AMD Linux kernel update, reveals that upcoming CDNA2 GPU will feature two dies

Marees

[H]ard|Gawd
Joined
Sep 28, 2018
Messages
2,039
In the most recent Linux kernel update, AMD engineers have confirmed that the upcoming CDNA2 GPU codenamed Aldebaran will feature two dies.

We have seen rumors that Aldebaran features two compute dies on the package, which were referenced in the Linux updates as die0 and die1, leaving a possibility that there might be more. Today AMD engineering confirmed that only the primary die will handle the power data, which should not be set through ‘secondary die’. This ultimately confirms there are two dies.

the primary die will regulate the power consumption and power limits for the whole compute part of the package. It is not clear if the power of the HBM2 memory will also be regulated by this die, or by a new I/O module.

AMD has never publicly confirmed that CDNA2 will be MCM design, however.

https://videocardz.com/newz/amd-con...-instinct-mi200-has-primary-and-secondary-die
 
Just because they build in driver support doesn't mean the product will come to light, at least in a retail manner. If they are producing a custom product for industrial use or something, it is good to have kernel support. It is even worth adding support just for internal testing of products that never end up being announced.
 
Just because they build in driver support doesn't mean the product will come to light, at least in a retail manner. If they are producing a custom product for industrial use or something, it is good to have kernel support. It is even worth adding support just for internal testing of products that never end up being announced.
wouldn't that lead to a lot of irrelevant code bloat, if they committed back to kernel mainline for internal products that are never released?
 
Just because they build in driver support doesn't mean the product will come to light, at least in a retail manner. If they are producing a custom product for industrial use or something, it is good to have kernel support. It is even worth adding support just for internal testing of products that never end up being announced.
How many times have we found something in the Linux kernel that turns out to never happen?
 
This a good step towards an MCM GPU, this could operate similarly to the register on their CPU in how it dispatches jobs to cores. But I doubt AMD has their interconnects at a stage yet where they could handle multiple CCX’s on a GPU with out introducing unacceptable levels of frame time delays. I would be thrilled to be wrong on that though.
 
Back
Top