That blisteringly quick storage expertise discovered within the next-gen consoles is coming to PCs too, debuting first with the RTX IO expertise in Nvidia’s new GeForce RTX 30-series graphics cards. Microsoft simply pulled again the curtain a bit extra on the way it works.

Yes, the creator of Windows is explaining how SSD expertise works in a graphics card. No, it’s not as weird because it sounds.

Both the Xbox Series X and Nvidia’s RTX IO faucet into Microsoft’s DirectStorage, a brand new DirectX API. Microsoft teased that it would be coming to PCs after the Xbox Series X announcement. This week, the company revealed a bit extra about how the expertise helps your SSD and GPU work extra intently collectively to cut back (and presumably get rid of) loading instances—although you’ll want a speedy NVMe drive to benefit from it.

“With Nvidia RTX IO, vast worlds will load instantly. Picking up where you left off will be instant. This is a very big deal for next-generation gaming,” Nvidia CEO Jensen Huang mentioned whereas introducing the expertise. Instantaneous loading can also be a key promoting level for the Xbox Series X and PlayStation 5 launching later this yr.

How Microsoft DirectStorage and RTX IO work

“Games have pushed PC IO and file systems to the breaking point,” Huang mentioned. DirectStorage was constructed to smash previous that. Traditionally, CPUs have each referred to as sport belongings out of your storage and decompressed them, passing the info by means of the system reminiscence over to your graphics card. Microsoft’s Andrew Yeung defined why that labored effectively earlier than, however not in an period of blazing-fast PCIe 4.0 NVMe drives:

“Previous gen video games had an asset streaming finances on the order of 50MB/s which even at smaller 64okay block sizes (ie. one texture tile) quantities to solely lots of of IO requests per second. With multi-gigabyte a second succesful NVMe drives, to benefit from the complete bandwidth, this shortly explodes to tens of 1000’s of IO requests a second. Taking the Series X’s 2.4GB/s succesful drive and the identical 64okay block sizes for example, that quantities to >35,000 IO requests per second to saturate it.

Existing APIs require the [game] to handle and deal with every of those requests separately first by submitting the request, ready for it to finish, after which dealing with its completion. The overhead of every request is just not very massive and wasn’t a choke level for older video games working on slower exhausting drives, however multiplied tens of 1000’s of instances per second, IO overhead can shortly develop into too costly stopping video games from having the ability to benefit from the elevated NVMe drive bandwidths.”

In right now’s world of 100GB-plus video games with large file textures and ludicrously quick PCIe 4.0 SSDs, that conventional CPU handoff has develop into the bottleneck.

But whereas CPU threads want to finish a activity earlier than transferring onto the following one, GPUs excel at executing many duties in parallel. DirectStorage takes benefit of that by letting ultra-fast NVMe SSDs ship information on to the ultra-fast devoted VRAM in your video card. It’s primarily reducing out the pokey middle-man, whereas additionally releasing up your CPU to do different work.

Nvidia

Yeung says DirectStorage affords a number of instruments for builders to maximise storage efficiency: “by reducing per-request NVMe overhead, enabling batched many-at-a-time parallel IO requests which can be efficiently fed to the GPU, and giving games finer grain control over when they get notified of IO request completion instead of having to react to every tiny IO completion.”