We picked 4 GPUs to compare, the RTX 4090, RTX 6000 Ada, RTX 3090, and RTX A6000, which are the most discussed within the professional fields. The RTX 3090 and RTX A6000 were the most popular GPU in the last generation (Ampere architecture), whereas the RTX 4090 and the RTX A6000 could be their descendants. With the found information, hopefully, we could give the broad consumers a better view of which GPU fits your scenario. Note that all these test results in this blog are based on a single GPU.
|
|
RTX 4090 |
RTX 6000 Ada |
RTX 3090 |
RTX A6000 |
|
GPU |
AD102 |
AD102 |
GA102 |
GA102 |
|
CUDA core |
16,384 |
18,176 |
10,496 |
10,752 |
|
Tensor core |
512 |
568 |
82(Gen 3) |
336(Gen 3) |
|
RT core |
128 |
142 |
328(Gen 2) |
84(Gen 2) |
|
Memory |
24GB GDDR6 |
48GB GDDR6 |
24GB GDDR6 |
48GB GDDR6 |
|
ECC |
Yes |
Yes |
No |
Yes |
|
Memory
Bandwidth (GBPs) |
1008 |
768 |
936 |
768 |
|
TDP (W) |
450 |
300 |
350 |
300 |
|
System interface |
PCIe 4.0 x16 |
PCIe 4.0 x16 |
PCIe 4.0 x16 |
PCIe 4.0 x16 |
|
MSRP (USD) |
1,599 |
6,800 |
1,499 |
4,650 |
The RTX 4090 and RTX 6000 Ada are based on Nvidia’s Ada Lovelace architecture, but the RTX 3090 and RTX A6000 are on the last generation Ampere architecture. In addition, the Ada Lovelace GPU adopts TSMC’s 4nm manufacturing process compared to Ampere’s Samsung 8nm process. The amount of CUDA cores increases significantly. Furthermore, the Ada GPU carries fourth-generation Tensor Cores and third-generation RT cores.
In terms of deep learning, the performance between RTX A6000 and RTX 3090 can say pretty close. According to lambda, the Ada RTX 4090 outperforms the Ampere RTX 3090 GPUs. So, we may infer the competition is now between Ada GPUs, and the performance of Ada GPUs has gone far than Ampere ones. However, so far, we have not obtained much information about the deep learning benchmarks of the RTX A6000. We will stay tuned for the latest benching results and update you anytime. Another consideration is the price. A single RTX 6000 Ada costs $6,800, which is more than 4x more expensive than RTX 4090. Hopefully, RTX 6000 Ada could have an even more powerful performance than other Ada GPUs.
Let us take a look at the difference in RT cores. The third-generation RT cores that the Ada GPU carries optimize the ray-tracing performance with the new Opacity Micro Map (OMM) Engine and new Displaced Micro-Mesh (DMM) Engine. These two new features allow the ray tracing job to run more real-time efficiently with a faster build time and 20X storage-saving power. All Ada and Ampere GPUs do decent ray tracing work, but the two specific parameters significantly heighten the efficiency of rendering an image. Below are some rendering benchmark scores.
Below are some rendering benchmark scores. This time, a single RTX 4090 outperforms the Ampere GPUs in rendering by around 155 ~ 200%. We can say the performance of the Ampere GPUs is far behind.
Benchmark results retrieved from bizon-tech.com
People still using the RTX 3090 for rendering can consider purchasing the new Ada-based GPUs if pursuing a higher render efficiency.
Last, we gathered some gaming performance data from the internet of the graphics cards we listed.
Benchmark results retrieved from gpucheck.com
This graph shows RTX4090 performs about 50% better than RTX 3090 on multiple-game performance. There is an improvement from Ampere GPUs to Ada GPUs (RTX 4090). Unfortunately, we still have not got much gaming data on the RTX A6000 (Ampere and Ada), but after all, we would not expect people to use the RTX A6000 for gaming due to its seven grand price tag. Again, the Ada GPUs (RTX 4090) win in gaming performance. However, the TDP comes to an extremity with 450 W, meaning that most people would need an even larger power supply unit to support this monstrous power consumption. The cooling design of the environment also needs to be considered, which may further drain your wallets.
From all the information we gathered, we see that the Ada GPUs perform better than the last-generation Ampere GPUs in every aspect.
Other references
https://bizon-tech.com/gpu-benchmarks/NVIDIA-RTX-3090-vs-NVIDIA-RTX-A6000/579vs585#benchmarks
https://www.techpowerup.com/gpu-specs/geforce-rtx-4090.c3889
https://lambdalabs.com/blog/best-gpu-2022-sofar
https://lambdalabs.com/blog/nvidia-rtx-4090-vs-rtx-3090-deep-learning-benchmark