Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.
Nvidia RTX 4060 Ti 16GB AI Performance
GPUs are also used with professional applications, AI training and inferencing, and more. Along with our usual professional tests, we've added Stable Diffusion benchmarks on the various GPUs. AI is a fast-moving sector, and it seems like 95% or more of the publicly available projects are designed for Nvidia GPUs. Those Tensor cores aren't just for DLSS, in other words. Let's start with our AI testing and then hit the professional apps.
We're using Automatic1111's Stable Diffusion version for the Nvidia cards, while for AMD we're using a recent Nod.ai Shark variant — we retested the AMD cards for this review using build version 20230713_819 for testing, as results have improved substantially compared to our previous testing. The Nvidia GPUs were tested after replacing the default CUDA DLL files with newer versions, as recommended by Nvidia.
Intel meanwhile has finally added an OpenVINO fork of Automatic1111's webui, which shows about a 50% improvement over our previous testing. Except, we had to run with the Stable Diffusion v1.5 model as the v2.1 model (base and otherwise) didn't work or generated garbage outputs. Also, going beyond a 720x720 resolution on the A770 16GB card resulted in various problems (this is a known issue), so we still don't have 768x768 results for Intel's GPUs.
While the AMD and Intel GPUs improved in performance, the overall standings remain unchanged. Nvidia's RTX GPUs rank at the top, with the Arc A770 16GB landing just behind the 4060 Ti now, and AMD's GPUs occupy the bottom portion of the charts. Note that we haven't seen any major improvements in Nvidia GPU performance since our previous testing, so other than adding tests with the 16GB card, those results haven't changed.
As we saw with many of the gaming benchmarks, the new 4060 Ti 16GB often falls behind its 8GB sibling. Whether that's due to the difference in cards, VBIOS, VRAM, or something else isn't clear. We'll see on the next page that, despite the higher boost clocks on specs, the Gigabyte card actually runs at generally lower clocks than the 4060 Ti Founders Edition, and that could explain the minor gap in performance.
There are other AI workloads, some that utilize very large LLMs (Large Language Models). ChatGPT and its ilk for example need hundreds of gigabytes of memory. Locally running chatbots meanwhile often have to sacrifice quality due to VRAM constraints. It's harder to quantify how much doubling the VRAM can help, but the 16GB card at least has the option to use larger models. That's a pretty niche market, but if you want an Nvidia card with 16GB or more VRAM, most other qualifying GPUs cost twice as much as the 4060 Ti 16GB.
Nvidia RTX 4060 Ti 16GB Professional Workloads
SPECviewperf 2020 v3 consists of eight different benchmarks, and we use the geometric mean from those to generate an aggregate score. Note that this is not an official score, but it gives equal weight to the individual tests and provides a nice high-level overview of performance. Few professionals will use all these programs, however, so it's often more important to look at the results for the application(s) you plan to use.
Nvidia's RTX 4060 Ti 16GB generally matches the 8GB, as expected. There's one instance where the added memory does seem to help: 3dsmax-07 has the 16GB card performing 16.5% better than the 8GB card. Most of the remaining tests slightly favor the 8GB model, due once again presumably to the slightly higher boost clocks. Generationally, it's also worth noting that the RTX 3070 still typically beats both 4060 Ti variants.
If you're looking to use consumer hardware on these apps, AMD's GPUs generally provide a much better value. They likely wouldn't match actual professional GPUs running the pro / Studio drivers in some tests (i.e. something like the AMD Radeon Pro W7600), but there's not a single case where the RTX 4060 Ti 16GB can beat the RX 6750 XT, never mind the RX 6800 and RX 6800 XT.
Moving on to 3D rendering, Blender repeats the same story we've seen throughout this review. The RTX 4060 Ti 16GB comes in slightly behind the 8GB card, though this time it's at least able to surpass the RTX 3070 by 10%, and it's 2.35 times as fast as the RX 6800 XT. As good as the AMD GPUs look in SPECviewperf, then, they look equally poor in Blender.
Our final two professional applications only support ray tracing hardware on Nvidia's GPUs. OctaneBench puts the RTX 4060 Ti 16GB a few percent behind the 8GB card as well as just behind the RTX 3070, yet again. V-ray on the other hand shows some performance gains using the RTX path, where the 16GB card is 8% faster than the 8GB card. That doesn't apply when using the CUDA path, however.
- MORE: Best Graphics Cards
- MORE: GPU Benchmarks and Hierarchy
- MORE: All Graphics Content
