The first performance benchmarks of NVIDIA's RTX A6000 which is the flagship workstation Ampere GPU-based graphics card have leaked out. The benchmarks were leaked by AdoredTV and show us what to expect from the fastest workstation graphics card in NVIDIA's stack which will be replacing its Quadro RTX 6000 solution from 2018.
NVIDIA RTX A6000 Flagship Workstation (GA102 GPU) Graphics Card's Benchmarks Leak Out, 11% Faster Than Its Predecessor
The NVIDIA RTX A6000 was announced earlier this month and after dropping the Tesla branding, NVIDIA has decided to drop Quadro branding from its professional workstation graphics cards too. The card will simply be called the RTX A6000 and is aimed at the workstation and content creation market segment.
NVIDIA RTX A6000 Graphics Card Specifications
Coming to the specifications, the RTX A6000 is powered by the full GA102 GPU core with 84 SMs or 10752 CUDA cores. We don't have the exact clock speeds since NVIDIA did not share the TFLOPs or Tensor numbers for the card yet but we do know that the Quadro RTX A6000 will offer higher clocks than the A40 due to its blower-fan design which offers higher cooling performance.
In terms of memory, the NVIDIA RTX A6000 features 48 GB of GDDR6 memory. The A6000 offers 768 GB/s speeds with its 16 Gbps memory dies. The RTX A6000 cards support vGPU with various graphics configurations starting at 1 GB up to the whole 48 GB VRAM buffer. Power is provided through the new EPS 12V 8-pin connector which is featured on the back of the card and delivers up to 300W of power to the GPU. Interconnect comes in the form of the latest NVLINK which offers 112.5 GB/s (bi-directional) speeds while the native PCIe Gen 4 interface provides a 16 GB/s link. Display options include four DisplayPort 1.4 on the A6000.
NVIDIA Workstation Graphics Card Lineup:
Graphics Card | RTX 6000 | RTX A6000 | Quadro RTX 8000 | Quadro RTX 6000 | Quadro GV100 |
---|---|---|---|---|---|
GPU | Ada Lovelace GPU | Ampere GPU | Turing GPU | Turing GPU | Volta GPU |
GPU Process | 5nm | 8nm | 12nm | 12nm | 12nm |
Die Size | 608mm2 | 628mm² | 754mm² | 754mm² | 815mm² |
GPU Cores | 18176 Cores | 10752 Cores | 4608 Cores | 4608 Cores | 5120 Cores |
Tensor Cores | 568 Cores | 656 Cores | 576 Cores | 576 Cores | 640 Cores |
Boost Clock | 2.50 GHz | 1.80 GHz | 1.77 GHz | 1.77 GHz | 1.62 GHz |
Single Precision | 91.1 TFLOPs | 38.7 TFLOPs | 16.31 TFLOPs | 16.31 TFLOPs | 16.66 TFLOPs |
Ray Tracing Spec | 210.6 TFLOPs | 75.4 TFLOPs | 10 GigaRays/Sec | 10 GigaRays/Sec | N/A |
VRAM | 48 GB GDDR6X | 48 GB GDDR6 | 48 GB GDDR6 | 24 GB GDDR6 | 32 GB HBM2 |
NVLINK VRAM | N/A | 96 GB With NVLINK | 96 GB With NVLINK | 48 GB With NVLINK | N/A |
Memory Bus | 384-bit | 384-bit | 384-bit | 384-bit | 4096-bit |
Memory Bandwidth | 960 GB/s | 768 GB/s | 672 GB/s | 672 GB/s | 870 GB/s |
TDP | 300W | 300W | ~225W | ~200W | 250W |
Launch Price | $6800 US | $4650 US | $10000 US | $6300 US | $9000 US |
Launch Date | Q1 2023 | Q4 2020 | Q4 2018 | Q4 2018 | 2018 |
NVIDIA RTX A6000 Graphics Card Benchmarks
In terms of performance, the benchmarks were reported within the SPECviewperf 2020 benchmark suite at a resolution of 3800x2120. Both the Quadro RTX 6000 and RTX A6000 were compared head on. The RTX A6000 scored a lead in almost all benchmarks with a maximum reported lead of up to 25%. On average, the RTX A6000 turned out to be 11% faster than the Quadro RTX 6000 which is a bit disappointing for a graphics card that features much better specifications and a better graphics architecture.
There's no explanation why the lead is just 11% over the predecessor but it could be due to several factors that range from drivers or poor optimization in SPECviewperf for the graphics card. NVIDIA will be releasing the RTX A6000 graphics card through its partners in mid of December with a wider GPU availability planned for early next year.