[News] AMD's Ryzen Chips 10% Smaller Than Comparable Intel Skylake Dies

[StyM]

https://www.techpowerup.com/230446/a...l-skylake-dies

At the International Solid State Circuits Conference (ISSCC), AMD presented a whitepaper in which they demonstrated how its upcoming Zen x86 core fits into a 10 percent smaller die area than Intel's currently shipping second-generation 14nm processor. According to reports, analysts and Intel engineers in the session said the Zen core is clearly competitive, though many as-of-yet unknown variables will determine whether the die advantage translates into lower costs for AMD. That said, one thing is clear: the chip will have to perform in addition to being smaller, if AMD wants to ever capitalize on the potentially higher margins a smaller die could grant them.

One of the ways AMD improved upon its ZEN core in comparison to its previous products has been on switching capacitance for their new chips, with a reported overall 15% improvement. In addition, AMD has apparently moved on to a metal-insulator-metal capacitor design, thus achieving lower operating voltages as well as as more fine-grained per-core voltage and frequency control (on to become part of their SenseMI technology suite). Looking at the image, which pits an AMD ZEN chip to a comparable Intel solution, we see that AMD saves additional die space by making do with only 12 metal layers as well as overall lower L2 and L3 cache footprints.

[Lightman]

This table shows how much better Intel process really is in the real world.
Despite SRAM cell size deficit somehow AMD engineers managed to create smaller overlay cache sizes. Of course cache includes a lot of supporting circuity and one could say it is there where AMD made it's savings.

I'm looking forward to the launch and hope for competitive product at a reasonable price which will offer performance good enough for excuse to upgrade from i7 4790K

[vario]

Im not sure there is something to be really proud of.
Yes they packed it denser, while using one less metal layer.That could be the reason why it clocks lower...
AMD on their first attempt tends to go too optimistic, and later on they have to fix it.

[zanzabar]

This table shows how much better Intel process really is in the real world.
Despite SRAM cell size deficit somehow AMD engineers managed to create smaller overlay cache sizes. Of course cache includes a lot of supporting circuity and one could say it is there where AMD made it's savings.

I'm looking forward to the launch and hope for competitive product at a reasonable price which will offer performance good enough for excuse to upgrade from i7 4790K

the samsung 14nm that global foundries uses does close transistor only like a gpu. intel gives more space between logic circuits and has much smaller node sizes so they can get the cache really close. if you measured node+gate like intel does amd would be around 18-20nm according to the data we got from samsung when they said they were comparable node to gate ratios with TSMC.