Hardly a fail; again pretty perfect IMHO. I will tell you, though, my OCD wants those mins to be 60/45 so badly, and really is my question if this product (in whatever power variation with tweaking) can do that.
Is it slightly anal-retentive when things like VRR exist? Sure, but it's the principle of the thing...especially if using FG...I think when you boil the 9070 xt down to it's core as a product it should (be able to) do that.
If nothing else, within driver improvements perhaps? I'm not blaming the benchmark, as this appears to track with many other games as well (~58-59fps mins before tweaking it).
I kind of just want to buy one and loop this bench until I could keep it that way (60 mins in software, if not preferably hardware; whichever is possible). I find things like that oddly satisfying.
I keep telling myself to wait until there is a somewhat reasonably priced card that can keep 1440p60 with up-scaling to 4k, but for many and/or perhaps most, that is perfectly good, certainly for right now imho.
Thank you very very much for your kind words!
@miguel1900 (
)
You are very welcome.
Thank you again for your work, as I think it will come in handy as easily digestible as more people begin to understand the limitations of cards wrt lumen; rt/pt in general wrt bandwidth/buffer etc.
I am curious, how difficult would it be to implement FSR1-4 and DLSS2-4 support into this (even with an Optiscaler-like hack)? Perhaps also the path tracer renderer of UE 5.5 as an option across resolutions?
IMHO, this would be the literal perfect bench for the future of gaming if those were there, given their general use-case. Being able to test essentially everything, even if some things (right now) appear a little nuts.
Or sometimes, scaler may appear outdated. You never know when the perf/quality trade-off might be worth it and this might be a quick way to check (and maybe see how far we've gone in each direction).
1080pRT->1440 'quality' FSR3/4 on a 9070 xt for example, as that is the intended market. 4090 and 1440p->4k RT or 1080p->4k PT using DLSS3/4 (which may slip more with new DLSS versions).
Maybe 1080p->4k in some combination for others, etc.
It could not only show the difference in performance between the companies up-scaling in a RT load, but also perhaps change in performance between versions a la FSR3/4 or CNN/transformer (and future iterations).
(That's something I'm always attempting to explain and would be great if I could show it in a ~2.5min bench, especially if people could run it themselves).
It would be cool if this was easily 'solvable' (by looking at something like general 1%,.1% lows; minimums) and I could point someone towards a chart for each in those scenarios.
Maybe factoring in OCing, etc, and the version of up-scaler used. So many people are new to these scenarios that I think it would help them to understand them and each product's general capability.
As I say, I'm one of those guys that's really waiting for 4K60 raster, 1440p60RT, and 1080p60PT mins to line up with up-scaling of the later two to 4k. I'm trying to figure out exactly what that would take.
I'm sure there are people that are targeting one step up, while one step step down may or may not continue to be a ~9070 XT (could trend slightly downward over time, similar to 4090 and it's general goals).
Or perhaps some slightly less orthodox combination.
It would appear they are generally intended to be mostly proportional (I don't know general sample ratio; you probably could figure it out fairly easily), granted that's not always exactly the case in practice.
Ofc scaling from 1440p->4k needs a little extra perf and 1080p->4k quite a bit more, but it could give people an idea of what they need if scaling from your current native Lumen setup (which I think is great.)
To me, at least, being able to find some kind of bench to test limits for all those things would be wonderful, as I think that could help people find their 'perfect card' for themselves given the increase of such scenarios.
I don't mean to ask you to do extra work, and I think this bench is wonderful as it is for native rez software and hardware lumen, only that this could make it truly shine above any other I've seen as a one-stop shop.
IMO we
need those scenarios to be
tested, scrutinized, and understood (by both reviewers and consumers, including myself as the latter, not just developers), and the way you display the info is fantastic.
It already perfectly showcases 5080's 1440p limitations (I attempt to explain lows; why stability important; especially as a product ages and/or higher DLSS requirements over time), and I appreciate it greatly.
Honestly, if that package existed, I would scream bloody murder until reviewers put it into their suites and ran the full gamut of native and scaling settings on every card.
I think all of those things are that important.
Truly, perhaps
the most important factors and/or limitations of any card right now and certainly going forward into the future (especially wrt the next generation of consoles and scaling from their innate capabilities).
Does what I'm saying make sense and/or do you agree? Is something like this possible and/or is it something you could/would consider implementing?
a subtle whisper ....upgrade now....
