© alexgo.photography / Shutterstock.com
Slated for a December release, Nvidia’s new RTX 6000 Ada Generation is set to join its family of workstation GPUs: the RTX A6000, RTX A5500, and RTX A5000. Announced at Nvidia’s GTC 2022 conference, along with the GeForce RTX 4080 and 4090, the RTX 6000 Ada is said to provide a performance boost of up to 4x compared to its predecessor, the A6000.
While certain specs and pricing remain elusive, we’re going to discuss everything we know so far about this GPU from the king of graphics, including whether it’s a smart choice for you to upgrade.
RTX 6000: Overview
Mostly intended for professional workflows, such as content creation, simulation, and GPU rendering, the RTX 6000 is the first workstation GPU based on Nvidia’s new Ada Lovelace architecture. As with their GeForce RTX 40 Series, this architecture offers third-generation RT cores and fourth-generation Tensor cores, both allegedly providing at least 2x the throughput of previous generations.
The RT cores aim to speed up the rendering of ray-tracing motion blur, as well as the workloads. These include rendering movie content and architectural designs. Meanwhile, the Tensor core delivers faster AI computation, such as the execution of mixed floating point and integer calculations.
Along with simulation and content creation, the RTX 6000 also brings big benefits to data science, as the huge 48 GB DDR6 memory allows the user to handle massive datasets with ease. The RTX 6000 also comes with full compatibility with the latest AMD and Intel CPUs, thanks to the fourth-gen PCIe support.
Just like all their GPUs based on the Lovelace architecture, the RTX 6000 incorporates DLSS 3, also known as DLSS Frame Generation. This technology advanced over DLSS2 by using AI to construct entirely new frames, rather than just pixels. It analyzes the difference between the old frame and the generated frame in order to determine how a scene is changing.
- NVIDIA Ampere Architecture-based CUDA Cores - Double-speed processing for single-precision floating point (FP32) operations and improved power efficiency provide significant performance improvements...
- Second-Generation RT Cores - With up to 2X the throughput over the previous generation and the ability to concurrently run ray tracing with either shading or denoising capabilities, second-generation...
- Third-Generation Tensor Cores - New Tensor Float 32 (TF32) precision provides up to 5X the training throughput over the previous generation to accelerate AI and data science model training without...
- Third-Generation NVIDIA NVLink - Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate graphics and compute workloads and tackle larger datasets.
- 48 Gigabytes (GB) of GPU Memory - Ultra-fast GDDR6 memory, scalable up to 96 GB with NVLink, gives data scientists, engineers, and creative professionals the large memory necessary to work with...
This allows the generation of sharper images at a lower resolution. And while there’s no official word on how this will benefit professional use, it’s likely it could have positive implications in applications such as computer-aided design (CAD), where the processes tend to be limited by the CPU.
Nothing is confirmed as far as pricing goes, but we can expect to shell out a small fortune for this GPU since even the previous RTX A6000 GPU still retails at around $5,000. Given the general price increases with the Ada range of GPUs over their predecessors, the Ampere GPUs, it’s likely the RTX 6000 will command a mammoth cost even higher than the A6000.
Nvidia’s Workstation GPUs: Side-by-Side Comparison
| RTX 6000 | RTX A6000 | RTX A5500 | RTX A5000 | |
|---|---|---|---|---|
| Architecture | Ada Lovelace | Ampere | Ampere | Ampere |
| CUDA cores | 18,176 | 10,752 | 10,240 | 8,192 |
| Tensor cores | 568 | 336 | 320 | 256 |
| RT cores | 142 | 84 | 80 | 64 |
| GPU memory | 48GB GDDR6 | 48GB GDDR6 | 24GB GDDR6 | 24GB GDDR6 |
| Graphics bus | PCIe 4.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 |
| TDP | 300W | 300W | 230W | 230W |
| Size | 4.4″ x 10.5″ Dual slot | 4.4″ x 10.5″ Dual slot | 4.4″ x 10.5″ Dual slot | 4.4″ x 10.5″ Dual slot |
As we can see from the table, several parameters are unchanged, such as the size, power consumption, PCI support, and memory, so factors like data transfer speed and 3D rendering will not see any improvement. The CUDA core, Tensor core, and RT core counts are almost double that of the A6000, though.
That’s why we expect to see vast improvements in virtually all professional applications without an accompanying increase in size or power consumption, which is definitely a big bonus.
Performance boosts will mostly be seen in any situation where the workstation is carrying out multiple calculations, as well as adapting calculations dynamically. These include 3D CAD and CAE (computer-aided engineering), as well as virtual prototyping and complex content creation.
How Does the RTX 6000 Stack up Against the RTX 40 Series?
| RTX 6000 | RTX 4080 | RTX 4090 | |
|---|---|---|---|
| Architecture | Ada Lovelace | Ada Lovelace | Ada Lovelace |
| CUDA cores | 18,176 | 9,728 | 16,384 |
| Tensor cores | 568 | 304 | 576 |
| RT cores | 142 | 76 | 144 |
| GPU memory | 48GB GDDR6 | 16GB GDDR6X | 24GB GDDR6X |
| TDP | 300W | 320W | 450W |
CUDA Cores Make the Biggest Difference
- Has 16,384 NVIDIA CUDA Cores
- Supports 4K 120Hz HDR, 8K 60Hz HDR
- Up to 2x performance and power efficiency
- Fourth-Gen Tensor Cores that offer 2x AI performance
- Third-Gen RT Cores
- AI-Accelerated Performance: NVIDIA DLSS 3
- NVIDIA Reflex low-latency platform
At a glance, we can see that, while the RTX 40 Series is based on the same architecture as the RTX 6000, pretty much every other aspect of the specs is different. While the amount of tensor and RT cores is actually slightly greater with the RTX 4090, the sharp increase in the number of CUDA cores will lead to significantly better performance and precision, especially for multithreaded applications.
The intended usage of the RTX 6000 heavily relies on multithreaded applications, i.e. content creation, CAD, CAE, and simulation, whereas the RTX 40 GPUs were designed with running games in mind, which tend to be mostly single-thread intensive.
While the RTX 6000 likely has a lower clock speed than its RTX 40 counterparts, the power consumption is also lower, meaning the power efficiency achieved will be much better, particularly considering the higher core count. This will be appealing to professionals, who will likely encounter heavy workloads and usage.
GDDR6 vs GDDR6X: Which Is Better?
It’s also worth discussing the memory, as this varies between all the Ada-based GPUs. Since GDDR6X memory is technically superior to GDDR6, it would be tempting to say the RTX 40 Series wins in this regard. However, most games can be run at 4K resolution with both GDDR5 and standard GDDR6 memory. So, for gamers, GDDR6X might be overkill.
While GDDR6X theoretically offers greater rendering efficiency, oftentimes workflows are limited by the overall graphical memory of their GPU. That’s because rendering uses a lot more memory than standard gaming. Therefore, the 48GB memory available with the RTX 6000 will be vastly more useful in intensive applications, such as video editing, than a much lower amount of GDDR6X memory.
Who Is the RTX 6000 Ideal For?
Generally, gamers aren’t going to get much out of upgrading their system with the latest RTX 6000. This is mostly because it’s very rare for even newer games to use more than 12GB of VRAM. Considering this, the 16 GB and 24 GB available with the RTX 4080 and 4090, respectively, will be more than enough to handle gaming at even the highest resolutions.
Clock speed is also a huge factor in gaming performance, and this is very likely to be lower with the RTX 6000 when compared to the RTX 40 Series. Games also tend to be mostly single-threaded applications, so won’t suffer that much in terms of performance without the extra cores the RTX 6000 provides.
Where the RTX 6000 will shine, however, is in professional and enterprise workflows. Especially those involving CAD, CAE, elaborate video editing, and content creation, as well as virtual workstations and remote computing utilizing AI.
These applications are mostly multithreaded and graphically complex, so their performance and accuracy will be greatly enhanced by the increased number of CUDA cores and the huge GPU total memory, as previously discussed. Video editing, in particular, will massively benefit from it, since the RTX 6000 is reported to have 3x the video encoding performance of the RTX A6000.
If you’re on the lookout for your next workstation GPU where efficiency and precision cannot be compromised, the RTX 6000 should be on your radar, particularly if you’re a fan of Nvidia‘s pre-existing workstation GPUs.
