PC enthusiasts know that if you want the best gaming CPU, you can't make a mistake with the Ryzen7 7800X3D.Its powerful gaming chops have been helped a lot by the addition of 64MB of 3D V cache, so when AMD announced the new Zen5 architecture, many wanted to have information about the Ryzen9000X3D model. It never happened, but AMD says it's "actively working on some really cool differentiators" to make 3D V-cache "even better.""

At Computex2024 in Taiwan, we had a chat for a while with Donny Woligroski, senior Technical Marketing Manager at AMD. When we talked about the plans for the Zen5 chip's 3D V-cache version, he was particularly keen on this technology.

"X3D stuff, we have to say a lot about it. The best part about it is that we are not just resting on laurels. We're improving what we can do with X3D, it's really exciting and I'm very looking forward to talking to people about it.

Woligroski could not get into the details, but continued to talk about what AMD is doing in terms of 3D V cache. "Can only fashion, hi, we are also a chip to X3D. We are actively working on really cool differentiation to make it even better. We are working on X3D and are improving it.

It may all sound like a blaster about nothing, but when AMD says, "No, just use the same thing as last time," and so on, at least, it's said to be working on making it better. The obvious question to ask is what these differentiators are and how we can improve our current 3D V cache.

One of the possible answers to the first question is that AMD may be working on implementing 3D V-cache of various sizes to increase the level of differentiation between different X3D models. At the moment, whether it's the Ryzen5 5600X3D, the Ryzen9 7950X3D, or the EPYC9684X with 1GB of L3 cache, the size of the 3D V-cache slice is always 64MB Products that are particularly lacking in V-cache, like the RYZEN8040U and the new Ryzen AI300 series. This is the Apu. These processors have decent integrated GPUs, but their performance is somewhat hampered by the relative lack of memory bandwidth. Having an extra level of cache between the GPU and VRAM, as shown on RDNA2 and RDNA3 graphics cards, is very helpful here.

Alternatively, AMD may be considering coupling smaller L3 slices to the underlying Ryzen X3D chips to make the ryzen9 version even more prominent, but in general, ryzen and EPYC Cpus require additional slices of the L3 cache to be placed on the surface of 1 or more Ccd (Core Complex Dies). It is combined. The extra transistors in this cache not only increase power consumption, but also increase the thermal resistance of the entire chip, forcing AMD to reduce the boost clock of the X3D version of the normal Cpu.

The first generation 3D V-cache used on the Zen3 processor has an area of about 41 square millimeters, which AMD was able to reduce slightly to the Zen4's 36 square millimeters. Both SRAM chiplets are manufactured by TSMC on its N7 process nodes, and while SRAM is not much expanded on smaller nodes, AMD may choose to use N3 for the 3rd generation.

Doing so could cause a larger amount of cache to be crammed into the same area or 64MB to still die thin. This means that you can set the boost clock a little higher than normal.

Whatever AMD is up its sleeve for the third generation 3D V-cache, you probably won't hear anything about the Ryzen X3D model until later in the year. The Zen4 was first announced in 2022/8, and the first X3d model, the Ryzen9 7950X3D, appeared almost 6 months later in 2023/2.

Assuming the Zen5 follows a similar timeline, the first Ryzen9000X3D chip should be available at the end of the year. Also, the Zen5 was announced a little earlier than expected, so the ryzen7 9800X3D could be launched by the fall and we're really looking forward to it.