It's rare, but it's not different than any other game. You pointed out yourself in this thread that Sonic Forces does this on One X. And Mr. Morgan detected the same scene-specific rather than load-specific resolution changes previously in Batman: Return to Arkham.Especially in a case like this where the resolution adjustments work differently than any other game (ie - apparently tied to specific sections of the track rather than adjusting by load).
No, Digital Foundry make mistakes all the time. Many of these are less obvious than resolution--missing or not mentioning graphical changes in Snake Pass, FF XII, etc.--but they do also commit errors on resolution too. For example, they said the PS4 Pro performance mode of Horizon: Zero Dawn was 1080p, and didn't announce the correct value of 1368p until, it seems, Guerrilla told them. (They've never updated their original article, though they've used the true number in later articles.) Or in Resident Evil VII, where they apparently relied on dev information that it was 1260p on Pro, and didn't check themselves to notice that a temporal pass was increasing resolution of some screen elements higher than that.I would understand this line of thinking if DF was wrong a lot and misleading about things often but this is the first major miss they have had that I can think of.
Of course, we have to allow them room for mistakes, and I must applaud their quick and professional handling of this one (especially in the face of wholly unwarranted extreme vitriol from the developer). But their results should be taken not as final authority, but as what I'm sure even DF would agree they actually are: best effort by earnest, trained individuals working under time and resource constraints. They should always be treated as tentative, and subject to enlargement, revision, or retraction. This is not a strike against them, but a positive trait of their work: it is concerned foremost with approaching truth, given inevitable human error.
We can't say that. I believe we can reasonably assume that the screenshots they decided to pixel count were all 1080p, yes. But if they played at least one lap on every course while analyzing--which it certainly seems they would--then they saw segments running at far higher resolution. They just didn't notice them.DF staff played the game and it stayed at 1080p the whole time.
Notice they do not refer to total pixels, but say they're scaling to 50-90%. Unless otherwise specified, such scaling numbers are taken to apply to both axes. You can check this convention yourself in Photoshop or any other image editor. Scale a 3840x2160 image by 50% (keeping aspect ratio) and the result will be a 1920x1080 image, only a quarter as many pixels.And 34BigThings are still claiming on their company website that their game renders at 50% to 90% of native 4K, when it actually is only 25% to 81%, even after they were made aware of that:
http://34bigthings.com/an-official-response-to-digital-foundry-analysis-of-redout-on-xbox-one-x/
What is going on here?
Thanks so much, I trust DF's drive to identify the ground truth. I assume the team is going to re-examine the Pro output as well? It seems you'd want to eliminate the possibility that there's similarly missable location-based resolution adjustment there as well.Dave will work on a new video first thing tomorrow.
I'll be back to my home office later this week and then promptly seeking help.
Last edited: