• 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.

AMD RDNA3 unveiling event

humbug said:
These GPU's are on TSMC's most expensive node and they are large, the coolers are expensive, the PCB's are expensive, they are expensive.

You want more than 40 FPS RT is Cyberpunk? That's expensive.

This is not the first time Nvidia have made fat expensive GPU's and over saturated games with a feature deliberately to make you think you need it.
Click to expand...


While i agree that the 4nm TSMC process will be expensive the actual gpu itself is not that large at 378.6 mm² and is probably the smallest £1200 gpu chip ever. The die size alone tell us it's a mid high chip as it's way smaller than usual high end chips.
 
TheRealDeal said:
While i agree that the 4nm TSMC process will be expensive the actual gpu itself is not that large at 378.6 mm² and is probably the smallest £1200 gpu chip ever. The die size alone tell us it's a mid high chip as it's way smaller than usual high end chips.
Click to expand...

The 7900XTX is 300mm2 and on 5nm.
---------

I've been saying for years the more we let Nvidia get away with the worse it will get, for all of us.

I'm still waiting for that to sink in.
 
Last edited:
humbug said:
The 7900XTX is 300mm2 and on 5nm.
---------

I've been saying for years the more we let Nvidia get away with the worse it will get, for all of us.

I'm still waiting for that to sink in.
Click to expand...


The actual full 7900xtx chip is 533 mm². We can't really compare it to normal Gpu's as it's the first of it's kind. The 7900xtx has 58 Billion transistors compared to the 4080 16gb at 45.9 billion. Yea it was game over when Titan sold well and a perfect storm through covid for the master plan to be brought forward quicker. Hopefully it backfires.
 
Last edited:
TheRealDeal said:
While i agree that the 4nm TSMC process will be expensive the actual gpu itself is not that large at 378.6 mm² and is probably the smallest £1200 gpu chip ever. The die size alone tell us it's a mid high chip as it's way smaller than usual high end chips.
Click to expand...

This misinformation needs to be corrected. 40x0 are NOT on 4nm, they are 5nm. The 4N is clever naming of an evolution of the the 5nm process.
 
Last edited:
TheRealDeal said:
The actual full 7900xtx chip is 533 mm². We can't really compare it to normal Gpu's as it's the first of it's kind. Yea it was game over when Titan sold well and a perfect storm through covid for the master plan to be brought forward quicker. Hopefully it backfires.
Click to expand...

The Chiplets contain cache and the IMC's, 32MB and 64Bit each, they are on 6nm, they cost very little, literally a couple of $ each.

This was the design choice AMD made to bring the costs down for themselves and it seems us, analogue IO and memory doesn't scale with node shrinks, so they spent the R&D to get it off the logic die, not on paying game developers to cram a billion unnecessary rays in to a game to make you think you need their GPU's, the only reason a company would chose to do that instead of getting around the end of Mores Law is because we tell them that's what works.
 
Last edited:
ICDP said:
This misinformation needs to be corrected. 40x0 are NOT on 4nm, they are 5nm. The 4N is clever naming of an evolution of the the 5nm process.
Click to expand...

Yea you are correct. Just read this "TSMC ramped its 5-nanometer node (N5) early last year. The N5 node is said to deliver around 1.8x improvement in density along with 15% higher speed at iso-power or, alternatively, 30% lower power at the same speed. Since then TSMC also introduced an enhanced successor to N5 called N4. N4 is a derivative of N5 (which is TSMC includes it under the “5nm family”) very similar to the N6 node they introduced to enhance their N7 node. N4 is said to provide a small 6% die shrink through “standard cell innovation” and design rule changes that help realize better area efficiency. There are also some performance and power improvements through various BEOL enhancements. TSMC says that N4 will use EUV for more layers thereby reducing the number of steps and masks, thus offering a cost advantage."
 
This is why Intel's margins on Xeon are so low, 4%, they are trying to compete with this, look at the cost scaling difference between 8 cores and 16 Monolithic vs MCM, now take that to 32 cores, 64 cores, 96 cores, 128 cores.

Its perhaps not that extreme for GPU's, or it might be, they don't say. all i know is the expensive bit, the logic die is on 5nm and half the size of the 4090, its not just that you're getting 2 for one, because of the way that yields work its more than that, the smaller the die the higher the yield scaling.

L2t9uQF.png
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom