AMD Graphics Driver Gets "More New Stuff" For Linux 6.9: Continued RDNA4 Enablement

[phoronix]

Phoronix: AMD Graphics Driver Gets "More New Stuff" For Linux 6.9: Continued RDNA4 Enablement

Following the initial AMDGPU driver updates targeting Linux 6.9 that were submitted to DRM-Next one week ago, another batch of AMDGPU feature updates were sent out today ahead of this next kernel cycle kicking off in March...

https://www.phoronix.com/news/Linux-6.9-AMDGPU-More-New-Stuff

[theriddick]

Maybe RDNA4 can be what RDNA3 should have been.,

[cl333r]

Maybe RDNA4 can be what RDNA3 should have been.,

Care to explain?

[Nocifer]

Care to explain?

Better power efficiency, better compute & AI performance and better RT performance compared to the competition. Also, better rasterization performance compared to the equivalent previous gen cards, i.e. a notable increase in frame per $.

On their own, the 7xxx series cards are quite good (I own one), and they'll be even better if/when AMD improves the software stack (mainly the FSR upscaler quality AFAIC). Compared to 6xxx, though, they're overpriced and not enough of a performance boost to be called a real new generation.

[theriddick]

Care to explain?

Only the 7900 XTX really pulled ahead in performance from previous gen. The lower model cards really didn't give a big reason to upgrade from the 6000 series.

Of cause NVIDIA had some duds also, but did bring to the table some new features for their newer models while AMD really didn't.

[schmidtbag]

Better power efficiency, better compute & AI performance and better RT performance compared to the competition. Also, better rasterization performance compared to the equivalent previous gen cards, i.e. a notable increase in frame per $.

On their own, the 7xxx series cards are quite good (I own one), and they'll be even better if/when AMD improves the software stack (mainly the FSR upscaler quality AFAIC). Compared to 6xxx, though, they're overpriced and not enough of a performance boost to be called a real new generation.

I agree with all of this - the 7000 series is nice but the efficiency is underwhelming and pretty much the entire lineup is $50 too expensive for what you're getting. I'll likely get one when the prices drop.

[theriddick]

I am surprised that the RDNA4 cards will be for the most part monolithic dies.
NVIDIA could beat them to the punch with chiplet based GPU's at this rate.

[Anux]

I am surprised that the RDNA4 cards will be for the most part monolithic dies.
NVIDIA could beat them to the punch with chiplet based GPU's at this rate.

To my knowledge Nvidia has no plans for chiplet on desktop.

AMD had RDNA 3 chiplets only for the memory controllers and that didn't work out as planned and is probably of much less use than having the compute die split. But splitting the GPU die is hard to make compatible with games that weren't build with split dies in mind.

[theriddick]

But splitting the GPU die is hard to make compatible with games that weren't build with split dies in mind.

I don't believe that is a issue or how it works. The problem is internal bandwidth and latency between the dies, not software compatibility. This is a hardware level multi chip system, not SLI or CF.

[Teggs]

Better power efficiency, better compute & AI performance and better RT performance compared to the competition. Also, better rasterization performance compared to the equivalent previous gen cards, i.e. a notable increase in frame per $.

On their own, the 7xxx series cards are quite good (I own one), and they'll be even better if/when AMD improves the software stack (mainly the FSR upscaler quality AFAIC). Compared to 6xxx, though, they're overpriced and not enough of a performance boost to be called a real new generation.

You may be familiar with the idea that this completely a naming issue and falls squarely at the feet of Jensen Huang. If you shift everything below the 4090 down a full or in some cases a half-tier and price accordingly, all of your generational improvement graphs fall into place. AMD is responsible for following this move.