AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion

Yes, that is true. Though it'll probably depend on games, OS, background processes... But I think 16 (2x8) GB is becoming standard RAM amount for new gaming PCs. And I believe main reason for that is people don't want their PCs to have less RAM than their phones :d

 

But when 4 GiByte graphics memory are starting to become too limiting for mainstream use, it is much easier to upgrade your system memory (best case: just add another 2 sticks) than your graphics memory.

 

That is also correct

Also, if Vega 32 is successor to RX 580 (36 CUs and ~1400 MHz), how faster will be? As it seems, GloFo will start production using 12LP in Q1 (not risk prod. as it was first thought), so if that one is used clocks will for sure be higher than current ones (1600 MHz). But I doubt that alone would be enough, even if Vega 11 reaches ~1800 MHz

 

Could be. But this only effects people keeping a card for a very long time, but for those it always made sense to buy the card with the bigger VRAM. So I still do not see the advantage of 4GB HBM2 HBCC over 8GB GDDR.

 

I don't see it either. I also don't see it as disadvantage, but most people will. Just because they see 4GB vs 8GB, not knowing the actual difference

 

It is risk production even though it will be used for chips going straight to market (same slide set as that slide is from sees GloFo considering H1/18 as risk production)

 

Not 100% Vega-only, but nonetheless:
https://www.pcworld.com/article/322...brand-as-multi-gpu-gets-more-complicated.html

Apparently, Crossfire is not a marketing term that's gonna live on.

 

How come this comes up now? It's been "multiGPU" in all their communication for a long time already (at least since 17.2.1 drivers in february, possibly earlier)

edit: checked: 16.12.x drivers say MultiGPU, 16.11.x CrossFire

 

1 - Substantially better power efficiency, which they need for laptop solutions.
2 - AMD has been praying the "4GB + HBCC is great!" idea, but all Vega 10 cards have at least 8GB (which makes HBCC practically useless for gaming for now).
3 - Vega 11 with GDDR5 would bring it too close to Polaris 10 cards, making it redundant.


I could see Vega 28/32 existing solely in laptops, offering e.g. GTX580 performance at ~75W power consumption. That would also go around the consumers who aren't willing to buy 4GB cards.

 

Polaris 10 with 28 CUs, 8GB of VRAM and ~75W Power consumption
https://pro.radeon.com/en/product/wx-series/radeon-pro-wx-5100/

AMD cards would be very energy efficient if their clock were just a little bit lower. AMD knows that and miners know that

 

It's quite evident Vega heavily missed it's performance targets.

 

But outside of a hardware defect in Vega 10, a 50% Vega will not do better.

 

Maybe they'll fix in Vega 11 what was/is broken in Vega 10

 

They already had all the time in the world to fix broken hardware with a respin. Vega10 passed the tapeout ages ago.

 

I can't speak for anyone else, but my GPU guy (Nate Oh) picked up on this when looking at AMD's announcement. This is the first time in some months where we've had any reason to look into CF, and.as AMD has never formally announced an end to the use of Crossfire, we've never noticed it before now.

 

People should really read the driver release notes, they're not that long and this has been in plain sight since last december

 

Soo many so much more important things to keep track of. Like state of DSBR, Prim. Shaders, Launch Prices, Availability, HMB2 temps and packaging... Good for you if you never fell behind the curve on any of these.

 

True, though crossfire/multigpu-switch happened long time before Vegas

 

https://videocardz.com/72934/rumor-amd-matisse-picasso-vega-20-and-ryzen-5-pro-mobile

Per slide, Vega 20 Q318.

 

Perhaps mGPU is more for Ravenridge CPUs with Vega GPUs.