That's about what the Intel A380 is. Mine is quiet at least (in my rig anyways, YMMV). Uses less power than a 5050 too.
I haven't seen a current GPU in that price range since 2020 I think. Didn't even know that they existed because I thought that AMD and nVidia got out of that business. I'll have to do some research on that card but it looks like a good candidate.
Intel is probably a worthy option. If you aren’t worried about raytracing, machine learning performance, or the latest and greatest game optimizations, the Arc GPUs compete fairly well — if you can get them near (a decent) MSRP), of course.
My production stats: GPU core 607 Mhz, 35 degrees, fans at 1925 RPM on the 1050 Ti.
My design goals for a GPU replacement: get rid of the black screens, run more efficiently than the 1050 Ti, run cooler than the 1050 Ti. I think that a 2050 would solve all of the requirements but my understanding is that the 4000 series had a big efficiency improvement. I can get rid of the black screens by running a 1660 Ti XC that someone gave me but there's more fan noise on that card.
The 1050 Ti is 21,534 for OpenCL. 5050 is 96,551. 1660 Ti is 65,879. 4060 is 101,040.
I suspect that I need a number of about 30K. A 5030 would probably be enough if they made such a thing.
I don’t have anything less than “70” tier cards, but maybe the following will provide some useful comparisons, nonetheless.
First up, a 4070 Ti (non-Super) Asus ProArt OC, ~2.5-slot card.
While testing with Geekbench 6.4, the majority of scores landed between 202,000 and 212,000. There were only a few outliers. The initial run after install was 176,167 but also a 198,215 and a 196,###. In the 80-100 percent power (limit) range, the average score was ~207,000. At 50% power limit, the scores fell between 196,000 and 206,000.
Next is an RTX 5070 Founders Edition, 2-slot with a length shorter than a triple fan but longer than a typical third-party dual fan card.
Geekbench scores had a range from 225,000 to 231,000. The average was ~228,400. Somewhat surprisingly, the GPU performed overall slightly better at 70% power limit.* In the same number of runs, the scores landed between 225,000 and 232,000 with an average of ~228,700. Also, there were no outliers to mention.
Regarding power and thermals, here are some HWiNFO 64 screenshots from after each test, spanning power limit equivalent runs, to show comparisons of max GPU power, thermals, and fan(s) speed.
50-percent power limit:
70-percent power limit:
Geekbench is a shorter test. As such, I decided to add Superposition data to further demonstrate power usage as well as the benefits of adjusting/tuning power limits. For example, the 4070 Ti had ~3% less performance with 25% less maximum power available.
RTX 4070 Ti - 100% (stock) power cap
1080p Extreme score: 11718
RTX 4070 Ti - 75% power limit
1080p Extreme score: 11385
RTX 4070 Ti - 50% power limit:
1080p Extreme score: 10248
*** Mis-captured screenshot ***
RTX 5070 - 100% (stock) power cap
1080p Extreme score: 14794
RTX 5070 - 75% power limit
1080p Extreme score: 14784
RTX 5070 - 70% power limit*
1080p Extreme score: 14779
RTX 5070 - 70% power limit*
4K Optimized score: 19235
* The RTX 50 series GPUs don’t have as low of a power floor versus previous generations. This has been noted by others, unfortunately, I don’t recall exact video titles to locate and reference.
Other testing information: ambient temp range of ~27 to ~28ºC. Case fans were at ‘idle’ (i.e., <=1000 RPM). Systems featured the AM5 platform, Windows 11, latest drivers and BIOS/firmware. Tool used to adjust GPU power limit was ASUS GPU Tweak III.