3 ms·
Here comes another knockdown punch for Intel. Now they got a solution to add hundreds of megs of cache for cheap. More importantly SRAM can be binned/tested/K
by baybal2 5y ago
Here comes another knockdown punch for Intel.
Now they got a solution to add hundreds of megs of cache for cheap.
More importantly SRAM can be binned/tested/KGDed separately!
And it can be fabbed on its own customised process!
And also, they can throw in an eDRAM there instead on a moment notice.
And as they already have silicon interposer here, adding HBM2/3 will be also a triffle.
- fomine3 5y agoSRAM / for cheap?
- baybal2 5y agoYes, it would certainly be slower than an on-die cache, but very cheap. They can seriously reduce the L3 cache area, which is like 50% of the die currently, or get rid of it alltogether. Imagine, 2 times more dies per wafer at near no cost except for extra packaging.
- fomine3 5y agoSeparating SRAM cache would improve chip yield so possibly decrease some cost, but I don't believe it make SRAM "very cheap" or "at near no cost".
- baybal2 5y agoSee how the defect rate work: If a particle kills a repairable part of the die (which can be downgraded, or cut-off) you still get from a quarter, to a half of die area dead, and a second grade chip. But if you can make 2x dies from wafer even with the same defect rate (it will usually be less.) Even if the die will completely die from a local defect, you still get much more perfect dies in total, which is more important for your bottom line. Standalone SRAM can be very repairable, and high yield with a custom process. Adding few spare columns, or SRAM banks should be covering for much more defects than if binning was done per entire CPU die.
- kingosticks 5y agoThis. A great way to reduce the cost of your chip is to add SRAM to it. I'm not joking.
- xondono 5y agoI don’t know if it’s “cheap” but It’s probably cheaper than the monolithic alternative right?
- b9a2cab5 5y agoFor mobile (the cash cow) the extra power of off-die communication kills it before it's even out of design.
- tester34 5y agocash cow or not, the world is calculating on servers
- baybal2 5y agoIt's not off-die, it's pretty much on top of it, and I believe they are using very wide, parallel I/O
- throwaway2048 5y agoselling mobile phones is a cash cow, but selling mobile chips really isn't, unless you are Qualcomm and have monopoly rights on the concept of a cellular radio.
- pdimitar 5y agoEven selling mobile phones isn't the cash cow it was some short 3 years ago. The demand is plateauing and people hold on to their phones for longer time which makes perfect sense in a world where almost every mobile company is constantly inflating the prices. Plus there's an upper limit to the computing power that people need from their phones as well, so the sleek ads for phones simply don't work on many people nowadays. They are content with what they have.
- kllrnohj 5y agoIntel's been showing off a lot of 3D stacking tech & mixed dies on packages recently, too. This isn't really a kockdown punch. It'll depend on who can ship what and when, not just who can show off demos. They both have shown off demos. See Intel's Foveros https://www.anandtech.com/tag/foveros https://www.anandtech.com/tag/foveros or Ponte Vecchio that they've been talking about for 2 years now ( https://www.anandtech.com/show/16453/intel-teases-ponte-vecchio-xehpc-power-on-posts-photo-of-server-chip https://www.anandtech.com/show/16453/intel-teases-ponte-vecc... )
- O5vYtytb 5y agoSure, but Intel hasn't even been able to ship 10nm or 7nm in meaningful quantities for years. I have my doubts that stacking is yielding well for them...