Home » GPU » GPU FAQs » RDNA 3 ray tracing

RDNA 3 ray tracing

One of the drawbacks of the company, it might be improving in the new generation

Updated: Nov 23, 2022 2:30 pm
RDNA 3 ray tracing

WePC is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Prices subject to change. Learn more

The new RDNA 3 architecture, vastly different from RDNA 2’s architecture inside RX 6000-series GPUs, powers both the 7900 XT and 7900 XTX. Even though the two GPUs have an identical design, it’s interesting to look at when considering AMD’s most recent innovations.

Most notably, chipsets, the innovation that helped Ryzen win the top spot in 2019, have lastly reached GPUs. That’s major news, but the RX 7000’s chipset design differs significantly from that of Ryzen and other AMD CPUs in several respects.

RX 7000 GPUs have one die with all the cores and several dies with one GDDR6 memory controller and 16MB of Infinity Cache on each, as opposed to several dies for cores and an additional die for I/O and other tasks.

The Graphics Compute Die (GCD) is the semiconductor with all the GPU cores, and Memory Cache Dies are the chips with memory controllers and cache (MCD).

This means that an RX 7000 GPU, more precisely the 7900 XTX, can have up to 96MB of Infinity Cache and a bus width of 384 bits, allowing for a maximum bandwidth of 5.3 TB/s on the 7900 XTX.

The 7900 XT, on the other hand, has 4GB less VRAM and just 80MB of cache in addition to a 320-bit bus and what looks to be one fewer MCD.

RDNA 3 performance
RDNA 3 ray tracing performance, source: AMD

Lower-end RX 7000 GPUs will probably employ a different GCD and even fewer MCDs, but we will know something once those are unveiled early next year. A new node is also included in RDNA 3.

The RX 6000 GPUs from the previous generation were produced on TSMC’s 7nm node; the RX 7000 GCDs are now built on 5nm, which has up to 1.8 times the transistor density of 7nm and can increase clock speeds by 15% while using the same amount of power.

Or consume 30% less power while maintaining the same clock speed. According to reports, 5nm enables RDNA 3 to be 54% more efficient than RDNA 2 when combined with architectural advancements.

TSMC’s 6nm node, which is essentially 7nm but emphasizes efficiency, low power, and
low cost is used to fabricate the MCDs for the RX 7000.

Every generation makes it harder to shrink I/O and cache components, such as memory controllers. Therefore AMD is sticking with 6nm for these MCDs; 5nm is too expensive and offers no benefit in this area.

Improved ray tracing performance

AMD promises up to 67% more performance in ray tracing scenarios. It showcased these in the RDNA 3 announcement and with it hope for the upcoming cards.

Radeon cards have struggled to keep up with Nvidia when it comes to ray tracing and other drivers. So this improvement can help them close the gap. Especially if their benchmarks are to be trusted. As it makes the games more playable.

With a background in engineering and PC gaming, Seb is a staff writer with a focus on GPU, storage, and power supplies. Also one of tech supports in the office he likes helping and solving problems.

Trusted Source

WePC’s mission is to be the most trusted site in tech. Our editorial content is 100% independent and we put every product we review through a rigorous testing process before telling you exactly what we think. We won’t recommend anything we wouldn’t use ourselves. Read more