Haswell vs. Ivy Bridge Overclocking

[ShintaiDK]

Mmm....sammich!

[Puppies04]

Originally Posted by Idontcare

So how can we, the six of us who care,

Make that 7

[Haserath]

Originally Posted by Ferzerp

It won't be a very effective promotion if it gets eaten.

They're giving away the left over fruit cake.

[Computer Bottleneck]

Originally Posted by bunnyfubbles

2. IIRC, it has been mentioned that Intel is going back to solder

Do you have a link?

P.S. In the January 6th Anandtech Podcast episode 13 @ 43:58 --> http://www.anandtech.com/show/6544/t...ast-episode-13 Anand does say he knows what material is under the haswell heat spreader, but can't say.

[Computer Bottleneck]

Originally Posted by Idontcare

Given that the development resources for Haswell were devoted to prioritizing performance scaling at the low-end of power and clockspeed, I very much doubt we are going to see much added clockspeed headroom over that of what we already see with IB.

Yep, that is definitely something to consider. (especially in light of Haswell ULT having a lower TDP despite voltage regs on package and a larger iGPU)

[isamu99]

I would really like to get a new processor and I'm hoping Haswell will easily overclock to 5.0Ghz on any of the Corsair watercooling systems. Do you think it'll be likely?

[bigsnyder]

Don't care how well it overclocks as long as it is a nice jump in CPU performance vs everything else, like how core2 crushed everything when it came out. If it is only 10-15% might as well get an Ivy setup.

[Idontcare]

Welcome to the forums isamu99

Originally Posted by isamu99

I would really like to get a new processor and I'm hoping Haswell will easily overclock to 5.0Ghz on any of the Corsair watercooling systems. Do you think it'll be likely?

Easily? Not likely. Even with a delidded Ivy Bridge you are looking at pushing cpu-murdering voltages at the chip to get to 5GHz with an H100.

Haswell will use differently optimized finfets, but if you look at where Intel is wanting to optimize Haswell (i.e. low power applications) we can safely assume these new finfets are more geared towards lowering power usage at low clockspeeds more than raising clockspeed.

[LagunaX]

Haswell will have the same 3D structure so again the heat generated in such a small space with perhaps a sloppy thermal paste again like Ivy makes me doubt great overclocks. But if it is 10-15% more efficient, then achieving the same overclock would not be so bad...

[2is]

I'm sure Intel has learned a thing or two about the 3D transistors since IB

[Idontcare]

Originally Posted by 2is

I'm sure Intel has learned a thing or two about the 3D transistors since IB

Of course they have, but not in a way that impacts 22nm. The learning that has gone on since IB debuted has all been towards 14nm and 10nm. This is all on a 4 yr timeline, learn something new last year and it won't make an impact for 3yrs yet.

[Computer Bottleneck]

Originally Posted by bigsnyder

Don't care how well it overclocks as long as it is a nice jump in CPU performance vs everything else, like how core2 crushed everything when it came out.

http://www.anandtech.com/show/6355/i...architecture/6

Conroe was a very wide machine. It brought us the first 4-wide front end of any x86 micro-architecture, meaning it could fetch and decode up to 4 instructions in parallel. We've seen improvements to the front end since Conroe, but the overall machine width hasn't changed - even with Haswell.

fast and narrow vs. slower and wide?

In simple terms I would think at some point widening/beefing up/enhancing Core would actually be more efficient that trying to further increase clockspeed?



But at what point will Intel feel the need to do this?

The power wall doesn't appear to be holding them back. (re: Mainstream quad core + iGPU are well below 130 watts.) Maybe storage speeds (or something else) needs to catch up first?

Or maybe Intel never makes another huge jump in IPC? And just keeps the generational gains very modest even at 5nm?

[isamu99]

Originally Posted by Idontcare

Welcome to the forums isamu99


Easily? Not likely. Even with a delidded Ivy Bridge you are looking at pushing cpu-murdering voltages at the chip to get to 5GHz with an H100.

Haswell will use differently optimized finfets, but if you look at where Intel is wanting to optimize Haswell (i.e. low power applications) we can safely assume these new finfets are more geared towards lowering power usage at low clockspeeds more than raising clockspeed.

I see. Thanks for the info and the warm welcome Idontcare. I'll keep that in mind