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

But I guess it's better to have Zen2+Vega in January than Zen2+Navi in July.
Exactly.
Unless Renoir skips LPDDR again, a significantly larger/faster iGPU than Raven Ridge's Vega 11 will be running circles around Ice Lake U with the Iris Pro G7, which is Renoir's sole competition.
 
Unless Renoir comes up later than expected, the difference should be half a year.
Regardless, I can't think of any reason to not adopt RDNA other than development time, or maybe AMD being concerned with the iGPU"s compute capabilities.

Since AMD is a big proponent of HSA I suppose that could make sense. The downside to iGPU's is the way they become wasted silicon once you add in a dedicated GPU. With a Vega ISA onboard you guarantee a compute focused resource which can also do some light graphics work as an iGPU. Suddenly it's not a waste, which sounds great in theory. But wouldn't it require more every-day use cases to become a valid sales argument? I.e. some tangible speedup in users everyday lives? Especially since iGPU's are historically aimed towards systembuilders uninterested in dedicated solutions. The rest have compute resources on hand through dedicated GPU's.

I don't work in the field though. Maybe Windows or MacOS already use compute resources for OS functions?
 
Has AMD come up with actually new chip, or does their new investor slide deck have a mistake in it?


The last known Excavator-based architectures were Bristol Ridge and Stoney Ridge and they have GCN3, not Vega/GCN5
 
Has AMD come up with actually new chip, or does their new investor slide deck have a mistake in it?


The last known Excavator-based architectures were Bristol Ridge and Stoney Ridge and they have GCN3, not Vega/GCN5

My first thought was that this had to be a typo, but the A6 and A4 logo do make me wonder whether this might be some new low-end offering. I can't say that I really see the point of an Excavator-based APU over a dual-core Ryzen, however.
 
My first thought was that this had to be a typo, but the A6 and A4 logo do make me wonder whether this might be some new low-end offering. I can't say that I really see the point of an Excavator-based APU over a dual-core Ryzen, however.

Shoehorning Vega into A-series might be cheaper than keeping support up for < GCN4? Maybe? At some level it's either a typo or someone at AMD figured the ROI is worth it over Ryzen. Though I'd be very curious to look at that calculation.
 
There's no Excavator on 14nm, nor Vega on 28nm.

I can't see how that thing is anything other than a typo. If they exist, future A4/A6 APUs are either rebrandings of older chips or cut-down Ryzen APUs.
 
do you remember this ?

  • Ron Bauman
- Senior Engineer: Lead Floating-Point Unit at Bulldozer microprocessors. I have
developed high-level-design concepts for the SIMD FPU architecture, detailed logic constructions,
controlling the pipelines, defining critical custom data-path macros, and supervising
placements/layouts. I have performed RTL coding for the latest Bulldozer series.I performed the timing closure tasks and revised the design to meet the performance/power commitments. I was a focal-point for tool issues and was always constant communicating with the EDA team.
- Custom Register Files: I accomplished successful designs and deliveries of the Register File
macros for 40nm and 22nm microprocessors with different generations of AMD's 22nm, 32nm, and 45nm SOI Technologies. Currently, I am responsible for the RF-macros at Memory Controller and Instruction Fetch Unit for Bulldozer with AMD 14nm FIN-FET Technology, including multiple Read/Write ports with Content-Addressable-Memory, and Parity generation. I have performed the VHDL set-up, custom circuit designs, the timing critical-path analysis, and the supervision of the layouts.
 
do you remember this ?
I don't remember it, but I know that quote is at least 4 years old. Perhaps the first 14nm tapeouts from AMD+GF did use bulldozer cores because Zen wasn't ready until a year later.
I wouldn't count on future products using Bulldozer. At least I don't see the point in it.
 


"Microsoft Surface", not Surface Laptop. Could mean nothing and it's still about the laptop though.
I don't think Microsoft would equip a Surface tablet with DDR instead of LPDDR, so it's most probably the Surface Laptop because it can give preference to hibernation instead of sleep.

Though it's weird that they're going with AMD's APUs now, when they could use Ice Lake across the board with similar performance.
I guess the Surface team just doesn't care about being late on hardware adoption.
 
Last edited by a moderator:
Though it's weird that they're going with AMD's APUs now, when they could use Ice Lake across the board with similar performance.
I guess the Surface team just doesn't care about being late on hardware adoption.

Perf per watt per dollar?
 
Considering other similar notebook's price, I think that they could have a decent margin even if they put in a pure gold heatsink.
 
What have the unit sales been like for Surface?

edit:

hm... how does the iGPU performance compare?

Anyways, it's a business procurement deal, so it has to come down to only so many variables regardless if they can afford to take a hit on the margin.
 
Perf per watt per dollar?

Yes, a Picasso + DDR4 2400MT/s should cost only a fraction of what an Ice Lake with Iris Plus + LPDDR4 3733MT/s does.
But then again, the Surface line is mostly made of premium products with premium margins, so shouldn't they be looking for the best performance per watt anyways?


hm... how does the iGPU performance compare?
Intel says the G7 Iris Pro with 128bit LPDDR4X 3733MT/s performs similar to a Vega 10 in the 3700U with DDR4 2400MT/s when both are at 25W:


aUDf7eU.jpg



I'm yet to find a review of an Ice Lake laptop with the G7 GPU. In fact as far as Ice Lake laptops go, there's only notebookcheck's review of the latest XPS 2-in-1 but for some odd reason they tested the Core i3 + G1 variant.
 
Yes, a Picasso + DDR4 2400MT/s should cost only a fraction of what an Ice Lake with Iris Plus + LPDDR4 3733MT/s does.
But then again, the Surface line is mostly made of premium products with premium margins, so shouldn't they be looking for the best performance per watt anyways?

It may just be Microsoft getting ready for an AMD product further down the road and this is just them putting out a product to gauge longer term viability of AMD CPUs in the Surface lineup. It's also possible that there are custom alterations in the xx50 lineup to suit Microsoft's needs. Perhaps it'll support LPDDR4 allowing for greater battery life.

I also wonder if this might be their first step in working closer with AMD for a future APU/SOC with higher GPU performance in order to replace using a discrete NV GPU in some of their other models, notably the Surface Book line.

There's also the possibility of Microsoft getting some kind of kick back for the Project Scarlett APU by using AMD APUs in their Surface line. Basically AMD saying we'll throw in a little something extra in the Project Scarlett SOC if you use our Ryzen APUs in your Surface line.

This does make one wonder if the Ryzen APUs will co-exist with Intel CPUs in the same Surface Lines or if a new Surface Lineup will be created for the AMD APUs or if the AMD APUs will be replacing Intel CPUs?

Regards,
SB
 
Back
Top