Apple Silicon Mac mini Predictions

[thunng8]

Crypto is like 1 out of 3 areas http://support.primatelabs.com/kb/geekbench/interpreting-geekbench-5-scores I don't use GB very often as the free version uploads system info. I usually use Cinebench + Unigine Heaven to sanity check my computer builds.

Cinebench is a bad benchmark. It basically Shows 1 workload, while Geekbench covers a wide range of workloads. Geekbench is much better than cinebench at prediction performance.

note that every single sub test is a real world workload. Geekbench is not bad at all and correlates well to the industry standard spec benchmarks that is used to guage server and workstation performance.

Blog — NUVIA

Industry insights, NUVIA articles and blogs

nuviainc.com

 

[thunng8]

Yes, that sounds about right, so 1/3 of the score is biased towards a platform

crypto accounts for 5% of the score, hardly 1/3. Here is a white paper that describes the workload.

https://www.geekbench.com/doc/geekbench5-cpu-workloads.pdf

 

[theSeb]

Yes, quite right. I was too influenced by Linus T’s views on the matter. However, I’ve been reading through that document today and i am not convinced by many of the tests.

 

[thunng8]

Yes, quite right. I was too influenced by Linus T’s views on the matter. However, I’ve been reading through that document today and i am not convinced by many of the tests.

Why are you not convinced? It is not a perfect benchmark by any means, but given that it correlates well with spec which is an industry standard shows it has its uses. Btw, Linus view were of a very old version of Geekbench which had quite a few issues which mostly has been rectified. One of which is that encryption has a large weighting which is now reduced to 5%.

 

[theSeb]

Why are you not convinced? It is not a perfect benchmark by any means, but given that it correlates well with spec which is an industry standard shows it has its uses. Btw, Linus view were of a very old version of Geekbench which had quite a few issues which mostly has been rectified. One of which is that encryption has a large weighting which is now reduced to 5%.

The nature of the tests favours devices/CPUs with bursty performance and does not take into account the thermals/performance over longer units of work.

As an example, take a look at the compiler test. It’s compiling a project with around 1000 lines of code. That’s not a meaningful test. That’s a first year lab assignment.

I‘ll believe Geekbench when I see RDR 2 running on my iPad Pro better than on my PC, because according to Geekbench it should.

 

[theluggage]

I thought the CPU horsepower of AS is already there? Am I reading https://browser.geekbench.com/ios-benchmarks correctly? 1500+ single core score that no x86 chip has archived on average?

I'd take benchmarks like that with a moderate pinch of salt, especially when you're trying to compare across platforms.

Even "real world" benchmarks can vary enormously with different software versions, different compiler settings and gritty details (like whether the compression settings you use make the best use of the available hardware accelerators).

However, it depends what you think it demonstrates: does that benchmark show that an A12Z is really faster than a 15" Mac Pro? Dubious. Does it make it more believable that its Mac-focussed successor could comfortably beat the current MacBook Air or entry-level 13" MBP? Heck, yes - and that's all ASi gen 1 really needs to do.

Apple might have to come up with a new wrinkle when they come to replace the "Pro" machines and I suspect that they'll depend as much on what extra bells and whistles Apple builds onto the SoC as much as raw speed. However, speed isn't everything and a lot of people working in (say) audio production "fast enough - and silent" might be more important.

After all, if CPU (or GPU) horsepower is your primary requirement then you should have switched to PC years ago.

 

[thunng8]

The nature of the tests favours devices/CPUs with bursty performance and does not take into account the thermals/performance over longer units of work.

As an example, take a look at the compiler test. It’s compiling a project with around 1000 lines of code. That’s not a meaningful test. That’s a first year lab assignment.

I‘ll believe Geekbench when I see RDR 2 running on my iPad Pro better than on my PC, because according to Geekbench it should.

Well, spec takes hours to run and in the end the results of Geekbench and specCPU correlates very well.

 

[sanfrancisofont1984]

After all, if CPU (or GPU) horsepower is your primary requirement then you should have switched to PC years ago.

I'm a multi-device person. I recently did a Ryzen build running Linux but not giving up Apple ecosystem any time soon.

 

[theSeb]

Well, spec takes hours to run and in the end the results of Geekbench and specCPU correlates very well.

specCPU does not run on iOS. I must say that I don't see the correlation when using AMD CPUs.

I think though that this does illustrate the reasons behind my scepticism.

Dell Inc. PowerEdge R6525 vs HPE ProLiant DL385 Gen10 Plus - Geekbench

The only way I can explain it is that it looks like it's GPU related, but then it still makes no sense.

Edit: I should add that I truly hope that AS Macs are everything to everyone and their dogs or kittens. I hope that they have better thermals with quieter fans, or no active cooling at all. I dream about them being refreshed every year with the latest and greatest performance cores. I just have trouble believing that my (admittedly wonderful) iPad pro is quite as wonderful as some of these benchmarks would have me believe. I guess we shall see very soon.

 

[thunng8]

specCPU does not run on iOS. I must say that I don't see the correlation when using AMD CPUs.

I think though that this does illustrate the reasons behind my scepticism.

Dell Inc. PowerEdge R6525 vs HPE ProLiant DL385 Gen10 Plus - Geekbench

The only way I can explain it is that it looks like it's GPU related, but then it still makes no sense.

Edit: I should add that I truly hope that AS Macs are everything to everyone and their dogs or kittens. I hope that they have better thermals with quieter fans, or no active cooling at all. I dream about them being refreshed every year with the latest and greatest performance cores. I just have trouble believing that my (admittedly wonderful) iPad pro is quite as wonderful as some of these benchmarks would have me believe. I guess we shall see very soon.

In terms of how powerful the ipad pro is (2018 model) - I can edit/export 4k videos and work in Lightroom editing my raw files faster than my 2017 macbook pro (in fact a lot faster).

There's a whole article on the correlation of geekbench vs SpecCPU. It looks like geekbench predicts specCPU results very well, much better than any other benchmark out there.

Those Geekbench comparisons are user submitted - you don't know what else was running on the server at time of the test. That could explain the differences. Although the <10% variance you reported is not that significant. It could be just the thermal settings of the server under test - one is turbo boosting more than the other.

Blog — NUVIA

Industry insights, NUVIA articles and blogs

nuviainc.com

Also SpecCPU does run on iOS but there are no official submissions from Apple. SpecCPU comes as source code - the user needs to compile it. Anandtech compiled it has ran it and has reported the results. The A14 results should be reported by anandtech in the next week or 2.

 

[theSeb]

Cheers. I could not find any results on the results browser. I shall check it out. It’s been a long time since I read Anandtech.

 

[thunng8]

Cheers. I could not find any results on the results browser. I shall check it out. It’s been a long time since I read Anandtech.

Here it is for A13:

The Apple iPhone 11, 11 Pro & 11 Pro Max Review: Performance, Battery, & Camera Elevated

www.anandtech.com

Here is another recent one which includes Tiger Lake:

Intel’s Tiger Lake 11th Gen Core i7-1185G7 Review and Deep Dive: Baskin’ for the Exotic

www.anandtech.com

Here is the relevant chart:

https://images.anandtech.com/graphs/graph16084/111168.png

The other big new is that the i9-10900k uses 40W, the Tiger Lake i7 (28W) uses 20W and the Apple A13 uses 4W running this single threaded benchmark. We are looking at a massive perf/watt advantage for Apple. If A14 is approx 20% faster in SpecCPU as predicted by geekbench, we are looking at a core that is faster than the i9-10900k while using 10X less power. That is an exciting prospect for future macs.

Anandtech will definitely run specCPU for the Zen3, A14 and upcoming Apple Silicon macs in the future

 

[EugW]

My main wish list for an Arm Mac mini:

1) It’s lower priced than the Intel Mac mini, and I want to be able to get 16-32 GB RAM without breaking the bank. I’m not optimistic about this.

2) The fan isn’t loud. I’m not optimistic about this either but I can hope.

3) It can support both my dual-link DVI 30” Cinema HD Display (via USB dongles), and 4Kp60 4:4:4 DisplayPort with HDR (again via USB-C dongles), and has an HDMI 2.1 port. I expect all of this to be supported. (I already know that my 30” Cinema HD Display works with current USB-C Intel Macs like my 12” MacBook, using a USB-C to mini-DisplayPort dongle, which is in turn attached to an Apple mini-DisplayPort to dual-link DVI adapter.)

The nature of the tests favours devices/CPUs with bursty performance and does not take into account the thermals/performance over longer units of work.

That’s likely a less important concern with an Apple Arm Mac mini since the Mac mini will likely have a fan.

 

[thenewperson]

The fan isn’t loud. I’m not optimistic about this either but I can hope

Really? I feel like with all their 'performance per watt' talk this is the thing I'd be most optimistic about.

 

[macjustin]

I am looking to upgrade from my 2015 27" iMac and seriously hoping a new AS Mini will do the trick. I have never been a fan of AIO systems since they do not allow the separate replacement of things like display and I prefer a more modular ecosystem. So my fingers are figuratively crossed. Very, very crossed.

 

[jazz1]

I am looking to upgrade from my 2015 27" iMac and seriously hoping a new AS Mini will do the trick. I have never been a fan of AIO systems since they do not allow the separate replacement of things like display and I prefer a more modular ecosystem. So my fingers are figuratively crossed. Very, very crossed.

I to own a 2015 iMac with an i5 CPU and 32GB memory driving two Dell monitors. It actually performs very well.

I am seriously thinking about an ARM Mini for a second workstation in the house. It struck me that potentially the ARM MacMini might outperform the 2015 iMac. I suppose some of that supposition is that what ever graphics engine the ARM MacMini has would outperform what is my 2015 iMac. Then of course the 2015 iMac‘s i5 should trounce the ARM.

All the rumors of when the ARM Macs will be released, and the model type release schedule is driving us Apple fans crazy. But I love the suspense most days!

 

[macjustin]

I to own a 2015 iMac with an i5 CPU and 32GB memory driving two Dell monitors. It actually performs very well.

I am seriously thinking about an ARM Mini for a second workstation in the house. It struck me that potentially the ARM MacMini might outperform the 2015 iMac. I suppose some of that supposition is that what ever graphics engine the ARM MacMini has would outperform what is my 2015 iMac. Then of course the 2015 iMac‘s i5 should trounce the ARM.

All the rumors of when the ARM Macs will be released, and the model type release schedule is driving us Apple fans crazy. But I love the suspense most days! View attachment 974494

I have a similar setup and I am hoping the graphics processor can exceed the 2015 iMac. The suspense...oh, the suspense.

Edited the date of my iMac as I am not a time traveler.

 

[mr_roboto]

I am seriously thinking about an ARM Mini for a second workstation in the house. It struck me that potentially the ARM MacMini might outperform the 2015 iMac. I suppose some of that supposition is that what ever graphics engine the ARM MacMini has would outperform what is my 2015 iMac. Then of course the 2015 iMac‘s i5 should trounce the ARM.

Why do you think the i5 should trounce the ARM? Apple's A14 phone chip contains a high performance ARM core design which benchmarks significantly better than Intel's 2020 cores, which are in turn better than Intel's 2015 cores.

 

[jazz1]

Your right. I don’t know what I was thinking. Just hoping if the ARM is also the graphics engine it can attain the fullest ARM goodness, as well as great graphics performance. When all is said and done I’m looking forward to some multiple Geekbench comparisons between the Intel line and the new ARM Macs. Should be fun.

 

[OldGreyGuy]

I like the Mac mini format and I hope Apple continue with the format. It allows me to plug in a couple of 28" 4K screens for all of the screen real estate I need and with iCloud I can happily walk out of my office, go home and everything is there in case I decide to WFH that day and as I am largely a freelancer this gives me maximum flexibility.

An AS Mini with good specs (1Tb SSD, 32Gb RAM) and good thermal performance would be in my shopping list, perhaps not the first release as I upgraded to exactly that as the pandemic hit in March 2020, upgrading from a late 2012 model to the 2018 model because I needed additional screen space.

 

[MalcolmH]

I am not sure if you have ever used Geekbench, but when you do you may notice that it's doing a lot of cryptography tests. This skews the results because iDevices have hardware acceleration for many of these specific operations, whilst x86 does not. Back in the day one of the previous versions did not even stress a Mac CPU to cause throttling, so mobile Mac scores were biased against desktop Macs. Using one synthetic benchmark as a data point and trying to extrapolate that to real world applications is generally not a good idea.

Regarding the other benchmarks you've linked, this really shows the benefits of running on a cut down and super optimised operating system vs having to deal with the bloat of a "full" operating system such as Mac OS or Windows and the advantages of close cooperation between the software and the hardware.

Having said that there is no denying that Intel has been sleeping for a decade allowing AMD to catch up and become a player again. Meanwhile Apple has been super busy in the background creating some amazing chip designs. The fact is that since they announced it along with pretty concrete dates for the transition means that everything is planned and pretty much known to them. They are clearly super confident. You don't run a trillion dollar company by steering blindly into the fog and hope for the best and they aren't going to start doing that now.

Intel have had crypto instructions for at least ten years. https://en.m.wikipedia.org/wiki/AES_instruction_set

 

[Fomalhaut]

I have a similar setup and I am hoping the graphics processor can exceed the 2105 iMac. The suspense...oh, the suspense.

Unlikely...

The 2105 iMac model (announced mid-2105 in the new Apple Shiapparelli conference center on Mars, by the recently re-booted AI CEO "Cook-1000") has a 64000-core Apple Q5000 quantum-CPU array with 11-dimensional holographic memory providing 8 Exabytes of infini-mode storage. The Dilithium batteries will last for at least 20 years on a single charge.

 

[theSeb]

Intel have had crypto instructions for at least ten years. https://en.m.wikipedia.org/wiki/AES_instruction_set

Yes, but Apple’s hardware is superior in this because it’s more important in mobile

as per Geekbench:

’Superior AES performance can translate into improved usability for mobile devices’

 

[Fomalhaut]

I to own a 2015 iMac with an i5 CPU and 32GB memory driving two Dell monitors. It actually performs very well.

I am seriously thinking about an ARM Mini for a second workstation in the house. It struck me that potentially the ARM MacMini might outperform the 2015 iMac. I suppose some of that supposition is that what ever graphics engine the ARM MacMini has would outperform what is my 2015 iMac. Then of course the 2015 iMac‘s i5 should trounce the ARM.

Consider yourself trounced already...

Assuming you have the best 2015 27" i5 iMac (with Intel Core i5-6600), it scores 902/3100 in Geekbench 5 - similar to an iPad Air 3.

Even the first Apple Silicon offering is likely to beat the brand-new Intel 11th Gen i7 Tiger Lake laptop CPUs, and will probably replace at least the mid-tier 2020 Intel MBP13.

Expect nearly twice the performance of your iMac.

 

[theSeb]

We will see very soon I guess. I would be shocked if the performance differences are truly as large as Geekbench on iOS devices predicts