So........when are the applications that use more cores coming?
Because it makes perfect sense to spend on more cores that aren't able to be used.
Eh, yeah chicken or the egg sort of problem. If more cores isn't the norm, developers can't very well make use of them. Can't utilize something that most people don't have. But that really does go both ways.
It starts making sense when they become more affordable. People may not need the cores, but to me if I can have more for what I used to be happy paying for less, then why wouldn't I take the freebie alongside performance boosts in other areas? And that's what a lot of people are starting to think. Enough people get on that, your average PC has more cores, and those can then be put to better use.
Or at least that's the hope. There is the problem of many tasks inherently being serial. I don't know of a way around that without rebuilding a whole lot from the ground up. But I suppose it's always possible that more parallel solutions may arrive, simply because the option is there... ...it gives devs a reason to think outside of the box. Though realistically it would be a fundamental shift in how we accomplish base-level tasks. Not happening anytime soon. Too many problems simply aren't solved that way. Don't get me wrong, I think we're seeing more than 4 cores used a lot more often these days, but it'll still be a while before we see the gains we saw with speed increases back in the day, if we ever do. Optimization can only go so far.
The thing I don't get about this whole "no more cores" thing is... ...what else should we be expecting instead? What would the next revolution be? What else is there to do currently? Gotta sell those CPU's mann... ...can't rely on node shrinks and clock speed boosts so much anymore. Now all that's left is to tweak existing architectures and work on better multicore packages. That's just where we're at as far as I can see.
I guess the real question is what these extra cores should really be worth vs things like IPC or just straight performance on serial loads. With AMD's additions, the price per core/thread is going down significantly and I really don't see how that could be a bad thing. But it does raise the question of where the point of competition really lies. Is adding more cores really competitive? I mean, for most folks... ...a faster CPU with fewer cores will still serve them better. Most of us here get it, I think, but a lot of folks still don't see how its apples to oranges between high clock speed, low core count chips and low clock speed high core count chips. AMD is bringing more IPC now, while Intel is working in more cores, but I've never seen AMD and Intel as competing factions.
I will say it is getting to a point where for Joe Shmoe, AMD is starting to look like the better preposition. They don't perform as well as Intel stuff, but for Joe Shmoe they really don't have to, so long as they cost significantly less. That they sometimes wind up with more cores than they need is incidental. And I think for a lot of people buying mid-range Ryzen machines right now, the core/thread count isn't the main selling point. It's the fact that they are fast
enough and cost significantly less to step up to.
You won't hear me complaining about more cores. I'll tell you, nothing bogs down a midrange dual or quad core CPU like running a bunch of live audio tracks in Reaper, many running double-digit vst plugins with high quality filtering alongside complex vsti synths and other instruments. I can now live track and monitor two guitars at once with all of that going on in massive projects because of this $160 6c/12t CPU. For me, that's a damned good value! I look back on all of those years of having entire sessions lost to overload instability, or having to wait for tracks to do a temporary bounce so that I could then track more or work with others without crazy latency.... ....breaking projects up and smushing together bounced tracks that are mixed by proxy.
But that is a very specific use case. Even most audio stuff still doesn't benefit - Reaper is kind of the exception in that it seems to really like spreading larger loads across threads. VST instruments are going that way little by little. In my case it was money well spent and saved - this will serve me well for some time to come, maybe even more later. I guess if you need more cores now, you already know you do. And if you're not sure, take a few anyway... ...but only if they throw themselves at you. It's nice to have a little bit of overhead.
All that said, I still find it hard to justify 8C/16T right now, though I think it's cool to see more of them popping up all the same. Again, if you really need that, you shouldn't have to ask if you do. I see both sides of it. I get not wanting to pay for silly theoretical performance, and the obsession with core count is pretty much exactly what marketers want right now, even though it's still more pragmatic to have maybe 6 at most. But I mean... ...the ship has sailed. Ultimately elevated core count is just the latest of many little inches forward. This is the progression now, until something better comes along.
I say don't pay extra more for cores, but if the most cost-effective option happens to have more cores, then lucky you! I guess...
