iMac Pro - First Hands On Thread

[bplein]

OK, time to move this out of the order thread.

I'm not going to show you an unboxing. If you are thinking of buying this, don't watch any unboxing videos, enjoy the quality packaging for the first time yourself (grin).

Ships with 10.13.2.

Basic "throughput plumbing test". Not a real benchmark but capable of telling us what idealized sequential throughput these are capable of:

Next, Geekbench CPU and GPU

https://browser.geekbench.com/v4/cpu/5708336
https://browser.geekbench.com/v4/compute/1629884

By the way, the GPU section completes in about 26 seconds. The app warns you it will take 2-5 minutes (grin).
[doublepost=1514080734][/doublepost]I'll continue to post first impressions here. Everyone else with one, feel free to jump in, I'm not claiming this thread exclusively for me
[doublepost=1514080833][/doublepost]Next impression: My iMac (late 2012 3.4GHz i7) is moving Lightroom original files to an external drive. It's much louder right now than the silent iMac Pro. But I haven't pushed the Pro at all yet. More to come later after I try a few games and import Lightroom and have it rebuild all the previews and look for faces.

 

[frankmcpartland]

Thanks for sharing your thoughts. Curious to hear if the iMac Pro can cool itself under stress without getting too loud.

 

[maxq]

That's my exact spec, so it's fascinating to see how it goes!

 

[Trebuin]

OK, time to move this out of the order thread.
View attachment 743482
I'm not going to show you an unboxing. If you are thinking of buying this, don't watch any unboxing videos, enjoy the quality packaging for the first time yourself (grin).

Ships with 10.13.2.

Basic "throughput plumbing test". Not a real benchmark but capable of telling us what idealized sequential throughput these are capable of:
View attachment 743481
Next, Geekbench CPU and GPU

https://browser.geekbench.com/v4/cpu/5708336
https://browser.geekbench.com/v4/compute/1629884

By the way, the GPU section completes in about 26 seconds. The app warns you it will take 2-5 minutes (grin).
[doublepost=1514080734][/doublepost]I'll continue to post first impressions here. Everyone else with one, feel free to jump in, I'm not claiming this thread exclusively for me
[doublepost=1514080833][/doublepost]Next impression: My iMac (late 2012 3.4GHz i7) is moving Lightroom original files to an external drive. It's much louder right now than the silent iMac Pro. But I haven't pushed the Pro at all yet. More to come later after I try a few games and import Lightroom and have it rebuild all the previews and look for faces.

Could you download AmorphousDiskMark & run that as well as Cinebech R15 I think is current. You'll want to post the CPU & GPU individually. If you want to check the power status, download Intel Power Gadget to see what the CPU does during the Geekbench & Cinebench tests.

 

[Bradleyone]

My first thought was I really shouldn't keep refreshing this thread, I've got work to do. Then I thought, "This will save me time when mine arrives!" So I will!

My iMac (late 2012 3.4GHz i7) is moving Lightroom original files to an external drive. It's much louder right now than the silent iMac Pro. But I haven't pushed the Pro at all yet.

When you do start loading it, do you think you could notice how the fan ramps, how frequently it changes rpm/pitch?

One of the things that annoy me is the way some fans are constantly ramping up and down depending on load. I had last years' MBP Touch Bar and it was constantly changing depending on what I was doing. Whereas, Mac Pro 2013 eases in and out over longer duration.

 

[Trebuin]

The benchmark trends on Geekbench are starting to turn interesting, more 10 core single threads are matching the 8 cores:
https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=iMacPro1,1

 

[Bradleyone]

Probably exceeding it, on average.

But that's a pretty huge variance even within the core: 4544 - 5462. Is that normal?

(Never been a big fan of Geekbench, except as a predictor of how fast my apps will run Mac vs Mac, relatively speaking.)

I hope that's not within chip and is dependent on whether user shut down their other processes.

 

[Trebuin]

Probably exceeding it, on average.

But that's a pretty huge variance even within the core. Is that normal?

(Never been a big fan of Geekbench, except as a predictor of how fast my apps will run Mac vs Mac, relatively speaking.)

I hope that's not within chip and is dependent on whether user shut down their other processes.

No, it's not supposed to be normal in real world results. I keep thinking that the reason we see this is because the CPU has a large uptick on the clock (4.5 on the 10 core) & if the tester is actually running some software that causes the CPU to start the test hotter, it will actually clock down to the base temp & give us a more accurate result, which should be closer to no difference between single thread 8 & 10 core. Some have already brought this up, but I think this was done to help sell the 10 core for those who know about the importance of the single thread performance (os & gaming performance). I used to use intel burnintest to heat up the CPU prior to testing to assure I know what the real world performance would be.

 

[bplein]

Thanks for sharing your thoughts. Curious to hear if the iMac Pro can cool itself under stress without getting too loud.

I ran a few GPU benchmarks (older ones) from Unigine, both the Heaven and Valley benchmarks. The iMac Pro was 2-3x faster than my late-2012 3.4 GHz i7 iMac. Neither machine made too much extra noise: The GPU on the Pro hit 90-95C and the fan barely had to turn to keep it from going higher.

Anyone have any good benchmarks to run to push the temps up?
[doublepost=1514086597][/doublepost]

Probably exceeding it, on average.

But that's a pretty huge variance even within the core: 4544 - 5462. Is that normal?

(Never been a big fan of Geekbench, except as a predictor of how fast my apps will run Mac vs Mac, relatively speaking.)

I hope that's not within chip and is dependent on whether user shut down their other processes.

The new Firefox eats up 30-40% of a single core just sitting idle on the Macrumors page LOL. So yes, it could be how people have run different things in the background.
[doublepost=1514086644][/doublepost]

Could you download AmorphousDiskMark & run that as well as Cinebech R15 I think is current. You'll want to post the CPU & GPU individually. If you want to check the power status, download Intel Power Gadget to see what the CPU does during the Geekbench & Cinebench tests.

I downloaded Cinebench and will do the AmorphousDiskMark too.

 

[Trebuin]

I ran a few GPU benchmarks (older ones) from Unigine, both the Heaven and Valley benchmarks. The iMac Pro was 2-3x faster than my late-2012 3.4 GHz i7 iMac. Neither machine made too much extra noise: The GPU on the Pro hit 90-95C and the fan barely had to turn to keep it from going higher.

Anyone have any good benchmarks to run to push the temps up?
[doublepost=1514086597][/doublepost]

The new Firefox eats up 30-40% of a single core just sitting idle on the Macrumors page LOL. So yes, it could be how people have run different things in the background.
[doublepost=1514086644][/doublepost]

I downloaded Cinebench and will do the AmorphousDiskMark too.

Thanks. If you want to run up the temps, see if you can find a Mac compatible intelburnin test. Heaven should have been able to up the GPU tests, but let me see if I can find something better.

Edit: Intelburntest was apparently a windows app I used to use. I'll have to find something better. Intel Power Gadget may not work with modern CPUs, but it's worth the try

 

[bplein]

Cinebench R15 results. I tried the Intel Power Gadget but it drops frames while running other benchmarks in the foreground. I've seen this before, it's a fairly useless tool IMO.

 

[Trebuin]

THOSE ARE FANTASTIC!

https://www.notebookcheck.net/Mobile-Graphics-Cards-Benchmark-List.844.0.html

Set it to show desktop cards & ignore the SLI (multiple tied cards). This thing actually crushes nVidia

Edit:
Ok, so the nVidia GTX still beats it, but this is still very good.

 

[bplein]

Amorphous Disk Test with all default settings:

[doublepost=1514088332][/doublepost]I am seeing some weirdness with my iMac Pro bluetooth keyboard. Every once in a while it gets behind my typing and then essentially gets stuck. Like it's lagging on a 9600 baud dialup. I can fix it quickly by turning it off and on.

At this desk I currently have a BT mouse on my old iMac, a BT Magic Mouse on my laptop, a BT keyboard on my laptop, and the keyboard and mouse for the iMac Pro. And my phone.

So I am not sure if I'll see this in the long run. If I do, Apple's going to hear about it.

 

[Trebuin]

What all these benchmarks people don't mention is the 4K results. 158.5/56.26 is really good
Attached is Samsung's 950 Pro m.2 numbers. The app you're running isn't perfect as it is ported, but run on windows might be more accurate. I'll test mine when I get things up & running.

Reason I care about 4k: that's the majority of your file operations & the main real world performance index of hard drives. People have been pushing sequential which is great if you're transferring a large file from computer to computer or computer via USB drive, but not if you're running operations & loading 2GB of map files constantly.

I also am aware of a bug in the port that will show on the second test on some devices & show abnormally high readings.

Thanks for posting these!

 

[bplein]

OK, I used a trivial way to chew up CPU, first 8 threads and then 16.

MacFanControl doesn't know how to measure the fan speed and gets it all wrong. Ignore those numbers. Also included are screen shots of smcfancontrol and how it is confused too.

Core temp ramped up north of 90C and the fan didn't become audible over the slightly elevated fan of my adjacent iMac which was running at 47C/2051rpm (smcfancontrol default settings).

This iMac Pro really is a lot quieter than the 2012 iMac and certainly from what I've heard about the 2014 i7 iMac.

Sorry about the screen captures, this is my first Retina Mac and the files were too fat! I had to shrink the dimensions and the DPI to display OK, probably shrunk them too much.(CORRECTED LATER: FIRST RETINA DESKTOP)

 

[Trebuin]

deleted
[doublepost=1514091107][/doublepost]

OK, I used a trivial way to chew up CPU, first 8 threads and then 16.

MacFanControl doesn't know how to measure the fan speed and gets it all wrong. Ignore those numbers. Also included are screen shots of smcfancontrol and how it is confused too.

Core temp ramped up north of 90C and the fan didn't become audible over the slightly elevated fan of my adjacent iMac which was running at 47C/2051rpm (smcfancontrol default settings).

This iMac Pro really is a lot quieter than the 2012 iMac and certainly from what I've heard about the 2014 i7 iMac.

Sorry about the screen captures, this is my first Retina Mac and the files were too fat! I had to shrink the dimensions and the DPI to display OK, probably shrunk them too much.(CORRECTED LATER: FIRST RETINA DESKTOP)

View attachment 743497 View attachment 743496 View attachment 743495 View attachment 743499
View attachment 743498

Sweet. Could you post the memory tab in "about this Mac?"

Also, that's ironically the way I temp the Mac when I can't find the right software.

So this is showing that the CPU isn't even dropping to the bottom at that temp? That's fantastic. I'd love to see what the 10 core does. If anyone has a 10 core, can they do the same thing?

 

[bplein]

i was able to get the fans working by running War Thunder (Steam) on 5K resolution, full screen, on "High" settings. Wonderful and smoother than 1440p on my 2012 iMac. No hard numbers. The fans were running but it wasn't that loud. Really quieter than my iMac under a similarly taxing load.

The system comes with 4x16GB modules

 

[phobos]

Keep the tests coming! Awesome stuff.
It’s amazing to hear that it’s as silent or even more silent than the late 2012 imac.
For me this is also the iMac it’s going to replace so it’s also going to be my benchmark. And since the late 2012 imac is 99% silent I’m ecstatic!
Looking forward to get my order!

 

[Bradleyone]

Yes, great tests, thank you for taking the time to do them and to post them.

Well worth storming the FedEx compound!

 

[thedeske]

Well Done and thanks to
bplein

 

[FredT2]

The benchmark trends on Geekbench are starting to turn interesting, more 10 core single threads are matching the 8 cores:
https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=iMacPro1,1

It appears that all of the fastest 10-core scores are from the machines with 128 GB, yes?

 

[AlexJoda]

If this thing is for double the money not a least twice as fast as my old trusty non maxed out 5K iMac (4 core i7 with only a 4GB Radeon 575 GPU) it seems too slow or too expensive or both.... (no offense...don't take it personally...we still have Christmas Eve and best wishes for all new iMac Pro owners...)

The Cinebench CPU Test is only faster for multi-core jobs. Single-core is faster with the 5k....

 

[driftless]

It appears that all of the fastest 10-core scores are from the machines with 128 GB, yes?

Which is probably why those were the specs of the iMac Pro's that were given to the early reviewers.

 

[FredT2]

Which is probably why those were the specs of the iMac Pro's that were given to the early reviewers.

Could very well be. On average it looks like a 4-5 percent difference. There doesn't seem to be any difference between 32 and 64 GB for the 8-core machines.

Did anyone notice that 14-core entry from December 20?

 

[c0ppo]

For everyone who has their hands on iMac Pro, well, first off, congrats. I envy you
But I'm still holding off on purchasing one. Mostly because it's way overpowered for my needs (app developer, game dev, but not AAA+ titles, so...), and lastly, I'm waiting for some decent reviews.

Could anyone just post on how the fans work? On iMac 27, 2017 with an i7 CPU, fans are quite annoying. At least to me. Can you hear them on iMac Pro during some normal stuff like browsing, music, and some simple apps? And how about under full load?