NVIDIA’s Titan RTX is a monster Turing card that costs US$2,499
NVIDIA’s Titan RTX is a monster Turing card that costs US$2,499
The NVIDIA Titan RTX is here, just a day after some prominent personalities and YouTubers teased the card in a handful social media posts and videos. This is NVIDIA’s most powerful Turing card yet, and the company even has a fitting nickname for it, fondly dubbing it the “T-Rex”.
But what a price the card comes with. If you thought the US$1,199 GeForce RTX 2080 Ti was expensive, the Titan RTX absolutely blows it out of the water. It costs a whopping US$2,499, which puts it well out of reach of the average consumer.
This is a prosumer card through and through, and it’s targeted at applications like AI, data science, and rendering. On the software side, NVIDIA is also supporting it with RAPIDS open-source software libraries.
The Titan RTX uses a fully-enabled TU102 GPU with 72 dedicated RT cores for real-time ray tracing and 576 tensor cores to help with machine learning. That's a step up over the GeForce RTX 2080 Ti, and the Titan RTX will also have more CUDA cores than its GeForce counterpart.
Here’s an overview of its specifications:
NVIDIA Titan RTX | NVIDIA Titan V | NVIDIA GeForce RTX 2080 Ti Founders Edition | |
---|---|---|---|
GPU | TU102 | GV100 | TU102 |
Process | TSMC 12nm | TSMC 12nm | TSMC 12nm |
CUDA cores | 4,608 | 5,120 | 4,352 |
Tensor cores | 576 | 640 | 544 |
RT cores | 72 | - | 68 |
Texture units | 288 | 320 | 272 |
Base/boost clock | 1,350MHz/1,770MHz | 1,200MHz/1,455MHz | 1,350MHz/1,635MHz |
Memory | 24GB GDDR6 | 12GB HBM2 | 11GB GDDR6 |
Memory clock | 14Gbps | 1.7Gbps | 14Gbps |
Memory bus width | 384-bit | 3,072-bit | 352-bit |
Memory bandwidth | 672GB/s | 653GB/s | 616GB/s |
Tensor performance | 130 TFLOPS | 110 TFLOPS | 57 TFLOPS |
Transistor count | 18.6 billion | 21.1 billion | 18.6 billion |
TDP | 280W | 250W | 260W |
Price (USD) | $2,499 | $2,999 | $1,199 |
The Titan RTX also utilizes the same dual-axial cooler design as the GeForce RTX Founders Edition cards, but the aluminum shroud is now drenched in gold, instead of silver. It's certainly a more luxurious look, and the Titan RTX is fully equipped to tear its way through a range of machine learning tasks. In fact, it's so powerful that NVIDIA says that it can handle real-time 8K video editing.
However, there's also one key aspect that sets it apart from the GeForce RTX 2080 Ti, even though the two share the same TU102 GPU. NVIDIA's GeForce Turing cards are limited to half-speed throughput in FP16 mode, even though Turing can accumulate at FP32 for still greater precision. This limitation doesn't exist on the Titan RTX, so it can handle full-speed FP32 accumulation throughput on its tensor cores.
Another area where the card stands out is its memory capacity. The 24GB of GDDR6 memory is more than double that on the GeForce RTX 2080 Ti. It's also far more than gamers will ever need, so it's clear that this is a card built for handling large datasets instead of gaming.
Finally, the card gets the same selection of display outputs as the Turing GeForce cards. This means three DisplayPort 1.4, one HDMI 2.0b, and one USB-C connector that supports DP alternate mode and VirtualLink.
The Titan RTX draws power from two 8-pin connectors, so you can definitely install one in your gaming rig, if you have the budget for it. That said, most gamers will probably still want to stick to the GeForce RTX 2080 Ti.