirement and the new 3 channel memory controler, Personally I'm going to be holding out for news of that.
HEHE, 3 dedictated memory channels for each processor with enthusiast grade unbuffered DDR3. And each CPU running 8 threads
we still don't know if overclocked Yorkfield's will come out on top of stock Lynnfield yet. Remember the boost to memory bandwidth/latency will be large, and with the revamped implementation of SMT this will be an effective 8 threaded chip as opposed to 4. Turbo mode, dynamic single core frequency changes for single threaded apps, will also most likely come in to play just like it has been implemented on the new penryn based mobile platform.
I have an inspiron 1525 with a T7250 in it. I loaded up prime 95 on it the other day on just one core, then locked the affinity to core 0. I literally saw a 1/2 second jump up to an 11x multi, but could never get it to repeat. Why, do you ask? Because the stupid computer kept tricking itself into thinking that both cores were being loaded. Idontcare and I had a fairly long discussion in the AT forums about this, with the tentative conclusion being that the computer doesn't know HOW to use just one core most of the time. I understand that nehalem will be able to jump up 3x instead of 1x, but if they don't figure out a way to improve the feature then it will continue to be nearly useless.