Post History

You can't really say anything absolute about discrete vs. onboard vs. APU. It all depends. You would look at performance / $ on a site like https://www.videocardbenchmark.net/ and go by that. Usually discrete gets better performance, because it's marketed on performance, is more specialized, and being separate can be cooled more effectively.

It's wise to figure out how much performance you actually need. Maybe you encode a video in 2 min, and if you could do it in 1 min it would be good enough. Then you don't need a card that is 10x faster and can do it in 12 seconds, even if it's good value. The way to do is to consider the benchmark score on your existing hardware and extrapolate/guesstimate from there.

>Can I expect my old graphics card to work with modern budget CPUs? What specific compatibility issues might I have to look out for, or how can I research this?

Generally, yes. The CPU has very little impact, which gets exaggerated by technical writers who want an excuse to write in depth articles. The main type of GPU today is PCIe which most normal motherboards have supported for over a decade. However, don't guess, check compatibility on a site like https://pcpartpicker.com/.

Note: Video cards often require their own power line, so you might need to get a new power supply unit as well.

>How would the power of integrated graphics on these CPUs likely compare to my old graphics card? I can easily find benchmarks for separate video cards and (by a different metric) the actual CPU performance of CPUs, but not for the video performance of integrated GPUs.

Usually, the best new mid or low-end hardware is better than old high-end software because technology advances rapidly in this field. Cards at the upper end of mid-range tend to give the most bang for buck. High end cards are overpriced because they compromise on efficiency for the sake of marketing (winning the "fastest card" race). Low end cards similarly compromise on performance to win the "cheapest card" race.

You should be able to find *some kind* of benchmark for iGPUs. It might not be specifically video, but it doesn't matter. It's technically true that some cards may do a bit better at one type of tasks than other. But if you compare different score between some popular cards, the difference is not very interesting... Unless your job is to write overly long articles about the minutiae of each product. They're all an array of a couple 100k processors, it's kind of hard to bias too much towards one task or the other - especially by accident.

>Notwithstanding that, if I get a CPU with integrated graphics, would keeping my existing video card plugged in likely make a noticeable difference in performance? If it wouldn't, I think I might prefer to save the power draw.

Usually, only one card is active at a time - the one your monitor is plugged into. To make multiple cards work together, you have to do a bit of advanced set up. But it's really only worth it if the cards are similar power. For example, maybe the fastest card on the market is still not enough for you, so you try to double them up. The new card will often be much faster than your old one, 5x or 10x. Jumping through all the hoops of configuring the multi-card setup correctly might not be worth 10-20% more performance, plus there could be bugs that make the cards slower when running together and erode the tiny gains.

I don't think the power draw is a worry either way. Modern cards throttle themselves down when not active. Even if not, mid range cards are actually pretty energy efficient so there won't be any catastrophic loss. If you get inefficient top of the line cards, then you might want to watch your power usage.