Kunal Khullar Social Links Navigation News Contributor Kunal Khullar is a contributing writer at Tom’s Hardware. He is a long time technology journalist and reviewer specializing in PC components and peripherals, and welcomes any and every question around building a PC.
cknobman The issue isnt that Nvidia makes these technologies to encourage unoptimized slop, but lazy developers realize they can utilize them to cut corners and make crap code run "acceptably". Reply
GenericUser2001 From what I have seen, I would call DLSS 5 an Uncanny Valley filter; while it might make human figures and faces look more "realistic" on a technical level, doing so without paying attention to small details can easily end up making the people look off putting or creepy. Which is exactly how I would describe many of the human models ran through DLSS 5. It may be beneficial for non-human objects, though I question whether the minor visual improvements it seems to create on those are worth the large performance hit. cknobman said: The issue isnt that Nvidia makes these technologies to encourage unoptimized slop, but lazy developers realize they can utilize them to cut corners and make crap code run "acceptably". That would be an issue for the older DLSS upscaling and frame gen stuff, which would at least get you more FPS, at the cost of some image quality & response time. This new DLSS 5 filter drastically REDUCES the fps you get at the same resolution. So if game devs want to use it, well they will need the base game to run much better before anyone turns this on. In theory anyways, there seems to be no shortage of sloppy devs out there. My concern would be lazy devs using it as an excuse to ship out less detailed base models and expect this filter to fill in the gaps. Reply
Roland Of Gilead Reduced performance (n)Increased Power consumption (n) It's not worthy of a release yet. NVIDIA are clearly trying to dangle a carrot of some sort here, and hope that the market will accept this crap. For me at least, I don't play games where I can stop and look at a characters facial features and think, WOW! That's amazing. Photo realism isn't the big deal NVIDIA thinks it is. Better quality games (and I don't mean GFX, I mean games worthy of playing regardless of GFX shiny bits) are what we want to play. Properly optimized and able to play across a fair selection of hardware setups. Reply
King_V We've gone from "I want the newest Nvidia feature, it gives me more FPS!" to "I want the newest Nvidia feature, it gives me less FPS!" Reply
Roland Of Gilead King_V said: We've gone from "I want the newest Nvidia feature, it gives me more FPS!" to "I want the newest Nvidia feature, it gives me less FPS!" Brilliant! :grinning: You're bang on though. Reply
timsSOFTWARE Game rendering is and always has been fundamentally an illusion, and ironically optimization is often about coming up with even more clever ways to maintain that illusion while doing less work compute-wise. So while I think there are some good reasons to prefer graphics rendered the traditional way – input lag, preserving the original intent for the art style and feel of the game, etc. – I don't think using DLSS/frame gen makes it any more "fake" than it already was. If you like how the game looks and feels to play, then I don't think there is any point in trying to be a "raster purist" – whereas if you don't, then obviously that's a different story. Reply
GenericUser2001 timsSOFTWARE said: Game rendering is and always has been fundamentally an illusion, and ironically optimization is often about coming up with even more clever ways to maintain that illusion while doing less work compute-wise. So while I think there are some good reasons to prefer graphics rendered the traditional way – input lag, preserving the original intent for the art style and feel of the game, etc. – I don't think using DLSS/frame gen makes it any more "fake" than it already was. If you like how the game looks and feels to play, then I don't think there is any point in trying to be a "raster purist" – whereas if you don't, then obviously that's a different story. The thing is with this new DLSS 5, it needs a significant INCREASE in the amount of computation & power being used, not less, for a frankly questionable gains in visual quality (personally I think the human models, especially faces look worse with DLSS 5 applied, non-human things do look a bit better). It is the exact opposite of prior DLSS versions in that regard. I think you could probably get the same or better gain in visual quality by just using better quality models and textures, at a similar or less performance cost. The only "advantage" of DLSS 5 would be on the dev side, letting them get sloppy on model creation. Reply
hotaru251 GenericUser2001 said: So if game devs want to use it, well they will need the base game to run much better before anyone turns this on. …you new to the area?devs will not do that they will just make to hit 30fps. Game devs have shown time and time again they will always skimp out of doing the work and just use technology as a crutch when possible becasue their bosses dont care and just like deadlines & less $ spent on project. timsSOFTWARE said: I don't think using DLSS/frame gen makes it any more "fake" than it already was. "fake" in this sense is easy to describe. If you run the same game on same hardware and the same settings the "real" stuff is replicated perfectly down to the very pixel every single run. "fake" is stuff that varies run to run (i.e. your framgen and upscaling) becasue its not "real" and is made up" hence "fake". Reply
tntdyno you all needed an article to realize this? i have been speaking this truth for years now. next time listen up when the truth is being spoken. Reply
usertests GenericUser2001 said: The thing is with this new DLSS 5, it needs a significant INCREASE in the amount of computation & power being used, not less, for a frankly questionable gains in visual quality (personally I think the human models, especially faces look worse with DLSS 5 applied, non-human things do look a bit better). It is the exact opposite of prior DLSS versions in that regard. I think you could probably get the same or better gain in visual quality by just using better quality models and textures, at a similar or less performance cost. The only "advantage" of DLSS 5 would be on the dev side, letting them get sloppy on model creation. Raytracing and DLSS 1.0 were also deployed and used for marketing before they were actually worthwhile. The eventual 60-series throwing more tensor cores at it could solve or alleviate the performance issues. Efficiency will also be improved, likely from moving to a TSMC N3 node. 50-series Blackwell was already disappointing at launch in part because it used the same 4N node as Lovelace. RTX 5090 was the standout because it was the only tier of card that got a die area increase (23% larger than 4090). Whether or not gamers actually like DLSS 5 is another story, but it could fare better if developers are taking it into account during development. AMD will probably be forced to respond. They may have already planned to with "FSR Diamond" or other features tied to RDNA5 and the new console generation. Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/video-games/pc-gaming/SPONSORED_LINK_URL
- https://www.tomshardware.com/video-games/pc-gaming/rpcs3-emulator-devs-slam-nvidia-dlss-5-as-ai-slop-generator-says-industry-pushing-more-upscalers-and-frame-generation-to-hallucinate-games-and-hide-their-lack-of-optimisation#main
- https://www.tomshardware.com/membership
- Universitas Gadjah Mada, Indosat and NVIDIA Open Indonesia’s First University AI Center to Develop Local AI Talent
- This 32GB of Team Group DDR5 RAM is the cheapest DDR5-6400 kit on the market – only $405 with promo code, buy the least expensive DDR5-6400 kit around with envi
- Tape companies ship 160 exabytes of storage in 2025, AI data demands drive 'unprecedented data growth' — capacity shipped in Q1 2026 rose 57% YOY driven by AI,
- Nvidia's RTX Spark N1X launches in October for laptops and desktops — 18 or 20 CPU cores, paired with 5,120 or6,144 CUDA cores, up to 128GB of unified memory
- AMD unveils Threadripper Halo Station, an AI workstation packing 96 cores and dual liquid-cooled MI350P accelerators — 'the most powerful workstation in the wor
Informational only. No financial advice. Do your own research.