For much of the 2010s, we were stuck with mainly dual-core and quad-core CPUs in PCs. However the arrival of Ryzen shook the PC industry, causing a rapid increase in core counts. At the time, there was fervent discussion on this matter, with many questioning if more cores were worth it, and how many cores are more than enough?
So how do things stand today? The latest Intel and AMD consumer processors top out at 24 and 16 cores respectively. What extent of modern software can take advantage of all those cores? What modern workloads are still bottlenecked by single threaded performance?
depends on your usage, for gaming single thread is still king, but for tasks like video editing and 3d rendering, multithreading is crucial. everyone has different needs!
Single thread is far more important these days
Yes, but that’s starting from 6 cores.
Because 2 cores are not enough even if they are Raptor Cove 6 GHz cores.
In the sense that we take having 6 or 8 cores for granted
depends on your usage, for gaming single thread is still king, but for tasks like video editing and 3d rendering, multithreading is crucial. everyone has different needs!
depends on your specific use case and the software you’re running. different workloads benefit from different performance metrics.
Once you reach a certain number of cores (~6-8) it depends on the workload.
Windows/File Explorers/Browsers/Games/General Use are all better off with single thread performance at this amount of cores.
Multimedia/Editing/Rendering/etc are better off with even more cores.
There is a balance between the two which nobody can solidly answer since it varies by use case.
Single-core is more important for 99% of normal consumers. Most office productivity apps or web browsers are only lightly threaded.
Also, scaling cores efficiently is hard both in software and in hardware. 10 cores at 1x performance are going to be a lot more efficiently used than 20 cores at 0.5x performance.
First statement is definitely true, second statement is generally true but is super software and implementation dependent. Generally the issue isn’t fast vs slow cores it’s keeping all of those cores fed with data with memory and the OS scheduler handling events that makes a larger difference.
Obviously software makes the biggest difference of all in how many threads it allocates and where. This is an especially difficult issue as optimizing software for a server and for say a gaming PC are totally different problems with different optimal solutions
Once you reach a certain number of cores (~6-8) it depends on the workload.
Windows/File Explorers/Browsers/Games/General Use are all better off with single thread performance at this amount of cores.
Multimedia/Editing/Rendering/etc are better off with even more cores.
There is a balance between the two which nobody can solidly answer since it varies by use case.
Probably (it widely depends), it’s very useful to increase core counts to about 8 cores in most cases. After that, the actual practical returns are greatly diminished. So, moving from 8 to 16 grants only 15-20% better performance, etc.
I would much rather take 8 cores with Zen 5 performance per clock and frequency of 8 GHz rather than 128 cores with Zen 1 performance per clock and (all-core) frequency of 3.8 GHz. In a vast majority of workloads, the former would outperform the latter.
Zen 5 performance per clock and frequency of 8 GHz
May the Liquid Nitrogen be with you!
depends on your specific use case and the software you’re running. for gaming, single-threaded performance is still key, but for tasks like video editing and rendering, multi-threaded performance is crucial. it all comes down to what you need your CPU to do!
depends on your specific use case and the software you’re running. for gaming, single-threaded performance is still key, but for tasks like video editing and rendering, multi-threaded performance is crucial. it all comes down to what you need your CPU to do!
A medium amount of core count with excellent memory cohesion.
A medium amount of core count with excellent memory cohesion.
for me, it really depends on what you’re using your PC for. for my music production and video editing, multi-threaded performance is key. but for everyday stuff like web browsing, single-threaded performance matters too. different strokes for different folks, i guess!
CAD or at least Solidworks is still really only able to use 1 core
This gets repeated every time, and it’s nonsense. SolidWorks is mostly single-threaded, but it’s also not a demanding application. Even PCs from a decade ago can handle assembling medium complexity models just fine. The few features that are demanding, like rendering and analysis tools like FEA, are in fact multi-threaded.
And if you actually need model assemblies so complex that single-threaded performance in the base app would become a problem, you’d be running a Quadro card anyway, and CPU performance wouldn’t matter.
Source: Industrial designer who worked with SolidWorks, PT Creo and Rhinoceros for a while after uni before transitioning to a different area of design.
Depends on your workload.
For most of what I do with sound, video and textfiles of various formats, single thread is not very relevant.
Occasionally you may want to convert a long audiobook that comes in a single file and audio is generally single threaded. But that is still only a minor inconvenience because it also converts in the hundreds of real time speed on a single core.
And if it was important enough, then it’s totally possible to write scripts that divide the file in small portions to distribute amongst workers, exactly like it has been done for video conversion where much more compute is needed and this kind of scenario is more realistic.
This is the crux of the issue. People talk about “ST workloads” vs “MT workloads”, but the reality is that single-threaded workloads largely do not matter, it’s all stuff that, at the very worst, is done after a few seconds. MT performance, on the other hand, can save you hours of productivity depending on what your work is.
We are long past the point where ST performance matters for the “snappiness” of systems. Zen 3, Alder Lake, M1 and newer are all more than perfectly “snappy” in any modern system. Gaming is the last use case where ST matters so long as you have a minimum amount of cores, but for professional use cases there’s nothing to even discuss, MT is the only thing that matters.
Good MT performance is contingent on good ST performance. Doesn’t really matter how many threads you have if they’re slow individually. Which will dig a hole faster – 64 toddlers with gardening spades, or one excavator?
Depends on your workload
What about web browsing?
Web browsing RAM speed/size and crypto acceleration makes the biggest impact generally which is why Apple CPUs do so well there. That said web browsing is a pretty broad spectrum as they can run essentially arbitrary code depending on the website.
ANYTHING thats sold as new stock nowadays should do the job MORE than fine.
You would think so, but there are some really bad SoCs out there, using only Arm A5x series cores that have no business rendering a website and will give you an authentic early 2010s mobile web browsing experience.