Thanks, what do you think the real world difference is in use, in particular with a video editing programme like Final Cut, or when editing RAW files in Lightroom?
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac mini i5 Tests
- Thread starter F-Train
- Start date
- Sort by reaction score
Thanks, what do you think the real world difference is in use, in particular with a video editing programme like Final Cut, or when editing RAW files in Lightroom?
That makes sense. Personally, I don't edit fast enough to keep up with these write speeds
I suppose it might make a difference to someone who is batch editing a lot of data.Let me try to draw an analogy. The ability to type 100 words/minute isn't much use if one is writing a business letter, let alone a novel.
Or am I missing the point?
https://forums.macrumors.com/threads/mac-mini-2018-128gb-ssd-speed.2153608/#post-26785124
This link shows how SSD size affects write speeds.
This link shows how SSD size affects write speeds.
Thanks, good link.
As discussed in posts 52 and 53 above, the question is what real world difference it makes.
I think that most people should be more interested in the fact that the read speeds of the i5/256GB/8GB and the i7/1TB/32GB, according to Blackmagic at least, are nearly identical.
Last edited:
Intel Power Gadget results for the i5 have been posted here: https://forums.macrumors.com/threads/mac-mini-i3-vs-i5-temps.2153401/page-3#post-26792869 I’m running my own test this morning, but won’t post unless my results are different, which I think unlikely.
In my case, the issue is whether to stay with the i5 or exchange it for an i7. I’ve posted my conclusion about whether temperature should be a factor in this decision here: https://forums.macrumors.com/threads/mac-mini-i3-vs-i5-temps.2153401/page-2#post-26792035
In my case, the issue is whether to stay with the i5 or exchange it for an i7. I’ve posted my conclusion about whether temperature should be a factor in this decision here: https://forums.macrumors.com/threads/mac-mini-i3-vs-i5-temps.2153401/page-2#post-26792035
When I first got it the room was silent. Nothing in the room running. Window and door closed. It was completely silent. Most of the time the fan is at about 1800rpm and you can really only hear it north of 3500rpm.
@archer75 has an i7
[doublepost=1542041415][/doublepost]Blackmagic Speed Test on a Samsung T5 500GB flash drive > Mac mini i5 Thunderbolt 3/USB-C port. I have this drive configured for two AFPS volumes. This drive was on sale in September/early October, and I'll be on the lookout for any discount on the 1TB and 2TB versions on Black Friday:
Last edited:
Blackmagic Speed Test on a Western Digital My Passport Pro 1TB Thunderbolt 1 5400RPM drive > Apple Thunderbolt 2 to Thunderbolt 3 adapter > Mac mini i5 Thunderbolt 3/USB-C port. The Apple adapter, at US$49, is expensive; but I have two of these drives (originally configured for a RAID) as well as Thunderbolt 2 drives, so the adapter is worth it to me:
Last edited:
I am going with i7. I want my Mini to last until Apple doesn't support it anymore.
And I am gonna try the liquid metal option what can go wrong,
I am having some doubts whether i7 can actually perform better than i5 due to thermal limitations, but it seems that i7 is not that thermally constrained, and outperforms the i7 by significant amount.
And I am gonna try the liquid metal option what can go wrong,
So, your conclusion is: get the i7?
I am having some doubts whether i7 can actually perform better than i5 due to thermal limitations, but it seems that i7 is not that thermally constrained, and outperforms the i7 by significant amount.
Liquid metal can't be used on aluminum heat sinks. I don't know what the heatsink in the mini is made of but something to consider.
I'm assuming that you meant i5 when you wrote "outperforms the i7 by significant amount".
The only conclusion that I've come to is that I'm not taking alleged thermal issues into account when deciding whether to keep the i5 or exchange it for an i7. I think that it's rabbit hole stuff, I have more confidence in Apple and Intel engineers than in MacRumors forum commenters, and in any event I have a three year AppleCare+ warranty
Last edited:
I just ordered an i7 / 16GB / 1TB.
Everyone has their own priorities but here are my thoughts:
i7:
I will let AppleCare take care of any thermal issues. I expect the performance gain i7 vs i5 to primarily be for short-duration tasks (e.g. compiling), prior to thermal throttling narrowing the gap. $200 for a few more smiles about fast response.
16GB:
I am replacing a '12 iMac with 24GB of RAM, and at first didn't want to get "less" than I had. But I went to an Apple store with an i3 8GB demo, and I ran every app from the launch bar, Terminal with yes>/dev/null, Safari with lots of tabs, etc. Memory pressure (vs "use") was still easily in the green. So twice that at 16GB is probably fine for me and this is the one thing I can change my mind on later.
1TB SSD:
I do a lot of file I/O so a fast internal SSD is the most important feature for me. Bigger SSD's provide the more storage, faster write speed, and with wear-leveling should last longer. Apple's SSD price to performance is a good deal (unlike RAM), so this was a no-brainer upgrade.
To everyone else, enjoy!
Everyone has their own priorities but here are my thoughts:
i7:
I will let AppleCare take care of any thermal issues. I expect the performance gain i7 vs i5 to primarily be for short-duration tasks (e.g. compiling), prior to thermal throttling narrowing the gap. $200 for a few more smiles about fast response.
16GB:
I am replacing a '12 iMac with 24GB of RAM, and at first didn't want to get "less" than I had. But I went to an Apple store with an i3 8GB demo, and I ran every app from the launch bar, Terminal with yes>/dev/null, Safari with lots of tabs, etc. Memory pressure (vs "use") was still easily in the green. So twice that at 16GB is probably fine for me and this is the one thing I can change my mind on later.
1TB SSD:
I do a lot of file I/O so a fast internal SSD is the most important feature for me. Bigger SSD's provide the more storage, faster write speed, and with wear-leveling should last longer. Apple's SSD price to performance is a good deal (unlike RAM), so this was a no-brainer upgrade.
To everyone else, enjoy!
Last edited:
Well...this (and other) threads (no pun intended) led me to do some tests, and I've learned something really interesting - and surprising!
It may be relevant for anyone considering the 6-core i5 versus the 6-core/HT i7 (like me).
I have an Rstudio script that runs a batch of CPU intensive jobs (8 jobs) in parallel. Because it is R, I can specify how many cores to spread the jobs across (i.e. how many instances of R). Up until today I have always been specifying the maximum CPU cores that my system presents (e.g. 8 for a 2011 i7 quad-core mac mini). I had always assumed that using the hyper-threading cores was a good thing to do...
Here is the crazy (to me) result.
Specifying 4 cores for 8 jobs: time for task = 1m 28 seconds
Specifying 6 cores for 8 jobs (i.e. two virtual cores) = 1m 38 seconds (!)
Specifying 8 cores for 8 jobs (i.e. 4 real+4 virtual cores) = 1m 28 seconds.
Specifying 2 cores for 8 jobs = 2 min 20 seconds.
Here is the activity monitor log: The big chunks of activity are those three tests. Ignore the small spieks in between. As you can see, the correct number of cores are being used in each case.
So, what I have learned is that hyperthreading (for these tasks) is completely useless - and potentially detrimental! For the 6 core specification, it looks like 6 jobs were done slightly faster than 8 jobs, but then the remaining 2 jobs had to be started afterwards, giving the worst overall time.
What amazes me is that the 4 core test gave identical times to when 8-cores (4 real/4 HT) were tested.
I've literally just today ordered the 6-core i7 mac mini, and I really wonder if it a bit of a waste!
The only benefit I can potentially see is from the higher turbo boost speed (4.6 for i7 versus 4.1 max for i5). Hyper-threading doesn't seem like it is going to help at all.
Comments?
It may be relevant for anyone considering the 6-core i5 versus the 6-core/HT i7 (like me).
I have an Rstudio script that runs a batch of CPU intensive jobs (8 jobs) in parallel. Because it is R, I can specify how many cores to spread the jobs across (i.e. how many instances of R). Up until today I have always been specifying the maximum CPU cores that my system presents (e.g. 8 for a 2011 i7 quad-core mac mini). I had always assumed that using the hyper-threading cores was a good thing to do...
Here is the crazy (to me) result.
Specifying 4 cores for 8 jobs: time for task = 1m 28 seconds
Specifying 6 cores for 8 jobs (i.e. two virtual cores) = 1m 38 seconds (!)
Specifying 8 cores for 8 jobs (i.e. 4 real+4 virtual cores) = 1m 28 seconds.
Specifying 2 cores for 8 jobs = 2 min 20 seconds.
Here is the activity monitor log: The big chunks of activity are those three tests. Ignore the small spieks in between. As you can see, the correct number of cores are being used in each case.
So, what I have learned is that hyperthreading (for these tasks) is completely useless - and potentially detrimental! For the 6 core specification, it looks like 6 jobs were done slightly faster than 8 jobs, but then the remaining 2 jobs had to be started afterwards, giving the worst overall time.
What amazes me is that the 4 core test gave identical times to when 8-cores (4 real/4 HT) were tested.
I've literally just today ordered the 6-core i7 mac mini, and I really wonder if it a bit of a waste!
The only benefit I can potentially see is from the higher turbo boost speed (4.6 for i7 versus 4.1 max for i5). Hyper-threading doesn't seem like it is going to help at all.
Comments?
Attachments
Last edited:
There's a few reasons why hypertheading can be slower. The main ones are a) cache thrashing and b) synchronization overhead. It's impossible to tell from what you've listed here which (or both) are responsible.
But today you've learned why one does not just assign threads to logical cores willy nilly without testing
Don't forget the i7 also has a larger cache in addition to the higher boost clock. Even at identical clocks it will be faster than the i5.
The one thing that I know for sure is that if I decide now to exchange the i5 for an i7, it will be based on the chance that the i7 will perform better on some tasks.
I do not have any actual facts at the moment to support that belief, at least on the uses that matter to me, which are 4K video and RAW photo editing; nor do I know, if i7 advantages exist, exactly what they are and what practical difference they would make.
I should add that I have decided that I will not be taking alleged thermal issues into account in making my decision, so it all comes down to computing performance.
I suspect that there are a lot of people in this position who ultimately resolve the issue by saying to themselves "It's just another $200". Of course, human nature being what it is, they'll invent some reason that makes it sound like they know what they are talking about
I do not have any actual facts at the moment to support that belief, at least on the uses that matter to me, which are 4K video and RAW photo editing; nor do I know, if i7 advantages exist, exactly what they are and what practical difference they would make.
I should add that I have decided that I will not be taking alleged thermal issues into account in making my decision, so it all comes down to computing performance.
I suspect that there are a lot of people in this position who ultimately resolve the issue by saying to themselves "It's just another $200". Of course, human nature being what it is, they'll invent some reason that makes it sound like they know what they are talking about
Last edited:
The one thing that I know for sure is that if I decide now to exchange the i5 for an i7, it will be based on the chance that the i7 will perform better on some tasks.
I do not have any actual facts at the moment to support that belief, at least on the uses that matter to me, which are 4K video and RAW photo editing; nor do I know, if i7 advantages exist, exactly what they are and what practical difference they would make.
I should add that I have decided that I will not be taking alleged thermal issues into account in making my decision, so it all comes down to computing performance.
I suspect that there are a lot of people in this position who ultimately resolve the issue by saying to themselves "It's just another $200". Of course, human nature being what it is, they'll invent some reason that sounds less arbitrary
Some of the hype about the i7 vs. the i5 has always been about Intel's legacy marketing. The i7 brand has always been marketed towards businesses, home businesses and desktop professionals. The i7 CPU does have more onboard cache which should lead to improved performance regardless of the application or whether HT is involved or not. Then there are the "professional" features such as HT that the i7 offers that are supposed to appeal to professionals.
One other issue not discussed is in regards to future support of the hardware which probably doesn't effect Mac users as much as it does Windows, Linux and other OS users (or users who run VMs requiring Windows support). Intel releases long term driver updates and security update support for their CPUs and system boards. In the past the i7 CPUs have tended to receive updates longer into the future than other lower tier models. Most professional grade software developers use the i7 as the industry standard for it's feature set and legacy (past software and driver) support. This usually doesn't effect Mac users as much since Apple usually provides it's own frequent MacOS driver updates and hardware support.
Therefore that extra $200 is not really all about raw performance but you are paying partially for a legacy Intel branding that you know will have full industry wide support. This doesn't mean the i5 won't also be fully supported as well. It's just that the Intel i7 has become an industry standard bar that has been set for desktop computer users.
Last edited:
Or $180 on EDU! I am thinking that if I pull the trigger on a mini the i5 or i7 decision at this point will be based on how badly I want to spend that extra $180 at that time.The one thing that I know for sure is that if I decide now to exchange the i5 for an i7, it will be based on the chance that the i7 will perform better on some tasks.
I do not have any actual facts at the moment to support that belief, at least on the uses that matter to me, which are 4K video and RAW photo editing; nor do I know, if i7 advantages exist, exactly what they are and what practical difference they would make.
I should add that I have decided that I will not be taking alleged thermal issues into account in making my decision, so it all comes down to computing performance.
I suspect that there are a lot of people in this position who ultimately resolve the issue by saying to themselves "It's just another $200". Of course, human nature being what it is, they'll invent some reason that makes it sound like they know what they are talking about
One other question. Does anyone believe the i7 will add longevity to the machine over the i5? If I am thinking this will be around for 5-8 years, will the i7 age any better?
If you are an Mac user who likes to upgrade every three to five years then the i7 should hold it's resale value better than an i5. If you are someone who likes to keep your older Macs then this is a non-factor.
Last edited:
Yeah - what I find a bit misleading is the activity monitor. With four cores active it shows as if there is still only half the CPU being used, with another four full CPU cores available - yet at shown above, in most cases, when I load them up with another 4 instances of R, they max out, but no extra work is actually getting done! It seems a total con...
There seems to be only one instance where the HT cores were helpful: that was when there were 6 jobs. In this case, specifying 6 threads finished significantly faster than 4 threads.
But given that result, I cannot fathom why, with 8 jobs, why 8 threads doesn't finish faster than 4 threads.
It seems like HT threads are only worth deploying if the total number of jobs isn't divisible by the number of real cores.
EDIT: Some more numbers:
10 jobs: on an i7 4 core/8Thread CPU:
4 instances of R: = 1m 42 sec
5 instances of R: = 1m 26 sec
6 instances of R: = 1m 26 sec
8 instances of R: = 1m 38 sec
10 instances of R: = 1m 30 sec (I only have 8 simultaneous threads available to the CPU, so each R thread was running below 100%, even when CPU was maxed out at 800%).
Here are CPU usage charts. With 4, 5, 6, 8 and 10 R instances, you can see the point when some of the jobs finish and the CPU usage drops off (even though the entire task is not completed).
So...none of this really makes much sense.
There seems to be only one instance where the HT cores were helpful: that was when there were 6 jobs. In this case, specifying 6 threads finished significantly faster than 4 threads.
But given that result, I cannot fathom why, with 8 jobs, why 8 threads doesn't finish faster than 4 threads.
It seems like HT threads are only worth deploying if the total number of jobs isn't divisible by the number of real cores.
EDIT: Some more numbers:
10 jobs: on an i7 4 core/8Thread CPU:
4 instances of R: = 1m 42 sec
5 instances of R: = 1m 26 sec
6 instances of R: = 1m 26 sec
8 instances of R: = 1m 38 sec
10 instances of R: = 1m 30 sec (I only have 8 simultaneous threads available to the CPU, so each R thread was running below 100%, even when CPU was maxed out at 800%).
Here are CPU usage charts. With 4, 5, 6, 8 and 10 R instances, you can see the point when some of the jobs finish and the CPU usage drops off (even though the entire task is not completed).
So...none of this really makes much sense.
Last edited: