Hi. I've been reading the thread and understand some of the technical stuff but most goes over my head. I have a 12 core 5.1 mac pro. I am wondering if somebody could be kind enough to explain in simple terms how a 12 core chip today is faster than what I already have? I'm looking to upgrade at some point.
Three general factors.
1. General Speed. A rough, "back of the envelope" metric with primarily same core instruction sets and implementations is to just multiply the cores by the base clock speed. That's a max theoretical throughput.
An 2010-2012 era 6 core Xeon X5650
https://ark.intel.com/products/47922/Intel-Xeon-Processor-X5650-12M-Cache-2-66-GHz-6-40-GT-s-Intel-QPI-?q=xeon X5650
12 * 2.66 = 32.92
An 8 core Xeon W 2155
https://ark.intel.com/products/126707/Intel-Xeon-W-2145-Processor-11M-Cache-3-70-GHz-
8 * 3.7 = 29.6 ( 4 cores less and relatively close )
a 10 core Xeon W 21
https://ark.intel.com/products/125042/Intel-Xeon-W-2155-Processor-13-75M-Cache-3-30-GHz-
10 * 3.3 = 33.0 ( ahead with 2 less cores)
At this point is usually where someone jumps in with a X5687 and a 3.6 base clock ( = 43.2 )
The 8 core probably beats the 5650 (and faster 5600 series options) on many real world workloads that actually involved getting substantive amounts of data from memory.
5650 memory clock ... 1333 .... tops ( a bit slower if use all of the memory DIMM slots ).
Xeon W -- 2666
So any time the 8 core needs to get something from memory it will arrive about up to twice as fast as the approximately 10 year old technology in the older Mac Pro. While the 12 core is waiting it will process a bunch of No-op ( no operation) instructions faster than the 8 core but it won't be getting anything pragmatically done. "do nothing" faster is baseless bragging about core count. If have more cores than can keep fed with instructions and data then not particularly buying much. Core count by itself isn't a measure of performance.
The 8-18 cores in the Intel W also have faster intercommunication between the cores than the two package 12 Xeon 5600 class. Any workload that involves swapping data between caches in the cores will generally be slower.
2. Implementation of instructions
Intel W can get closer to that theoretical max because have better branch prediction, cached decoded versions of instructions ( don't have to reinterpret the instructions from scratch every time) , better caches , etc. Even with similar clock rates ( say slap in a higher clocked Xeon 5600 series ). When you code is in a reasonably sized loop doing work the Intel W can get to the execution stage of the code quicker.
The ability to shift between base clock and turbo modes is substantially better and faster. ( the processor has better tools to dynamically engage and disengage overclocking as workloads change. ). If have code that goes into and out of parallel computation modes the Intel W series will do that better.
The thermal/power management is better.
3. Instructions the 5650 doesn't even have.
Vector math of AVX2 with more than substantially better than SSE4
Using the virtualization instructions? More than substantially better than stuff from 10 years ago.
Workloads that match those areas it isn't even close.
About the only advantage the 5600 series has at this point is that it is cheap ( in part because it is all old, used components in de-support status. )
Also do you think there will be a class of mac pro. What I mean is if it is module could there be one range with bluray drive and video caption cards for those that work with video, while another class with specific stuff for audio, and yet a third type for a basic mac pro that can do a little of both?
First, there really wasn't before ( class meaning some substantive product case variation and not merely component configurations). So not particularly likely now. (Actually less likely since market being targeted is smaller now) . Apple isn't going to be everything for everybody.
If talking like class in the sense of " Mac Pro" and "Mac Pro server" (basically the same core system but with different software and hardware components ) then still probably not. I'll loop back to this.
What was there before was more flexibility in configurations. The flexibility primarily there for older components that the market (largely) and Apple (specifically) have moved away from probably won't be there. Optical Drives as in the same bucket a SCSI ports, RS 232/422 , serial keyboard/mouse ports , etc. It is highly unlikely to be there. Empty 5.25" drive bays? Extremely likely no (see previous answer). FireWire? Nope. Empty 3.5" bays? Doubtful ( Mac Mini and iMac Pro revisions over last 2 years dropped HDDs all together). 2.5" bays? Maybe ( there are certainly more cost effective 2.5" SSDs out there in 2019+ time frame that can expand a Mac Pro flash storage capacity ) . However, Apple may simply just go to PCI-e drives only ( and walk away from SATA drives); essentially performance matters more than internal capacity.
Video capture cards (plural ),... I wouldn't bet on it. An empty slot for a video capture card would be more than prudent for Apple, but not a sure thing. [ I think that is grounded in just how fixated Apple is on making the Mac Pro a literal desktop machine. If they are then seriously in maybe status. ]
If there is a slot where folks can put in their own third party stuff to skew the system toward video , audio , Machine Learning , Gaming , etc. then Apple doesn't have to sell those "class" systems directly. Apple will sell the basic system that works and folks will add to it so it does something more. Apple isn't going to be closed (or extremely custom ) Apple systems to do audio , video etc.
[doublepost=1548869277][/doublepost]
Hopefully this could be a turning point and they would realise that the Mac Pro doesn't need to be a ground-braking new design, just a box that runs MacOS.
Dropping optical drives is a policy Apple pragmatically explicitly outlined about 6-7 years ago. There is nothing practically "ground breaking" there at all.
FireWire ports? Probably gone too. that wouldn't be ground breaking either.
Apple is unlikely to take inventory of all of the old stuff that folks have and basically design a container for it. Some things will probably make the cut and some stuff won't. Things that the overall market is largely walking away from will be far more likely to go than stay.
They concentrate on reliability, expandability and serviceability, and hopefully produce something that isn't that far removed from the competition (HP Z, etc). The benefit of the Mac for most people is the OS and hardware that is designed to work with it. Everyone else knows this, I just with Apple could understand this, and not assume this hardware also needs to be thin and pretty.
Dropping optical drives doesn't nothing to impede reliability or serviceability at all. It has minuscule impact on expandability in the context of contemporary ports ( various flavors of USB Type C ports ). Mechanical, spinning drives improve reliability how over solid state alternatives?
Maximizing the number of "race to the bottom" components tossed into an ultra 'expandable' container doesn't necessarily make it more reliable.
Those three things are not 100% orthogonal to each other. To some extent there are trade offs. I think folks are fooling themselves that Apple doesn't understand the trade offs. That Apple picks a different balances doesn't mean they don't understand your personal balance points. It is just different weightings.
Apple should realize that went a bit too far with the Mac Pro 2013 design. ( The iMac Pro is indicative that they are not doing a 180 degree reverse course). They need to find a better balance, but I think it also delusional that Apple is going to concede 99% in the other direction also. Incrementally even the older Mac Pro design wasn't primarily tasked with just being a container for commodity parts.
