• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

RDNA 3 rumours Q3/4 2022

Status
Not open for further replies.
A little underwhelmed on raster (was hoping for at least 4090 raster). RT looks to be even more disappointing.

But

BUT...

They did crush the pricing.

Nvidia will *have* to respond at the 4080 tier I think because these cards should slot above or alongside for substantially cheaper.

4090 and 4090Ti look to be untouchable though and that means prices will continue to be sky high at that tier.
They didn't really show much if any raster performance charts?
 
They can quite literally double the size of the raster core whenever they want, they're just choosing to take pity on Nvidia my dude.

The 5NM core is smaller than the TSMC 4NM core used in the AD103 based RTX4080 16GB. It wouldn't surprise me if yields are better on Navi 31 than AD103. The memory chiplets are tiny.
 
Last edited:
The funniest part of the presentation was when they said unlike out competitor you won’t need a new power supply (you don’t)

Few minutes later : you will need a new monitor with display port 2.1 :cry: :cry:
 
The funniest part of the presentation was when they said unlike out competitor you won’t need a new power supply (you don’t)

Few minutes later : you will need a new monitor with display port 2.1 :cry: :cry:
DisplayPort is backwards compatible. So you'll only "need" a DP 2.1 monitor if you want to use that bandwidth.
 
Last edited:
Buying 7900XTX to replace my 3080 if review benchmarks match AMD’s claims

Yup time to dump that 10GB card like it was a hot potato.

lBaOXl2.gif



@Nexus18 now :-

FKmjLUm.gif

^... make sure to click


y3ZIOK7.gif


;) :p
 
No one is scalping £1000+ radeon in 2022/24
I wouldnt rule it out, but the handy thing even with the sizeable price difference between the 7900s & 4090s means the margins for scalping must be limited because demanding £1500 is gonna force folks to retailers where they have 4090s sitting on shelves already.
There was no limit for day1 4090s, but the cap for a 7900xtx is surely around £1500 otherwise you'd buy the more powerful 4090 and a fire extinguisher.
 
It's not a generational gap at all.

its a divergence.. at the moment the 2 cards cant be compared because youd get a completely different picture depending on what you choose
- pure raster
- raster + dlss/fsr
- raster +rt
- raster + rt +dlss/fsr

even dlss/fsr settings are not comparable given how the internet doesnt stop arguing about iq differences
my estimate is nvidia is a couple generations ahead of amd right now because they have chosen to move away from traditional raster performance, nothing was stopping them from filling up CUDA cores to the brim or using the die space for more ROPs
 
Sorry for not wanting to trawl through the last dozen pages, but have they revealed performance numbers (even if they're AMD's own numbers).

I've just started catching up so wanted to know if I'll eventually come across it in some of the media coverage.
 
Sorry for not wanting to trawl through the last dozen pages, but have they revealed performance numbers (even if they're AMD's own numbers).

I've just started catching up so wanted to know if I'll eventually come across it in some of the media coverage.
Mostly for RT, not much for rasterization (which is still the vast majority of games).
 
Lmao doesn't that cost more new

Hardware Unboxed reviewed both and said the RX6950XT matched the RTX3090TI:

So how does an RX7950XTX only match an RTX3090TI?
its a divergence.. at the moment the 2 cards cant be compared because youd get a completely different picture depending on what you choose
- pure raster
- raster + dlss/fsr
- raster +rt
- raster + rt +dlss/fsr

even dlss/fsr settings are not comparable given how the internet doesnt stop arguing about iq differences
my estimate is nvidia is a couple generations ahead of amd right now because they have chosen to move away from traditional raster performance, nothing was stopping them from filling up CUDA cores to the brim or using the die space for more ROPs
Nvidia is at the reticule limit for TSMC 5NM now. AMD isn't - the core die is literally half the size of the AD102 on a slightly worse version of TSMC 5NM. The memory dies are tiny and made on a lagging node.

The next jump to TSMC 3NM won't be as big as the jump from Samsung 8NM(which was a 10NM process node),so Ada Lovelace looks big because the transistor jump was much bigger than normal. If anything AMD has far more room to improve performance over the next few generations.

Now if Nvidia can move to their own chiplets soon then it might be different,but otherwise they will face a scenario like Intel is now. Intel has to throw die space and transistors at the problem.

Edit!!

Another thing - if AMD is comparable in performance/watt or even better than Ada Lovelace,that means the interconnects are more efficient than I expected. In Zen1 the IF,was a major contributor to CPU power consumption.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom