With the RX 7900 fresh out of the oven and the rest of the graphics cards with RDNA 3 GPUs still in the works, it is too early to talk about RDNA 4. We know that there is already some “information” on the net that with the eagerness of visits “they filter » Information that later turns out to be false. At this time we are talking about an architecture that is not even finished and, therefore, continues to be developed within the Radeon Technology Group laboratories. However, this will not serve to introduce you to the development stages of a graphical architecture.
AMD will have to improve the voltage-frequency curve
One of the challenges that AMD has faced from one generation to the next is to increase the efficiency of its GPUs through the so-called performance per watt. However, since consumption does not increase linearly, but rather in the form of a logarithmic curve, this is tricky. That is, the performance is not the same at all powers. So the manufacturer usually takes the segment of information that interests him and shows it to you. They are not lying, but it is not reliable information either.
What we do have to take into account is that at the moment, despite the fact that the RTX 4090 has a much higher overall consumption, it has a much better voltage and clock speed curve than that of the RX 7900 XTX. This is one of the points to stir, since it will allow AMD to not only make RDNA 4 reach higher clock speeds, but it will also leave a greater margin to be able to install much faster video memories.
When a manufacturer tells you that one architecture is more efficient than another, they are really improving the relationship between voltage and frequency. That is, get more clock speed for less consumption or maintain the MHz, but with less voltage. In this way, superior performance is achieved, even if the number of drives is not increased.
Using the new 12VHPWR connector
AMD’s goal for RDNA 4 would be to at least match the voltage to clock speed matching the RTX 4090. Especially since, by 2024, the 12VHPWR connector will already have become commonplace in a lot of places. of power supplies. This will be essential to not be left behind by an NVIDIA that will use said connection.
On which node will GPUs with RDNA 4 architecture be manufactured?
As of today, what we officially have is an AMD roadmap as the only official information. The only thing we know? The use of a more advanced TSMC node for its manufacture than the 5 nm one, which suggests that we could expect to see the next generation of AMD graphics cards under 3 nm, but it may also be that they bet on the same node than NVIDIA today. That is, the 4 nm one, which is compatible in design with the 5 nm one and would make things much easier for them when launching a new architecture.
On the other hand, we have the current economic situation, which has slowed down sales and with it the release to the market of the different graphic architectures. AMD can’t afford to have multiple generations competing with each other. Hence, the date for the launch of RDNA 4 is at the end of 2024. If we count that the last year is used to finish designing the chip. So we find ourselves with a clear situation, it is nonsense to talk about final technical specifications, since they can undergo changes. Precisely the information “leakers” already caught their fingers with the RX 7000 and the last-minute changes that AMD made in response to the rise in wafer prices.
AMD may not make use of TSMC’s 3nm node
The reason for this is that while the N3 node is much more advanced, its costs are so high that it almost eliminates the benefits of higher density. Hence, AMD talks about a more advanced node, since they could perfectly bet on the N4 variant, which is compatible with the one they currently use and also has a much lower manufacturing cost.
However, we must clarify what we base for such a statement and it is the fact that we will see Zen 5 in both nodes N4 and N3. The reason is that they will seek not to manufacture large chips in said node. The most certain thing is that only the CCX of the processor where the cores and the cache are located. However, on larger chips such as laptop processors that tend to be monolithic and RDNA 4 GPUs could pull the 4nm node for that purpose.
We will see more chiplets in RDNA 4
One of the things that the rumors talk about and that I do agree with is the fact that AMD is going to add more chiplets. What’s more, it’s going to split the core GCD into several smaller chips. What do we base it on? In a recently released AMD patent titled DIE STACKING FOR MODULAR PARALLEL PROCESSORS. However, the existence of this patent does not ensure anything for RDNA 4, it could be perfectly applied to CDNA or be facing a much more distant architecture.
What does the patent tell us?
What interests us at this point is the separation of the GCD into two parts called SED (Shader Engine Dies) and CP (Command Processor), which would be found in the chip that is in the base and would not be a chiplet on top of it. . Look at the amount of SED on the 602a base that would be the interposer
In a GPU, the command processor is located in the central part and is in charge of distributing the tasks between the different clusters or sets of shader units or cores. However, for communication, this is a problem if we divide it among several chiplets, especially with regard to wiring if we use traditional interfaces, from there to move it to the interposer or the base for greater distribution and lower cost.
This, at the same time, allows the central chip to be broken up into several much smaller and more symmetrical chips. Which are cheaper to make and AMD doesn’t have to make different chips for different configurations. In the RX 7000 we will see that there will be different GCD configurations, but the SED chips will always be the same in different numbers within the RDNA 4 based RX 8000.
The other chiplets, potential changes in the MCD
The MCDs have the task of communicating with the video memory, however, and continuing with the patent, the chip that serves as the base is not only designed to carry out the communication between the different chiplets, but also all the work of interacting with the memory, be it GDDR7 or any other standard, it would carry over to what we call the Interposer. Hence, the cache is mentioned in the diagram as a separate element. That is, the Infinity Cache will continue to exist, but it will not be the MCD we know, but rather all high-density SRAM memory and the capacity to be used as a last-level cache.
What do we see most realistic for the future of AMD GPUs?
Leaving the patent aside, which we will repeat has no reason to be related to RDNA 4 and could be about its family of cards for computing in supercomputers. We believe that what AMD is going to focus on for RDNA 4 will be to solve the problems with RDNA 3. Especially in its voltage and frequency curve, which has room for improvement, since to achieve the same results in games than an RTX 4080.
Not surprisingly, if we look at how many points the RX 7900 XTX gets per watt, in benchmarks like FireStrike Ultra we will see that it stays at 52 points per watt, without changing the clock speed or undervolting either. However, if we do the classic voltage drop exercise, then the RTX 4080 can reach 77 points per watt and the RX 7900 XTX stays at 58. Hence, at the beginning of the article we told you that one of the Points that AMD will play in RDNA 4 will be precisely that and it is not a minor change, but something that requires a complete redesign of the entire chip, from the first to the last transistor.
This doesn’t mean that certain system units are going to stay the same, and we may be wrong, but we don’t think RDNA 4 will be much different from RDNA 3 in the sense that we will continue to have a large GCD as the core element. Of course, with improvements in the units for Ray Tracing, one of the weak points of the current architecture and with the ability to achieve higher results in certain consumptions. That is, we don’t see AMD making a complex chip out of multiple chiplets.