Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

SFjohn

macrumors 68020
Original poster
Sep 8, 2016
2,225
4,526
I calculate the 18 core would be approx 20% faster than the 10. However that’s just multiplying numbers... Does anyone know how the larger L3 memory would factor into speed? Single treaded it looks like the 10 core is 5% faster than the 18 core... For arguments sake, assume cost is not a factor. But overall performance is. Uses are Final Cut X, Logic, Lightroom, some Photoshop (not much). Thoughts?
 
I calculate the 18 core would be approx 20% faster than the 10. However that’s just multiplying numbers... Does anyone know how the larger L3 memory would factor into speed? Single treaded it looks like the 10 core is 5% faster than the 18 core... For arguments sake, assume cost is not a factor. But overall performance is. Uses are Final Cut X, Logic, Lightroom, some Photoshop (not much). Thoughts?
The cache actually makes a significant difference in different jobs. Because of that, no one can completely accurately estimate the performance difference. The closer the cache, the closer the estimate; the larger the difference in cache, the larger the error in the estimate, and the larger the differences in different benchmarks.
 
I calculate the 18 core would be approx 20% faster than the 10. However that’s just multiplying numbers... Does anyone know how the larger L3 memory would factor into speed?..Uses are Final Cut X, Logic, Lightroom, some Photoshop (not much)...

Even with perfect buses, infinitely fast caches, zero multi-core hardware contention, and heavily multi-threaded code, the code itself will constrain multi-core speedup. Each execution thread is not totally independent of others and must periodically synchronize work with them. Besides that a certain % of the execution path cannot be parallelized. Only a small % of serialized code will cap maximum multi-core speedup; this is called Amdahl's Law: https://en.wikipedia.org/wiki/Amdahl's_law

Using the base Xeon frequency and assuming 100% parallel code, the 18 core might be 27% faster. Using the turbo boost frequency it would be more. But who knows the exact turbo boost behavior for a given workload? This varies with the exact CPU and many other factors.

There is supposedly a max "all core active" boost but I don't know the number for the specific Xeon-W "B" variants used in the iMac Pro: https://www.pugetsystems.com/blog/2...-What-You-See-Is-Not-Always-What-You-Get-675/

Before the iMac Pro, the fastest Apple computer any FCPX user ran was (depending on workload), a 2017 top-spec iMac 27 or a 12-core D700 Mac Pro, or some modified older Mac Pro or Hackintosh.

Momentarily setting aside the 18-core iMac Pro, the 10-core Vega64 version is likely about 2x faster than any prior Mac anybody has run, esp. on FCPX. So the question is how much faster than 2x improvement would the 18-core machine be on a FCPX, Logic and Lightroom workload? And if it was (say) 25% faster would it be worth it?

For certain highly parallel code it makes a difference as has been shown here on the latest benchmarks on Apples iMac Pro page. However these are carefully selected -- see the footnotes. For some benchmarks they don't show the 10 vs 18 core versions: https://www.apple.com/imac-pro/
 
Even with perfect buses, infinitely fast caches, zero multi-core hardware contention, and heavily multi-threaded code, the code itself will constrain multi-core speedup. Each execution thread is not totally independent of others and must periodically synchronize work with them. Besides that a certain % of the execution path cannot be parallelized. Only a small % of serialized code will cap maximum multi-core speedup; this is called Amdahl's Law: https://en.wikipedia.org/wiki/Amdahl's_law

Using the base Xeon frequency and assuming 100% parallel code, the 18 core might be 27% faster. Using the turbo boost frequency it would be more. But who knows the exact turbo boost behavior for a given workload? This varies with the exact CPU and many other factors.

There is supposedly a max "all core active" boost but I don't know the number for the specific Xeon-W "B" variants used in the iMac Pro: https://www.pugetsystems.com/blog/2...-What-You-See-Is-Not-Always-What-You-Get-675/

Before the iMac Pro, the fastest Apple computer any FCPX user ran was (depending on workload), a 2017 top-spec iMac 27 or a 12-core D700 Mac Pro, or some modified older Mac Pro or Hackintosh.

Momentarily setting aside the 18-core iMac Pro, the 10-core Vega64 version is likely about 2x faster than any prior Mac anybody has run, esp. on FCPX. So the question is how much faster than 2x improvement would the 18-core machine be on a FCPX, Logic and Lightroom workload? And if it was (say) 25% faster would it be worth it?

For certain highly parallel code it makes a difference as has been shown here on the latest benchmarks on Apples iMac Pro page. However these are carefully selected -- see the footnotes. For some benchmarks they don't show the 10 vs 18 core versions: https://www.apple.com/imac-pro/

Any thoughts about the 14 vs. 10 core for FCPX?

I find it interesting that the 8 - 10 - 14 - 18 core processors are $800 away from their siblings that are immediately faster or slower.

Post-Script: I ask the question about cores as that is the only part of my configuration that I need clarification on. I currently shoot 2 camera 4K but I want to future proof myself for 8K.
 
Last edited:
  • Like
Reactions: SFjohn
Even with perfect buses, infinitely fast caches, zero multi-core hardware contention, and heavily multi-threaded code, the code itself will constrain multi-core speedup. Each execution thread is not totally independent of others and must periodically synchronize work with them. Besides that a certain % of the execution path cannot be parallelized. Only a small % of serialized code will cap maximum multi-core speedup; this is called Amdahl's Law: https://en.wikipedia.org/wiki/Amdahl's_law

Using the base Xeon frequency and assuming 100% parallel code, the 18 core might be 27% faster. Using the turbo boost frequency it would be more. But who knows the exact turbo boost behavior for a given workload? This varies with the exact CPU and many other factors.

There is supposedly a max "all core active" boost but I don't know the number for the specific Xeon-W "B" variants used in the iMac Pro: https://www.pugetsystems.com/blog/2...-What-You-See-Is-Not-Always-What-You-Get-675/

Before the iMac Pro, the fastest Apple computer any FCPX user ran was (depending on workload), a 2017 top-spec iMac 27 or a 12-core D700 Mac Pro, or some modified older Mac Pro or Hackintosh.

Momentarily setting aside the 18-core iMac Pro, the 10-core Vega64 version is likely about 2x faster than any prior Mac anybody has run, esp. on FCPX. So the question is how much faster than 2x improvement would the 18-core machine be on a FCPX, Logic and Lightroom workload? And if it was (say) 25% faster would it be worth it?

For certain highly parallel code it makes a difference as has been shown here on the latest benchmarks on Apples iMac Pro page. However these are carefully selected -- see the footnotes. For some benchmarks they don't show the 10 vs 18 core versions: https://www.apple.com/imac-pro/
Thanks for the Wikipedia and Pugetsystems link. I’ve never read about an “all core active” boost before. I’ve always assumed if all cores were active I would be looking at the minimum rated speed. If it was say an overall 25% faster machine it might just be worth it.
 
I think the 10 core is the sweet spot for the iMacPro.

I disagree, the similar, non Xeon chips clocked almost the same on the single thread (that actually baffled me)
& the multi thread clicked 17% faster. I just couldn’t justify 800 for a 17% boost, even if the majority of my work requires it. Near doubling over the 4 core, yes for around 2k more. 17% for 800, no.
 
I disagree, the similar, non Xeon chips clocked almost the same on the single thread (that actually baffled me)
& the multi thread clicked 17% faster. I just couldn’t justify 800 for a 17% boost, even if the majority of my work requires it. Near doubling over the 4 core, yes for around 2k more. 17% for 800, no.

If you keep the iMac pro for 5 years, 800$ mean 160$ a year, so about 0,80$ a day if you work 200 days in 1 year. It means 0.10$ per hour.
With this view, are you still sure it doesn't worth it to spend the money for a 17% boost?
 
from where come the 17% boost? i’m currently asking myself if it’s interresting to have for a machine i will keep 3 years (accounting time to amortize) . I’m not sure yet how to simulate the use of them over the time. also what is the diff with the top iMac?
 
Any thoughts about the 14 vs. 10 core for FCPX?....I ask the question about cores as that is the only part of my configuration that I need clarification on. I currently shoot 2 camera 4K but I want to future proof myself for 8K.

Since no 14 and 18 core iMac Pros have been evaluated by 3rd parties, it's almost impossible to say. But looking at the clock speed vs cores, it would appear the 14 core might not be that much faster than 10: 10 x 3Ghz vs 14 x 2.5Ghz. Of course that doesn't account for turbo boost, but I'm not sure if that changes things much if both 10 and 14 core machines were in an "all core boost" situation. If they both have the Vega64 GPU, no difference there.

Apple's benchmarks on https://www.apple.com/imac-pro/ show 10 and 18-core configurations, but no 14-core. If you assume the 14-core numbers are halfway between, the degree of improvement from 10 to 14 core varies with each benchmark, but on the FCPX benchmark would be quite modest.

The good news is the 10-core Vega64 machine looks very fast -- maybe about 2x faster on FCPX than the top-spec 2017 iMac. In an era when 15%-20% generational improvement is a typical, 200% improvement is a lot.
 
If you keep the iMac pro for 5 years, 800$ mean 160$ a year, so about 0,80$ a day if you work 200 days in 1 year. It means 0.10$ per hour.
With this view, are you still sure it doesn't worth it to spend the money for a 17% boost?

Can someone that works 200 days in 1 year push out CPU binding task faster then the CPU can complete them? :D

Obviously I can only speak for myself however this is probably something to keep in mind, a faster CPU in this case is only beneficial if you are actually waiting for it. And there are plenty of reasons you are waiting or need faster (deadlines, last minute corrections requiring another export, program is unusable during exports, etc) however for me I'm working on other things at that point and 800 dollars could be used on something else to increase productivity, like making a resume for a job I only work 200 days a year!
 
  • Like
Reactions: RuffDraft
Since no 14 and 18 core iMac Pros have been evaluated by 3rd parties, it's almost impossible to say. But looking at the clock speed vs cores, it would appear the 14 core might not be that much faster than 10: 10 x 3Ghz vs 14 x 2.5Ghz. Of course that doesn't account for turbo boost, but I'm not sure if that changes things much if both 10 and 14 core machines were in an "all core boost" situation. If they both have the Vega64 GPU, no difference there.

Apple's benchmarks on https://www.apple.com/imac-pro/ show 10 and 18-core configurations, but no 14-core. If you assume the 14-core numbers are halfway between, the degree of improvement from 10 to 14 core varies with each benchmark, but on the FCPX benchmark would be quite modest.

The good news is the 10-core Vega64 machine looks very fast -- maybe about 2x faster on FCPX than the top-spec 2017 iMac. In an era when 15%-20% generational improvement is a typical, 200% improvement is a lot.

Thanks, joema2.
 
Does anyone have examples of workflows which would benefit from 14 or even 18 cores compared to 10 cores?

Telestream episode engine outputs. Of all the software I use, it provides a fairly linear advantage to having more cores. Telestream doesn’t even sell it anymore so I can’t imagine it’s a popular choice, but it takes full advantage of cores.
 
  • Like
Reactions: SFjohn
I think the 10 core is the sweet spot for the iMacPro.
It really depends on what you do and your work flow. For me an 8-core with 64GB of RAM is the sweet spot. I run scientific simulation that do not benefit from GPU, and about half of them benefit from cores, and half from clock speed...I can see the demo configuration used 10-core, Vega 64, 64GB of RAM being one of the most popular configurations though.
 
I calculate the 18 core would be approx 20% faster than the 10. However that’s just multiplying numbers... Does anyone know how the larger L3 memory would factor into speed? Single treaded it looks like the 10 core is 5% faster than the 18 core... For arguments sake, assume cost is not a factor. But overall performance is. Uses are Final Cut X, Logic, Lightroom, some Photoshop (not much). Thoughts?

In general more cores will always mean greater performance overall. There are two exceptions with one being unrealistic

  1. Some workloads will actual cause performance to decrease with an increase of cores. However, this normally will only happen when you start throwing 100s of cores at a parallel task that has a very small serial piece of code.
  2. The unrealistic case for most casual users is that running a task that is single threaded will perform best on a processor offering the highest GHz. However, in macOS there are 100s of small processes always requesting CPU cycles so this is why its unrealistic.
Ignoring cost to a large degree, then always max out your computer. Doing this will stop any future angst about having not done so.
 
In general more cores will always mean greater performance overall. There are two exceptions with one being unrealistic

  1. Some workloads will actual cause performance to decrease with an increase of cores. However, this normally will only happen when you start throwing 100s of cores at a parallel task that has a very small serial piece of code.
  2. The unrealistic case for most casual users is that running a task that is single threaded will perform best on a processor offering the highest GHz. However, in macOS there are 100s of small processes always requesting CPU cycles so this is why its unrealistic.
Ignoring cost to a large degree, then always max out your computer. Doing this will stop any future angst about having not done so.

The last two iMacs that I purchased were totally maxed out. The iMac Pro, is a different beast, though. I have two different configurations in my Apple cart. And, TB3 really changes the game with external storage.
 
This thread has been extremely helpful. I ordered a maxed out iMac Pro. Due date is mid to late February. Special thanks to joema2 for all the info. This will likely be a 10 year investment, with chip speed increases slowing down I think it will hold its own in years to come. Please keep your comments coming, I’d love to stay informed on the differences between the 10 & 18 core iMac Pros coming out. I think the idea of a possible 25% speed bump was enough for me. It doesn’t sound like there is a ton of single threaded software these days...
 
Wow... what a great iMac Pro you will have! Here’s to hoping it provides you with that ten year scope!

I went for a 10-core, 64GB RAM and 16GB Vega for FCPX and 3 streams of 4K. For me, I went with the maximum of what I could afford, so I am really happy with that configuration and hope that it lasts me for six to seven years. I may update to 128GB RAM if I ever need that much.

As it stands though, the 10-core is plenty fast enough for FCPX and I’m really pleased as to how well optimised it seems to be for the new machine. This seems to be a much better purchase for FCPX than the Mac Pros were, given that FCPX now seems to be extremely optimised for the iMac Pro, and yet a maxed out 2017 iMac could transcode video faster than the Mac Pros.

Joema2 is such a helpful poster and he was able to guide me in my purchase before the iMac Pros came out. Massive thanks to Joema2!
 
This thread has been extremely helpful. I ordered a maxed out iMac Pro. Due date is mid to late February. Special thanks to joema2 for all the info. This will likely be a 10 year investment, with chip speed increases slowing down I think it will hold its own in years to come. Please keep your comments coming, I’d love to stay informed on the differences between the 10 & 18 core iMac Pros coming out. I think the idea of a possible 25% speed bump was enough for me. It doesn’t sound like there is a ton of single threaded software these days...

I'm very sure the processing power will be able to last 10 years. I'm using the 2013 Mac Pro and it's as fast as it is on day 1. But I'm not sure if the screen can last that long though. My monitors usually start to turn yellow in year 4-5 with intensive use, and with some colour calibration they still work, but the screen durability is the main thing to worry about on the iMac.
 
Well, in seven years Apple will announce the iMac Pro obsolete, and you get no repairs and replacement parts.

(RX Vega 64 with 300 Watt I'm looking at you)

Good luck! Especially for the ones with maxed out iMac Pro.

If I'm still using an iMac Pro in 7 years either:

1. I've been unemployed for years.
2. I've mismanaged my time/money.
3. I've lost the ability to work as a professional ADR person.
4. On a 2020/2021 version of the iMac Pro.

Typically I replace my machine every 3-4 years with the exception of last year when I got the 2016 touch bar MBP.

I do understand that everyone has different needs/wants. But this is just over the top, not everyone holds on to their machine for years. Some of us replace them often, or buy what is available when we need more power for current and upcoming work.
 
If I'm still using an iMac Pro in 7 years either:

1. I've been unemployed for years.
2. I've mismanaged my time/money.
3. I've lost the ability to work as a professional ADR person.
4. On a 2020/2021 version of the iMac Pro.

Typically I replace my machine every 3-4 years with the exception of last year when I got the 2016 touch bar MBP.

I do understand that everyone has different needs/wants. But this is just over the top, not everyone holds on to their machine for years. Some of us replace them often, or buy what is available when we need more power for current and upcoming work.

You know that's just what people say, you know to justify it in their own heads! (Breaks down to be much cheaper per year that way.) I got a 2016 MBP that I thought I'd be using as my main machine for years to come, and here we are a year later... Oh well got a lovely 4K and 5K screens to plug into an iMac pro now ;-)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.