AMD 65nm doesnt seem to be too hot

[nyker96]

I was reading a bit on Brisbane from various sites and it seems to me the process besides higher energy efficiency cannot allow any higher clocks than the 90nm. Since it is just a die shrink this points to the 65nm process been unable to allow any higher clocks. I wonder if the K8L using the exact process can do any better (perhaps a bit better tuned in Q3). It just does not look too mature comparing to Intel's 65nm. I wonder if this will handicap K8L to a high degree. What do you guys think?

 

[Yellowbeard]

I have an FX-62 and a new 5400+ on my bench right now. With sane voltages (1.525v or less on water) they both crap out at almost the same point, 3.2ghz or less. It seems the same with others I see posting also.

 

[aka1nas]

AMD's first few batches at a new node usually don't clock great. They tweak the "recipe" between production runs, so it will probably get a bit better quickly.

 

[DrMrLordX]

What you want to look at will be Rev G2 Brisbane CPUs. Those are the ones that could potentially hit clock speeds higher than 3.0-3.2 with sane voltages. They may not be out until March, unfortunately.

 

[nyker96]

Just read from another thread claiming IBM has a good 65nm process to sell to AMD that allows 5GHz on Brisbane core. Not sure if it is true.

 

[mountcarlmore]

sure, and i have a bridge to sell to whoever wants one...

 

[DrMrLordX]

If IBM has such a process, expect AMD to reserve that for K8L.

 

[mountcarlmore]

amd already has the process, although it is nowhere near allowing 5ghz clocks, thats just absurd. it already uses it on 90nm, thats how fx74 and the upcoming 3.2ghz chip can exist on 90nm, its called sige, or silicon germanium, adds 40% transistor performance, which ofcourse does not translate to realworld 40% clockgain, but it does improve things quite a bit. it will come to 65nm eventually, probably same time as star comes, amd just doesnt like to throw too many things into the mix on a new process, thats why 90nm saw sige first.

 

[Conky]

I'm afraid that by the time AMD figures out the new process and lets users overclock accordingly that Intel will already jump to the next smaller process and beat them silly again. I would really like to see AMD remain competitive.

They became rather complacent when they had the lead and had to know Intel was going to want their market back. I guess they expected Prescott2 and got C2D instead... in the words of Homer, Doh!

Anyway I really hope AMD fires back soon. These serious imbalances are not good for the consumer.

 

[dmens]

the reason ibm can clock power6 that high is due to circuit design methods techniques and uarch. they both use the same process.

 

[IntelUser2000]

the reason ibm can clock power6 that high is due to circuit design methods techniques and uarch. they both use the same process.

Very true. Even then there are rumors Power 6 does sacrifice per clock performance to achieve the extraordinary high clocks(which is no problem for a CPU that clocks almost 2.5x the predecessor, unlike initial Pentium 4). Then there are worries about power consumption. IBM reports under 100W with power sensitive apps, which in translation means pretty close to 100W at near idle.

Do you want a CPU that runs close to what it runs at load at idle, just the CPU??