- Aug 19, 2019
- 2,934
- 11,447
It depends what he's got, but in general, yes, you're right. It really depends on the tasks at hand. For instance, for most of my gaming, my Ryzen 9 3900X is absolutely beastly, it runs like a solid champ. When it comes to something like Star Citizen, though, I'd be better off with a 5600 3D or whatever it is. It can't be clocked up as well as my chip, and doesn't have the raw computational output, but that huge cache drastically boosts performance over most standard chips.I generally consider replacing the CPU to be pointless, especially if it has the penultimate generation of top-end Ryzen / Core. In each generation of new CPUs, we get only a small increase in performance, Moore's law is dead, and even more so, as you said, the main work is done by GPUs. It would be an extra expense for no purpose.
I still have the old Haswell gen Core i5, it's only 4/4 and it handles most tasks, even modern games.
Most games, not unlike what we all play here, are quite adequately run on much older hardware. My last desktop was an i7-7700k, the one before that (which I gave to my housemate) is an i7-970. Six cores, 48GB of DD3 in triple-channel mode. The base clocks are lower, but despite that, it is still a beast of a processor. Very few programs that we run on our machines are really running into significant hardware bottlenecks. Prior to AMD pushing out the Ryzen 2 series, despite many revisions in their CPU lineups, processor improvements had been practically dormant for a decade.
Given a choice, though, I would absolutely still do all the rendering with a machine that was at least no more than 4 years old on the CPU/mobo architecture if I had a choice, and that's where DPC is at right now. He's got a choice. As long as the new GPUs don't need better than what he's got (PCIe 4.0, for instance) then I suspect he'll be just fine doing only GPU upgrades for his rendering work.
If they come out with a significant memory architecture change/boost, that would be cool. Anything that speeds up the moving of the data is a good thing, but it also needs to be cost effective and semi-practical. I don't expect anyone, gamer or developer, with any sense to pay 3x the cost for only a 10% conditional improvement.
But since we won't see the new episode until next year now, I guess we'll see what DPC gets himself for Christmas.