86 Processors compared over at Tomshardware

[eelw]

People with Nehalems are still getting by just fine, lol.

Yup, happy with my Gulftown still.

 

[AtenRa]

Hmm? Says in the image they are using Prime95...but still strange to see 3770K use more than 2700K. I remember from the initial reviews all Ivy's using less than their Sandy counterparts.

Hhhmm, i was reading IDCs topic for the FX8350 at the time.

 

[OVerLoRDI]

They reviewed 86 CPUs. They have to draw the line somewhere...

Yes but it would have been more interesting to take the top of the like cpu from each generation all the way back to athlon/pentium 3 days or something and it test ~15 cpus

 

[Pottuvoi]

Yes but it would have been more interesting to take the top of the like cpu from each generation all the way back to athlon/pentium 3 days or something and it test ~15 cpus

Especially if each would be forced to same Mhz..
Single thread performance increase between generations would be interesting to see.

 

[Idontcare]

28nm will not be any better for AMD most likely. They will be losing SOI and only making a half-node drop. Likely a draw for them...

Yep, just like it was for their 45nm->32nm shrink, the xtor tradeoffs IBM/GloFo made resulted in not much improvement in performance or power but did result in smaller dies...the shrink from 32nm to 28nm will result in smaller chips that are less costly to manufacture but the underlying xtor parametrics are such that they likely won't realize a clockspeed or power improvement.

For that they need gate-last HKMG, not the gate-first integration path they are tied to. TSMC knew this and chose well. See many 28nm chips from GloFo lately?

 

[AtenRa]

According to AMD, they can have up to 30% reduction both in area and power consumption in the same node by using High-Density cell libraries. What we dont know is how much they will loose by going bulk.

 

[peonyu]

Amazing when you look at some of those charts in that it shows just how much of a step backwards AMD took with the bulldozer microarchitecture.

If only they had kept with the optimization and shrinks of the thuban lineage

As an FX-8350 owner, it is not surprising all to see piledriver dominate the following chart though.

Who would have guessed that making the 32nm equivalent of Prescott would be history repeating itself? Somehow Hector didn't see that coming, wonders never cease...

Well, atleast they are dominating somewhere [besides high end video cards atm].

 

[ShintaiDK]

According to AMD, they can have up to 30% reduction both in area and power consumption in the same node by using High-Density cell libraries. What we dont know is how much they will loose by going bulk.

Shouldnt they start using it now? Or did it actually instead increase hotspots so it couldnt be used in performance production?

 

[inf64]

High density libraries are(or were) targeted for post SR core. So not even Kaveri and Opteron version with SR core will see this change.