AMD is confident it has a secret weapon that can beat Nvidia – but we’re worried it might be getting a little cocky.
That secret weapon (which admittedly isn’t that secret, as AMD is bragging about it) is the use of GPU chiplets in their upcoming RDNA 3 graphics cards. Instead of using a single ‘monolithic’ chip for your GPU, the AMD will use multiple chiplets that combine to make a large – and very powerful – chip with a high number of GPU cores.
AMD actually does something similar with its Ryzen CPUs, but this will be the first time a chiplet design has been used for GPUs – something the company thinks will give it a big advantage over its main competitor Nvidia, which maintains a monolithic design for its CPUs. upcoming Lovelace RTX 4000 GPUs.
As Tweaktown Reports (opens in new tab), Sam Naffziger, AMD senior vice president, corporate member and product technology architect, explained in an interview with VentureBeat that “Nvidia certainly hasn’t jumped on the chiplet bandwagon yet. We have a big advantage there and we see big opportunities with that. They will be forced to do so. We will see when they deploy it.”
That’s certainly a bold statement and points to AMD being pretty confident that its chiplet design will give it the performance advantage over Nvidia with its high-end GPUs.
The future of GPUs
In fact, Naffziger clearly thinks that chiplet designs are the future of GPUs, and it’s a matter of when – not if – Nvidia adopts chiplets.
With the physical limitations of single chips clearly a concern – after all, chips can only get so small and contain so many transistors – AMD is looking at ways to keep making more powerful components.
Multi-chiplet designs are certainly an answer to this question, but as Naffziger explains in the interview: “We’ve seen this happen for a long time. As I said, delivery times are long. We’ve been investing in things like Infinity Cache, chiplet architecture and all these approaches that explore new dimensions to keep the gains coming.”
Analysis: Don’t get too cocky
While AMD’s confidence in chiplet designs and the performance they can bring to GPUs is certainly exciting for the future of PC gaming in particular, we don’t want them to get too cocky.
Nvidia should never be underestimated, and AMD has a long way to go if it wants to seriously disrupt Team Green’s dominance when it comes to GPU sales.
On the power front, we can definitely see the potential of multi-chiplet designs, and the prospect of AMD beating Nvidia when it comes to power and performance is certainly exciting.
But if AMD gets complacent, it could stumble at this crucial juncture, and while everyone loves a good underdog, AMD hasn’t been able to convince most gamers to adopt their GPUs. An arrogant attitude may not help (although some may also find it attractive).
Additionally, AMD and Nvidia are joining the GPU fight for Intel, and while Team Blue hasn’t made much of an impact, it can still shake things up substantially.
And while Nvidia may be ignoring chiplets, Intel certainly isn’t. As Naffziger points out, “Intel has certainly leapfrogged [chiplet design]. Ponte Vecchio is the poster boy for chiplet extremes.”
Intel’s Ponte Vecchio architecture will power Intel’s high-performance GPUs and use a chiplet design that will use multiple parallel processors to boost performance.
Could we get into a situation where AMD vs Intel fight for the GPU performance crown while Nvidia is left behind?
We don’t think so – again, never underestimate Nvidia. We’re sure it will still produce some of the best graphics cards when the RTX 4000 series launches later this year (even if it’s delayed, as the rumors suggest).