Separate names with a comma.
Discussion in 'Mac mini' started by EugW, Nov 2, 2018.
Core i3-8100B quad core 3.6 GHz 4728 / 14375
Core i7-8700B hex core 3.2 GHz 5512 / 23516
The i7 outperforms the top-end 5K iMac (disregarding the GPU of course). Very impressive.
Yes. Which makes the lack of an iMac update even more significant.
Basically, the Mac mini is the guts of a 21.5" iMac, using the same class of CPUs, so this is the performance one would expect from an updated 21.5" iMac. The 27" models would be even faster... if they updated them that is.
I guess we could look at i5-8500 scores for a comparison.
4826 / 18231 from Generic Geekbench scores.
In theory the i7 looks a very solid update but the sustained turbo heat on that is going to be a concern.
I'd be happy with the i3, to be honest.
My 2017 iMac Core i5 7600 is not much faster.
The i7-8700 is still a 65 Watt class chip, at least in same. If Apple's cooling solution is decent, it might be OK. I would just hope the noise isn't too bad.
However, I'd probably buy the i5-8500. I suspect that is the best power utilization to performance balance. It still has 6-core performance, without the increased power utilization of HyperThreading, so it probably wouldn't get too hot quickly, meaning the fan would stay quiet for extended periods of sustained workload.
That is only a single result. There are currently only 3 results reported for the 2018 i3 Mini, so we need to wait for a few more results to get a more reliable picture.
That said, the average of those 3 results is 14259, which is only very slightly lower, and is still 25% faster than the previous most powerful Mini (2012 quad i7 2.6GHz).
Which is nice.
I think it's going to be a given that the i3 will run cooler than the i5 or i7. The i7 isn't out of the question on price alone but I'm waiting for results on sustained FCPX exports (real world).
In real terms, the i5 will heat up when it turbos (the i3 doesn't but has higher base clock for its fewer cores). The i7 gets even hotter because of the extra threads (and the turbo).
This explains why the i3 is fairly close to the i5 for single core but obviously the i5 and i7 thrash it for multicore.
Geekbench is only good for showing how well a system runs Geekbench. You shouldn't rely on it for anything else.
Yes Geekbench is a "synthetic" benchmark, but when I upgraded both CPUs in my Mac Pro 5.1, my score doubled, and my encoding time for videos was cut in half.
It's not entirely useless like you seem to think, it runs what it's creators thinks are the most common use of a CPU I think, so it give a relatively good idea of performances.
What did you upgrade from and to?
Yes, actually it is. For example I could pull one of the processors from this system (a dual 8 core / 16 thread Z620), run Geekbench and HandBrake encodes, reinstall the second processor, run the same benchmarks again and get double the Geekbench score but not double the HandBrake score. Why? HandBrake cannot fully utilize all of the threads this system has.
I went from the base CPU in a 5.1 dual CPU, to 2 x x5690. I don’t remember the single core score, but multi core I had around 12000 before, went to 25000 with the x5690.
I sold that 5.1 a while ago, so I can’t do the same test you're speaking about (removing a CPU).
I looked around a bit about what you’re saying (Handbrake not using all threads) and apparently it's not inherent to handbrake itself, but to which codec you encode your file with. Some are more « threads friendly » than others. In my case, I was ripping my dvd and blu-ray collection to h.265. Some people are saying the h.264 encoder is less multi-core/thread friendly and/or less « optimized ».
Some updates have been posted on GeekBench...
Core i7-8700B hex core 3.2 GHz 5958 / 26619
If legit, this makes the Mini faster than my maxed out 2010 cMP.
I'll find out in 3 days when my new Mini is delivered.
I'm in this camp. It seems Geekbench isn't the best measurement of real world power, or sustained performance under load.
--- Post Merged, Nov 4, 2018 ---
I agree that multi threading is a good way to inflate the scores, but not always leveraged by real world tasks.
--- Post Merged, Nov 4, 2018 ---
Based on those numbers, the i3 outperforms my 2013 quad-core Mac Pro trashcan.
Whatever the reason the reality is, at least for the particular parameters I used, the Handbrake transcode time did not halve despite the fact the Geekbench scores would have doubled.
This is the problem with Geekbench...it only shows how well a given system will run Geekbench.
Wow... those multi-core scores are exactly what my 12-core X5690 system got as well. Single-core scores are twice what my 5,1 scored. Amazing... because it's only half the cores.
Now that's really depressing (or awesome) depending on if you have one of those. Those 12 core Mac Pros were very expensive.
The X5690 was actually a faster processor than anything offered from Apple for the 5,1 Mac Pros.
And still are, relatively speaking. I wonder if this new Mini will put downward pressure on the Mac Pro systems...especially the lower core count 6,1 systems. They've remained stubbornly high despite their age.
You do wonder, following a bit of analysis with benchmarks, if some classic Mac Pro owners on a budget might view the 2018 Mac mini as a viable upgrade path. After all these years the advent of Coffee Lake with extra cores means that the horsepower is now able to compete with 2012 Mac Pros. Anyone needing GPU can now add that on, there's a 10 Gig Ethernet option for power users, the PCIe NAND Flash is presumably as fast as anything in the current Mac range and Thunderbolt 3 will take care of other needs.
Maybe this generation of Mini is not looking at PC switchers but classic Mac Pro switchers?
At some point technology will pass the classic Mac Pro by. We're close to that in a lot of situations but there are still some areas where the classic Mac Pro has benefit.
At least we know it will be a great geekbench machine . . . but there wasn't really any doubt about that.
And for some, that's all they need. But for those who need sustained throughput, cinebench (of the widely available benchmarks) does a better job of estimating performance.
Problematic sustained performance measurement with Geekbench is mainly a problem with fanless Macs. While I can see it being a problem with the small Mac mini too, the risk of that is significantly lower than with say the MacBook, because the Mac mini has a fan. I think the main problem here is the Mac mini i7-8700 will likely be very loud under heavy load.
That’s why without having experienced these first hand yet, I’d guess the best model to get for my type of usage would be the i5-8500. It would be fast, but not anywhere as hot as the i7-8700, and thus not anywhere near as loud as the i7-8700.
My guess is the i5-8500 will be the best balance between power vs fan noise, and that the Geekbench score will reflect its performance reasonably well vs other Macs.
There are several problems with Geekbench, you've hit on one.
Most of the bigger problems with Geekbench were fixed with Geekbench 4.
Its main problem now is it’s too short.