AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion
https://community.amd.com/community...ng-amd-radeon-software-adrenalin-2019-edition
Radeon Software Adrenalin 2019 Edition out, includes tons of new features and a choice of automatic core/mem OC or automatic undervolting for Vegas
Radeon Software Adrenalin 2019 Edition out, includes tons of new features and a choice of automatic core/mem OC or automatic undervolting for Vegas
I have another take on this Vega2 logo....
One of AMD's recent presentation streams, AMD made a claim I thought should've stole the show, or had industry impact, that seemingly went unnoticed… and it was that they have achieved unified memory across their new interconnect.
I believe that Vega 2 logo, might be Trademarked for a new 7nm Navi GPUs, that have two 7nm Navi dies stitched together using Infinity Fabric 2. Therefore you could have Navi for gaming @ 150w, then you have Navi with a second GPU as Vega2 cards @ 300w. (rem: Vega is the marketing name for their Radeon Product, not Nav. So V2).
??
One of AMD's recent presentation streams, AMD made a claim I thought should've stole the show, or had industry impact, that seemingly went unnoticed… and it was that they have achieved unified memory across their new interconnect.
I believe that Vega 2 logo, might be Trademarked for a new 7nm Navi GPUs, that have two 7nm Navi dies stitched together using Infinity Fabric 2. Therefore you could have Navi for gaming @ 150w, then you have Navi with a second GPU as Vega2 cards @ 300w. (rem: Vega is the marketing name for their Radeon Product, not Nav. So V2).
??
Vega 20 (Vega 12 with 20 CUs) is far more likely for iMac than Vega 48
D
Deleted member 2197
Guest
Wonder why some features (ie, Virtual Super Resolution) were excluded from R9 series. Is it a hardware issue?
That would be a performance regression from Polaris, already in the current iMac.
Well the small Vega could fit into iMac 21.5 (currently using polaris 21) whereas the big one could fit into iMac 27 (currently using Polaris 20). No idea if that's the case, but it would be an upgrade for both (and from the TDP values both should be doable, although Vega 48 would probably have to be clocked quite low).
iMacmatician
Regular
Given that Cascade Lake is scheduled to be released in early 2019, I think it's likely that the iMac Pro will be updated at WWDC 2019. Given Apple's stated timeframe of 2019 for the new "modular" Mac Pro and the fact that the original Mac Pro and the cylinder Mac Pro were both announced at WWDC, I would be shocked if there wasn't a Mac Pro announcement at WWDC 2019. (That being said, the actual release may be later.)
Vega 20 is probably a safe bet for both pro desktops.
The problem is that a Vega 48 CU is likely to be lower end than the current Vega 56 CU in the base iMac Pro (even if the 48 CU is somehow ahead in FLOPS etc., the name suggests a lower end model), so if the 2019 iMac Pro uses Vega 48 CU then I expect it to start at a lower price than the current iMac Pro. The base iMac Pro ($4999) is over 2x the price of the highest end standard configuration of the 27" iMac ($2299) so there's room for the gap to shrink.
A Vega 10 chip in the regular iMac seems a bit out of place given that the highest end GPUs in previous iMacs were GK104/Tonga level or lower, but perhaps it could happen if the Vega 48 CU is a relatively low end variant and Apple wants to extend the regular iMac upwards.
If the performance numbers in the AdoredTV Navi rumor are anywhere near accurate, I'm not sure what is the point of a Vega 48 CU in the 2019 Apple lineup, since Navi 12 and Navi 10 are strong contenders for a regular iMac.
Last edited:
I don't understand how anyone can take those "rumors" seriously? What company would build 1 chip to cover ~45% difference between 2 models and then another chip that's barely 30% faster?
Ditto! [emoji1]Already have my Vega with 16gb of hbm2
Source..? Or are you talking about a previous CEO and team?
Anarchist4000
Veteran
It might actually make sense along the lines of Rome. That's more or less what Vega already does, just within a single chip. Using IF internally to connect GPU to controller. Concern is getting the singular front end running as fast as possible and communicating with CUs over IF efficiently while using an older node for costs.