According to a post by well-known leaker Kepler on X, AMD will be using “slow” 18Gbps GDDR6 memory modules in its next generation RDNA 4 GPU line (RX 8000). If true, this will be the fourth RDNA architecture in a row to use GDDR6 memory. However, if RDNA 4 only uses 18 Gbps chips, it will actually be a step backwards.
As with all leaks, this information should be taken with a grain of salt. It’s hard to believe that AMD will cap its RDNA 4 graphics cards at a maximum speed of 18 Gbps, considering that some of the current RX 7000 series GPUs already use faster modules with 20 Gbps bandwidth.
It’s possible AMD will opt for slower modules if it simply doesn’t need more bandwidth, but I suspect that won’t apply to at least some GPUs.
AMD has already abandoned faster GDDR6X solutions in favor of more power-efficient (but slower) GDDR6 memory modules. The company was also the first to use a large L3 cache, called Infinity Cache, in its last two GPU architectures to compensate for the lack of raw throughput improvements.
However, there are good reasons to upgrade to GDDR7, regardless of how much cache is available. GDDR7 is the industry’s next-generation graphics memory architecture that will potentially double the memory bandwidth of GDDR6 (up to 40 Gbps vs. 20 Gbps) while also having 50% greater data transfer efficiency.
AMD’s decision to go with GDDR6 will depend on how much memory bandwidth it needs and how much money it can save on design by using the existing memory standard.
