NVIDIA TITAN RTX, The Titan of Turing Graphics Cards Unveiled – 24 GB GDDR6 Memory, 130 TFLOPs AI Learning and $2499 US Price

Author Photo
Dec 3

NVIDIA has officially unveiled their Turing GPU based TITAN RTX graphics card. The NVIDIA TITAN RTX is the flagship prosumer graphics card that will be featuring loads of memory and 130 TFLOPs of Compute performance.

NVIDIA TITAN RTX, The Titan of Turing Graphics Cards Unveiled – $2499 US Price For 24 GB of GDDR6 Memory

Just like all of the recent TITAN series graphics cards, the TITAN RTX is aiming the prosumer market. The card is aimed at content creators, developers, and scientists who can utilize the cards 130 Tera-flops of performance for AI research, data science, and creative applications. The TITAN RTX utilizes the familiar Turing GPU architecture which we have seen on the GeForce RTX and Quadro RTX GPU lineup.

nvidia-geforce-gtx-max-q-laptops-ogimageRelated NVIDIA GeForce RTX 2060 Mobility Specs Leak, Features 6 GB GDDR6 VRAM – RTX 2080 Ti, RTX 2080 (Max-Q) and RTX 2070 (Max-Q) Mobility GPUs Leak Out Too

“Turing is NVIDIA’s biggest advance in a decade – fusing shaders, ray tracing, and deep learning to reinvent the GPU,” said Jensen Huang, founder, and CEO of NVIDIA. “The introduction of T-Rex puts Turing within reach of millions of the most demanding PC users — developers, scientists, and content creators.”


NVIDIA TITAN RTX Feature Highlight:

  • 576 multi-precision Turing Tensor Cores, providing up to 130 teraflops of deep learning performance.
  • 72 Turing RT Cores, delivering up to 11 GigaRays per second of real-time ray-tracing performance.
  • 24GB of high-speed GDDR6 memory with 672GB/s of bandwidth — 2x the memory of previous-generation TITAN GPUs — to fit larger models and datasets.
  • 100GB/s NVIDIA NVLink can pair two TITAN RTX GPUs to scale memory and compute.
  • Incredible performance and memory bandwidth for real-time 8K video editing.
  • VirtualLink™ port provides the performance and connectivity required by next-gen VR headsets.

Moving on to the specifications, the TITAN RTX uses the full TU102 GPU configuration with 6 GPCs, 36 TPCs, 72 SMs and 4608 CUDA Cores arranged within those SMs. There’s also 576 Tensor Cores and 72 RT cores that handle the bulk of AI/DNN and Raytracing workloads. The clock speeds will be maintained at 1350 MHz for the base and 1770 MHz for the boost frequency. The card features 24 GB of GDDR6 VRAM along a 384-bit bus interface that is clocked at 7.00 Gbps (14.00 Gbps effective) clock.

nvidia-gddr5x-memory-3Related NVIDIA GeForce GTX 1070 To Be Equipped With GDDR5X Memory, AIBs Custom Graphics Cards Leak Out

The card pumps out 672 GB/s of bandwidth and additionally comes with 6 MB of L2 cache. Power is provided through dual 8-pin connectors with a rated board TDP of 280W. The card also packs in the latest display connectivity with 3 DP, 1 HDMI, and a single USB Type-C port.

NVIDIA Titan Series

Graphics Card Name NVIDIA GeForce GTX Titan NVIDIA GeForce GTX Titan Black NVIDIA GeForce GTX Titan X NVIDIA Titan X NVIDIA Titan Xp NVIDIA Titan V NVIDIA RTX Titan
GPU Family Kepler Kepler Maxwell Pascal Pascal Volta Turing
GPU Core GK110 GK110 GM200 GP102 GP102 GV100 TU102
Process 28nm 28nm 28nm 16nm 16nm 12nm 12nm
GPU Cores 2688 2880 3072 3584 3840 5120 4608
Base Clock 836 MHz 889 MHz 1000 MHz 1417 MHz 1405 MHz 1200 MHz 1350 MHz
Boost Clock 876 MHz 980 MHz 1089 MHz 1531 MHz 1582 MHz 1455 MHz 1770 MHz
FP32 Compute 4.7 TFLOPs 5,6 TFLOPs 6.7 TFLOPs 11.0 TFLOPs 12.1 TFLOPs 15.0 TFLOPs 16.2 TFLOPs
Memory Bus 384-bit 384-bit 384-bit 384-bit 384-bit 3072-bit 384-bit
Memory Bandwidth 288.4 GB/s 336.0 GB/s 336.6 GB/s 480.4 GB/s 547.6 GB/s 652.8 GB/s 672 GB/s
TDP 250W 250W 250W 250W 250W 250W 280W
Price $999 US $999 US $999 US $1200 US $1200 US $3000 US $2499 US
Launch Year 2013 2014 2015 2016 2017 2017 2018

If we talk about performance, the card rocks 16.2 TFLOPs of FP32 compute which is higher than the Titan V’s 15.0 TFLOPs compute. It also comes with 11 Gigarays per second of ray tracing prowess, again, which is slightly higher than the 10 GRays/s of the Quadro RTX 8000 solution. NVIDIA states that the card would provide a 100 GB/s, full range NVLINK solution when two cards are paired for together in a multi-GPU environment.

We are looking at around 10% of graphics horsepower uplift versus the RTX 2080 Ti and slightly more to around 5% with the added memory bandwidth increase. The TITAN products were first aimed at the GeForce market but that changed a while ago when NVIDIA decided to turn them more appealing for the AI and research audience. This still didn’t stop the gaming masses from getting hands on these cards as their driver stack includes GeForce driver benefits.

In terms of design, the card features the same NVTTM cooler that we saw on the GeForce RTX 2080 and GeForce RTX 2080 Ti series cards. There are some slight changes which include a gold-colored shroud and more shiner edges which is obviously taken from the TITAN V. The shroud itself is a custom forged and die-cast aluminum cover with diamond-cut edge detailing which offers a really beautiful design with smooth edges.

The card also makes use of dual 13-blade axial fans which produce 3 times more airflow than the standard fan designs. The card makes use of vapor chamber cooling with 2 times more density and the same black coating for the heatsink fin array underneath the shroud. Finally, we have the 13-phase iMON DrMOS power supply which NVIDIA states is designed to overclock and provide the most stable operation under heavy workloads.

NVIDIA TITAN RTX Graphics Card Gallery:

Pricing wise- the card is expected to launch later this month in the US and Europe. Pricing is suggested at $2499 US which is slightly less than the $3000 US Titan V we got last year. Pricing is around twice as much as the RTX 2080 Ti ($1199 US) but you have to consider more than twice the amount of VRAM (11 GB vs 24 GB) and a full unlocked Turing TU102 GPU. Also, the Quadro RTX 6000 costs around $6000 US while offering similar specifications.