AMD Licensing RDNA Graphics IP To Samsung For Smartphones & More

[numacross]

So GCN ISA, but new microarchitecture? I didn't really fully understand that comment. So the next generation (Navi 2?) will use new ISA and that new mircoarchitecture?

The ISA stays the same - it's still GCN. But the microarchitecture running this ISA changes.

Let's make a crude CPU analogy. It's like for amd64 ISA going from Excavator ("GCN microarchitecture") to Zen (RDNA microarchitecture). The amd64 ISA changed very little (added a few instructions) in this transition but the actual implementations in the microarchitectures are vastly different.

[shmerl]

Let's make a crude CPU analogy. It's like for amd64 ISA going from Excavator ("GCN microarchitecture") to Zen (RDNA microarchitecture). The amd64 ISA changed very little (added a few instructions) in this transition but the actual implementations in the microarchitectures are vastly different.

I got that part. That's what some described as "hybrid" - GCN instruction set (ISA), but new hardware implementation of it (microarchitecture). But do they plan to change ISA with next iteration of Navi though? That wasn't clear.

[numacross]

I got that part. That's what some described as "hybrid" - GCN instruction set (ISA), but new hardware implementation of it (microarchitecture). But do they plan to change ISA with next iteration of Navi though? That wasn't clear.

In the tread I linked before bridgman didn't say, but I doubt that they will change the ISA drastically.

[wizard69]

This makes me wonder about Apples GPU and the possibility that there is some AMD IP in there via secrets agreement. I know Apple hired a large number of AMD engineers a few years ago.

[anth]

I wonder if the sale of Imageon from AMD to Qualcomm included a 10 year non-compete clause.

[Veto]

I got that part. That's what some described as "hybrid" - GCN instruction set (ISA), but new hardware implementation of it (microarchitecture). But do they plan to change ISA with next iteration of Navi though? That wasn't clear.

Why should they?

Unless the GCN ISA is becoming a serious hindrance for evolving the micro-architecture it would be discarding a huge SW investment (drivers, shader compilers, LLVM, ROCm etc.). Furthermore it seems like GPUs have evolved from fixed-function graphics accelerators towards unified, general purpose SIMD machines useful for all kinds of massively parallel computation (Physics simulations, block-chain, neural networks, etc.).

Apparently AMD has already evolved the GCN ISA throughout the GCN generations, so I would bet AMD to extend the ISA with raytracing and neural network oriented instructions and implement the capability in a more generic way (e.g. by adding BFloat16 support). That compared to other companies, who seem to prefer the opaque fixed-function approach, which often require software to be highly tailored towards that particular company and their architecture.

[smitty3268]

The ISA stays the same - it's still GCN. But the microarchitecture running this ISA changes.

Let's make a crude CPU analogy. It's like for amd64 ISA going from Excavator ("GCN microarchitecture") to Zen (RDNA microarchitecture). The amd64 ISA changed very little (added a few instructions) in this transition but the actual implementations in the microarchitectures are vastly different.

It's still unclear how big of a change in the architecture rdna is, and it will be so for at least 1 more week until AMD releases more info.

The architecture has changed every time they put out a new version of GCN, so the question is whether rdna is completely different, or just another regular update (like GCN 2 -> GCN 3) tuned more towards gaming performance rather than compute and given a spiffy new marketing name.

Lisa Su's presentation at Computex implied the former, but there are zero details out there in the public so it would be foolish to believe it's any more than the latter until we hear more.

The few leaks and rumors that have been out have hinted that there may be some clockspeed and power consumption issues with Navi early on, so let's hope AMD has gotten those worked out.

[Teggs]

Making some money is better than making no money, which is what AMD's previous non-participation in that market amounted to.

It's tempting to think they should go for full-bore participation in the massive and expanding mobile graphics market. On the other hand that hasn't worked out for everyone who tried, and AMD seems to have enough challenges from laptop graphics on up as is. Intel and Nvidia don't seem to be that worked up about it, either.

[RussianNeuroMancer]

Is there Snapdragon or Exynos-based boards similar (performance / price) to Odroid-N2, Khadas VIM3, HiKey 960/970?

[theriddick]

Wonder how many watts a mobile gpu needs to conform within, 50w?