There's no denying that AMD's new Threadripper 9980X is insanely fast. With 64 cores, 350 watts of power, and the latest Zen 5 architecture, it's hard to go wrong. The problem with a 64-core Threadripper like the 9980X isn't having enough computing power. It's about finding the workloads that can actually take advantage of all those cores.

Threadrippers are great for things like virtualization and demanding multi-tasking, but I wanted to dig in to find the workloads that can actually take advantage of the full chip all on their own. There are several workloads where the 32-core Threadripper 9970X will perform just as well as the 64-core Threadripper 9980X. But, if your work often requires one of the workloads below, spending up on more cores is the way to go.

7 Blender

Blender isn't the last rendering app on this list; I'll let you know that upfront. Rendering is a heavily parallelized workload, so it makes sense that you'd see clear scaling from a 32-core CPU to a 64-core one. It's not one-to-one, but the Threadripper 9980X managed an average score that was 72% ahead of the Threadripper 9970X. For context, in a more benchmark-focused rendering app like Cinebench, the Threadripper 9980X managed less than half of that increase, even looking at multi-threaded performance.

This is an app that really wins on two fronts with the 9980X. Not only does it see a larger-than-average generational improvement, it also scales very well with the higher core counts available on Threadripper chips. GPU-based rendering in an app like Blender is still the way to go — the advantage of 64 cores, although present, will go down if you're using the GPU as the primary render engine — but rendering remains one of the clearest workloads where a high-core count CPU shines.

6 LAMMPS

You've probably heard of Blender, but I'd wager that most folks reading this have never heard of LAMMPS, or Large-scale Atomic/Molecular Massively Parallel Simulator. Everything you need to know is right there in the name. LAMMPS is an open-source molecular dynamics simulator that's designed for parallel operations. And whenever you have a demanding workload that's designed with parallelization in mind, you'll probably find an app that scales well with a 64-core Threadripper CPU.

Compared to Blender, the jump isn't as stark, but the Threadripper 9980X still manages 36% better performance in this workload compared to the 32-core Threadripper 9970X. That's about twice as high as the generational improvement of AMD's latest Zen 5-based Threadripper chips.

5 LLVM Clang

Code compilation is one of the workloads that's massively accelerated by higher core counts. LLVM Clang is just an example here, included as the compiler of choice in SPECworkstation 4 for building PyTorch, but you can bet that building just about any software from source will see a speed-up with a 64-core Threadripper. In this specific workload, the Threadripper 9980X posted results 34% ahead of the Threadripper 9970X, once again scaling beyond what the generational improvement offers with Zen 5 Threadrippers.

It isn't just building PyTorch with Clang, though. As you can see on Open Benchmarking, the Threadripper 9980X has already claimed the third-highest slot for compiling the Linux kernel, only outclassed by dual AMD Eypc CPUs packing up to 192 cores each. Taking both generational improvements and core count increases, the Threadripper 9980X manages to compile the kernel in about half the time of the Threadripper 7970X, which is a massive upgrade.

4 LuxCoreRender

I told you Blender wasn't the last renderer on this list. LuxCoreRender is an open-source, free, physically-based rendering engine that scales with higher core counts available on Threadripper chips, and you can actually use it as the renderer in Blender. Based on the results from SPECworkstation — which itself uses the LuxMark benchmark — the Threadripper 9980X came out about 48% ahead of the Threadripper 9970X. That's not quite as stark as the jump you see with Blender, but that's likely due to the way LuxCoreRender is designed.

Unlike Blender, LuxCoreRender is primarily a render engine, and it's designed around heterogeneous computing. You don't use a single, primary source of compute for rendering. Instead, LuxCoreRender leverages all the CPUs and/or GPUs available to it in order to massively speed up rendering. When looking at core count advantages, they show less clearly, as the GPU was constant between benchmark runs.

3 NAMD

NAMD, or Nanoscale Modecular Dynamics, or Not Another Molecular Dynamics Program, is, as you can probably deduce, another molecular dynamics program for scientific simulation. The tool, first released 30 years ago, was built by the Parallel Programming Laboratory at the University of Illinois Urbana-Champaign, and if you've been paying attention, you already know that "parallel" should set off alarm bells. Higher core counts scale incredibly well with this type of workload.

How well? According to SPECworkstation, the Threadripper 9980X is a clean 50% faster than the Threadripper 9970X. I'm not going to pretend to be a molecular scientist, but according to Puget Systems, NAMD is a heavily CPU-limited program. Although it's accelerated by CUDA, the majority of the workload still lands on the CPU. That makes a high core count all the more important.

2 Finance and options pricing

I'm using the sub-score for "options pricing" from SPECworkstation here because there aren't a ton of commercialized applications that run these types of financial simulations. Most of the applications, at least from what I can tell, are proprietary tools built by and for financial institutions. In options pricing, which looks at things like the Monte Carlo algorithm and Black-Scholes model, the Threadripper 9980X managed to outclass the 9970X by a massive 59%. And, believe it or not, that's not even the best showing for a 64-core CPU when it comes to financial workloads.

Looking at the Poisson distribution model, which attempts to find the probability of a given event within a specific timeframe, the 64-core CPU was a staggering 183% ahead.

1 SRMP

SRMP, in this case, stands for surface-related multiples prediction, which is an algorithm written by Evgeny Kurin that's used to process seismic data, particularly in the oil and gas industry. Seismic data is collected by recording the acoustics of the Earth's subsurface in order to find things like oil reserves, as well as analyize the ground. The problem with seismic data, as you can probably guess, is that it's messy. There's a ton of data that's recorded, and the raw data isn't perfect out of the gate for analysis.

That's where an algorithm like SRMP comes into play, cleaning up multiples in a predictive fashion, so the data is actually usable. Naturally, with such large amounts of data, you need a lot of compute for seismic data processing. That shows through clearly in my benchmark results, where the Threadripper 9980X ended up 48% ahead of the 9970X.