AMD's RDNA2 architecture, on which the upcoming graphics card generation, also known as Navi 2x, is based, will dominate hardware-side ray tracing. With announcements for the Xbox Series X and PlayStation 5, this has been known for a long time.
According to a current rumor, not all accelerators of the latest generation should support real-time beam tracking, but only the most expensive models. What's in the latest information and why does it make sense?
What do the rumors say and how trustworthy are they?
The information comes from user AquariusZi from the forum of the Chinese website PC Shipping, which has already been able to draw attention to itself through accurate leaks in the past and to which the latest statements on the chip size of Navi 2x can be traced:
more on the subject
Twice as fast as RX 5700 XT? Leak AMD Big Navi
So it should three versions of the Navi 2x graphics chip give:
- Navi21 505 – Navi 21, high-end, Big Navi, 505 mm² chip area
- Navi22 340 – Navi 22, probably for the middle segment, 340 mm² chip area
- Navi23 240 – Navi 23, probably for the entry-level segment, 240 mm² chip area
* Chip area can vary by + – 5mm² according to the leak
In the same text AquariusZi points out that AMD is probably like Nvidia and Raytracing exclusively for its high-end products, better known as AMD Radeon RX Big Navi (Navi 23 with 505 mm² chip area).
Nvidia only offers hardware ray tracing for models from the Geforce RTX 2060 (Turing). The algorithm can also be executed by driver on certain models of the GTX 16 (also Turing) and GTX 10 series (Pascal), however (even greater) performance losses can be expected.
AquariusZi also writes that smaller navigation chips are apparently designed in such a way that they can "cancel" ray tracing. However, errors in machine translation from Chinese may have occurred here. Presumably it is meant that corresponding Cores either deactivated or not integrated into the chip at all will.
Why does that make sense?
According to the post, the focus of the smaller Navi chips (presumably Navi 22 and 23) is from the start on maximum efficiency of the RDNA2 architecture in order to be able to compete with Nvidia's GTX 16 series.
more on the subject
New from AMD to Radeon graphics cards & Zen 3
In addition, ray tracing requires a lot of power and the corresponding hardware resources. Naturally, fewer dedicated RT cores can be accommodated on a smaller chip area than on a larger chip.
Nvidia, for example, was criticized in the first few months after the Turing release at the end of 2018 that the slower models with fewer RT units and in particular the RTX 2060 were not fast enough to be adequately armed for the loss of fps with activated ray tracing be.
A certain mitigation of fps loss is possible with Nvidia's Deep Learning Super Sampling (DLSS) technology, but it is still an upscaling method.
It is not yet known whether AMD-like upscaling technologies such as Nvidia's tensor cores and DLSS are integrated into the RDNA2 architecture.
Leave a Reply