Intel details Haswell overclocking

Read this earlier. Pretty interesting.

I'd like to see how GT3e is going to affect temps.
 
So they are bringing in Strap clocks to the mainstream platform. So a locked CPU will only be able to strap to 25% or 67% above stock: no inbetweens... It's better than nothing, but it ain't good.

Edit: I'm going to get a Haswel Celeron on strap the shit out of it.
 
On-die voltage tregulation doesn't sound all that wonderful to me.

The VRs usually get pretty toasty in the current designs, with beefier ones usually allowing for slightly higher overclocks.

With the voltage regulation on-die, I am wondering exactly how much extra heat is going to be produced by the CPU itself.
 
all that math is making my head spin....but I am interested in the 8ghz max part :)

probably only possible with ln2 of course lol
 
Only part of the VRM is on the die. The mosfets (hottest part of the VRM system) is still on the motherboard. Besides, Haswell has a 84 watt TDP, only 7 watts up from the IB 77 watt TDP.

Also, I wonder if Intel will limit the clock straps to certain models, like they did with 2011 processors (only the 3820, 3930k and 3960x got straps, none of the Xeon models got straps).
 
aphexcoil said:
Woah woah woah ... where did you see 8ghz??
Click to expand...

It's like the sixth or seventh paragraph in the linked article. The 8GHz is just the max overclock without overclocking BCLK at all if the information Intel provided is correct and unlocked multipliers go up to 80. He's not saying he's actually witnessed a part running there.
 
cyclone3d said:
On-die voltage tregulation doesn't sound all that wonderful to me.

The VRs usually get pretty toasty in the current designs, with beefier ones usually allowing for slightly higher overclocks.

With the voltage regulation on-die, I am wondering exactly how much extra heat is going to be produced by the CPU itself.
Click to expand...

But, in a different way that's also shifting part of a more difficult to cool part to a more easy to cool part.

But there is the other side of putting a load of stuff into one package. Motherboards usually fail before CPUs, so having more motherboard in the CPU might be a bad thing, plus not give the options some of the high end boards give.
 
Man that article makes me feel guilty for wanting to dump AMD for Haswell. LOL
 
I don't think anyone but LN2 guys will ever even come close to 80x multiplier. They'll probably use multi + BLCK strap combo.
 
DeathPrincess said:
80 seems kind of random... Unless they were releasing a 8.1ghz extreme edition and didn't want it competing..:D why not 100 or something more practical like 128/64?
Click to expand...

So that when they release Broadwell, they can bump that artificial cap to '90' and advertise the shit out of it. Then when they release Skylake, they bump it to '100' and advertise the shit out of it. Then when they release Skymont, they bump it to '110' and advertise the shit out of it. Then when they......


You get it.
 
KazeoHin said:
So that when they release Broadwell, they can bump that artificial cap to '90' and advertise the shit out of it. Then when they release Skylake, they bump it to '100' and advertise the shit out of it. Then when they release Skymont, they bump it to '110' and advertise the shit out of it. Then when they......


You get it.
Click to expand...

I do not think this could work after the first iteration. Just like AMD can not do an additional LN2 overclocking stunt.
 
KazeoHin said:
So that when they release Broadwell, they can bump that artificial cap to '90' and advertise the shit out of it. Then when they release Skylake, they bump it to '100' and advertise the shit out of it. Then when they release Skymont, they bump it to '110' and advertise the shit out of it. Then when they......


You get it.
Click to expand...

You may be right. They'll need something to make us want to buy a new CPU from them and since they're going to BGA with Broadwell and they're only doing ~5% IPC improvements it seems, there isn't as big a rush or "gotta have it" sentiment with each release. They gotta have something else to market to us as a big enough improvement over the last gen to make us want to upgrade.
 
DejaWiz said:
Man, Haswell OC'ing could be pretty awesome. I just hope they have removed their collective heads from theirs butts and will implement fluxless solder under the IHS with all those VR parts on the die.




Fear not! Intel has debunked the BGA only rumor.
Click to expand...

when are we going to know about the solder part for Haswell?
 
Last edited:
This looks like it has potential, but I'd really like for them to put out data on Haswell-E or (bah, sigh) IvyBridge-E etc.. . if Haswell overclocks well, its going to be even more of a kick in the teeth for Intel to push out new "enthusiast" chips that are IB-E, built on old tech etc.. It would be nice to be able to appraise the price, performance, features, and release date differences between Haswell and Haswell-E (and/or IB-E) to make an informed choice.
 
Xaeos said:
if Haswell overclocks well, its going to be even more of a kick in the teeth for Intel to push out new "enthusiast" chips that are IB-E, built on old tech etc...
Click to expand...

I don't think HWL will encroach too much on the enthusiast platform. The purpose of LGA2011 is to provide those with deep enough pockets the capability to overclock an unlocked hexacore (otherwise the Xeon home user segment would be larger).

As long as LGA1150 is limited to quad core, there will always be hexacore buyers and users, regardless of the platform's age. As an example, there is still decent demand for the last generation of hexacore processors and people are willing to pay a healthy sum to get them.
 
You must log in or register to reply here.
Back
Top