Home Creators Posts Import Register Favorites Logout

Content

Hi everyone, Rich here. Today I finished off a sponsored video for Asus talking about the evolution of the 70-class GPUs from 2018's 2070 to today's 5070. It contains a brief appearance from GTX 1080 - the GPU that was most price-comparable to the 2070 at launch. It was great to do a video like this where a pitch was made and the sponsor just let us get on with it. Their input amounted to providing the TUF 70-class GPUs from the last few generations as the hardware being tested.

Part of this project was a look at latency. "Input lag" has been a key criticism of frame generation. There was talk of 120fps FG experiences "feeling" like 60 frames per second. More lag is a bad thing, right? But Nvidia Reflex is the mitigating factor, so I turned off Reflex on a bunch of games and on all of them, 4x frame generation on delivered lower latency than non frame-gen, Reflex off scenarios. The Doom TDA shot above shows how frame-gen lag kind sits between Reflex on and Reflex off. And then I thought to myself... well, what about all those games where Reflex isn't turned on by default? And what about GPUs that don't support Reflex and have no alternative anti-lag solution in place?

My own view on input lag is that it is not an issue until you actually start to notice it as an issue and that threshold is likely different for each and every one of us. I think we need to do more work in this area, but while the image above is interesting... well, if you have an Nvidia GPU and a copy of Cyberpunk, load up the Nvidia overlay, get it to display PCL (PC Latency) and then turn Reflex on and off. The results may surprise you.

Still, just something that's been on my mind a bit. Let's talk #content!

Rich (ie me) has indeed finished that video. Time to do more work on our website, complete DF Clips for the week and perhaps find time to take a look at a new AMD CPU.

John finished up collaborating with Oliver on the GRID Legends review and is back to working on the Xbox 360 Launch DF Retro episode. The second and concluding part. Obviously, plans are afoot for the next episode of the DF Retro Super Show.

Alex is working on a DLSS Preset L piece. Does it improve on Preset M quality? What about ray tracing/denoising. And how about in its preferred Ultra Performance Mode set-up?

Oliver is likely working next on his iPad Pro Apple M5 review!

Tom is actually taking some time off this week.

Lots of irons in the fire though, and it's a pleasure to once again work with Illusion on modernising our capture tools. Our full OBS port for performance analysis capture is now in beta and we're now working on improving some of the algorithms for fewer false positives. You can't blindly rely on tools, there always needs to be a sanity check, but the better the analysis in the first place, the less review work we'll need to do.

Anyway, as usual, I welcome your input, suggestions for projects, questions etc. Have a nice week!

Files

Comments

Chris Munch

I love the comparisons using old hardware, just to see how things have changed. Also, I can't be the only one who gets motion sick (literally sick) when frame generation is used. Obviously most people don't have that issue since I never see it brought up, but something about the disconnect between what you see and feel is unsettling.

TheBrainninja

“Does DLSS have the humility to hold this L? Find out [date TBD] on Digital Foundry!”

Marcus

Latency is really interesting because a lot of the criticism around frame generation seems deeply disingenuous. At the same time I have criticisms of the technology and I especially dislike how it’s being marketed. But I really do believe in that Tom Peterson future where resolution and frame rate is basically just filled out by your GPU using stuff like frame generation and upscaling. And then you can really use all of those really fancy monitors that are 6K or 5K or 500 Hz. I’m also happy to see an M5 iPad Pro review, and especially a preset L analysis. Considering you are looking into latency and gaming is quite expensive right now have you considered looking at something like the M4 Mac mini paired with GeForce now Ultimate? Mac and Nvidia devices are the only ones that get access to VRR with GeForce now so getting a cheap good computer you can use for work and web browsing as sort of a Home computer in the traditional sense and then you can rely on a service like GeForce now for streaming seems like an interesting video in sort of these budget constrained times. Maybe you could take a look at some of the shortcomings there like the weird AirPlay lag issue that seems to happen on Mac devices

Gaboou

The problem with Reflex for me is that it sometimes causes a pretty bad micro-stutter (on my config at least) - Horizon Forbidden West and Hogwarts Legacy come to mind. And even Frame Gen can't smooth this out. But as Reflex is mandatory when using DLSS Framegen I can't enjoy its otherwise great effects in these titles, sadly... 😔

SL0MAR

As a new owner of a 5070FE, I'm super excited for this video! Looking forward to the hints and suggestions on finding generalized sweet spots for current gen. A little early yet, but what about 4.5 dynamic MFG? When it lands, what's the best use case scenarios for a 5070? Considering a 60hz display, would 40 fps settings (dMFG off) be a good target for 60FPS (dMPF on)? 50? 30? How useful is it? Thanks for reading! *edited dynamic, not variable...

speedlemur

It's a problem in arc raiders too (I'm not even using frame gen there, just want lower latency).

Filip

I once posted a comment under one of your videos (not sure which one now) stating the exact same thing - that frame gen + reflex has tangibly lower latency than no frame gen + no reflex. They almost ate me alive... first, they didn't make the connection that there are millions of computers without any lag mitigation, millions of consoles without any lag mitigation that have drastically higher input lag than vast majority of users using frame gen, multi-frame gen, doesn't matter... and they're tolerating it just fine. It's just now that suddenly every dork that learned "latency bad when fg on" says it while he has 200ms lag in RDR2 or 150ms lag in Doom Eternal at 60fps (not much lower in 120hz because it's triple buffered with no frame queuing, just fully saturated Vsync buffer). While I dare to say almost everyone that uses frame gen sanely has no more than 30-70ms input lag, doesn't even have to be 60fps; 50fps and above feels just fine for me. Input lag is not an issue if the tech is used properly... some frame gen artifacts here or there but nothing like Lossless frame gen (LSFG is not good enough for me). DLSS FG feels much more sophisticated than that. Then they moved the goal post and said "who would want to play without reflex if it's available"... well, that's true but that's not the point... frame gen has lower latency than (almost?) any game running natively without any lag mitigation, at the same frame rate. And that's an interesting finding IMHO!

Fractal Flow

The input lag of 4xFG with Reflex On being lower than 0xFG with Reflex Off is gonna blow some people's minds. I anticipate the Denial Squad will come out in full force

Fractal Flow

Hmm. Interesting, I have never heard of FG motion sickness. To my great surprise, I use FG everywhere, when my native frame rate is at least 65-70, and love the smoothness. Maybe I'm getting old, but I don't feel any disconnect in input response, even when using a mouse. What framerates are you using FG at that gives you motion sickness?

Matt_Rant

I’d love to see the Asus Z13 tested under long duration load (1 hour of Doom TDA, Control, Cyberpunk, etc) against other Ryzen 395+ Max tablets/portables. All have clams of decibels and vapor chambers, but I’d love to see a detailed numbers breakdown. Throwing in Asus’ RTX 5090 eGPU from CES as a follow up would be icing on the cake.

Daniel Suttor

Really looking forward to that 5070 piece! Here a video idea I had floating in my head for a while: I'd love a video about what it takes to get absolutely pristine image quality, so DLAA vs FSR Native vs DLDSR vs straight-up downsampling. When I upgrade my GPU, I always love going back to older games and cranking them up to ridiculous settings. It would be great to see if its worth it and where the trade-offs lie. Cheers, have a great week!

David Ruppelt

> millions of consoles without any lag mitigation that have drastically higher input lag I'd say consoles don't count as the gamepad controls hide quite a lot of lag. Same for gamepad on PC. For those FG also should never be an issue latency wise. > Then they moved the goal post and said "who would want to play without reflex if it's available"... well, that's true but that's not the point... Is it not? For me it is. I am quite annoyed that some coverage of FG claims that FG latency effectively is better than no FG because of Reflex, when that is not the case as we have had Reflex at least in same games before FG was introduced. You could argue that less games would have gotten Reflex if it was not mandatory for FG. But I think that is less an argument for FG than it is an argument against the game developers. > frame gen has lower latency than (almost?) any game running natively without any lag mitigation, at the same frame rate. And that's an interesting finding IMHO! On that we agree. Especially interesting, since it doesn’t match my perception. While some old games really have perceivably bad latency, most felt good to me. With FG on the other hand so far I’ve encountered a roughly equal number of games that feel good or bad from a latency / directness perspective. I probably should spend some time measuring latency to back up my perception. Or let DF do that for me :)

Filip

good points 👍