Computex 2019: NVIDIA Unleashes Turing-Based RTX Quadro Mobile Workstation GPUs

NVIDIA brings RTX technology to creative professionals on-the-go with new Quadro RTX graphics for mobile workstations.
In what is surely the first of many announcements from graphics giant NVIDIA at Computex in Taipei, Taiwan, NVIDIA is introducing a new lineup of mobile Quadro RTX graphics and mobile Turing-based (sans ray tracing) Quadro graphics cards that will be integrated into mobile workstation solutions from its OEM partners (Dell, HP, and Lenovo among others) throughout this year. The lineup includes the Quadro RTX 5000, 4000, and 3000 as well as Quadro T2000, T1000, P620, and P520 GPUs. The Quadro RTX GPUs are the full Turing setup with RT cores for ray tracing and Tensor cores for AI and compute acceleration while the non RTX Quadro for Mobile Workstation graphics appear to be similar in nature to the smaller GTX 1660 Ti in that they lack the extra RT and Tensor cores but do offer up the newer (and slightly more) Turing-based CUDA cores than their existing P200, P1000, P600, and P500 predecessors.
Further, while the newer Turing-based lineup is paired with the same memory type (GDDR5 for non RTX) and memory bus width, memory bandwidth has been increased suggesting NVIDIA is using faster clocked memory. The Quadro RTX graphics do get GDDR6 memory and the top-end Quadro RTX 5000 gets 16 GB of GDDR6 on a 256-bit bus (448 Gbps) making it well-suited to handling large data-sets and mobile rendering, design, and visualization workloads working with physically-accurate models in real time for things like architecture and interior design. Engineering, design, and traditional workloads like oil field mapping along with AI/deep learning, and other GPGPU tasks should run on these new mobile GPUs.
On the desktop side of things, NVIDIA already has the Quadro RTX 6000 (TU102) and Quadro RTX 5000 (TU104), and the new mobile Quadro RTX 5000 appears to match up well with its desktop-based namesake offering up the same number of CUDA, Tensor, and RT cores and paired with the same GDDR6 memory on a 256-bit bus (448 GBps) with the only difference being clockspeeds resulting in reduced performance (e.g. 9.4 vs 11.2 TFLOPS SP) but at a much lower 80-110W TDP (vs 230W) better suited for a portable system. The Quadro RTX 4000 doesn’t match up with a [consumer] desktop part (yet?) but the Quadro RTX 3000 is, at least on paper specifications-wise, a lower clocked (and likely better binned) RTX 2070. Well, okay, that’s not entirely true, there is a desktop Quadro RTX 4000 but the mobile variant announced today that is closest spec’d to it would be the Quadro RTX 3000 but even then the amount of memory is different… and let’s just back away slowly from the naming cluster that is desktop and mobile graphics SKUs!
In any event, the new Turing chips appear to offer up a bit more performance at similar TDPs to their Pascal-based predecessors across the mobile workstation lineup with the new mobile Quadro RTX further sprinkling in some real time ray tracing and AI acceleration hardware as well as a respectable increase in CUDA core count for the new vs old top end part. NVIDIA’s new mobile Quadro cards will be featured in systems launching throughout the year from various OEM/SI partners. Exact pricing and availability was not disclosed, but expect enterprise-level pricing on these systems with professional graphics and the costs that come with that support. There may well be systems launched later this week as Computex gets under way, so stay tuned if you are interested in tracing your rays on the road!