Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP All Models How many Mac Pro cores would you get?
- Thread starter Jordan XP
- Start date
-
- Tags
- @mac.com
- Sort by reaction score
For 3D rendering, the 28 core will kick ass with apps that render via the CPU.
(Cinema 4D)
But the 28 core mMacPro makes for a very poor choice as a rendering box because it's waaaaaaay too expensive.
(Cinema 4D)
But the 28 core mMacPro makes for a very poor choice as a rendering box because it's waaaaaaay too expensive.
Each core is a single processor capable of processing data. Many Mac laptops in use today are dual core. So imagine 14 times that in processing power.
As opposed to the $150,000 Win10, 56 core 1TB RAM rendering machines that are already being used by the film industry? I don’t think so.
If you read the TechCrunch articles, you’ll see the film industry is the target customer.
[doublepost=1565942043][/doublepost]In the Keynote, Apple demonstrated 1,000 VIs with plugins on each track running in Logic. Thy didn’t state how many cores in the demo, however.
Since a 10 core iMac Pro can handle 300 — there are YouTube demos of this, I suspect that high end scoring studios are also eying the 28 core.
And even worse if we're looking at "false cores" (AKA SMT). Not only do you have an issue with the software being unable to scale to more cores - but for most all workflows SMT2 is not twice as fast as physical, SMT4 is not twice as fast as SMT2, SMT8 is not twice as fast as SMT4...
I use some big systems, and usually turn SMT2 off - it's slower than physical cores. I will turn it on for a few highly threaded workflows that may get a 10% to 30% boost from SMT2 - but those are uncommon.
as has been true since the first dual-core Macs in 2005. Multi-core Macs have been around in some form for 14 years.
There is no such thing as a single-core Intel Mac, is there?
The apps that will benefit support multiple cores. If the apps you use don’t, then you aren’t the intended customer.
Is there a single pro AV app that doesn’t take advantage of multiple cores? None that I know—certainly none that I use.
There’s a lot of incorrect info on some Adobe apps in this regard. Actually, ancient history that keeps being brought up as if current.
[doublepost=1566152763][/doublepost]
So... more physical cores are better for these apps that you don’t name.
Got it!
Cost/Benefit analysis required. I'd take as many as I could afford for Audio use.
The goal of SMT isn’t to double/quadruple speeds. It’s that the CPU fan pipeline more efficiently in that configuration. If one thread is stalled waiting on memory or something it can jump to another.
Nice theory, but real world numbers don't offer much support unless you have crazy numbers of runnable threads.
See https://superuser.com/questions/1166529/performance-impact-of-hyper-threading for some interesting analysis.
I'd welcome configuration suggestions for video editing using the Adobe CC suite (Premiere Pro, After Effects, Media Encoder, etc),. Will eventually replace my upgraded 5,1 as 4K becomes more common. Seems there is likely a "sweet spot" as well as a tradeoff between spending on the processor vs video card & RAM. Back in 2010, the 6-core 3.33GHz seemed to fit that sweet spot (good combo of high clock & decent number of cores vs the other options at the time) ... not the fastest clock or the greatest number of cores, but sort of a Goldilocks "just right." Any similar option for 2019?
Most of the key content creation software packages have added more multithreaded operations as they iterate. That said, there's a lot of variance - even within a particular package. If, for example, you do a lot of NR (noise reduction) and your software of choice is updated to leverage multi-core performance for that operation, more cores are likely worth the bigger spend. If compositing is your wheelhouse but your primary package doesn't scale well past a single core, or does most of it's compute on the GPU, then not so much. (These are just arbitrary examples to make a point.)
Without drilling down into the specifics of key workflows, or, having real world performance metrics for unreleased hardware, as of today I'd consider us more at the "educated guess" stage.
If you made me select the best price/performance option based on what we know or can infer, it looks like the 16c, 96G RAM, MPX Vega2 Duo with 1TB SSD in the T2/3 garden would keep the cost semi-reasonable with enough guts to qualify as a legit workstation.
Without drilling down into the specifics of key workflows, or, having real world performance metrics for unreleased hardware, as of today I'd consider us more at the "educated guess" stage.
If you made me select the best price/performance option based on what we know or can infer, it looks like the 16c, 96G RAM, MPX Vega2 Duo with 1TB SSD in the T2/3 garden would keep the cost semi-reasonable with enough guts to qualify as a legit workstation.
It would be hard to suggest any Apple system if your workflow is only multi-platform apps - especially considering the steep entry price for the mMP.
While I'm not thrilled with the high entry price, it's a long-term business investment for the next 5+ years, central to how I make a living. So reliability (first and foremost), an OS that makes sense to me, upgradeability are all critical. So if the mMP can save me even a few hours per year over the next few years vs. dealing with a new OS or less reliable hardware, the mMP will earn its keep. But I do want to make sure that I make good initial choices for components that are not easily (or economically) swapped out and upgraded, which I assume is the case for the CPU. Which is why I welcome suggestions.
The reliability of the MP7,1 is unknown. The MP6,1 had some serious issues.
Apple will make you haul a broken system to a store and leave it for service. Dell and HP will come to your home/office the next day for repairs, and offer 5 year service warranties.
What my machine would be doing would be my first piece of the decision making.
Example, I have owned the last Mac Pro and now the 18Core iMac Pro.
I would never own either of these machines for strictly personal home use, my reason is because for me, the machine is a tool to do a job and at the higher prices of the iMac Pro/Mac Pro, it needs to be producing income rather than simply consuming it.
But to stay on topic... 28 CORES!!!!
Example, I have owned the last Mac Pro and now the 18Core iMac Pro.
I would never own either of these machines for strictly personal home use, my reason is because for me, the machine is a tool to do a job and at the higher prices of the iMac Pro/Mac Pro, it needs to be producing income rather than simply consuming it.
But to stay on topic... 28 CORES!!!!
Again, it's not meant to double or quadruple speeds. It's to allow the CPU to stage more efficiently. The benchmarks are exactly in line with that.
It needs to be exposed as a separate core to software for pipelining, but in no way is it architected to act as a core. Trying to infer it's there to be a virtual core is completely misleading.
There may be situations where a CPU is never waiting on memory (which is rare, but possible) in which case things would be even. But generally it's a small optimization that allows the CPU to work on one thing while waiting on another to load it's data.
I work on this sort of stuff. It's not a "theory." It's actually how the CPU was designed. If you want to know more about the "theory", here's a whole article on CPU pipelining:
https://en.wikipedia.org/wiki/Instruction_pipelining
I'm not sure why you think that pipelining and SMT are related. Pipelining is mostly "per thread".
But 10% throughput improvements on massively multithreaded apps just aren't that exciting - unless you have a web server cluster with a million active requests.
I'd like to see actual application (not synthetic like Geekbench) benchmarks comparing Xeon CPUs with and without SMT2 enabled, and the same for SMT4 and SMT8 CPUs.
And I'd like to see an analysis comparing the microarchitectures - since some of the POWER9 CPUs are "12 core/96 thread" (SMT8) and others are "24 core/96 thread" (SMT4).
10% is about the intended speed difference.
You just said it benchmarked about a 10% difference. That's about the correct number. We're in agreement about what the performance boost is. You just think Hyperthreading is something extra that it is not supposed to be.
The only thing we're in disagreement about is saying that it's about "massive" multithreading. It's about scheduling, not really multithreading. The scheduling is about 10% more efficient. Threading is just the tool the CPU uses to get more instructions at once to schedule more efficiently.
If a CPU has a wider or deeper pipeline, more threads per core will be more efficient. Nothing to do with massive multithreading, although you'd need at least that number of threads to see the improvement.