Recently I have been searching around the WEB in the topic of CPUs. I came across one of the specs which is lithography. I reaearched it and receiveda load of scientific definitions were which a bit too complicated for me to come to terms with. Could one of you help me understand how the lithography length affects the performance of the CPU or computer in general. For example how does the difference in the sandy bridge 32nm and the ivy bridge 22nm change performance?
Currently I am using an i7 Quad Core 3.6ghz 3820 (32nm).
Simply put the change in lithography isnt going to add much to the performance of the processor. The reason why is that shrink down the die size it gets easier to make additional processors but it is other changes that end up making the performance better.
February 12, 2013 3:14:31 PM
Thank you! It cleared things up for me. Baically, smaller the size, the more room to expand in other areas which play a big part in the performance.
Well that's a bit simplistic really. The main bar for performance is heat. This is why people who set overclocking records use LN2, because it keeps the chips below the point where thermals start to cause issues.
Using a smaller node or lithography means a couple of things right out of the gate.
First you can get more transistors into what is usually a smaller die space. Secondly these transistors can run faster before they create the same heat as larger ones. So a die shrink absolutely will result in the ability for increased performance.
That's not to say you will get it, performance depends on the company in question what they do with the space on the die, how fast they clock the chip how many more transistors are there on the chip.
As IntelEnthusiast says there are architectural changes, tweaks and improvements that will account for more of the larger gains we see, however to say that the die shrink isnt going to add much to the increase is just wrong.
February 12, 2013 6:03:35 PM
Thank you! So then the Ivy Bridge will out perform a Sandy bridge with identical specs except for the node size (already knew this, just like confirm any information I plan to use later on). 22nm is extremely small, a bit curious how they go about operating at such a size.. But that is something for another topic!