The Battlefield V Tides of War Chapter 2: Lightning Strikes Update #3 patch, beyond sounding like a Final Fantasy title, has quite a few major improvements. The headlining feature is improved RTX support, which we will discuss shortly, but fans of the game may appreciate the other bullet points, too.
But first, because we are a computer hardware site, the RTX stuff. DLSS, which was recently added to 3DMark and greatly improved the image quality, has been added to Battlefield V. This setting uses machine learning to produce a best guess at antialiasing, versus calculating it with a direct algorithm (such as with TXAA or FXAA). Now that MSAA is somewhat uncommon, because it is incompatible with certain rendering processes, we’re stuck with either antialiasing via post-process or super-sampling. Super-sampling is expensive, so it’s usually either FXAA, which tries to find edges and softens them, or TXAA, which gives neighboring frames different sub-pixel positions and blends them. Both cases have issues. TXAA is considered the “higher end” option, although it gets ugly when objects move, especially quickly and visibly smooth. Because DLSS is basically a shortcut to provide something that looks like super-sampling, it should avoid many of these issues.
DXR raytracing performance was also improved.
Okay, now the tech enthusiasts can stop reading – it’s time for the fans.
Vaultable object detection is said to have a major improvement with this release. DICE acknowledges that Battlefield V movement wasn’t as smooth as it should be. There were a lot of waist-high barriers that players can get stuck behind, which the vaulting system should propel them over. It should be much easier to move around the map after this update, which is good for people like me who like to sneak around and flank.
DICE has also discussed several netcode changes, such as adding more damage updates per packet and fixing some issues where damage should be ignored, or healing should occur but would be ignored, and so forth. Basically, all of the netcode improvements were related to health or damage in some way, which is a good area to focus on.
Also, the Rush game mode, introduced in the Battlefield Bad Company sub-franchise, will return on March 7th "for a limited time"… whatever they mean by that.
The update should be available now.
Every implementation of DLSS
Every implementation of DLSS so far was first and foremost upscaling technique.
I.e. upscaling from 1440p to 4k using AI mojo. That is fine in itself. But marketing this as AA (with performance improvement to boot) is deliberate consumer confusion from nVidia side, and I say this is RTX 2080 owner.
Anyways, I believe if you make good analysis of DLSS implementation and compare image quality and performance, it will get lot of interest.
That’s why AMD needs to get
That’s why AMD needs to get that Patent for some FPGA programmable compute right on the HBM2 stacks into a product form that’s market ready ASAP. It’s for some in VRAM Frame Buffer FPGA implemented Tensor Cores action doing the very same sorts of DLSS like AI upscaling/AA while AMD works on getting that Tensor Core IP onto some Post Navi GPU Die in ASIC form.
Here is an article from Photo District News(1) describing the process of how that AI upscaling training is done.
(1)
“How AI “Learns” to Upscale Low-Resolution Images”
https://www.pdnonline.com/gear/how-ai-neural-networks-upscale-images/
So far I’ve been unimpressed
So far I’ve been unimpressed with any anti-aliasing technique except MSAA. I’d rather use no anti-aliasing at all than something like FXAA or TAA/TXAA/whatever, hands down. And with some of the visual artifacting of DLSS, I’ll probably not use that either (but I’ll give it a shot at least). Does anyone actually think these bad anti-aliasing techniques look good? In most cases it’s like smearing vaseline on your screen.