YES

I hope:
- it still has the dithering for 16-bit color support (even if windows doesn't do the latter anymore)
- the linux drivers are good
- it's not sidelined by some slop ambitions (powervr already has had dedicated neural cards released)
- it properly supports a framebuffer texture effect and not send the framerate to single digits from a scene render-to-texture tile recursion
- it supports a depth buffer (kyro didn't, neon250/clx2 didn't, pcx1/2 didn't)
- it has a glsl compiler that doesn't suck *cough*AMD*cough*
- it has very competitive price ranges (kinda difficult with the ram supply for lake-drying plagiarism/surveillance centers situation)
- not suddenly ghosted by a manufacturing partner/backstabbed by nvidia/banned in the US
- DirectX 11 support clearly showing a dedication for game compatibility, so maybe they'll also think about DirectX9/8/7 and GL 1/2/3 and SGL1 & 2

leileilol wrote on 2026-03-24, 23:29:

I hope:
- it still has the dithering for 16-bit color support (even if windows doesn't do the latter anymore)
- the linux drivers are good

I too would like proper Linux support and also 16-bit dithering.

The big reason why I even got this Intel NUC was because of being told Intel Integrated Graphics still dither 16-bit graphics.

10+ years old performance and graphics features. Nuff said. I also wouldn't put any hope in legacy features, because those chips definitely have zero relations with old PowerVR.

DracoNihil wrote on 2026-03-25, 20:05:

The big reason why I even got this Intel NUC was because of being told Intel Integrated Graphics still dither 16-bit graphics.

Intel HD on a haswell's amazing as that 16-bit post dither exposed more filtering precision than just using 32-bit color, so it's actually better for games like Q3 that shift some bits around for overbrights where 16-bit color would be a significant loss.

The Serpent Rider wrote on 2026-03-25, 20:38:

10+ years old performance and graphics features. Nuff said.

I'm still gaming on a 9-year-old AMD card at the latest. Shit's fine, meaningful GPU progress have been glacial with the abusive neural novelties that CEOs want, and it's very easy to forget the barrier of entry when you're already spoiled.

Oh, and I also wouldn't mind a tech demo in the vein of Ultim@te Race Pro for this as-of-yet-vaporware gaming GPU 😉

Modern ImgTech chips already exist in the Chinese domestic market with cards from Moore Threads and InnoSilicon, both brands are available on AliExpress (ie. MTT S80 from Moore Threads, Fenghua 2 from InnoSilicon) and there's numerous videos on Youtube for the MTT S80 specifically. There's plenty of other Chinese domestic brands (ie. http://www.graest.com/Prolist_T270.html), but the only other one I know of with definitive lineage to an existing/old company is the Glenfly Arise cards, which are S3/VIA-derived - the 2CU version of the Glenfly chip is also available as an iGPU with Zhaoxin CPUs.

I do have a Fenghua 2, but I haven't spent any time on it yet because the Windows drivers are for Win 10/11 and the Linux drivers are closed-source.

https://youtu.be/6fugxs0YdNQ
So MTT S80 performance varies from Radeon 7970 to RX 580 level (if you're really lucky, like in Crysis 🤣 ). And that's with relatively popular games. Outside of something like CS2 it drops to Radeon HD 5000 levels.

The last decade PowerVR IPs into the GMA series of SoC for the netbooks and mini PC were not bad idea beside lacking a different logic of driver development actors and the logic around such complex needed task. The GMA500 and GMA3600 PowerVR SGX5 platforms may have been integrated into a low end desktop AGP/PCI video card if were used in their full version running all unified pipelines the architecture could beside not for netbook costs/power demand.
If a desktop video card would ever be released even lowest end sector market, I'd buy it if Linux driver would be released.