Next-Gen Nvidia Data Center GPUs
Known Nvidia leaker kopite7kimi—who originally revealed the name of the company’s next GPU architecture a few years ago—says Nvidia’s next-gen data center GB100 Blackwell series graphics cards will have a similar core count as Ada Lovelace chips but will have “significant changes in its unit structure.”
An earlier tweet (via Hardware Info) indicates that the GB100 GPUs will use a chipset-based multi-chip module (MCM) instead of the current monolithic designs.
MCM is a sophisticated manufacturing technique that divides GPU components across distinct dies. The leaker doesn’t tell us what the configuration will look like—such as whether it will contain multiple GPU compute dies or just isolated memory dies alongside a single GPU core—nor does it indicate whether the GB100 will be the only Blackwell GPU with a multi-die design.
This departure from a monolithic focus would offer Nvidia’s data center GPUs a lot of breathing room for potential chip customization in order to compete with AMD’s Instinct MI300 series processors.
The red team’s own data center processors are packing a whole Zen 4 CPU inside an extremely enormous GPU.
AMD also already has some chipset gaming GPUs that we like so far, such as the AMD Radeon RX 7900 XTX, though its inaugural processors still don’t have multiple compute dies, just distinct compute, I/O, and memory dies.
Anyways, the processors rumored here are exclusively data center GPUs, which might not be particularly intriguing to us PC gamers. Though the Blackwell architecture is reportedly coming to our GeForce cards under the GB200 appellation, if Nvidia has discovered a method to make multi-compute dies function within a traditional game rendering pipeline, then we might still have something here to get very enthusiastic about.
Because that’s the difficult part. When it comes to data center GPUs, doing a whole load of serious number processing rather than traditional frame rendering is a lot simpler to do when distributed across multiple compute dies.
That’s why SLI and Crossfire multi-GPU solutions kept on going in the spirit of the computation space long after they were promptly executed in the gaming space.
That’s essentially what we’re looking at when it comes to multi-compute MCM graphics cards: an SLI/Crossfire solution embedded into a single processor. And there’s explicitly no mention here of whether Nvidia is planning to send that out to its next-gen GeForce cards.
We can be optimistic, but we’d presumably imply it’s unlikely in this next GPU iteration. But it’s been rumored that the next iteration of Nvidia gaming GPUs won’t happen until at least 2025—plenty of time to sort that all out, right Jensen?