I’m planning on buying a new Mac, and am confused about the issue of cores. Here’s a paragraph from Apple:
“The Mac Pro offers two advanced processor options from Intel. The Quad-Core Intel Xeon processor is available in a single-processor, quad-core configuration at 3.2GHz. For even greater speed and power, choose a 6-Core Intel Xeon processor. Since the Mac Pro comes with either one or two processors, you can have a 6-core Mac Pro at up to 3.33GHz, or, to max out your performance, a 12-core system at up to 3.06GHz.”
In what respect is, say, a 12-Core 240GHz faster or more powerful than a Quad 320GHz?
And can processor cores be upgraded like memory? Can I buy a 6-core and at some point upgrade it to 12-core? Or would this be like buying a whole new computer?
“Core” is short for “CPU core”. Back in the “good old days” 1 CPU = 1 CPU core, but now through layering and other techniques most CPUs consist of multiple cores. Basically, a dual-core CPU is two “old fashioned” CPUs wired together so they share input/output and memory all in a single chip. If that explanation helps at all.
To make a really simplified analysis, you could multiple the number of cores by the clock speed to determine a rough performance metric, then compare the result. Better would be to find your favorite tech site’s benchmarks.
Can processor cores be upgraded? Only by replacing the entire processor. You have to remove the 2-core processor to replace it with the 4-core processor. Typically this isn’t worth the cost/effort involved.
If I may, unless you’re using this computer for 3D rendering or something similar, you probably should focus more on RAM and disk capacity/speed than processor cores. I’ve seen a lot of people who have very simple server needs over-buy on CPU… I’ve seen 8-core web servers before that were feeding data from a 2-core database server! For example. Most server applications simply are all that CPU intensive.
They would only be faster if you are running apps that make extensive use of parallel processing - high-end CAD/CAM stuff, and perhaps Photoshop and InDesign (probably some video editing SW, but I’m not so familiar with that area). The lower-ghz 12-core CPU will actually be slower running single-threaded apps than a higher-ghz 6-core.
So unless you have those specialised needs the fastest option will be
While we’re on the topic of speed, getting an SSD for your boot/OS drive will make hugely more difference than a CPU boost from 3.20 ghz to 3.33 ghz.
I would say this analysis is so simple that it is entirely meaningless. Software has to be specifically written to take advantage of multiple cores. If you’re running a computationally intensive task that is not so written, it will max out one core and leave the other 11 sitting idle.
So the benefit will depend on what kind of software you’re running. A lot of high-end graphics and video stuff, engineering software, and some games are specifically designed to take advantage of multiple cores. Other things are not.
No one has mentioned the fact that all 2600 series Xeons support hyperthreading. So a 6 core machine with hyperthreading enabled with give you 12 concurrent threads.
I agree, but without knowing what the computer is for that’s all I can do.
Hell, personally, I’d spent 1/3rd the money on a decent PC server with the same amount of cores but, again, without knowing what it’s for I can’t recommend that either. Maybe it’s running something that can only run on OS X.
Yes you get 12 concurrent hardware threads, but your OS is already providing infinite (well, limited only by memory) concurrent software threads. Since every single “core” in the analysis is two hardware threads, that doesn’t affect the answer of the question.
What part of it don’t I know? That link is just marketing fluff.
I’m saying it’s not relevant to the question because you can’t buy a non-hyperthreaded Xeon, AFAIK. So all the CPU options are equivalent as far as hyperthreading. So hyperthreading isn’t relevant from a “which CPU option is best?” standpoint.
If he didn’t realize that 6 cores meant 12 threads and as a result he was considering a 12 core machine, I think it might be quite relevant.
edit: and btw, it’s hardly marketing fluff. Intel has spent many billions, with a ‘B’ developing the tech and I can tell you for a fact that it works. A 4 core hyper threaded machine from Intel will literally dance circles around one of AMD’s single threaded 6 core chips.
I didn’t say the TECHNOLOGY was marketing fluff, I said the WEBPAGE was marketing fluff. If you meant to demonstrate that I don’t understand the technology, you need to point me to something I said that was incorrect. You haven’t done that, so I’ll assume I wasn’t incorrect.
Although even for single-threaded apps, the OS will spread them around between cores if you have more than one process running (and just-booted Mac will have about three dozen processes running, so that’s always the case). Multiple cores/hyperthreading is almost always a benefit, even for single threaded apps. The Mac also has several API mechanisms that can “auto” multithread your apps, so there are more multi-threaded apps than you’d guess, including several of the ones provided by Apple itself.
I notice that you’re looking at the Mac Pros. Consider that these machines have not had a meaingful upgrade in three years, and the cost/performance on them is very high at the moment: high cost for very old tech. Tim Cook has promised new ones, but they’re still probably a year away. (Professional high-end Mac users are really miffed at the moment about the lack of attention paid to the Mac Pro. I wouldn’t buy one unless you absolutely need the expandability, which is available in no other Mac.)
In the meantime, you might look at the higher-end iMacs, if you don’t mind the form factor. For single-threaded apps, several of these beat the lower-end Pros in benchmarks, and there should be even faster new iMac models yet this year. Plus, they’re cheaper.
Timewinder: what socket/series Xeon is used? I’m going to guess and say the 5500 series Gainestown that use the 1366 socket. If that’s the case, can you upgrade to maybe a 5600 Westmere and sell the old chip. Xeons do tend to hold their market value.
edit: never mind - 6 core would have to be Westmere - maybe a newer model though?
Sure, but 12 cores over 6 on a faster CPU? I don’t believe the 12 cores will provide any compensatory benefit over the faster CPU speed - 6 cores is still heaps.
Completely agree, and I should have mentioned this too. Wait for the new iMacs and buy the fastest one of those.
A lot of the work I do involves large, complicated Photoshop, Illustrator and InDesign files. I may be getting into 3D rendering within the next year or two. So how does this affect what I need?
Perhaps the OP could give us an idea of his/her work load.
I run a 2600 series 2P octo core server (one of 4 mixed breed servers) at 3.2ghz. That’s 32 threads. I run it at 100% load 24/7 and it chews through cpu intensive work loads like nothing else. That’s probably the exception, but still, if the OP could give an idea of how the machine would be used, that would be very helpful in rendering advice.
My experience with graphics software is limited, but I think one of your main concerns will be access times. That will mean getting data from storage into memory of course, so obviously you will want a solid SSD and probably one that is at least 120gig although 256 and larger is probably going to be more comfortable.
You’ll also want to eliminate any memory paging to the greatest extent possible so you should probably find out what the maximum is for the new machine and plan on getting that. 8gig modules are available but still a little pricey. Most non-server boards will have 4 slots. Server boards with 2 processor sockets can have 4 times that or more. But the more RAM you have, the less likely any part of your image or the program you’re running gets pushed out of RAM to virtual memory (=disk drive)
You’ll also want to pay attention to the memory clock speed and especially latency. That will usually be expressed with a code like CL7 which means 7 clock cycles for a particular type of operation. The lower that number the better but the faster the memory, the higher the number tends to be. This is a long story and worthy of it’s own thread. Basically which is more important depends on the type of memory access which is most frequent. I generally opt for lower latency first but it’s a balancing act.
My guess is that graphics programs will benefit from multithreading but speed is also important so I would look for reviews of the machines you will be considering that include the software you will use. Look at the benchmarks, read the reviews and try to decide from there. I would probably favor speed over cores depending on the difference. But this will probably be another balancing act.
Software has to be specifically written to take advantage of multiple cores. If you’re running a computationally intensive task that is not so written, it will max out one core and leave the other 11 sitting idle.