Hell is Us is available now with a time-limited offer.
Hell is Us is available now with a time-limited offer.
Not my sort of game, but the specs raised some eyebrows in places. Demo is out on Steam so you can try it yourself if you want! Apparently it runs for 2 weeks. Full game release in September. There is some bug where it can crash after 1st cutscene. Lowering setting got me past that, then turn up again after. Some indications of perf. 7800X3D. 5070 Ti at 250W power limit, all at 4k Starting area, looking in a direction that seemed to give lowest fps: high 70+ fps, 9GB VRAM, default TAA very high 42+ fps, 10GB VRAM, default TAA ultra 27+ fps, 11GB VRAM, default TAA ultra 23+ fps, 11GB VRAM, DLSS ultra quality ultra 54+ fps, 10GB VRAM, DLSS balanced very high 90+ fps, 9GB VRAM, DLSS balanced very high 70+ fps, 9GB VRAM, DLSS ultra quality VRAM usage is reported by GPU-Z so will include other stuff. After this testing, I did try closing everything (Steam, Discord, Chrome) and that reduced VRAM usage by around 1GB. I don't know if further areas get much more demanding but it doesn't seem too bad to run on 8GB at lower settings. Before people get upset over Ultra, I did a quality comparison. Without moving the camera, I took screenshots at high, very high and ultra. I just compared the high and ultra ones. Very little difference, but on closer looking it seems the main difference is how much additional random foliage you had growing around. It is so minor I didn't spot it while changing the setting. There might be a very slight difference in some rendering but I can't be 100% on this one. Texture quality seems same to my eyes. No obvious difference in shadows. Basically I don't think anyone will miss out dropping below ultra. I'll let Digital Foundry dig deeper if they want. Didn't see anything about RT in settings. . The waterfall at the starting location shows obvious SSR like effect so that isn't RT.
I’m completely okay with Ultra delivering 4K at 30fps on a 4090 Game. Scaling isn’t about unoptimized games—it’s about what you expect. What concerns me most is the assumption that all high-end displays rely on DLSS without specifying settings. Is it really quality, balanced performance, or just marketing? Honestly, DLSS shouldn’t be the focus of these comparisons because it mainly masks how much resolution is actually rendered. Calling 4K quality 4K when it’s actually QHD is misleading. I’ll test it myself to see what my 3060 can handle at 1440p Medium with smooth frames. I’m hoping for reasonable performance without DLSS, even if it means a lower frame rate.
They don't rely on DLSS by default. Each preset uses its own resolution scaling together with TAA. I can't remember the precise figures, maybe around 90% for Ultra and another 10% reduction at higher settings. I noticed the performance dropped when I switched to DLSS quality, even though the visuals seemed lighter. Someone needs to figure out the exact render resolution needed for a more accurate impression. Yes and no—consider how it feels from your side. With a 4K screen, I aim for a 4K result. What happens behind the scenes to ensure that output matches what you see matters less than how it looks. Even if the intermediate image appears QHD, it won't match rendering directly at QHD with the same settings. Other options like LOD are tweaked to achieve a similar quality at a higher resolution.
On my RTX 3060 (and R7 7700x) I'm using 1440p at high settings, DLSS Quality. I keep it steady around 40fps, usually in the low 50s for 4K High. With DLSS Quality, the 20s 4K High preset gives me balanced performance in the mid to high 30s average, and about 1% lower in the 20s. In the 4K High preset, it drops to the low 30s. I aim for 40fps at medium preset DLSS, and I never drop below 35fps. Daniel's waterfall stream tests suggest this setup works well with Nvidia Reflex and a controller.