Throughout this year &rsquos Hot Chips conference in Cupertino, California, AMD has presented what they&rsquove been capable to obtain by using automated style and design tools ( application ) to rearrange the units inside its Bulldozer processors.
In advance of going ahead, you ought to study Part 1 and Element two of our AMD Hot Chips report.
Like we&rsquove just explained, such a layout &ldquooptimization&rdquo is mainly applied for really large chips such as GPUs and readers should retain in mind that the fastest GPUs these days hardly attain above 1. 2 GHz in typical conditions.
This kind of a frequency would be catastrophically minimal for a CPU like AMD&rsquos Bullzoder and regardless of the reduce manufacturing charges and power consumption, not a lot of would be interested in powering a personalized laptop with one thing like this.
The point is that the Bulldozer die doesn&rsquot have as numerous transistors as a Tahiti dies and as this kind of will be capable to attain considerably larger frequencies than the GPU dues to obtaining a smaller and significantly less complicated die with a a lot more modest power consumption.
Now we have a 2 GHz Bulldozer that has a modest and economic die dimension and an improved power consumption degree.
Interestingly adequate, AMD is not aiming for a total overhaul of its CPU die design.
Just like shown at the 2012 Hot Chips conference, the corporation is only redesigning components of the CPU using the automated design tools. The units in question turn out to be considerably smaller and manifest reduced energy consumption.
AMD is displaying a floating point unit (FPU) in the graphs created public at Hot Chips. The unit has been considerably reduced in size by utilizing a diverse style resolution ( software package working with a Large Density cell library).
A 30% reduction in die area is touted along with a 15% to 30% energy consumption and these are outcomes typically obtained by moving manufacturing from 1 node to another.
Receiving a design and style from 32nm to 28nm manufacturing could consider a complete year or even additional.
Including to the expected 20 % ~ 30% die shrink and energy consumption reduction that comes with such a move an extra 30% shrink and power improvement due to a tighter style and design would result in amazing final results.
Such effects would be comparable with a move to 20nm manufacturing.
The unknown is the frequency of this kind of a style, but the company could opt for a differentiated clock layout wherever some units perform at a certain frequency and other units have a a lot greater functional frequency.
Extrapolating from AMD&rsquos graphs, we could think about a CPU that has far more FPUs that are all kept fed by really quickly dispatch units with efficient branch prediction.
It is all about balance and obtaining the ideal recipe, but what AMD is fundamentally saying is that they are working with quite a few answers that present very good improvements and that mixed will provide amazing benefits.
No comments:
Post a Comment