How much can HKMG help AMD?

[zedx]

Sorry if this question is a bit late but really according to IBM http://www-03.ibm.com/press/us/en/pressrelease/23901.wss apparently using HKMG can give a 40% boost. Does that really mean you could get a 4ghz + Phenom II at 45nm? that might even beat Nehalem at the same Power Consumption. But doesn't Intel already use HKMG? This HKMG thing could mean a 4 ghz Phenom III(?) at 32nm consuming just 65W ! Just the HKMG and 32nm could mean 80% improvement? That would mean an 8core Bulldozer might be even close to 3x faster than K10.5 at the same TDP??!! Seemed outrageous to me so thought I'd ask here(I'm a bit newb and not that much techie so....)

 

[jimmysmitty]

It a vs 45nm though so its hard to say what it will do. Intels 32nm with its second gen HKMG shows abilities to get to higher clocks than Intels 45nm on much less voltage and a lower TDP.

TBH, this is from IBM themselves so it has to be taken with a grain of salt and as well is over a year old. AMDs 32nm is not expected till maybe end of 2010 or if something happens, 2011.

That means Intel will have about a year to push out more 32nm chips and work out more of its HKMG. Then in 2011 Intel will push out its 22nm.

Basically by the time this comes, Intel will have something faster than Nehalem.

 

[zedx]

But AMD's potential clocks seem tremendous compared to what Intel has achieved. Even if the improvement is 25-30% at 32nm, it's quite high. Even that much can give a 80-90% improvement in performance at same power vs 45nm no HKMG. Intel's 32 vs 45nm is not that radical though. If all goes well, Bulldozer might actually be able to level with Sandy Bridge. No choice but to wait for 2011(or leaked benches) though. Wonder how much this will help GPU's though.

 

[jennyh]

We just don't know tbh. 32nm HKMG could end up being a massive benefit for AMD. Maybe HKMG was the real reason why intel could take the lead back and they are about to lose it again big time.

We can hope, but we just don't know yet. What we do know is that there is a big improvement due from AMD's process and not quite so big an improvement from the intel one.

 

[someguy7]

One big problem with your theory. When the core based cpus came out they did not have HKMG. It came out with the 45nm core2s. Long after Intel took the lead.

 

[jennyh]

Ok maybe HKMG is the reason why intel could maintain the lead at 45nm against what is basically a very fast cpu in the Phenom II.

When AMD have their 32nm released in a years time, we'll see both on a fairly even footing for the first time in a very long time.

 

[someguy7]

Not really. You could say the 65nm c2q vs the original Phenom was on a fair footing. Both where 65nm and SOI. It is not the HKMG or the process node that make the Intel faster. Intel has just had a higher IPC and has been faster clock for clock since the introduction of the core based cpus.