NVIDIA presents 12 new GPUs based on the Blackwell architecture, including some with 96 GB of ECC GDDR7

Nvidia has presented a dozen new GPU in its GTC 2025 event, the largest launch of the company in two years.

Based on Blackwell’s architecture, newcomers use the nickname RTX Pro to differentiate themselves from previous generations (Ada Lovelace, Ampere and Turing) and, from their consumption breaths.

The flagship models are three variants of GPU RTX Pro 6000 Blackwell with 96 GB GDDR7 memory and up to 4000 AI tops Performance: twice the amount of memory in its former performance champion, the RTX 6000 and an amazing 4x of the RTX 5090, the best GPU it has to offer the consumption market.

NVIDIA GPU launches

Together with the standard work station edition, NVIDIA also introduced the Edition of Blackwell Max-Q work and server edition.

The latter is the successor of the L40 Data Center GPU series, which provides a very necessary consistency to the Nomenclature of the GPU.

As for the Max-Q work station edition, it is still a mystery. Nvidia launched Max-Q technology in 2017 and this is generally associated with portable GPUs that try to achieve maximum efficiency.

However, the PC of the workstation rarely point to optimal energy consumption, except perhaps in power environments, such as mini PC factor in a small form.

Three professional desktop GPUs were also introduced: PRO 5000, PRO 4500 and PRO 4000, which will be available from May 2025.

Given the launch cycles of previous products, I hope that more models focused on the input level and the main parts of the market are, which will be launched at the end of 2025.

Six new portable GPUs were also launched, all of them with the Naming RTX Pro Convention and, confusedly, some have the same name as their desktop counterparts.

The RTX Pro 5000 Blackwell is the new laptop flagp GPU with 24 GB memory of ECC GDDR7; Other models include series 4000, 3000, 2000, 1000 and 500 and must directly replace their respective “Ada Generation”.

All these pieces will be available in OEM Partners in mobile work stations as of June 2025.

We will strive to update this article when more details of the cards are published (including prices).

Nvidia’s key GTC note also saw the formal DGX Spark launch, previously known as Project Digits, DGX Station and Blackwell Ultra (or GB300), the most powerful GPU in Nvidia.

Slide to move horizontally

A GPU comparison of the Ada and Blackwell Professional generation
Header cell – column 0	Header cell – Column 1	CUDA NVIDIA PROCESSING CENTERS	NVIDIA RT Cores	Tensioner nuclei	GPU memory (GB)	Maximum memory bandwidth (GB/s)	Floating point performance Unique precision (tflops, peak)	Ai tops
L40S	Data center	18176	142 (3rd generation)	568 (4th generation)	48	864	91.60	1466.00
L40	Data center	18176	142 (3rd generation)	568 (4th generation)	48	864	90.50	724.00
RTX PRO 6000	Data center	Row 2 – Cell 2	Row 2 – Cell 3	Row 2 – Cell 4	96	Row 2 – Cell 6	Row 2 – Cell 7	4000
RTX 6000 ADO Generation	OFFICE	18176	142 (3rd generation)	568 (4th generation)	48	960	91.10	1457.00
RTX 5000 ADA Generation	OFFICE	12800	100 (3rd generation)	400 (4th generation)	32	576	65.30	1044.40
RTX 4500 ADO Generation	OFFICE	7680	60 (3rd generation)	240 (4th generation)	24	432	39.90	634.00
RTX 4000 ADA Generation	OFFICE	6144	48 (3rd generation)	192 (4th generation)	20	360	26.70	427.50
RTX 4000 Sff Ada Generation	OFFICE	6144	48 (3rd generation)	192 (4th generation)	20	280	19.20	306.80
RTX 2000 GENERATION ADA	OFFICE	2816	22 (3rd generation)	88 (4th generation)	16	224	12.00	191.90
RTX PRO 5000	OFFICE	Row 9 – Cell 2	Row 9 – Cell 3	Row 9 – Cell 4	Row 9 – Cell 5	Row 9 – Cell 6	Row 9 – Cell 7	Row 9 – Cell 8
RTX Pro 4500	OFFICE	Row 10 – Cell 2	Row 10 – Cell 3	Row 10 – Cell 4	Row 10 – Cell 5	Row 10 – Cell 6	Row 10 – Cell 7	Row 10 – Cell 8
RTX PRO 4000	OFFICE	Row 11 – Cell 2	Row 11 – Cell 3	Row 11 – Cell 4	Row 11 – Cell 5	Row 11 – Cell 6	Row 11 – Cell 7	Row 11 – Cell 8
RTX PRO 6000	OFFICE	Row 12 – Cell 2	Row 12 – Cell 3	Row 12 – Cell 4	96	Row 12 – Cell 6	Row 12 – Cell 7	Row 12 – Cell 8
RTX PRO 6000 MAX-Q	OFFICE	Row 13 – Cell 2	Row 13 – Cell 3	Row 13 – Cell 4	Row 13 – Cell 5	Row 13 – Cell 6	Row 13 – Cell 7	Row 13 – Cell 8
RTX 5000 ADA Generation	Laptop	9728	76 (3rd generation)	304 (4th generation)	16	576	42.50	681.80
RTX 4000 ADA Generation	Laptop	7424	58 (3rd generation)	232 (4th generation)	12	432	33.60	538.00
RTX 3500 ADA Generation	Laptop	5120	40 (3rd generation)	180 (4th generation)	12	432	23.00	368.60
RTX 3000 ADO Generation	Laptop	4608	36 (3rd generation)	144 (4th generation)	8	256	19.90	318.60
RTX 2000 GENERATION ADA	Laptop	3072	24 (3rd generation)	96 (4th generation)	8	256	14.50	231.60
RTX 1000 ADO Generation	Laptop	2560	20 (3rd generation)	80 (4th generation)	6	192	12.10	193.00
RTX 500 ADA Generation	Laptop	2048	16 (3rd generation)	64 (4th generation)	4	128	9.20	154.30
RTX PRO 5000	Laptop	Row 21 – Cell 2	Row 21 – Cell 3	Row 21 – Cell 4	24	Row 21 – Cell 6	Row 21 – Cell 7	Row 21 – Cell 8
RTX PRO 4000	Laptop	Row 22 – Cell 2	Row 22 – Cell 3	Row 22 – Cell 4	Row 22 – Cell 5	Row 22 – Cell 6	Row 22 – Cell 7	Row 22 – Cell 8
RTX Pro 3000	Laptop	Row 23 – Cell 2	Row 23 – Cell 3	Row 23 – Cell 4	Row 23 – Cell 5	Row 23 – Cell 6	Row 23 – Cell 7	Row 23 – Cell 8
RTX Pro 2000	Laptop	Row 24 – Cell 2	Row 24 – Cell 3	Row 24 – Cell 4	Row 24 – Cell 5	Row 24 – Cell 6	Row 24 – Cell 7	Row 24 – Cell 8
RTX Pro 1000	Laptop	Row 25 – Cell 2	Row 25 – Cell 3	Row 25 – Cell 4	Row 25 – Cell 5	Row 25 – Cell 6	Row 25 – Cell 7	Row 25 – Cell 8
RTX Pro 500	Laptop	Row 26 – Cell 2	Row 26 – Cell 3	Row 26 – Cell 4	Row 26 – Cell 5	Row 26 – Cell 6	Row 26 – Cell 7	Row 26 – Cell 8

Must Read

Leave a Comment Cancel Reply