I benchmarked my old MP5,1 vs my new Mac Mini 8,1+eGPU

[tu2thepoo]

I started a thread in the Mac mini forum: https://forums.macrumors.com/thread...ac-mini-egpu-against-my-2009-mac-pro.2160637/

Long story short - the Mac mini was faster in almost every way, with very few regressions in performance (and most of those didn't matter to my use case). So if you don't need more than 6 cores and you've got the cash, it's totally doable to build a Mac mini that's less hassle and faster than the ol' cheesegrater.

 

[AlexMaximus]

I started a thread in the Mac mini forum: https://forums.macrumors.com/thread...ac-mini-egpu-against-my-2009-mac-pro.2160637/

Long story short - the Mac mini was faster in almost every way, with very few regressions in performance (and most of those didn't matter to my use case). So if you don't need more than 6 cores and you've got the cash, it's totally doable to build a Mac mini that's less hassle and faster than the ol' cheesegrater.

Lets put this in perspective for you.

Sure, sure.. the mini is indeed 8 years newer and therefore a lot better, and it has to be. But that was never the point. The point is NOT to put that 5K+ on the counter and spend it instead for a dive trip to Indonesia while keeping the cheesegrater and do the 16th little bitty upgrade for Mojave with a BIG grin on your face....

That's the reason why the cheesegrater is so sexy.

But hey, let's talk about your new shiny mac mini from 2018 in a full whopping eight years from now in 2026. I humbly invite you to zoom yourself into the year 2026 with a non-upgradable fully enclosed mac mini with T2 chip for controlled product obsolescence. Will you be able to postpone the next computer upgrade for another two to three years well into the year 2029 and do a dive trip instead?

This is basically where we cheesegraters do stand at the moment in the year 2018 from the viewpoint of the year 2010.

The ugly truth here is this: From the viewpoint of Apple, they absolutely hate us 5.1's They do not want their users to be able to use a computer for such a long extended time and be cool with it. Over the last 6 years, they degraded their computer product portfolio to "utility device status" to quicken the sales cycle and shorten the usable lifespan of your price increasing Apple products. They deleted all upgrade possibilities. The intentional stopping of e-GPU support for the MP6.1 or the battery manipulation software on your iPhone is just an example. There should be law suits...

In your case, you did not have much choice, because a real great product such as the 5.1 is just not available any more these days, -and never will be again. The test of time will show, how happy you will be with your mac mini in the year of 2026. I hope you don't sit on a mountain of controlled product obsolescence landmines.

If you want to know how companies think these days, watch this down below. It gives you a good primer.

But then again, did you really have free choice in the first place?

 

[fastlanephil]

There’s a newer Logic Pro X track count test available at Gearslutz>DAWs. My 5,1 Mac Pro scored 116 tracks. Another 5,1 user with a similar configuration also scored 116 and another 5,1 user scored a lower count of 109 tracks. The best 2018 six-Core Mac mini score that I’ve seen is 91 tracks. Other Mac mini users with similar configurations are scoring 71 to 83 tracks. That much of a difference could indicate CPU heat throttling.

 

[tu2thepoo]

There’s a newer Logic Pro X track count test available at Gearslutz>DAWs. My 5,1 Mac Pro scored 116 tracks. Another 5,1 user with a similar configuration also scored 116 and another 5,1 user scored a lower count of 109 tracks. The best 2018 six-Core Mac mini score that I’ve seen is 91 tracks. Other Mac mini users with similar configurations are scoring 71 to 83 tracks. That much of a difference could indicate CPU heat throttling.

It depends - I could totally hit the heat/power limits causing it to lock at 3ghz rather than 4ghz (see the end of my post). However, it never clocked under 3ghz under 100% load - so people need to plan and provision appropriately. So in the sense that many people won't see a sustained 4ghz, that's true - but I've never seen mine block below the baseline 3ghz.

 

[tu2thepoo]

Lets put this in perspective for you....But then again, did you really have free choice in the first place?

I stopped trying to prove how much I tweaked/tuned/whatever my computer a long time ago, but if that stuff gives you satisfaction then do you, brohim.

 

[pl1984]

Those were interesting results. What I found most interesting is how pathetic eGPU performance is compared to using the same card internally on an eight year old system. I was also interested in that you did not perform the Handbrake benchmark on the Mini using macOS (presumably so you could use the processors Quicksync feature). I'd be interested to see that test run using macOS so we could see how the Mini compares without Quicksync. It's also interesting the Mini, using Quicksync in Windows, was only 25% faster than the Mac Pro. I would have thought hardware performance would have been much higher than that especially given the Mac Pro is using ten year old technology (though there's likely some difference due to Windows).

 

[tu2thepoo]

Those were interesting results. What I found most interesting is how pathetic eGPU performance is compared to using the same card internally on an eight year old system.

In older/simpler benchmarks (Heaven/GFXbench) the Mac Pro has a lot higher peak performance, that's true. On more modern engines/benchmarks (Superposition, the actual games I played) the Mac Mini either matched the Mac Pro or was faster. So, "it depends".

I was also interested in that you did not perform the Handbrake benchmark on the Mini using macOS (presumably so you could use the processors Quicksync feature). I'd be interested to see that test run using macOS so we could see how the Mini compares without Quicksync. It's also interesting the Mini, using Quicksync in Windows, was only 25% faster than the Mac Pro. I would have thought hardware performance would have been much higher than that especially given the Mac Pro is using ten year old technology (though there's likely some difference due to Windows).

Quicksync on Coffee Lake processors supports the entire h.264 profile, so it's a lot closer (but not matching) x264 (CPU encoding) on quality/compression features. A lot of hardware encoders fail to support the more advanced compression profiles available so that's why they end up going 5-6x faster. For example, NVenc (on my Geforce 1080) on handbrake will do h.264/h.265 at like ~200fps, but the file sizes are ludicrously large. On an older i5 Skylake CPU, I could get Quicksync to encode h.264 at something like 90-100fps, but the quality was much worse and file sizes 2-3x larger than using CPU encoding.

I am curious to re-test the encoding performance under MacOS. Maybe later this weekend!

[edit] re-tested under MacOS and posted in the original thread

 

[StellarVixen]

Of course Mac Mini is gonna beat that dinosaur. It has faster CPU, faster DDR4 ram...the only thing it is lacking is graphics, but eGPU can come to the rescue.

I think comparing it with trash can Mac Pro would be more fair. I think Mac Mini kills it in single core CPU score, but loses in multicore. But again, Mac Mini has faster DRAM. Regarding gfx, I think it is gonna be close call, powerful GPUs are kinda bottlenecked by Xeon's lower single core score, but where only GPU power is required, I think Mac Pro performs better, as it has graphics directly on logic board PCIe slot.

 

[tu2thepoo]

Of course Mac Mini is gonna beat that dinosaur...I think comparing it with trash can Mac Pro would be more fair. ...

I figured that too, but I wanted to quantify how different they were in performance. I think a lot of people who don't have a specific reason for sticking with their cMP (whether it be massively parallel apps or wanting to be smarter than the sheeple) might be looking for that kind of comparison.

Also, I didn't have a trashcan MP to test with

 

[namethisfile]

Lets put this in perspective for you.

Sure, sure.. the mini is indeed 8 years newer and therefore a lot better, and it has to be. But that was never the point. The point is NOT to put that 5K+ on the counter and spend it instead for a dive trip to Indonesia while keeping the cheesegrater and do the 16th little bitty upgrade for Mojave with a BIG grin on your face....

That's the reason why the cheesegrater is so sexy.

But hey, let's talk about your new shiny mac mini from 2018 in a full whopping eight years from now in 2026. I humbly invite you to zoom yourself into the year 2026 with a non-upgradable fully enclosed mac mini with T2 chip for controlled product obsolescence. Will you be able to postpone the next computer upgrade for another two to three years well into the year 2029 and do a dive trip instead?

This is basically where we cheesegraters do stand at the moment in the year 2018 from the viewpoint of the year 2010.

The ugly truth here is this: From the viewpoint of Apple, they absolutely hate us 5.1's They do not want their users to be able to use a computer for such a long extended time and be cool with it. Over the last 6 years, they degraded their computer product portfolio to "utility device status" to quicken the sales cycle and shorten the usable lifespan of your price increasing Apple products. They deleted all upgrade possibilities. The intentional stopping of e-GPU support for the MP6.1 or the battery manipulation software on your iPhone is just an example. There should be law suits...

In your case, you did not have much choice, because a real great product such as the 5.1 is just not available any more these days, -and never will be again. The test of time will show, how happy you will be with your mac mini in the year of 2026. I hope you don't sit on a mountain of controlled product obsolescence landmines.

If you want to know how companies think these days, watch this down below. It gives you a good primer.

But then again, did you really have free choice in the first place?

Actually, the 2018 Mac Mini in the year 2029 will function just as well as in 2018. It might even gain performance via Software Optimizations that come along with MacOS updates. eGPU support will mature. It will in the most likeliest scenario be even faster in the year 2029 running a year 2023-2029 MacOS iteration (we don't know MacOS support down the road; Mac Mini being "Metal" capable and an iGPU with System RAM availability, it can be supported indefinitely depending on how MacOS evolves).

And, since a Mac Mini is mini, it will be a lot more graceful when it does fade into obsolescence.

A big, fat, heavy, chunky cheesegrater, in comparison, like the PM G5's before it will fade into obsolescence a little harder. Its demise will be more of a burden; its death coffin, more pricey; its funeral, not even far from where it sat!

Spending 5K on diving because you own a cMP5,1? What? Does. Not. Compute. Do you even know how to delegate PC money from Vacay money, bro?

This is true for the "trashcan" Mac Pro, as well. The grace into obsolescence part that is!

PS--And, I own a G5 Power Mac (found in recycling center) and a cMP5,1 with RX580! I dread the day this becomes a Power Mac G5!

PSS--And, my cMP5,1 is turned off 99% of the time because I don't wanna browse the web and consume 160 watts doing so!

 

[tu2thepoo]

There’s a newer Logic Pro X track count test available

Is this what it's supposed to look like? I got a copy of Logic Pro X but I've never used it before.

 

[pl1984]

I figured that too, but I wanted to quantify how different they were in performance. I think a lot of people who don't have a specific reason for sticking with their cMP (whether it be massively parallel apps or wanting to be smarter than the sheeple) might be looking for that kind of comparison.

Also, I didn't have a trashcan MP to test with

I do, if anyone is interested I'll perform the benchmarks tomorrow. It's a six core, 16GB, 256GB, D300 configuration.

 

[Synchro3]

I think Mac Mini kills it in single core CPU score, but loses in multicore.

Loses only a bit: https://browser.geekbench.com/mac-benchmarks

 

[tommy chen]

i love my old cheesegrater +g*

 

[fastlanephil]

[

Is this what it's supposed to look like? I got a copy of Logic Pro X but I've never used it before.
View attachment 811039

That’s the older Evans Logic Pro X track test. The new Logic Pro X track test was posted by TNM.

https://www.gearspace.com/board/app...y-new-benchmark-replace-dated-evans-test.html

 

[tpivette89]

It would be educational to see multiple tests between a Mac mini (late 2018) 6-core (either i5 or i7) vs a 6-core nMP... both with 16GB of RAM (for equal testing grounds). Obviously, the nMP is going to beat the Mac mini interns of graphical performance. But with everything else being equal, both systems behave similarly.

Both have 6-core processors, both have external expandability, both have non-upgradeable graphics (ok... the Mac Pro can upgrade from D300 - 700, but nothing else internally). The kicker is, both are around the same price (albeit new for the Mini vs used for the Pro).

Thing is, I already have a new Mini and a cMP... however, I've been thinking that a nMP would be an awesome addition to my 2018 Mini for long term rendering. Seems as though the Mini will hit it's thermal limit much sooner than the nMP... anyone have any experience with this in either FCPX or Handbrake between the 2018 Mini and the 2013 MP?

My machine is a 2009 flashed to a 5,1 with dual X5650 (2.66ghz 6-core CPUs, 36GB of RAM, and a HD 4870). I could get around $700 for either the complete system or parting out... while there is a nMP for sale locally on craigslist for $1000, but I think he would go as low as $900. Would anyone else upgrade?

 

[reukiodo]

...there is a nMP for sale locally on craigslist for $100, but I think he would go as low as $900. Would anyone else upgrade?

Surely this $100 is a typo with a missing '0' ? Otherwise the question should be: Why wouldn't anyone else spend $100 for the upgrade?

 

[thornslack]

Can I haz $100 nMP?

 

[spacedcadet]

The new Mac Mini is smoking for some uses.
https://macperformanceguide.com/MacMini2018-MemoryBandwidth.html

 

[Pval]

TLDR

1. Spending $1000+ will get you a faster CPU & RAM
2. Running at high resolution (4K), the impact of CPU and RAM on performance is negligible on high end GPUs

Same can be found in other benchmarks of the latest GPUs.

 

[tpivette89]

Surely this $100 is a typo with a missing '0' ? Otherwise the question should be: Why wouldn't anyone else spend $100 for the upgrade?

Doh... yeah there was a 0 missing. It’s listed for $1000.

 

[pl1984]

It would be educational to see multiple tests between a Mac mini (late 2018) 6-core (either i5 or i7) vs a 6-core nMP... both with 16GB of RAM (for equal testing grounds). Obviously, the nMP is going to beat the Mac mini interns of graphical performance. But with everything else being equal, both systems behave similarly.

Both have 6-core processors, both have external expandability, both have non-upgradeable graphics (ok... the Mac Pro can upgrade from D300 - 700, but nothing else internally). The kicker is, both are around the same price (albeit new for the Mini vs used for the Pro).

Thing is, I already have a new Mini and a cMP... however, I've been thinking that a nMP would be an awesome addition to my 2018 Mini for long term rendering. Seems as though the Mini will hit it's thermal limit much sooner than the nMP... anyone have any experience with this in either FCPX or Handbrake between the 2018 Mini and the 2013 MP?

My machine is a 2009 flashed to a 5,1 with dual X5650 (2.66ghz 6-core CPUs, 36GB of RAM, and a HD 4870). I could get around $700 for either r the complete system or parting out... while there is a nMP for sale locally on craigslist for $100, but I think he would go as low as $900. Would anyone else upgrade?

I am happy to perform any free benchmarks on my nMP (6 core, 16GB RAM, 256GB SSD, D300). If anyone has a specific benchmark they'd like me to run let me know. I offered before but I did not see any responses expressing interest.

 

[h9826790]

In older/simpler benchmarks (Heaven/GFXbench) the Mac Pro has a lot higher peak performance, that's true. On more modern engines/benchmarks (Superposition, the actual games I played) the Mac Mini either matched the Mac Pro or was faster. So, "it depends".

Quicksync on Coffee Lake processors supports the entire h.264 profile, so it's a lot closer (but not matching) x264 (CPU encoding) on quality/compression features. A lot of hardware encoders fail to support the more advanced compression profiles available so that's why they end up going 5-6x faster. For example, NVenc (on my Geforce 1080) on handbrake will do h.264/h.265 at like ~200fps, but the file sizes are ludicrously large. On an older i5 Skylake CPU, I could get Quicksync to encode h.264 at something like 90-100fps, but the quality was much worse and file sizes 2-3x larger than using CPU encoding.

I am curious to re-test the encoding performance under MacOS. Maybe later this weekend!

[edit] re-tested under MacOS and posted in the original thread

For your info.

I never try NVenc in handbrake, but in FFMpeg, my 1080Ti works really well. The quality is good, file size is controlled by bitrate. I use the follow simple code to compress 4K video (or transcode from HEVC to H264, etc).

ffmpeg.exe -hwaccel cuvid -i input.mkv -b:v 12000k -c:v h264_nvenc -c:a copy output.mkv

It's 20x faster than CPU encoding (W3690), and I really can't tell the difference in quality.

 

[tpivette89]

I am happy to perform any free benchmarks on my nMP (6 core, 16GB RAM, 256GB SSD, D300). If anyone has a specific benchmark they'd like me to run let me know. I offered before but I did not see any responses expressing interest.

Awesome! Your 6,1 is as close to similar specs as my Mini (6-core/16GB RAM)... is there a standard performance test in handbrake that we could both utilize that would compare the two systems (I don't use handbrake so wouldn't know)?

 

[Pval]

Handbrake doesn’t utilize your CPUs 100%

But maybe Cpumark will give you a hint:
W3690 : 9160
nMP : E5-1650v2 : 12724
Mini : i5-8500 : 11909