Geekbench Scores: Single vs Mult-Core Scores?

Questions on how we spend our money and our time - consumer goods and services, home and vehicle, leisure and recreational activities
Post Reply
User avatar
Topic Author
RooseveltG
Posts: 660
Joined: Sun Sep 07, 2008 2:56 pm
Location: The Rust Belt

Geekbench Scores: Single vs Mult-Core Scores?

Post by RooseveltG » Tue Aug 13, 2019 8:45 am

I am in the market for a Mac mini.

I notice that it has an extremely high Single-Core Geekbench Score and a much lower (relative to other computers) Multi-core Score.

Can anyone explain the significance of this? When is the computer faster and when would it lag?

Thanks in advance for any clarification.

Roosevelt.

aednichols
Posts: 35
Joined: Mon Nov 21, 2016 4:44 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by aednichols » Tue Aug 13, 2019 9:01 am

Most consumer applications use only one core at once.

Unless you have workflows that specifically benefit from parallelization like audio/video encoding or compiling code, a fast single-core score is good enough.

If you have an Apple store near you, they are very willing to let customers test drive the computers in store for as long as they want.

User avatar
jhfenton
Posts: 4104
Joined: Sat Feb 07, 2015 11:17 am
Location: Ohio

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by jhfenton » Tue Aug 13, 2019 8:13 pm

RooseveltG wrote:
Tue Aug 13, 2019 8:45 am
I am in the market for a Mac mini.

I notice that it has an extremely high Single-Core Geekbench Score and a much lower (relative to other computers) Multi-core Score.

Can anyone explain the significance of this? When is the computer faster and when would it lag?

Thanks in advance for any clarification.

Roosevelt.
aednichols explained it well. For most folks, the single core performance matters more for a quick feel in day-to-day usage.

Which Mac mini are you looking at? I have a 2018 Mac mini with a 6-core 3.2 GHz i7, and it has excellent multi-core performance. I do do video encoding, so I wanted the multi-core goodness. (The version of the i7 processor in the 2018 mini also has hyper-threading with 12 virtual cores--2 per real core--so when I do video encoding, I see CPU usage reported at greater than 1100%. The versions of the i3 and i5 processors in the 2018 Mac mini do not do hyper-threading, so don't perform as well on multi-core tests.)

User avatar
whodidntante
Posts: 5867
Joined: Thu Jan 21, 2016 11:11 pm
Location: outside the echo chamber

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by whodidntante » Tue Aug 13, 2019 9:40 pm

*Nerd alert.*

Multi-core processors are extremely common today and are even used in some low-cost embedded devices. The reason is not higher performance as some believe, but higher performance for workloads that can exploit a multi-core design, at a given price point and power budget. And frankly, we were reaching limitations for single-core performance with known HW architectures and fabrication processes, and some of the steps taken to eek a bit more performance reached the point of absurdity. So system designers began favoring multi-core processors.

If you want to see a single-core processor that demonstrates this clearly, look at the Intel Pentium 4 (Netburst architecture). It's a power hog but had the best general-purpose computing single-core processing capability available at the time. It also did a reasonable job as a foot warmer for those cold winter nights.

A lot of people who ask will benefit more from faster single-core processing. But if you do video editing, simulation, or like to install gobs of malware from the Internet, good multi-core performance can help quite a bit. Processors do not perform 4x faster than a single core if they have 4 cores, for many reasons. Shared resources like floating-point units, bus contention, thermal limitations/throttling, synchronization overhead, context switching overhead, workloads that are not optimal, etc.

randomguy
Posts: 7968
Joined: Wed Sep 17, 2014 9:00 am

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by randomguy » Tue Aug 13, 2019 10:10 pm

aednichols wrote:
Tue Aug 13, 2019 9:01 am
Most consumer applications use only one core at once.

Unless you have workflows that specifically benefit from parallelization like audio/video encoding or compiling code, a fast single-core score is good enough.

If you have an Apple store near you, they are very willing to let customers test drive the computers in store for as long as they want.
I don't think that has been remotely true for a while now. Take your web browser. It will have a half dozen threads (granted most of them will be IO bound) downloading the various parts of the web page, doing imagine decoding, and running javascript. Most games these days use things like the unity engine which also exploits threads. And even for apps which aren't explicitly multithreaded, the OS and background apps can suck up those extra cores.

Now there is a definite limit on how many cores you want for most uses. Your average app doesn't need 112 cores like a server does. Somewhere in the 4-6 range seems to be the current sweet spot but it has been creeping up over the years. 5 years ago we would have said 2-4. Is it worth paying for those extra cores and slightly faster single CPU performance? That is hard to say and will depend strictly on usage.

mountains
Posts: 34
Joined: Sun Jun 17, 2018 10:06 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by mountains » Tue Aug 13, 2019 11:03 pm

randomguy is right.
Multi-core performance is what matters today even with basic workloads like browsing and email.
I think that 4 cores is still quite sufficient for most/many use-cases today though.

All that being said: unless you are doing some very heavy duty work the MacMini will be plenty fast. So I would make the decision based on whether you like Macs and its features.

(It's also worth considering if it's worth spending more and getting a MacBook Air or Pro. You can still use them with an external keyboard and monitor when at your desk but you can also take them without if you need to.)

User avatar
whodidntante
Posts: 5867
Joined: Thu Jan 21, 2016 11:11 pm
Location: outside the echo chamber

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by whodidntante » Tue Aug 13, 2019 11:21 pm

randomguy wrote:
Tue Aug 13, 2019 10:10 pm
I don't think that has been remotely true for a while now. Take your web browser. It will have a half dozen threads (granted most of them will be IO bound) downloading the various parts of the web page, doing imagine decoding, and running javascript.
For a lot of business/consumer apps threads are used primarily for software development convenience. It was easier to write the code that way than to have one thread that did everything. And it makes things more responsive to input along the way. It may or may not be faster, end to end. And if it's faster, it is not a linear speedup.

User avatar
Topic Author
RooseveltG
Posts: 660
Joined: Sun Sep 07, 2008 2:56 pm
Location: The Rust Belt

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by RooseveltG » Wed Aug 14, 2019 8:52 am

Thanks for all the responses.

I have an older Macbook Pro and see the spinning beach ball more than I would like. I also almost exclusively use my laptop as a desktop so I am considering the Macbook Mini. Not sure about the lack of a real video card in the mini so I am still on the fence.

Roosevelt.

02nz
Posts: 2107
Joined: Wed Feb 21, 2018 3:17 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by 02nz » Wed Aug 14, 2019 9:01 am

Geekbench only measures processing performance, not the performance of your storage (HDD or SSD). And if your older MacBook Pro has a mechanical (spinning) hard disk, that's probably what causes the beach ball. SSDs are so much faster than HDDs that going from an older computer w/ HDD to one with an SSD will feel like a quantum leap in performance.

BTW the integrated graphics is total fine if you're not a heavy gamer.

Jon
Posts: 32
Joined: Tue Feb 18, 2014 9:03 am

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by Jon » Wed Aug 14, 2019 9:15 am

Add me to the group of nerds who thinks that multi-core performance is increasingly important. For a modern operating system running modern software, I no longer recommend less than four cores, even for general-purpose use.

That said, all the Mac Minis will probably suit your needs. If you create media or compile code, you might want to opt for a 6-core model, but otherwise it isn't necessary.

Definitely don't buy a system with a mechanical hard drive. Solid state drives are the only way to go in 2019.

mmmodem
Posts: 1668
Joined: Thu May 20, 2010 1:22 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by mmmodem » Wed Aug 14, 2019 9:42 am

Looks like you're getting all opinions from the single-core versus multi-core speculators. Suffice it to say, some applications perform better with multi-cores. Some applications don't because of the way the program was written. Without knowing what programs you use normally, we can't say which is more of a benefit to you. It doesn't really matter what we think on this anyway since all you care about is minimizing the spinning beach ball.

And if that's the sole purpose then your next computer should have an SSD minimum 200 GB, minimum 8 GB of RAM. And you're going to get a lot of contrary opinions on what minimum is but I think everyone will agree "the more the merrier."

And finally Mac mini or MacBook pro? Simple, for the same dollar you get more performance from a Mac Mini, I'd get a Mac mini if there is no need for the portability.

Dedicated video? My opinion, you'll have to pay a pretty penny to get the option to buy it on an upper trim Macbook Pro. If you're not sure I'd save 2/3 the cost and get the mini. $2400 vs $800.

02nz
Posts: 2107
Joined: Wed Feb 21, 2018 3:17 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by 02nz » Wed Aug 14, 2019 12:17 pm

We don't know what age the "older MacBook Pro" is, but OP might have concluded that the Mac Mini isn't a big jump in multi-core performance if he's looking at the base Mac Mini configuration. With a higher configuration (more CPU cores) there's a bigger jump:

2012 MacBook Pro: 3574 single, 12003 multi
2015 MacBook Pro: 3810 single, 13101 multi
Current Mac Mini (i3-8100B, 4 cores): 4686 single, 14003 multi
Current Mac Mini (i7-8700B, 6 cores): 5661 single, 24327 multi
Intel NUC i7-8559U, 4 cores (this is the maybe the closest thing the PC world has to a Mac Mini): 5201 single, 17656 multi

But again, any of these systems will feel plenty snappy in normal use as long as it has an SSD.

User avatar
jhfenton
Posts: 4104
Joined: Sat Feb 07, 2015 11:17 am
Location: Ohio

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by jhfenton » Wed Aug 14, 2019 12:45 pm

02nz wrote:
Wed Aug 14, 2019 12:17 pm
But again, any of these systems will feel plenty snappy in normal use as long as it has an SSD.
+1 The 2018 Mac minis all come with PCIe-based SSDs. They scream. (Prices are high, though, for the larger sizes, so I only bought a 256GB internal SSD, and use a 2TB external SSD. The external SSD only gets 500-600 MB/s, a fraction of the speed of the internal, but it's hard to tell the difference for most purposes.)

Edited: Corrected my GB/MB typo.
Last edited by jhfenton on Wed Aug 14, 2019 7:07 pm, edited 1 time in total.

02nz
Posts: 2107
Joined: Wed Feb 21, 2018 3:17 pm

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by 02nz » Wed Aug 14, 2019 1:01 pm

jhfenton wrote:
Wed Aug 14, 2019 12:45 pm
02nz wrote:
Wed Aug 14, 2019 12:17 pm
But again, any of these systems will feel plenty snappy in normal use as long as it has an SSD.
+1 The 2018 Mac minis all come with PCIe-based SSDs. They scream. (Prices are high, though, for the larger sizes, so I only bought a 256GB internal SSD, and use a 2TB external SSD. The external SSD only gets 500-600 GB/s, a fraction of the speed of the internal, but it's hard to tell the difference for most purposes.)
Looks like the current Mac Mini has the SSD soldered on. If it were m.2 you could get a TB for around $100. Ironically the other offender on this is Microsoft with its Surface devices. The upcharge from 128GB to 256GB on the Surface is $200-300 - or more than 10X the retail cost of that extra storage!

User avatar
jhfenton
Posts: 4104
Joined: Sat Feb 07, 2015 11:17 am
Location: Ohio

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by jhfenton » Wed Aug 14, 2019 7:06 pm

02nz wrote:
Wed Aug 14, 2019 1:01 pm
jhfenton wrote:
Wed Aug 14, 2019 12:45 pm
02nz wrote:
Wed Aug 14, 2019 12:17 pm
But again, any of these systems will feel plenty snappy in normal use as long as it has an SSD.
+1 The 2018 Mac minis all come with PCIe-based SSDs. They scream. (Prices are high, though, for the larger sizes, so I only bought a 256GB internal SSD, and use a 2TB external SSD. The external SSD only gets 500-600 GB/s, a fraction of the speed of the internal, but it's hard to tell the difference for most purposes.)
Looks like the current Mac Mini has the SSD soldered on. If it were m.2 you could get a TB for around $100. Ironically the other offender on this is Microsoft with its Surface devices. The upcharge from 128GB to 256GB on the Surface is $200-300 - or more than 10X the retail cost of that extra storage!
Yep. It's really the only disappointment on the 2018 Mac mini. (The RAM is upgradeable, albeit a bit awkwardly.) You have to pay a sizable premium for the convenience of having a larger, super-fast SSD built-in. You can get similar performance (2500 MB/s) with a Thunderbolt 3 external SSD, but you pay through the nose for those too (˜$400 for 1TB). I'm content to settle for 500-600 MB/s with a much cheaper USB 3.1 interface.

randomguy
Posts: 7968
Joined: Wed Sep 17, 2014 9:00 am

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by randomguy » Wed Aug 14, 2019 9:21 pm

whodidntante wrote:
Tue Aug 13, 2019 11:21 pm
randomguy wrote:
Tue Aug 13, 2019 10:10 pm
I don't think that has been remotely true for a while now. Take your web browser. It will have a half dozen threads (granted most of them will be IO bound) downloading the various parts of the web page, doing imagine decoding, and running javascript.
For a lot of business/consumer apps threads are used primarily for software development convenience. It was easier to write the code that way than to have one thread that did everything. And it makes things more responsive to input along the way. It may or may not be faster, end to end. And if it's faster, it is not a linear speedup.
It is easier cause people want responsive apps.:) Nobody wants to stare at the beach ball anymore. And yes it definitely isn't linear. Adding a core for most tasks gets you 20-30% speed ups not 100% unless you happen to be something that splits up nicely.

For the OP, the beach ball is often driven by hard disk speeds (i.e. SSD versus spinning disks is very noticeable in terms of responsiveness. It isn't the throughput that really matters for most cases. It is the seek time) and sometimes by running out of ram (if the computer runs low it starts using the hard disk as RAM storage. That is horrible for performance). You might want to look at what the memory usage looks like when your computer slows down. 8 might be enough. 16 almost definitely for most users.

The base mac mini is a pretty solid. The question is do you want to pay 300 bucks more for something like 30% faster? That is very hard for anyone to say. It will come down to how often the computer isn't fast enough AND how much you value 300 bucks.

User avatar
Topic Author
RooseveltG
Posts: 660
Joined: Sun Sep 07, 2008 2:56 pm
Location: The Rust Belt

Re: Geekbench Scores: Single vs Mult-Core Scores?

Post by RooseveltG » Fri Aug 16, 2019 7:46 am

Valuable input appreciated.

I have a 4-year-old MBP with an SSD. I may wait until Mac OS Catalina arrives but am wondering if clean installs are worth doing in a Mac.

Roosevelt.

Post Reply