The programs only have to be able to use any of those cores. You could run 36 single-threaded programs, and each could run on its own virtual processor core.
But wouldn't these processes each be "residing" on one core, and avoiding being swapped out and back in to memory, as they would be on a processor with fewer cores?
Hi Bozo
Short answer is YES and NONE, for most desktops.
Although Timwhit's point that browsers use multiple threads is technically true (and the OS can run multiple processes across different cores), it doesn't lead to better performance once you have enough cores to avoid saturation of a single core. Queuing theory shows that by far the biggest improvement is gained when comparing a single core with a dual core. There are raw processing tasks (eg 7z archiving) that scale with more cores, but this soon peters out due to network or storage performance constraints. Considering a browser is limited by the speed of internet requests (and the web server), 18 cores seems a trifle excessive ...
If you mean the host OS, no.
It's simply Intel's response to AMD's high core count CPUs. Even the typical power user can't take advantage of either, but Intel already had the CPUs being sold as Xeons, so they disable ECC make an i9 SKU and sell them for big money to the high end enthusiast market. More money for Intel with virtually no additional effort and they don't leave that niche for AMD to exploit alone.
For what? I have a feeling your "need" doesn't actually exist. You're not running a 3-way or 4-way SLI setup, so...I just need more PCIe lanes not all those cores.
Of course it will work. However, I don't think you understand what you're talking about. The CPU has PCIe lanes and so does the chipset. For example, the Z170 chipset has 20 lanes and the CPU has 16 lanes. So the graphics card connects to the CPU's 16 lanes. That leaves 20 from the chipset for everything else. How does that not work, even with the peripherals you listed?
And the Z370 chipset for Coffee Lake has 24 more lanes. So there are 40 total to work with in a Coffee Lake system. The 16 from the CPU will go to the graphics card. The 24 from the chipset will support the rest of your crap.
Yeah, but what actual use case are you going to push more data than that simultaneously? Even a 10GbE copy to or from a NVMe x4 drive isn't going to hit that (presuming it actually all has to go through the CPU and can't be handled DMA style directly by the chipset).
Fools with money.
Threadripper 1900x. "Just" eight cores/16 threads, but boatloads of PCIe lanes.I just need more PCIe lanes not all those cores.
I think that you mean IPC, and, no, it's not as good as Coffee Lake. My feeling is that most people wouldn't notice the difference.
He's not going to notice that either.
Because that's all they put on the CPUs. You still can't seem to put together any sort of specific hypothetical use case where it's going to actually pose a noticeable limit to yourself though. Or even, how about some benchmarks showing where it's a real limit vs. the Intel CPUs with more lanes?
I guess the moral to the story is that if you basically tread water in your product development in terms of IPC performance for almost a decade even AMD can eventually catch up.
