Considering how much better Navi seems to be, it'd be strange for AMD to stick to Vega after moving the APUs to the 7nm node.

zen2 is better than zen1 and navi is a lot better than vega, if they don't use it, do they have a lot of defective chips that can be reused?

It makes sense if 7nm allocation is limited. Making as many chiplets as possible is probably more profitable than making small navi chips for APU.

Kaarme:

Considering how much better Navi seems to be, it'd be strange for AMD to stick to Vega after moving the APUs to the 7nm node.

There is maybe an reason: the HBM APU, if navi based, would have to share ram with the computer, and so pick up ram from the system (1st bad point), ram that is considerabely slower than regular HBM (2nd bad point). On other hand Vega isn't so good as dedicated GPU but is really nice as APU graphic, it fit all the need of an APU user, and have no concurence right now.

rl66:

There is maybe an reason: the HBM APU, if navi based, would have to share ram with the computer, and so pick up ram from the system (1st bad point), ram that is considerabely slower than regular HBM (2nd bad point). On other hand Vega isn't so good as dedicated GPU but is really nice as APU graphic, it fit all the need of an APU user, and have no concurence right now.

Don't iGPUs always use the system memory as video memory? An APU with Vega graphics would need to have such a memory controller, not an HBM controller. That part would be no different between Vega and Navi.

rl66:

There is maybe an reason: the HBM APU, if navi based, would have to share ram with the computer, and so pick up ram from the system (1st bad point), ram that is considerably slower than regular HBM (2nd bad point). On other hand Vega isn't so good as dedicated GPU but is really nice as APU graphic, it fit all the need of an APU user, and have no concurence right now.

While Navi is better it still suffer the same problems as VEGA, lack of bandwidth i am hoping memory speeds go to 3200 from 2400MT/s, better yet DDR4X 3733MT/s like Intel uses. Vega has more filtrate than Intels, but the latest G11 uses of DD4X and gives them 50% more bandwidth which is very important. I doubt HBM is on them because HBM is 1# expensive and 2# far larger than a chiplett, not sure if it is possible or will impact the pin layout but if the VEGA GPU is the same size as the chiplett then it would both of them on the right, the memory IO moved up and then the last chiplett at the bottom with 1x HBM joined to the VGA chiplett. Still don't know enough to make conclusions.

I'm curious for what will be the Picasso 15D8_REV_93 AMD Radeon(TM) Vega 1 Graphics 0x93 used. Can it play 1080p60fps ?

Dazz:

While Navi is better it still suffer the same problems as VEGA, lack of bandwidth i am hoping memory speeds go to 3200 from 2400MT/s, better yet DDR4X 3733MT/s like Intel uses. Vega has more filtrate than Intels, but the latest G11 uses of DD4X and gives them 50% more bandwidth which is very important. I doubt HBM is on them because HBM is 1# expensive and 2# far larger than a chiplett, not sure if it is possible or will impact the pin layout but if the VEGA GPU is the same size as the chiplett then it would both of them on the right, the memory IO moved up and then the last chiplett at the bottom with 1x HBM joined to the VGA chiplett. Still don't know enough to make conclusions.

There is no DDR4X for Desktop. Also, HBM3 will be much better and cheapter than both HBM1/2, so why not use that ?

Hilbert Hagedoorn:

Over at Reddit a user that goes by the named rogame posted a listing of what seems to be AMD's future AMD APU lineup, spotted in AMD Bootcamp Drivers. Albeit you'll not see anything frequency relat... Listing of AMD Renoir APUs Leaks - Possibly ZEN2 Based

What you're calling "Raven2" is actually Banded Kestrel, "officially" used in couple Ryzen Embedded SKUs but in reality it's used even in Ryzen 3000 -series as Ryzen 3 3200U and probably every other 2C/4T+3CU APU

Kaarme:

Don't iGPUs always use the system memory as video memory? An APU with Vega graphics would need to have such a memory controller, not an HBM controller. That part would be no different between Vega and Navi.

Oups my bad, it is the mobile version that have HBM, not the desktop (despite originaly it was intended to have it).

D1stRU3T0R:

There is no DDR4X for Desktop. Also, HBM3 will be much better and cheapter than both HBM1/2, so why not use that ?

Well considering there is no Intel Ice lake parts for Desktop at present, that does not mean it won't officially support DDR4 3733MT/s