New Upcoming ATI/AMD GPU's Thread: Leaks, Hopes & Aftermarket GPU's

Into the GPU Chiplet Era: An Interview With AMD's Sam Naffziger

"The demand for gaming and compute performance is, if anything, just accelerating, and at the same time, the underlying process technology is slowing down pretty dramatically — and the improvement rate. So the power levels are just going to keep going up. Now, we've got a multi-year roadmap of very significant efficiency improvements to offset that curve, but the trend is there.”...

Naffziger explained some of the improvements AMD has seen with its previous RDNA 2 architecture. For example, if it can run at 2.5 GHz and 1.0V instead of 1.2V, the latter would require 40% more power. By leveraging expertise from its CPU design teams, Naffziger said that AMD has been able to drive higher clocks with RDNA 3 while remaining efficient. AMD has long discussed its strategy of ‘cross-pollinating’ CPU and GPU design teams, bringing the best tech from both sides of the house to each new CPU and GPU design...

"Performance is king," stated Naffziger, "but even if our designs are more power-efficient, that doesn't mean you don't push power levels up if the competition is doing the same thing. It's just that they'll have to push them a lot higher than we will."...

Naffziger didn’t tip his hat on the next-gen RDNA 3 architecture, but our best guess is that AMD's GPU chiplet design will likely end up looking at least somewhat like the current CPU designs, where it will have GPU chiplets that house the compute units (CUs), shader cores, and some Infinity Cache. There would then be at least two I/O chiplet designs, one that could scale to higher chiplet counts with a wider memory interface, and one that might only support up to two GPU chiplets with a narrower interface. AMD would link the chiplets via an updated Infinity Fabric, and it would likely have a modest chunk of cache on the I/O chiplet to help optimize memory accesses...

 

"Performance is the king" thats an nvidia moto and i see amd too going with it. Although i see rdna3 being more efficient than Ada there is no question power consumption is going up substantially.

 

Ready for dual PSU PCs?

"Daddy, I want a new PC, and I need a 220V line installed in my bedroom." : Dad nods, proud that his little one has impressed him:

 

not yet needed, but a lot of people will need to upgrade their psu.

also,it's imperative that people with high end systems check their psu's opp values.
otherwise, they may experience shutdowns.

let's say someone gets a 450w card for a 850w psu. it will be fine,with something like a 5800x it should stay well under 600w for average power draw. but it will probably need 1000w opp to avoid tripping it.
https://www.tomshardware.com/reviews/seasonic-focus-plus-gold-850-psu,5247-6.html

 

I think it's all down to their strategy.
they're doing one gpu series per each new node. hence the power requirements. they're using everything a given process can literally survive.
if they planned another series on 4nm, there would be no 450/600w cards now.

 

AMD RDNA 3 ‘GFX11’ GPUs May Feature Hardware-Accelerated FSR 3.0 Tech Thanks To WMMA ‘Wave Matrix-Multiply-Accumulate’ Instructions

->https://wccftech.com/amd-fsr-3-0-gfx11-rdna-3-gpus-hardware-acceleration-wmma-instructions/?beta=1

AMD's Next-Gen FSR 3.0 Tech Could Feature Hardware Acceleration Through RDNA 3 'GFX11' GPUs Brand New WMMA Instructions

One of the key advantages of FSR 1.0 and FSR 2.0 compared to NVIDIA's DLSS has been that it does not rely on any hardware assistance such as dedicated Machine Learning (ML) blocks but that may soon be coming to an end.
While AMD has done an absolutely great job with FSR, offering not only a visual quality on par with NVIDIA's solution but also by making it more open-source,
it looks like in the coming generation, AMD might be going one step ahead & using dedicated machine learning blocks to further boost the performance and visual quality that FSR has to offer.

As spotted by @0x22h, the LLVM repository was recently updated with a new commit, introducing WMMA (Wave Matrix Multi-Accumulate) instructions on GFX11 hardware.
The GFX11 codename is internally used for AMD's RDNA 3 GPU family which will be featured in the next-generation Radeon RX 7000 and Radeon Pro graphics cards.

Similar to how NVIDIA uses matrix multiplactions for deep learning operations through its latest Tensor Core architecture,
the AMD WMMA instructions will be fused on a hardware level to help achieve better Machine Learning or DNN operations.
Now there aren't a lot of details provided but this recent update in the LLVM could be a hint at a major graphics pipeline overhaul in the RDNA 3 GPUs.

 

AMD RDNA3 Navi 31 GPU with six MCDs to feature 384-bit wide memory bus

This will probably mean a big increase of RDNA3 memory capabilities.
But that 384 bit bus is going to consume more power.

 

Navi 33 Only 8GB ?

 

NAVI 33 will be 7600xt.

AMD Navi 33 “RDNA 3” GPUs Allegedly Featured Within Mid-Range Radeon RX 7600 XT Graphics Card Around $400 US, Faster Than RX 6900 XT

https://wccftech.com/amd-navi-33-rd...cs-card-around-400-us-faster-than-rx-6900-xt/

 

If it has 8GB they can kiss my arse. Considering Intel Arc A770 16GB since today I don't need to play DX9 and older games.

 

Too good to be true.
If it was a 7700 XT, maybe I could believe it.

 

RIP the budget GPU for $200

 

AMD preparing AI-based Noise Suppression technology, an alternative to NVIDIA RTX Voice

 

AMD needs to be more creative here, instead of just reactive all the time. If they really want to make inroads into Nvidia's market share, not only does performance need to be as good or better ( which it currently is on par) they need to innovate instead of copying the opposition.

 

Rumour has it, Radeon PRO/7950 XT 32/48GB to fight off Ada_Lovelace Titan 48GB.

 

Here's hoping it's faster and with really great RT performance.

 

7900xt gonna be a beast
https://wccftech.com/could-amds-rdn...-offer-an-effective-bandwidth-of-over-6-tb-s/

 

Hype train is close to derailing now. Maybe we are going back to times when we had abundance of memory. Nvidia really played pandemic shortage right. Putting the SMALLEST possible amount of VRAM on all graphic cards (excluding rtx3090), because why not - it all sold anyway.

Intel ARC 1st gen is a joke, sadly. Maybe 4th gen will be worth something.

Right now GPU prices going down like 5$ each week. I was hoping for 300$ 6700XT by now . But I don't mind seeing current GPUs rot in warehouses.