Talk about Nvidia RTX Pro 6000 and its role in the future.
Talk about Nvidia RTX Pro 6000 and its role in the future.
I remember back in the day, about 2010 when 4 GB RAM was considered a huge amount and cards with 1 GB VRAM were high-end and extreme. 15 years later and those numbers have shifted to 32 GB system RAM and 16 GB dedicated VRAM. A difference of 8 times for RAM and 16 times for VRAM. If this does keep up we should expect 256 GB RAM and 256 GB VRAM in about 2040. Although, I personally believe we will not go there in terms of VRAM so soon.
What do you think about the RTX Pro 6000 with its 96 GB VRAM and slightly higher core-count than the RTX 5090? Would it still be exceptional and extreme 10 years into the future? When would we call such cards 'the norm?'
Thank you!
PS. This is entirely superstitious and within the exciting realm of prediction. Opinions may not come to fruition with the passing of the years, but they are intriguing intentionally as part of the communication in the discussion.
👍
Different GPUs for various tasks. A workstation GPU isn't the same as a gaming GPU.
Gaming (GeForce) GPU – versatile performer, specialized in gaming. It delivers top optical quality.
Workstation (e.g., Quadro) GPU – versatile performer, expert in all areas except gaming. It excels with wireframes and complex shapes.
Since Quadro GPUs are primarily employed in design and visualization (3D modeling, digital imagery), they demand significantly more VRAM than any comparable GeForce GPU from the same period.
Considering that a 32GB Quadro model launched in 2018 and a GeForce model with 32GB VRAM appeared in 2025, gaming GPUs are roughly eight years behind in terms of VRAM capacity.
By around 2032, it’s possible to expect GeForce GPUs with 96GB VRAM. At that time, workstation GPU VRAM could reach approximately 300GB.
Currently, the RTX 5090 stands out; it’s not the standard, but the most popular choice according to recent Steam hardware reports.
Your memory might be failing.
In 2010, GeForce released the 400-series and 500-series.
For the 400-series, 1GB VRAM was standard. Even the weaker models in the line had 1GB VRAM (GeFore 405 and GT 430).
There’s also a 2GB VRAM version of the GT 430. Details: https://www.techpowerup.com/gpu-specs/pa...2-gb.b5579
But in the 400-series and 500-series, no GPU offered 4GB VRAM. Gaming GPUs from that era (2010) typically had between 0.5GB and 3GB VRAM.
Yet, 2010 also introduced a 6GB VRAM Quadro model (Quadro 6000).
I recall that year very well, a bright morning. Back then, the 5000 generation was using the Radeon HD 5850 and 5870 cards, which typically had around 1 GB of VRAM. By the end of 2010, models like the HD 6970 were available with 2 GB of VRAM. Comparing 1 GB cards from 2010 to 16 GB cards from 2025 isn’t quite the same as comparing 2 GB cards from 2010 to 32 GB cards from 2025. Still, it’s reasonable to say they fit within the higher-end range for each respective year.
After that, the i7-990X was released in 2011 and really changed things. System RAM jumped from 4 GB to 8 GB during that period. The 8 GB version became the standard for enthusiasts. In 2010, 4 GB was considered a strong option.
Thanks for the sharp, precise feedback. That’s exactly what we need to keep these discussions very engaging!
Most of the demand for additional VRAM now comes from AI training. While running AI itself uses less, training still requires more. For GPUs designed for AI training, I believe we'll observe an increase in VRAM capacity. As previously noted, this also benefits workstations such as CAD and certain content creation tools, though the need will be significantly higher than before. The primary area where VRAM will rise is for nearly all workstations, which will mainly depend on CPU RAM.
Thank you for your message!
The more common RAM, the better it is. I recall back in 2001 on my first PC (Celeron Tualatin 1.3 GHz, GeForce2 MX 400 with 64 MB memory, 256 MB RAM, 40 GB HDD; running Win98 SE, Win2000 Pro and Linux RedHat 7.3) having 256 MB of RAM was quite sufficient. Just four years later, in 2005, the need for RAM increased significantly toward around 1 Gigabyte, and from that point on, the main issue with my system became insufficient memory to handle new software releases. Four years brought a major shift in RAM demands during those times. Like reaching 128 GB in 2029 (which won't occur, as trends are changing).
👍
It seems the outcome hinges on the development of local AI/LLM models. We could end up using AI as a service within games, shifting that workload to the cloud and keeping your device lighter in memory. However, creating single-player offline experiences powered by AI might require significant RAM, likely relying on the GPU now. Consumer CXL and memory cards could eventually step in. Progress is slow on 4K, but if 8K gaming emerges, we’ll need even more VRAM.
Back then I could launch several Chrome instances with many tabs open (30 to 40 per window) while running Microsoft Office on a modest 896 MB RAM system. Everything functioned smoothly. Now, when I open Chrome with numerous tabs and the machine is loaded with software applications, the RAM consumption reaches around 23.7 GB. Starting Windows and its drivers, plus the software that depends on the hardware, consume about 9 GB of memory. In my experience, Windows 7 used roughly 1.01 GB for the same workload.
In the past, software was relatively lightweight. It became slightly more demanding with Vista and 7. Today, we are hitting unprecedented levels of RAM and VRAM usage, along with CPU and GPU demands.
I’ve heard Crysis 1 was designed to use up to 2 GB of RAM. I’m certain that if developers had crafted better software, the hardware requirements would have changed significantly. Software creators used to prioritize quality; nowadays, most focus on profit, which often leads to subpar performance.
There’s a lot of code that runs slowly due to outdated mindsets. This is why powerful hardware is now essential for tasks that once worked well on older systems.
Even though this is just a rant, I really believe things are heading down a negative path. There will come a time when software becomes so inefficient that even the most advanced machines struggle to keep up.
Cool comment, by the way! Thanks for sharing your thoughts!
Websites became more intricate with numerous capabilities, yet they are also struggling financially, as the worth of ad space is a very small fraction of what it used to be. Nowadays, sites are mostly encrypted, and your browser isolates each tab in its own sandbox. More data needs to be kept in memory than just the content displayed on each tab. HTML 5 supports almost any programming language, reducing the need to write custom browser or web code. Many sites load large amounts of unused code that never gets executed. Since these are all executed at runtime, optimization for developers is not straightforward.
Back in the day, you could upgrade a game like Crysis to use more than 2GB of memory, which was common during the XP era to give 32-bit systems up to 4GB of available space.
I agree that without hardware limits, programmers can create somewhat inefficient code. However, this isn’t always true. Recent Doom games are surprisingly efficient and can run on older hardware.
In general, a program using just 1GB is quite excessive. The exact amount depends on what the program actually does and how much memory it uses. Applications consist of source files, audio, video, etc., but the core program itself tends to be relatively small. This capacity allows developers to significantly enhance features and visuals.