• 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

On paper, could the chiplet design give them the means to totally overtake nvidia in ray tracing capabilities? I mean nvidia are dedicating sections of the core to being tensor cores for RT, amd i suppose could have a dedicated chiplet or 2 for ray tracing operations. So they can presumably use more space to do this since it's not a part of the main core? Probably the only reason they don't is these chips were designed before rt really became a thing. Sorta reminds me of voodoo rampage back in the day where the plan was to have a t&l engine on a separate chip called "sage"
 
Last edited:
I don't think it's a question of can they I think it's a question of value.

As long as decent ray tracing costs 4 figures it's a tiny market to fight over.
 
Intel arc had better ray tracing performance relative to amd cards of the same caliber. It just comes down to how much hardware they want to allocate to the function. Clearly amd believes raster performance is more important currently

New reference design from amd looks really good
 
Gerard said:
On paper, could the chiplet design give them the means to totally overtake nvidia in ray tracing capabilities? I mean nvidia are dedicating sections of the core to being tensor cores for RT, amd i suppose could have a dedicated chiplet or 2 for ray tracing operations. So they can presumably use more space to do this since it's not a part of the main core? Probably the only reason they don't is these chips were designed before rt really became a thing. Sorta reminds me of voodoo rampage back in the day where the plan was to have a t&l engine on a separate chip called "sage"
Click to expand...


Even with chiplets, Navi31 is big. GCD plus MCD = 533mm2. And if they had to stick another GCD on there it would be 833mm2 and also either more MCDs or double density MCDs which will cost more. Yea chiplets are cheaper because each chip has a higher yield but total size still counts and based on what happens with Zen, it looks like AMD likes to benefit from having lower cost via better yields and keeping total area under the competitions monolithic chips and going to 833mm2 or 1066mm2 would not be keeping with the cost or value saving mantra.

As it stands today, the 7900xtx is probably cheaper to produce than a rtx4090 but if it had to double the GCD and MCD then it would likely be more expensive to make and have a higher TDP and AMD seems to care about cost, value and power draw more than overall performance.
 
Last edited:
The reference card looks cool but it is most likely power limited so the only way to squeeze more performance is to get a 3x8 pins card from AIBs and hope you don't have to wait for custom tools to unlock more power this time.
 
Grim5 said:
As it stands today, the 7900xtx is probably cheaper to produce than a rtx4090 but if it had to double the GCD and MCD then it would likely be more expensive to make and have a higher TDP and AMD seems to care about cost, value and power draw more than overall performance.
Click to expand...
A 7950XT with 2 GCD's doesn't seem a sure bet yet but I hope they do it...
 
I think there are still problems with the chiplets as they are right now, with the RDNA 2 the cards were released 20 days after they were announced, now there is a 40 days gap. Why do you need 40 days to bring some cards to the market? Either there are problems or Jensen already bought at least half of AMD. :)
 
DeliciousStorage said:
I wonder if the chiplet design will increase idle power consumption like it does with Ryzen CPU's, especially as there's 7 chips in total compared to 3 on the CPU's.
Click to expand...
Erm, I notice idle power on AMD Ryzen platform is way less than intels monolithic dies

I used to have Intel Gold 8500 with a B series board and a AMD Ryzen 3100 with B series board.

Despite the Ryzen having more cores and chiplet in design the total platform power usage is way down compared with Intelwhich was a monolithic die.

It was 30w vs 45w idle. I would be interested in knowing where you get your chiplet consume more power theory from?

From die power usage perspective a chiplet design should be lower power consumption in ALL use cases that’s full on work load as well as idle. Power efficiency is the name of the game for chiplets.
 
Last edited:
sparkymark75 said:
You mean like Ray Tracing? :D
Click to expand...
you mean the thing used in like 40+ games that is becoming a standard feature? maybe AMD users only play indie games though
2 x 8pin connectors, clearly as much performance as possible, but limited by the available power. Even the cooler looks reasonable.
Click to expand...
one of the AIB boards has 3 pins.


sounds like the 4090 series where they are maxed out and not limited by watts anyway though
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom