NVIDIA RTX PRO 6000 (MAX-Q): Blackwell with 96 GB GDDR7 at 300 to 600 Watts is officially here 33 comments
Nvidia is also introducing the new generation of Blackwell GPUs to its workstation portfolio. This is accompanied by the new RTX Pro branding. The professional graphics card fleet is led by the RTX Pro 6000 Blackwell (Max Q) with a whopping 96 GB GDDR7. These are also available for fanless servers. In notebooks, things are quieter.
Up to 96 GB of memory – rumors had already predicted such a capacity for the new Nvidia RTX Pro 6000 with Blackwell for the first time. NVIDIA RTX Pro for the Office Workstation
Like the GeForce RTX 5090 (test), the RTX Pro 6000 uses the GB202 GPU, but with a wider expansion stage (24,064 vs. 21,760 Fp32-Alus, +11%). The memory interface is still 512 bits wide and there is no longer a physical 96. Instead of 32 GB on the gaming graphics card, two adjustments are possible:
The RTX Pro 6000 is equipped with memory modules on both sides; two chips are each connected to a 32-bit controller using the shell process. Theoretically, this translates to 64 GB. Instead, 2 GB GDDR7 chips, like those in the GeForce RTX 5090 laptop GPU, are used on the 3 GB RTX Pro 6000 chips. This results in 96 GB.
The RTX Pro 6000’s cooling system is identical to that of the RTX 5090 (the “Double Flow-Through” principle), while the TDP is 25 watts higher at 600 watts.
Nvidia RTX Pro 6000 Blackwell Workstation Edition
The RTX 6000 Pro MAX-Q offers the same base model with a TDP of only 300 watts, which is a direct heat exhaust system with a radial fan.
Below the two new top models, NVIDIA, with the RTX Pro 5000, RTX Pro 4500, and RTX Pro 4000, are three other Blackwell workstation variants, the two smaller ones with GB203 GPUs (RTX 5080 / RTX 5070 TI, RTX 5090 Laptop GPU).
NVIDIA RTX PRO BLACKWELL FOR WORKSHOP RTX PRO 6000 MAX-Q RTX PRO 4500 RTX PRO 4000 GPU GB203 FP32-ALUS 24,064 14,960 FP32 Performance 125 Tflops 110 Tflops? ? ? RT Core Performance 380 Tflops 333 Tflops? ? ? AI exceeds 4,000 3,511? ? ? Interface Recording 512 Bit 384 Bit 256 Bit 192 Bit Memory 96 GB GDDR7 with ECC 48 GB GDDR7 with ECC 32 GB GDDR7 with ECC 24 GB GDDR7 with ECC Memory Throughput 1.792 GB/s 1.344 GB/896 GB/s ECC? TBP 600 W 300 W 200 W 140 W Dual Flow Dye D, 2 Radial Fan Slots, 2 Radial Fan Slots, 1 Slot Outputs 4 × Display Port 2.1b
The RTX Pro 6000 and RTX Pro 6000 Max-Q are expected to be available from PNY and TD Synnex as well as in workstations from Boxx, Dell, HP, Lambda, and Lenovo starting in May. NVIDIA RTX Pro for Server Workstations
A third variant of the NVIDIA RTX Pro 6000 is designed for use in servers. This variant features a fanless, two-slot design that can be configured from 400 to 600 watts. The server takes over its own cooling, and the memory is slightly lower (1.6 TB/s instead of 1.8 TB/s of storage bandwidth).
The server variant is expected to be available soon in systems from Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro. Later this year, the model is expected to be available in cloud instances from AWS, Google Cloud, Microsoft Azure, and Coreweave, as well as from other hardware manufacturers.
NVIDIA RTX PRO for Notebook Workstations
Later this year, the Nvidia RTX Pro is also expected to be included in the notebook. However, there will be no memory upgrade compared to the GeForce RTX 5000: to 24 GB, as the notebook’s GPU is no longer supported by the RTX 5090. Also in the case of shader configurations, RTX Pro 5000, RTX Pro 4000 and RTX Pro 3000 the RTX 5090, 5080 and 5070 TI GPUs are expected at the end of March, while RTX Pro 2000, 1000 and 500 should already give a perspective on the 5070, 5060 and 5050 GPUs.
Nvidia RTX Pro Blackwell for Laptop Workstations FP32-ALUs FP32-Power AI TOPS Memory interface through TGP DisplayPort PCIe RTX Pro 5000 Blackwell 10.496 49.8 1.824 24 GB GDDR7 (ECC) 256 Bit 896 GB/s 95-175W 2.1b 5.0 RTX Pro 4000 Blackwell 7.680 38.7 1.334 16 GB GDDR7 (ECC) 256 Bit 896 GB/s 80-175W 2.1B 5.0 RTX Pro 3000 Blackwell 5.888 29.1 992 12 GB GDDR7 (ECC) 192 BIT 672 GB/S 60-140W 2.1B 5.0 RTX PRO 2000 BLACKWEL 128 bit 384 GB/s 45-115W 2.1b 5.0 RTX Pro 1000 Blackwell 2.560 13.6 572 8 GB GDDR7 128 Bit 384 GB/s 35-115W 2.1B 5.0 RTX Pro 500 Blackwell 1.792 9.2 294 6 GB 5.0 Rtx 58 Bit 288 GB/s 35-75W 2.1B 5.0 9.728 42.6 682 16 GB GDDR6 (ECC) 256 Bit 576 GB/s 80-175W 1.4A 4.0 RTX 4000 ADA 7.424 33.6 538 12 GB GDDR6 (ECC) 192 Bit 432 GB/S 60-175W 1.4A 4.0 RTX 3500 ADA 5.120 23.0 36911 (ECC) 192 Bit 432 GB/s 60-140W 1.4a 4.0 RTX 3000 Ada 4.608 19.9 319 8 GB GDDR6 (ECC) 128 Bit 256 GB/s 35-140W 1.4a 4.0 RTX 2000 Ada 3.072 14.5 232 8 GB GDDR6 128 Bit 256 GB/s 35-140W 1.4a 4.0 RTX 1000 Ada 2.560 12.1 193 6 GB GDDR6 96 Bit 192 GB/s 35-140W 1.4a 4.0 RTX 500 Ada 2.048 9.2 154 4 GB GDDR6 64 Bit 128 GB/s 35-60W 1.4a 4.0 Themen: GDDR7 Graphics Card Nvidia Nvidia GTC 2025 Server Workstation
An engineer by training, Alexandre shares his knowledge on GPU performance for gaming and creation.
