AMD and NVIDIA have led the way, but they are more than likely behind in the race for MCM GPUs. And does Intel appear to be the latest player to arrive, but the first to move forward in this regard with their Xe graphics cards. Will this tile or tile or chiplet technique be similar to the now defunct SLI or Crossfire?
SLI, CrossFire, Dual GPU and MCM, different concepts that do not lead to the same performance
It must be clear that these four concepts are not the same, they are not made in the same way and above all, the performances are very different between them. For example, SLI and CrossFire have always needed the PCIe bus to exchange data, textures, synchronizations etc … Even via an interconnect bridge as we all know.
This creates synchronization issues due to render times between GPUs, so in many cases we have had the so-called Dual GPUs, which are two chips on a single PCB interconnected by it, each with its VRAM. dedicated and its resources. It was the most advanced stage of SLI or CrossFire, but in terms of consumption it was complicated and in terms of cooling, let’s not even speak, really a challenge for the engineers.
Now MCM is coming and as such there are other challenges to overcome, and of course they are not minor. We are talking about using different chips on the same substrate, not like in Dual GPU which each had its direct solder to the PCB, no we are talking here about a common interposer for the chips.
Size matters, consumption and data buses too
One of the paradigms facing MCM-based GPUs is precisely the overall size of the chip generated by the junction of the SoCs in the interposer. Intel has shown the way, with a single chip that incorporates a curious IHS to dissipate heat through direct contact.
And, the watts to be dissipated are going to be very high in all cases of high-end GPUs. It must be taken into account that the use of the interposer consists precisely in reducing the consumption of data generated at the interconnection of each chip, neither more nor less, in addition to being able to introduce the corresponding clear wiring.
The second issue is latency, as games are very sensitive in these cases and performance can crash. Game engines are going to need to be updated, therefore APIs will be of great importance to hardware development companies.
And it is that although the interconnection of modules / chips / tiles / SoC should be the most transparent for engines and APIs, as well as for operating systems, which is very important for correct performance. On top of that, the scalability with this type of GPU will be much higher, as we might even see multiple types of cores or chips in the immediate future, creating a new paradigm very similar to its appearance. big little.
So no, the four concepts don’t even look alike from a distance, MCM architectures are the future and whoever successfully implements them first can take the lead because of the full advantage it brings.
