So the multi-core results were 50% faster in the i7 which has 100% more cores? Why aren't the results closer to 100% faster?
Because the number of cores is only one of of the many factors that influence overall system performance. A single computer with 4 CPU cores is not functionally equivalent to 4 computers with 1 CPU each. Think about the job that you do for a living. What if someone could clone 3 more of you to do your job. Keep in mind that everything that you need to do your job, like an office, a desk, a box of tools, a computer, the car you drive to work, etc., now has to be shared by you and your 3 clones. Do you think your job would get done 4 times faster?
The reality of the actual speedup delivered by multiprocessing-capable general purpose computer systems is actually worse than what shows up in benchmarks. The vast majority of productivity software either isn’t optimized for multiprocessing and/or the algorithms used to solve the problems at hand would not benefit from multiprocessing,
On the other hand, when software is specifically written to take advantage of multiprocessing capabilities the speedup can be very good, but never linear with the number of cores because of shared resource bottlenecks and the overhead associated with making multiprocessing work, such as scheduling access to shared resources like the CPU cores, memory, storage, IO, etc. So having more cores almost always provides a net benefit to the overall computing performance and experience for a given system, but not what you may expect.
dewme said: On the other hand, when software is specifically written to take advantage of multiprocessing capabilities the speedup can be very good, but never linear with the number of cores because of shared resource bottlenecks and the overhead associated with making multiprocessing work, such as scheduling access to shared resources like the CPU cores, memory, storage, IO, etc. So having more cores almost always provides a net benefit to the overall computing performance and experience for a given system, but not what you may expect.
This is what I've experienced over the years, going from single-cores to multiple. Everything works more smoothly. Some of that is probably OS improvements, but I think a lot of it comes down to having more cores allows more of them not to get tied up with stuff that used to create stutter/lag. Plus, these days, a lot of us run multiple apps at once, where some might be doing background things.
Years ago, I had to be so careful when doing things like screen or audio recording. Now, I can run Folding@home on a couple cores, while a video encode is running, or be doing a screen recording of a game I'm playing or some CAD app that is also fairly resource heavy, and it all seems not to care that much. That's why I'll almost always take more cores over higher performance ones. (As I might have mentioned earlier, I even turn off Turbo Boost, as doing so seems to mostly keep my cores inside the thermal capabilities of my 2018 mini. So, I even prioritize quiet over all out performance.).
I’m surprised there was not an even larger boost with the i7.
It’s the thermals. Apple limits the voltage to stay roughly within the 10W TDP. By contrast, with the 13” Pro, even the base will boost up to about 25W despite the 15W nominal TDP because it has better cooling.
They are a bit all over the place, particularly the multi core. My scores are slightly lower than Apple Insider’s. But the really low scores are likely because Spotlight is running and heating up the machines. 1150-1200/3300-3400 is about normal.
So the multi-core results were 50% faster in the i7 which has 100% more cores? Why aren't the results closer to 100% faster?
These are the Y-Series chips. They are up to 10W now but have a maximum up-capacity of 12W. By contrast the chip in the base 13” Pro is nominally 15W but can go up to 25W. Thus the 13” Pro multi core score is about 4x as fast as the single core score, while the multi-score form is only about 2.5-3x as fast. All 4 cores are running, but likely only around 1.7GHz sustained. The single core score is based on the burst speed (3.2 GHZ on the i3 and 3.8 GHz on the i7).
So the multi-core results were 50% faster in the i7 which has 100% more cores? Why aren't the results closer to 100% faster?
What if someone could clone 3 more of you to do your job. Keep in mind that everything that you need to do your job, like an office, a desk, a box of tools, a computer, the car you drive to work, etc., now has to be shared by you and your 3 clones. Do you think your job would get done 4 times faster?
I’m surprised there was not an even larger boost with the i7.
Yes, going from 2 cores to 4 cores + hyperthreads would ordinarily be a bigger boost. The results suggest the memory controller may be handcuffing the i7. And why would someone splurge for the i7 and not also get 16GB RAM?
Why would anyone spend the money on an i7? It'a an Air. ¯\_(ツ)_/¯
Comments
On the other hand, when software is specifically written to take advantage of multiprocessing capabilities the speedup can be very good, but never linear with the number of cores because of shared resource bottlenecks and the overhead associated with making multiprocessing work, such as scheduling access to shared resources like the CPU cores, memory, storage, IO, etc. So having more cores almost always provides a net benefit to the overall computing performance and experience for a given system, but not what you may expect.
Years ago, I had to be so careful when doing things like screen or audio recording. Now, I can run Folding@home on a couple cores, while a video encode is running, or be doing a screen recording of a game I'm playing or some CAD app that is also fairly resource heavy, and it all seems not to care that much. That's why I'll almost always take more cores over higher performance ones. (As I might have mentioned earlier, I even turn off Turbo Boost, as doing so seems to mostly keep my cores inside the thermal capabilities of my 2018 mini. So, I even prioritize quiet over all out performance.).
These are the Y-Series chips. They are up to 10W now but have a maximum up-capacity of 12W. By contrast the chip in the base 13” Pro is nominally 15W but can go up to 25W. Thus the 13” Pro multi core score is about 4x as fast as the single core score, while the multi-score form is only about 2.5-3x as fast. All 4 cores are running, but likely only around 1.7GHz sustained. The single core score is based on the burst speed (3.2 GHZ on the i3 and 3.8 GHz on the i7).