Skip to main content
  1. Home
  2. Computing
  3. Features

I tested Nvidia’s new RTX feature, and it fixed the worst part of PC gaming

Add as a preferred source on Google

You wouldn’t suspect that a mod like Portal: Prelude RTX would debut such a major piece of kit for the future of PC gaming. But lo and behold — it’s the first game we’ve seen with Nvidia’s new RTX IO, a feature that was announced nearly three years ago.

It’s not as flashy as ray tracing or DLSS, operating in the background and offering a slew of benefits without drawing attention to them. But according to my testing, RTX IO may have a bigger impact on PC gaming than any other RTX features.

Recommended Videos

What is RTX IO?

Portal: Prelude RTX | RTX IO Off vs On Comparison – Cake Scene

First, let’s talk about what RTX IO even is. It’s GPU-accelerated asset loading and decompression that promises faster load times, smaller install sizes, lower CPU utilization, and less texture pop-in. It does this by offloading some work your CPU does to the GPU, taking advantage of the massive core counts available on graphics cards.

When games typically load, assets are taken off the SSD and loaded into system memory. They cycle out of memory onto your CPU for decompression before returning to the memory. Once the full asset is ready, it’s sent to the graphics card. This system worked for a long time, but there’s a problem with it now: the PS5 and Xbox Series X.

As I’ve written about previously, both the PS5 and Xbox Series X have dedicated decompression hardware that cuts the CPU entirely out of this process. Assets are decompressed faster, and therefore, arrive at the GPU faster. The storage inside PCs these days is fast enough to keep up with this level of streaming, but they lack the dedicated decompression hardware to make it work. That’s where RTX IO comes in.

Diagram showing Nvidia's RTX IO technology.
Nvidia

Compressed data comes off the SSD into system memory and it goes straight to the GPU. It’s passed through GDeflate, Nvidia’s open-source data compression scheme built for GPUs, and it’s immediately ready at the GPU.

If you keep up with this tech, you can probably spot the similarity to Microsoft’s DirectStorage. RTX IO works on top of DirectStorage, and it even supports Vulkan extensions, allowing it to work in games like Portal: Prelude RTX. None of this is proprietary, either; there’s been plenty of finger-wagging at Nvidia over its (now open-source) Deep Learning Super Sampling, but RTX IO works on any DirectX 12 graphics card.

But does it work?

Portal Prelude RTX image comparison.
Image used with permission by copyright holder

This is all great in theory, but GPU decompression is uncharted territory on PC — even with games like Forspoken that support DirectStorage. You don’t need to look further than the install size to see that RTX IO is working. If you were to download Portal: Prelude RTX right now, you’ll see it’s around 24GB (exactly 24.29GB at the time of writing). With RTX IO off, the total install size is 39.16GB. That’s a 38% reduction in install size for the exact same game.

Texture loading was much faster, too. In one scene, it took 3.13 seconds for textures to load in with RTX IO off, seeing a noticeable pop-in. With RTX IO on, they arrived in 1.36 seconds. In the most demanding scene I looked at, it took 6.34 seconds for textures to load with RTX IO off. Turned on, they arrived in only 2.51 seconds.

Jacob Roach / Digital Trends

Keep in mind the scope of Portal: Prelude RTX, as well. It’s a game that goes deep, not wide, offering full-on path tracing and highly detailed materials for every corner of a scene. In a larger game, it’s not hard to see the benefits of RTX IO for texture streaming. With the right optimization, and RTX IO cutting roughly two-thirds of the time it takes to stream textures, there are some clear applications in large, open-world games.

CPU utilization dropped, too. In the demanding scene I mentioned above, the utilization peaked at 37% (and that was with the monstrous 24-core Intel Core i9-13900K). Turn RTX IO on, and it dropped to 22%. From what I could tell, it seemed like Portal: Prelude RTX scaled to eight cores. Some games, particularly those built in Unreal Engine, could see a larger benefit if they’re limited to a few cores.

What this means for PC games

Ratchet from Ratchet & Clank: Rift Apart.
Image used with permission by copyright holder

Make no mistake: Portal: Prelude RTX is an appetizer for what games with the tech could look like in the future. Thankfully, we won’t have to wait too long to see what it can do in other games. Nvidia confirmed Ratchet & Clank: Rift Apart will support RTX IO on PC, allowing that game’s storage-intensive design to operate.

What makes me most excited is the optimization benefits that RTX IO (and DirectStorage, for that matter) can bring. I’ve been vocal about the fact that it’s a bad time to be a PC gamer and that largely boils down to persistent stuttering across major AAA releases. Hardware-accelerated decompression doesn’t solve the stuttering woes on PC, but it can help.

As we’ve seen in games like Star Wars Jedi: Survivor, Dead Space, and Redfalltraversal stutter is still a very real problem on PC. It’s particularly problematic when games are designed to run on only a few CPU threads (as is the case with Unreal Engine titles like Star Wars Jedi: Survivor and Redfall). Taking some work away from those cores in the decompression department can free them up to, well, run the game, hopefully avoiding some of the major traversal stuttering issues we’ve seen this year.

If that’s the case, RTX IO and GPU decompression is a huge swing to help PC games. Faster loading times and smaller install sizes don’t hurt, but the real benefit of something like this hopefully lies in its ability to reduce stuttering in games that are streaming massive assets in and out of the game world.

It seems that will be the case. Here’s how Nvidia explained it: “RTX IO alone cannot completely remove stutter, but it can be an aiding technology to help reduce stuttering. It can do so by scaling down the dependence on CPU compute when the need to load textures and geometries ‘faster’ is the cause of stuttering, and freeing up the CPU to work on other tasks.”

For now, we just have to wait until we see more games with RTX IO and other branded GPU decompression tech. Ratchet & Clank Rift Apart kicks that off on July 26, but hopefully, we’ll see more games sporting the tech throughout the rest of the year.

Jacob Roach
Former Lead Reporter, PC Hardware
Jacob Roach is the lead reporter for PC hardware at Digital Trends. In addition to covering the latest PC components, from…
Apple could offer MacBook Ultra in two sizes with one-of-a-kind OLED display
A new report reveals the MacBook Ultra's display size, OLED tech, and launch window
macbook-ultra-sizes

Apple's rumored MacBook Ultra is shaping up to be one of the most significant Mac redesigns in years, and a fresh industry report suggests it could land sooner than anyone expected. Research firm Omdia has released a new study on OLED adoption in laptops, and buried inside it are some very specific details about Apple's next MacBook.

What screen sizes will the MacBook Ultra come in?

Read more
You can now send emails directly from ChatGPT on the web
You can now draft, edit, and send emails directly inside ChatGPT on the web
Person using ChatGPT on a MacBook.

If you have ever drafted an email in ChatGPT and then had to copy it across to Gmail or Outlook to actually send it, you can now skip that extra step.

OpenAI has rolled out the ability to send emails directly from writing blocks in ChatGPT on the web, keeping the whole process inside a single conversation from start to finish.

Read more
Turns out, teaching games like Battleship can make small AI models a whole lot smarter
By turning Battleship into an AI training ground, researchers helped smaller models reason more efficiently.
AI Apps installed on iPhone Gemini DeepSeek Claude ChatGPT Auren

Small AI models just got a surprising boost from a very old game.

MIT researchers used a Battleship-style setup to test whether AI agents can improve how they gather information before making a move. The result was a sharp jump in performance for smaller systems, including one model that went from rarely beating humans to winning most of its games after researchers changed how it searched the board.

Read more