AMD Execution Thread [2024]

But not everyone needs applications requiring AVX/2. If that's the case, I believe Microsoft'd be already emulating that (it's not very difficult even without vector instructions, it's just a matter of performance). It's probably going to go like this: if enough people are buying these laptops for, say, Office, application vendors will notice and will port their applications. Rosetta 2 does not support AVX/2 either.
It's rumored that Rosetta 2 couldn't emulate AVX/2 due to Intel's patents. If so Microsoft can't emulate it either, even if it wants to.
 
Last edited:
It's rumored that Rosetta 2 couldn't emulate AVX/2 due to Intel's patents. If so Microsoft can't emulate it either, even if it wants to.
That's not true. There are already open source user space x86 emulators out there that implement AVX/2 support. The main blocker from the blog is as follows ...
As you can see a lot of new instructions are now implemented. This now leaves us with about thirty more instructions that need to be implemented before we can start avertising the features on SVE2-256bit supporting hardware. This is significant as we keep finding more and more games that are requiring AVX to run
And until ARM implementations change suit, they won't be able to run these applications ...
 
Instead of playing the renaming game, AMD should look at their market share, because it's not going well...

View attachment 11426


Not the most trusting in the various "analyst" unit sales, there's hardly a guarantee any of them are right.

But AMD can obviously do better, in terms of GPU sales and naming conventions.

I'd say the best is Apple's "A/MX" series. Just make it obvious. The letter denotes what size device it's for (M = tablet/laptop/etc. A = phone/etc.), then the higher the number is the newer it is, so damned simple and easy to communicate to people!

That being said, at least the "AI 300" series shows they can make slight improvements in naming. Maybe RDNA4 will drop a number and a few extraneous letters as well, and be distinct from the CPU naming. No more "AMD 5XXX" being both the CPU and GPU at the same time.
 
Last edited:
And AMD own financial numbers follow JPR study...

It could be lower or higher. But you might have to be an accountant familiar with the industry to know how to read financial statements from AMD, where do they book what? When do they book revenue? If RDNA4 gets released 2 months from now, does that mean they stopped producing RDNA3 dies months ago to clear their inventory for RDNA4, if they did do they stop booking revenue because all available inventory is shipped to board partners, how do they count revenue from shipping to board partners versus from final retail sales?

Financial statements from big companies are a mess that can be hard to divine timely facts from.
 
So, NVIDIA managed to sell 10% more units in Q1/24 vs Q4/23 (7.7m vs 7.0m), thus increasing their share to 88%. Despite this their gaming GPU revenue dropped by 8% in Q1/24 vs Q4/24. I am guessing this is due to them slashing prices with the Super series introduction?

Meanwhile, AMD fell 55% in unit sales in Q1/24 vs Q4/23 (1.0m vs 1.8m), while their gaming revenue dropped by 33% in the same period, the deficit could be explained by console sales picking up the difference of course.

I think it's more likely their unit sales fell off more in the low end. 7600 is not that compelling vs 4060, 6600 was much better versus it's competition.
 
Didn't we get these same kind of rumors last time? Somebody is always rumormongering about delays.

I dont really see any reason why there should be a wait. Especially for RDNA4. Blackwell could maybe have the excuse of GDDR7 yield/volume issues or whatever, but I'm still generally expecting things this year.
paper launch, maybe ...

The next generations of video cards from AMD and Intel will not appear until early 2025. Tweakers verified this with several sources during the Computex 2024 fair in Taipei. According to previous rumors, competitor Nvidia will soon release a new GeForce RTX 5000 series. It is possible that AMD and Intel will announce their new GPUs just before the end of the year to meet the aforementioned deadlines, but an actual release at the CES fair in early January is more likely. In any case, availability will only start in the new year.

https://tweakers-net.translate.goog...html?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en-GB
 
Last edited:
Tweakers verified this with several sources during the Computex 2024 fair in Taipei.
It's nothing but a rumor. Who on earth knows who they were talking to or what was specifically being said.

Especially when there's literally no reason given for why everybody's GPU's are getting delayed, it's not something I'm taking super seriously.
 
I dont really see any reason why there should be a wait. Especially for RDNA4.

Only reason I can think of is inventory levels. If AMD made too much RDNA3 and it hasn't sold yet, and RDNA4 doesn't have a high-end card that's above current RDNA3 cards, it might not make sense to release anything until the old stock is sold.

Would sure explain the weirdness with driver code having been very early compared to earlier releases. That is, if the product is basically ready to ship but isn't shipping because of business reasons.
 
You'd either need to release in the summer (likely not ready) or next year to avoid the potential disruptions from the US election.

I'm referring to the potential impact on the business climate especially in terms tech products produced in China.
 
Only reason I can think of is inventory levels. If AMD made too much RDNA3 and it hasn't sold yet, and RDNA4 doesn't have a high-end card that's above current RDNA3 cards, it might not make sense to release anything until the old stock is sold.

Would sure explain the weirdness with driver code having been very early compared to earlier releases. That is, if the product is basically ready to ship but isn't shipping because of business reasons.
There's a known trick that's been used for decades to get rid of excess inventory before a launch of a new, better product....
 
There's a known trick that's been used for decades to get rid of excess inventory before a launch of a new, better product....

And there's still at least 3-4 months to go, plenty of time to clear inventory, if any. These rumours don't seem very credible tbh, both Nvidia and AMD have pretty much been iterating on a 2 year cadence with a Q4 release. And given the mid-range nature of the new parts (small monolithic dies, similar GDDR6 memory and likely reuse of board designs) , shouldn't really be any reason for delays. We'll probably get some sort of official update at Gamescom in August I suppose.
 
https://wccftech.com/amd-peano-open-source-llvm-compiler-designed-for-ryzen-ai-npus-ai-accelerators/

On behalf of AMD, I’m pleased to announce the open sourcing of an LLVM backend for AMD/Xilinx AI Engine processors. These processors exist in a number of devices including RyzenAI SoCs. The repository currently focuses on supporting the AIE2 architecture implemented by the XDNA accelerators in “Phoenix” and “Hawk Point” devices.


Note that these accelerators include an array of processors, while the LLVM backend only supports a single processor. Support for devices as a whole is available in open source tools based on MLIR.
 
You keep bringing this x86 256-bit AVX issue on ARM but it's not really a problem. AVX or whatever extension, is a click on the compiler option and then the software runs faster if it can take advantage of the CPU instructions available
Yep, this is what happened with the latest Mac Game Porting toolkit 2, which added AVX2 instructions into the compiler. Now even emulated games are running on Mac.

 
Back
Top