Big changes are coming with AMD's RDNA 3 architecture, and it will give Navi 31 crazy

[WYP]

An almost 4x increase in FP32 TFLOPS over the RX 6900 XT. Wow!



Read more about AMD's rumoured RDNA 3 architecture changes and huge FP32 performance boost.

 

[AngryGoldfish]

Also hearing above 3Ghz clock speeds on some of the RDNA3 chips.

 

[Dicehunter]

IF this is true it'll be interesting to see how this TFLOP increase will translate to real world performance over RDNA2.

 

[WYP]

IF this is true it'll be interesting to see how this TFLOP increase will translate to real world performance over RDNA2.

It will be very workload dependent. Just like Ampere/Turing over Pascal. That said, this will be great for AMD in workloads outside of gaming, where the TFLOPS really matter.

With architectural changes like this, it is more important to think of per CU or per SM performance than shader counts and stream processor counts. Like I highlighted in the article with Turing VS Ampere.

 

[ImprovizoR]

So the rumor is that this thing will draw 450W. That would be insanely power efficient. It sounds too good to be true, honestly.

 

[NeverBackDown]

Yeah for gaming TFLOPs don't really mean anything. It's really more about how efficient the hardware is at managing it's resources so it's able to be as fast as possible.

Now for science and math workloads that's obviously an entirely different matter and TFLOPs actually mean something.


I'm excited for what the bring. I hope it releases alongside a bunch of new technologies like Nvidia has but that's actually competitive or innovative.

 

[robbiec]

My next GPU will be a 7900XT assuming they are available My 6800 has done well this last year

 

[Emu]

So the rumor is that this thing will draw 450W. That would be insanely power efficient. It sounds too good to be true, honestly.

450W would be a bit unbelievable if you consider that the specs stated show a 300% increase in shaders and a 20% clock speed increase over the 300W 6900xt. 600W total board power would be much more believable given a 30% power decrease for the jump to 5/6nm lithography. I think it mostly comes down to how much of a power efficiency improvement they are getting with the chiplet design over a monolithic die - using chiplets will give them the opportunity to cherry pick the best dies for the flagship model.

For what it is worth, I have a feeling that the flagship models for RDNA3 and the 40-series are going to have eye-watering prices and that the 7800/4080 models are going to be the card to go for if you want the best but don't want to spend a fortune.

 

[KingNosser]

They are going to be many things but cheap is not one of them.

I'm intrested from a technical point but atm i'm keeping my 6800xt as long as possible cause cards are going to get really expensive and nanometers only help so far the chiplet option while a better future still can only realistically go so far until we are at the end point and the tech wall is hit hard stop.

I think it's also fair to say how many chiplets are needed silicon isn't cheap and so the more used the higher the price, the one thing amd have is better power atm.

 

[AngryGoldfish]

For what it is worth, I have a feeling that the flagship models for RDNA3 and the 40-series are going to have eye-watering prices and that the 7800/4080 models are going to be the card to go for if you want the best but don't want to spend a fortune.

That's the way it's been a while though. The 1080Ti was somewhat of an exception where, for once, it was worth putting forward the extra money over the 1080 if you could technically do so. But up until now, the best cards have been the XX60-XX80 class and their AMD equivalents. They've usually been the best of what the company could do for the average consumer.

 

[Anon]

Every new generation is better, but I'm still waiting for good drivers and energy efficiency.