The year 2020 was an amazing time for both the PC and console world. We saw brand-new CPU and GPU releases from both AMD and NVIDIA, but also we got two next-gen consoles from Xbox and Sony.
TSMC already has the 5nm process in the works and AMD is planning to base its RDNA 3 and Zen 4 architecture on that. This means we can expect even further performance improvements.
Table of ContentsShow
Why should we be excited for RDNA 3?
There are a few reasons why. First, we have to acknowledge the fact that AMD finally brought something competitive enough to battle against NVIDIA’s finest. The RX 6900XT easily trading blows against the RTX 3090 in raw rasterization performance.
While lacking in ray tracing, these RDNA 2 GPUs still make for a good choice.
The second reason why we should all be excited about RDNA 3’s competitiveness is that AMD has not been present in the flagship/high-end part of the GPU market for years.
The last time AMD came up with a counter to NVIDIA was more than 10 years ago in 2011 with the HD 7970 against the GTX 580 and GTX 680. If AMD makes another competitive comeback with RDNA 3 to incite rivalry, we are going to see a considerable improvement in price-per-performance from both companies.
Having said that, it is time to focus on the information we have about RDNA 3 and talk about exactly what we can expect.
Currently, any kind of information regarding the release of RDNA 3, or sometimes referred to as Navi 3 is just a rumor.
Official details about the release date and specifications are expected somewhere in middle or late 2021. Although, at the time of writing, there are a lot of issues regarding chip and semiconductor manufacturing which means things could change, get delayed, etc.
Nothing is set in stone.
AMD Gaming GPU Architecture Roadmap
Considering that most of the GPUs from the RDNA 2 lineup were released around November 2020, it is safe to assume that we will not be seeing any high-end RDNA 3 releases from AMD at least until the end of 2021. Even that is a reach.
Based on AMD’s gaming GPU roadmap, they will have an advanced node ready for RDNA 3 before 2022. But, this roadmap is just for the development of the chip. This does not necessarily confirm a release of a consumer GPU.
The mid-end and low-end RDNA 2 SKUs scheduled to release throughout 2021 supports the roadmap above.. It is important to remember that these tech companies are making most of their profits from the more affordable products.
In Q1 of 2021, RX 6700XT released which is a direct contender against NVIDIA’s 6GB RTX 3060Ti (and 3070). But, a contender against the RTX 3060 is also expected (an RX 6700) and other lower-end GPUs.
There’s no official information regarding something more affordable, but an RX 6500 or at least an RX 6600 is anticipated in 2021.
All of this tells us that consumer RDNA 3 products are not bound for 2021.
The new 5nm fabrication process from TSMC sounds promising. Especially when you consider the performance per watt increase we saw from AMD’s old GCN architecture to its successor, RDNA 1.
The 5700XT and 5700 offered a lot of raw rasterization performance for a very competitive price. The 5700XT sometimes even surpassing the 2070 Super which comes with a much more expensive price tag (at least $100+).
But, that’s not the most impressive part of the 7nm fabrication process. The impressive part was the monumental 50% leap from RDNA 1 to RDNA 2. The longer the architecture sticks around, the more optimization it can get to increase performance and efficiency (example: Intel’s 14nm+++ arch).
To put it simply, the 5nm fabrication process will probably bring us even more efficiency and raw power. AMD claims that another 50% leap is expected.
For those that are interested in the details, numbers, and specifications, keep in mind that there isn’t a lot to share except a couple of rumors and assumptions. But, we understand that even that might be enough for some.
Multiple sources are rumoring that Navi 3 will be with a chiplet design which is something that we have not seen in the GPU industry before.
However, we have seen this same design in AMD’s Ryzen processors though. The chiplet design of their processors is their secret formula for the considerable increase in performance and physical cores. Maybe this secret formula will work with their RDNA 3 graphics cards too.
The problem with this kind of chiplet design is the increase in latency. This is probably why AMD Ryzen processors have higher latency than Intel CPUs and favor higher RAM clocks speeds.
Still, the gains from a chiplet design is worth it since it is effectively cutting down the price of one CPU almost by half.
The increase of latency in CPUs is not a huge problem, but an increase in latency with GPUs, that could be a serious downgrade in performance. Both SLI and Crossfire technologies always had problems with latency between multiple GPUs. Running GPUs in parallel is not optimal.
Because of this, AMD is coming up with a unique solution, an active bridge connecting all of the chiplets together. The first GPU chiplet would also be connected with the CPU directly with the Infinity Fabric.
Rumors say that a few of the GPUs in the RDNA 3 lineup will have two chiplets with 80 CUs. That is a total of 160 CUs. To put that into perspective, the current most powerful consumer graphics card from AMD has 80 CUs. With double CUs and streaming processors (SPs), more than double the performance is anticipated.
Another thing to note is that AMD’s response to the DLSS feature will probably be coming in 2021 which could bring additional performance over NVIDIA GPUs.
With the lack of official information for RDNA 3 and the 5nm fabrication process, it is impossible to predict the pricing of the Navi 3x lineup. Hopefully, there will be more details in mid-2021.
However, we can assume that RDNA 3’s price tag will not deviate too much from the current pricing. It seems like both NVIDIA and AMD have hit an MSRP sweet spot. They will find no reason to provide any cheaper offers.
There may not be a lot of details regarding RDNA 3, but the pattern is quite clear, showing huge performance and efficiency leaps. All that is left is to wait and see.