One of the problems of the system design is the energy cost of data transfer, which is calculated by the formula P = V2* C * f, where V is the voltage squared, C is the ability of transistors to hold a charge and f is the frequency or clock speed.
As we increase the clock speed, the voltage increases and so does the power consumption. This is a problem that RAM and VRAM designers face and which GDDR6 has had a negative impact, causing the number of chips that hit high speeds to be lower than those that come out in each wafer.
When GDDR6 was introduced, three different bandwidths were announced: 12 Gbit / s, 14 Gbit / s and 16 Gbit / s. Which ones The 14 Gbps setting has been standardized on a large number of devicess, because it is the most used, either in new generation consoles or in graphics cards of medium-high cards.
However, the question that comes to my mind is: What about the 16 Gbps GDDR6 setting? Being the fastest, it should be the one that gets used the most, but it’s almost completely new and apart from AMD’s RX 6000, we haven’t seen it used in other devices. The problem? The AMD RX 6000 experienced a huge stock shortage when it launched.
Considering that TSMC doesn’t seem to have any issues with its 7nm manufacturing node, then the only logic we have left is that high-speed GDDR6 memories have an issue that makes them scarce.
GDDR6X is also a sign of this
The fact that NVIDIA decided to do without high speed GDDR6 memory for the RTX 3080 and RTX 3090 gives us a lot clues about a problem GDDR6 may have and is that said VRAM could not support a clock speed in its communication interface.
GDDR6X has a much lower bus clock speed than GDDR6, therefore the energy consumed by the energy transfer is much lower. But the key is not only in this point but in the fact that it is very likely that GDDR6 memory will not be able to maintain a sufficiently high clock speed on its interface for long periods of time.
All GDDR6 memory chips are performed the same, but stress tests are performed to see if a memory chip is valid for a given speed, otherwise it is tested at a lower speed. The bottom line is that there are few GDDR6 16Gbps or higher memory chips, which affected the stock of the RX 6000 on the one hand and led NVIDIA to choose GDDR6X on the other.
AMD will eventually adopt GDDR6X memory
At the moment, GDDR6X memory is only found in the RTX 3080 and RTX 3090, but that will eventually spread to the rest of the NVIDIA lineup, at more modest clock speeds and eventually to the RTX 3000 Super. As for AMD, something similar will happen and that is precisely the amount of GDDR6X chips that they can get with a performance similar to a 16Gbps GDDR6 is higher, which would put an end to the stock issues.
The only obstacle? AMD has gone for higher densities, so until GDDR6X with 2GB of density per chip doesn’t show up they won’t make the leap, but it’s possible we’ll see a review of the RX 6000 with that. type of memory.
GDDR6 issues affect consoles
The consoles use GDDR6 at 14 Gbps, a configuration which is widely used and which has high demand and high manufacturing capacity.
It is said that it is the consoles that have limited the amount of GDDR6 memory for the graphics cards, given the imminent launch of these, but in the case of the AMD RX 6000 that are on the market, the lack of stock cannot not be due to it. because consoles do not use GDDR6 memory at the same speed.
GDDR6 is going to be VRAM for low-midrange
The conclusion we draw from all of this, although it may seem rash to many, is that GDDR6 is going to have a much shorter lifespan than GDDR5 on PC or at least we’ll see it sharing space with GDDR6X for a long time, but being used in smaller GPUs.
As for the evolution of GDDR6X, nothing will happen to it as long as the bandwidth does not force to place this memory at clock speeds that place it under the same problems as GDDR6.
Which means you have to find other ways to scale the bandwidth of VRAM, since the traditional is reaching the limit.