|
eames posted:Cannonlake to be a stacked/3D modular architecture? On-die coprocessors for native ARM code? The article pretty much explains why it isn't likely to be real in the first couple of paragraphs. Pretty much the only reason the writer thinks it could be real is that they used an acronym of an upcoming microarchitecture in one of the drawings. It isn't really feasible with how chips are currently produced. Anyone buying them would have to get a lot of them, and you wouldn't be designing them off a website. One of the purposes of patent drawings is to express some of the patent's embodiments (implementations). So saying a customisable chip could have the latest intel cores, FPGA... is for illustrative purposes. They need to say that it can everything, be customised however, and procured whenever to cover their bases. So going with a codename (which we've known for a year prior) for a something that isn't even out makes the most sense in that context. This is like seeing that PS4 patent where you say 'Mcdonalds' to skip an advert, and extrapolating that Mcdonalds is so specific that there must be some deal in place for it to happen. So therefore it will happen.
|
#
¿
Mar 3, 2017 04:26
|
|
|
|
| # ¿ Apr 26, 2024 18:17 |
|
|
silence_kit posted:Why isn't it feasible? I certainly don't think that some guy in this thread is going to be able to afford a custom multi-chip-module assembly for his gaming rig, but why would it be unreasonable for Intel to offer this kind of service for its larger enterprise customers? fishmech posted:It's not that it'll be infeasible forever. It's that it's infeasible for a thing Intel would start offering tomorrow or even next month. Yeah, I was a bit harsh with my words. It's not so much that it's infeasible (more like impractical), but it is not something that would be the basis for all, or most (if any) Cannonlake processors. They will have to add an assembly component to the production chain for a start. But there will also be other cost factors such as producing more photomasks for each component and the uncore, and having a larger volume of parts to test. There are also some technical considerations with regards to memory between different architectures, and thermal issues (memory gets hot if you have a GPU on the package). Not unsolvable problems, but complications. It makes a lot of sense if you have plenty of fab capacity on older processes, and very little of and expensive new process. I can't speak to whether they have overcome the cost and technical hurdles, but it's not the sort of thing they would hide up their sleeve until launch. It is the sort of thing they will need to transition to, and there's no reason to hide it from AMD. It benefits them to demonstrate the technology and shop it around to garner interest. It will be a matter of execution, rather than getting an advantage by using secret technology. There are two use-cases I can think of: HPC and EUV. I'm sure people would love to put a CPU and non-intel GPU on the same die, and cram more chips into their supercomputer. EUV is so behind and below projections that they would want to get as high of a yield as possible. If I had to speculate, Fab 42 would be a good opportunity to sort it out, but that is 3-4 years away. Intel hasn't disclosed the technology yet, though either option will likely have issues. It isn't as secretive as the article makes out. Many of the problems being faced today involved things discussed in my university physics classes a decade ago.
|
|
#
¿
Mar 3, 2017 08:20
|
|
|
FX-9900k Vishera Lake
|
|
#
¿
Oct 20, 2018 03:29
|
|
|
nice clock
|
|
#
¿
Oct 24, 2018 12:35
|
|
|
eames posted:Coffee Lake with the -F suffix (no iGPU) are inexplicably priced the same as the regular chips, or as one commenter put it: It is handy the one time when your GPU fails unexpectedly. That's got to be worth a few dollars at least.
|
|
#
¿
Jan 17, 2019 03:06
|
|