Which Mac Pro to get?

[Darthmnkyrpm]

Alright, so I just have a quick question. I've been looking at the MP page on Apple.
http://store.apple.com/us/buy-mac/mac-pro
I intend to use this primarily for big programs like Photoshop, AE, Premiere, Cinema 4D, etc and some gaming on the side.

Now my question is, which is better, something like a 3.7GHz Quad-Core Intel Xeon E5 or something like 2.7GHz 12-Core? I'm guessing it's the second one since it's most likely going to be more expensive, but I was wondering why? What effect does having a stronger processor versus more cores have? Sorry if this is a silly question, but I'm just curious.

Also, for the tasks that I'm wanting, which do you think would be best? I think I'm at least going to get the 6-core one, but I was wondering which one the other things I should focus on upgrading?

 

[kappaknight]

which is better, something like a 3.7GHz Quad-Core Intel Xeon E5 or something like 2.7GHz 12-Core? I'm guessing it's the second one since it's most likely going to be more expensive, but I was wondering why? What effect does having a stronger processor versus more cores have? Sorry if this is a silly question, but I'm just curious.

Think of it this way... you have about 1,000 checks of various denominations you need to add up. Your choice is to either ask 4 smarter than average people to help you, or 12 average people to help you.

In the first case, each person will need to add up 250 checks each. In the second case, each person only need to add up about 83 checks each.

Which group do you think will be able to finish the task faster?

 

[Celedral]

This of it this way... you have about 1,000 checks of various denominations you need to add up. Your choice is to either ask 4 smarter than average people to help you, or 12 average people to help you.

In the first case, each person will need to add up 250 checks each. In the second case, each person only need to add up about 83 checks each.

Which group do you think will be able to finish the task faster?

That's a great way to put it.

 

[haravikk]

Which group do you think will be able to finish the task faster?

Excellent comparison! It does depend heavily on your workload, but all the apps the OP mentions are getting better and better multi-threaded performance, so more cores is definitely the answer from a pure performance stand-point.

Plus, Turbo Boost lets the 12-core processor achieve the same high single-threaded speeds if that's what your workload needs at the time. I think the 12 core can also scale that pretty well, so it can "mimic" an 8-core, 6-core or 4-core processor too (to a degree), i.e - faster speeds for less parallel workloads if that's what you're doing.


Remember there's another factor though; those 12 cores won't come cheap, in fact most people are expecting it to be a $3,000 option! Also remember that each core has two hardware threads, so a 4-core can actually process 8 threads simultaneously, though it depends on what they're doing exactly. For a lot of real world cases it's a lot like having 8-cores though, so the 6-core has 12 threads, 8-core has 16 and 12-core has 24. So it's really up to you how parallel you think your workload needs to be, and how much you can afford to spend.

Finally, remember also that the Mac Pro has two powerful GPUs in there as well. With the right programs it's like having three processors, maybe even better than three depending on what they do with the GPUs, so you're not exactly short on power, it's just up to developers to let you actually use it.

 

[Darthmnkyrpm]

That's a great way to put it.

Agreed, that really helped me understand it for sure.

Excellent comparison! It does depend heavily on your workload, but all the apps the OP mentions are getting better and better multi-threaded performance, so more cores is definitely the answer from a pure performance stand-point.

Plus, Turbo Boost lets the 12-core processor achieve the same high single-threaded speeds if that's what your workload needs at the time. I think the 12 core can also scale that pretty well, so it can "mimic" an 8-core, 6-core or 4-core processor too (to a degree), i.e - faster speeds for less parallel workloads if that's what you're doing.


Remember there's another factor though; those 12 cores won't come cheap, in fact most people are expecting it to be a $3,000 option! Also remember that each core has two hardware threads, so a 4-core can actually process 8 threads simultaneously, though it depends on what they're doing exactly. For a lot of real world cases it's a lot like having 8-cores though, so the 6-core has 12 threads, 8-core has 16 and 12-core has 24. So it's really up to you how parallel you think your workload needs to be, and how much you can afford to spend.

Finally, remember also that the Mac Pro has two powerful GPUs in there as well. With the right programs it's like having three processors, maybe even better than three depending on what they do with the GPUs, so you're not exactly short on power, it's just up to developers to let you actually use it.

And thank you for that answer. So, for the tasks that I've mentioned, what do you think is most important? I'm planning on at least the FirePro W9000 graphics cards. I just don't know what is more important, ram, processor, etc?

 

[jwt]

That's a great way to put it.

Not really. Photoshop doesn't make good use of multiple cores. AE and Premier do. Not sure about Cinema 4D.

Again, not sure about C4D, but for the first 3 you'd be well served by spending your money on CUDA cores rather than processor cores. Take a look at barefeats.com for benchmarks.

 

[Celedral]

Not really. Photoshop doesn't make good use of multiple cores. AE and Premier do. Not sure about Cinema 4D.

Again, not sure about C4D, but for the first 3 you'd be well served by spending your money on CUDA cores rather than processor cores. Take a look at barefeats.com for benchmarks.

Fair enough, premiere will use all CPU cores when rendering/exporting though, which is my primary program at the moment.

 

[flat five]

Think of it this way... you have about 1,000 checks of various denominations you need to add up. Your choice is to either ask 4 smarter than average people to help you, or 12 average people to help you.

In the first case, each person will need to add up 250 checks each. In the second case, each person only need to add up about 83 checks each.

Which group do you think will be able to finish the task faster?

what happens when you have 1000 numbers to calculate but they have things other than a + or - in front of them?

+8 -2 ÷4 x6 (=9) is different than +8 ÷4 x6 -2 (=10)

there are a lot of calculations which can only occur on a single core

 

[kappaknight]

Not really. Photoshop doesn't make good use of multiple cores. AE and Premier do. Not sure about Cinema 4D.

Again, not sure about C4D, but for the first 3 you'd be well served by spending your money on CUDA cores rather than processor cores. Take a look at barefeats.com for benchmarks.

what happens when you have 1000 numbers to calculate but they have things other than a + or - in front of them?

+8 -2 ÷4 x6 (=9) is different than +8 ÷4 x6 -2 (=10)

there are a lot of calculations which can only occur on a single core

I think you guys have completely missed the point of the example.

Yes, some programs aren't utilizing all cores. Yes, some processes require tandem processing. The point of the example is to point out what resources are available to you when you buy a 4 core vs. 12 core computer. I didn't realize I needed to spell everything out. Geez...

Maybe a better example is that you have a bunch of chores around the house that needs to get done, and you can either employ 4 or 12 people to do the various tasks at the same time. There.

 

[flat five]

I think you guys have completely missed the point of the example.

no, i get the example and it's fine

Maybe a better example is that you have a bunch of chores around the house that needs to get done, and you can either employ 4 or 12 people to do the various tasks at the same time. There.

it's not 'there'.. i mean often times, yes, more cores is better.. but often also one fast core is better.. a lot better.. and way cheaper too..

it's an area where i think many people overspend while at the same time, get less performance because they've bought the wrong processors for their needs..

dunno, if i were going to buy a 12core, i'd make damn sure it's going to benefit my workflow/software.

many processes simply can't be divvied up across multiple cores.. as in, 9 women can't grow a baby in one month.. it takes one woman 9 months.

 

[macuser453787]

Think of it this way... you have about 1,000 checks of various denominations you need to add up. Your choice is to either ask 4 smarter than average people to help you, or 12 average people to help you.

In the first case, each person will need to add up 250 checks each. In the second case, each person only need to add up about 83 checks each.

Which group do you think will be able to finish the task faster?

That's an interesting analogy.

 

[Darthmnkyrpm]

Not really. Photoshop doesn't make good use of multiple cores. AE and Premier do. Not sure about Cinema 4D.

Again, not sure about C4D, but for the first 3 you'd be well served by spending your money on CUDA cores rather than processor cores. Take a look at barefeats.com for benchmarks.

So, where/how do I get those? Is that something through Apple or through third party? And how much will the new Mac Pro be customizable anyways? I thought you really couldn't customize it?

 

[iBug2]

So, where/how do I get those? Is that something through Apple or through third party? And how much will the new Mac Pro be customizable anyways? I thought you really couldn't customize it?

CUDA cores are GPU cores. You can't really get any CUDA cores into the new Mac Pro since CUDA is Nvidia only, but there are a lot of OpenCL computation units inside the new Mac Pro. Any app that makes use of those will run insanely fast.

 

[CaptainChunk]

So, where/how do I get those? Is that something through Apple or through third party? And how much will the new Mac Pro be customizable anyways? I thought you really couldn't customize it?

Well, you can add more RAM to it. And I would imagine that a company like OWC will come around and build a third-party PCI-E SSD upgrade for it, similar to what they've done for the rMBPs. Outside of those things, you can't really upgrade anything.

The graphics cards utilize proprietary sockets though. I'm not so sure how much of a demand there will really be for 3rd-party upgrades there. "Official" drop-in options were historically scarce even for the existing Mac Pros that are being phased out. This however may become less of an issue as OpenCL support achieves parity with CUDA in professional applications. Adobe and DaVinci (to name a couple) are already trying.

 

[haravikk]

many processes simply can't be divvied up across multiple cores.. as in, 9 women can't grow a baby in one month.. it takes one woman 9 months.

Interesting choice of analogy there

I pointed out above though that the 12-core can actually adapt to this sort of thing, as the Turbo Boost feature doesn't just work for a single core at 3.9ghz, it can also be used to run only 4 cores at 3.7ghz for example.

Only issue is that I haven't seen any numbers on how long the processor can maintain that kind of level of processing, as it's determined by the temperature of the processor, so for this reason it's unlikely you'll be able to run a single thread at 3.9ghz for long, but I would expect that the ability to balance number of active cores at faster speeds is fairly flexible, so when you buy the 12-core processor you're effectively getting the 8-core, 6-core and 4-core performance levels bundled in there too, to a degree that I doubt you'll notice the difference, but at one hefty premium.

It's also worth remembering that just because a single app can't take full advantage of twelve cores, that doesn't mean you can't run multiple apps. So if you're rendering in one program, you could still still be editing content in other without much slow-down; it's not just a case of doing one job faster, but about whether you can do multiple jobs simultaneously.


The question really isn't whether more cores is better, as I think the answer is a fairly simple yes in this case. The question really is whether you should spend your money getting a better CPU, or better GPUs, or even just save your money (as both options are likely to be extremely expensive).

Apple's opinion in this is clearly aimed towards OpenCL being the future of high-end computing with its two GPUs, and I fully expect that to hold true over time as more and more applications add better OpenCL integration. At the moment things are bit less certain thanks to apps that currently have poor (or no) support for it.

I know I'm not really giving a concrete recommendation here, so sorry about that, but it's going to be hard to do that until we know the price for all the build to order options. I fully expect the 12-core, the D700s and the 1tb SSD to cost a small fortune together, but if you go for only a single one these then you're specialising the improvement made to your system. Personally my thinking is that the 6-core with D500's should be good enough for most people, at which point the question is whether to add more capacity to the SSD, and if I were personally to go any further than that I'd buy D700s and hope apps improve OpenCL usage over time.

 

[flat five]

It's also worth remembering that just because a single app can't take full advantage of twelve cores, that doesn't mean you can't run multiple apps. So if you're rendering in one program, you could still still be editing content in other without much slow-down; it's not just a case of doing one job faster, but about whether you can do multiple jobs simultaneously.


The question really isn't whether more cores is better, as I think the answer is a fairly simple yes in this case.

that example sort of breaks down too because a single core isn't limited to a single process.. as in- you can use a spreadsheet and listen to an mp3 at the same time on a single core.. multiple threads and multicores aren't the same thing.. and intels hyperthreading of a single core adds to the amount of tasks which a given core can handle even more..

a fast quad is the best for 99% of computer use and very few workflows/applications are going to max it out..

again, i'm just stating that this is an area where i feel many people tend to overspend and people should make sure they're going to be using 12 cores in such a way that it will make up for the slower clock speed.. as in, those 12 cores are going to have to be running at least half the time in order to make up for the other half running slower than if you had a 4-6 core.. (numbers out of thin air..didn't really think it through)

----------

At the moment things are bit less certain thanks to apps that currently have poor (or no) support for it.
[...]
and hope apps improve OpenCL usage over time.

what type of applications do you use? and do you talk with the developers of your software?

every single one of my developers are at least messing around with openCL in applicable areas.. my main rendering program is already non-beta openCL.. thea's presto engine has recently been released and it runs entirely on the gpu (though cuda only at this stage).. i don't really see it as gpgpu's future being 'uncertain' at this stage.. maybe from a consumer's pov it is but talk to some devs and you might get a different picture.. they'll generally welcome the conversation.

 

[cosmos]

a fast quad is the best for 99% of computer use and very few workflows/applications are going to max it out..

again, i'm just stating that this is an area where i feel many people tend to overspend and people should make sure they're going to be using 12 cores in such a way that it will make up for the slower clock speed.. as in, those 12 cores are going to have to be running at least half the time in order to make up for the other half running slower than if you had a 4-6 core.. (numbers out of thin air..didn't really think it through)

----------

Flat Five sums up my thoughts very well. IMHO, it is going to take a year or more for developers to code applications for the nMP effectively.

Unless you have very specific multicore applications, you would be better spending your money on extra memory or external storage.

Yes, it would be great to have a 12 core monster, but other than bragging rights, what will it give you except a depleted bank account?

 

[VirtualRain]

a fast quad is the best for 99% of computer use and very few workflows/applications are going to max it out..

again, i'm just stating that this is an area where i feel many people tend to overspend and people should make sure they're going to be using 12 cores in such a way that it will make up for the slower clock speed.. as in, those 12 cores are going to have to be running at least half the time in order to make up for the other half running slower than if you had a 4-6 core.. (numbers out of thin air..didn't really think it through)

True.

I think in general, many computing tasks that lend themselves to more than a handful of parallel threads (eg. images, video) can probably benefit from thousands of cores, not just 10 or 12. Hence the rising value of GPGPU.

Truth be told... I've had a quad core CPU in my primary computer since the Q6600 in 2007 and almost 7 years later I'm waiting for software that can fully exploit all that compute power day-to-day. It's always been coming in the next release.

I just hope that 7 years from now, I'm not saying the same thing about dual GPGPUs.

many processes simply can't be divvied up across multiple cores.. as in, 9 women can't grow a baby in one month.. it takes one woman 9 months.

Interestingly, I've used this analogy a few times over the years in project discussions where someone invariably claims we can complete the project sooner if we just put more people on it. This gets a chuckle every time.

 

[haravikk]

that example sort of breaks down too because a single core isn't limited to a single process.. as in- you can use a spreadsheet and listen to an mp3 at the same time on a single core.. multiple threads and multicores aren't the same thing.. and intels hyperthreading of a single core adds to the amount of tasks which a given core can handle even more..

I know, but my example was rendering + editing; i.e - twelve cores could allow you to run two (possibly more) demanding apps alongside each other. Of course you're right that a single core can easily handle multiple, less demanding apps if all you're doing is browsing and e-mailing, though someone needs to tell Safari this, but I was speaking more about applications with more significant requirements.

Point being, if you have an app that has significant processing times where you're just waiting for it to finish, then the more cores you have the more feasible it is to throw some cores at the process, while leaving the rest to do something else, which means more time working and less time wasted on waiting.

All I'm saying is that the 12-core CPU is undoubtedly the best, the issue however is weighing up whether it's worth the (likely significant) extra cost for what it gives you; if yours is a workload that can take advantage of the extra cores then it may well be worth it. All I was really saying is that many people think that multiple cores means you need one app taking advantage of all them, but there are other use-cases for lots of cores, and also that although the 12-core clock speed is slower, it's not actually worse than the 4, 6 or 8 core options for different workloads, it's just going to be fiendishly expensive.

every single one of my developers are at least messing around with openCL in applicable areas.. my main rendering program is already non-beta openCL.. thea's presto engine has recently been released and it runs entirely on the gpu (though cuda only at this stage).. i don't really see it as gpgpu's future being 'uncertain' at this stage.. maybe from a consumer's pov it is but talk to some devs and you might get a different picture.. they'll generally welcome the conversation.

Oh I fully expect it to happen, my point really is that we're at that kind of mid-point where some apps can take advantage of it, but others are more closely tied to CUDA, some are still working on OpenCL support, and some simply don't support it at all yet. Point is that if you take the leap now for a Mac Pro with D500's or D700's, then unless the apps you use most already support OpenCL then you may not see the benefit right away, and even when apps do add OpenCL it may still not be fully utilised until future releases, so it could be some time before you really get the most out of the machine.

I've been bitten by pretty much the same thing in the past when I bought the high-end PowerMac G5 in the belief that 64-bit support was obviously going to take off, but although it was a great machine, I never really saw the full benefit in my G5's lifetime as 64-bit adoption was surprisingly slow. I know OpenCL isn't quite as new as 64-bit was back then, and it has clearer benefits, but it's a similar case that rushing in may leave you disappointed if you can't take advantage of the extra power right away, or the releases you need aren't forthcoming.


All this said in my particular case, even if I had the money, I don't think I'd consider more than the 6-core CPU, as it allows me to go for the D500's, maybe even the D700's while keeping the total cost somewhat reasonable. I'd only consider a better CPU if the price is right, but in general I do see OpenCL as the future.

 

[VirtualRain]

Point is that if you take the leap now for a Mac Pro with D500's or D700's, then unless the apps you use most already support OpenCL then you may not see the benefit right away, and even when apps do add OpenCL it may still not be fully utilised until future releases.

I've bitten by pretty much the same thing in the past when I bought the high-end PowerMac G5 in the belief that 64-bit support was inevitable, but although it was a great machine, I never really saw the full benefit in my G5's lifetime. I know OpenCL isn't quite as new as 64-bit was back then, and it has clearer benefits, but it's a similar case that rushing in may leave you disappointed if you can't take advantage of the extra power right away, or the releases you need aren't forthcoming.

Great discussion and good points. Different people have different philosophies on this, but I think living in the present when buying a computer is the best philosophy, and one that I admit to continually learning the hard way. Everyone wants to try and future proof their investment, but invariably that excess computer capability you invested in never gets fully utilized because of lagging software or some other unforeseen bottleneck that is revealed later or new technology comes along that instantly obsoletes what ever you thought was ahead of the curve at the time.

 

[MattDSLR]

I will be watching this as we will actually see how will they compare to 5.1 models

 

[flat five]

I think in general, many computing tasks that lend themselves to more than a handful of parallel threads (eg. images, video) can probably benefit from thousands of cores, not just 10 or 12. Hence the rising value of GPGPU.

right. if a process can be divided into many pieces then the more cores the merrier.. to a certain extent.. at some point, the overall process' speed isn't limited by the amount of cores available.. all of the data being generated by each core needs to be combined into one whole so eventually, you reach a point where this combining becomes the bottleneck.. i.e.- it's possible to have too many threads going at once.
i'm not sure when this tipping point occurs but it will (i assume) happen much sooner with gpgpu because there are so many threads going at once. with my specific use for multiple cores (3d rendering), i think it will take quite a bit of threads to bog down the overall process simply because there's not a heck of a lot of data in a final image.. as in- it can take 24hrs to create a 1MB image so there's not a whole lot of data which needs to continually be added together.. (though take that fwiw.. i'm not a developer so for all i know, there's a mega amount of number transfers which must occur throughout)

All I'm saying is that the 12-core CPU is undoubtedly the best, the issue however is weighing up whether it's worth the (likely significant) extra cost for what it gives you; if yours is a workload that can take advantage of the extra cores then it may well be worth it.

yes, i realize that's what you're saying and it's the point i keep responding to in this thread.. i guess my responses are sounding like i'm saying "12core is the best but most people won't use them so they're wasting their money"... that's not what i'm saying..
i'm saying that 12cores are actually worse because they are slower.. more specifically, they are slower during the times a user is actually interacting with the computer.. generally, when a process is maxing out all available cores, it's a point where the user is no longer sitting at the desk working.


idk, i wrote something in a different thread earlier which is pretty similar to what we're talking about here.. i'll just copy/paste it here.. think i'm done typing for the night

-----------------------------(from another thread)

I'm a tech geek and software developer, but I don't work in the video/3d modeling/media industry. I don't understand the point of this machine's focus. Can someone with the requisite experience and knowledge comment.

1)Why is Apple getting rid of the dual-CPU architecture of the last gen and moving to a single CPU? This seems completely counter-productive as much of the clientele using the machine wants as much CPU performance as possible for 3D rendering, pro video/audio compositing/rendering, etc.

just speaking of one of those.. the 3d rendering part--

what are you rendering? a model but who drew it? the most time consuming part for the user, by far, is the modeling phase.. a typical project for me would be something like 4 days to model/texture, 1/2 day to run previews, 4 days to render..

for the 4 days of modeling, the time i'm actually sitting at my computer and interacting with it, i want the highest clock speed available.. all of the drawing processes are linear and can't be multithreaded.. a faster clock speed will allow for higher poly models as they won't choke as soon during navigation etc.. transformations are faster.. booleans are faster.. everything about it is faster and smoother/less glitchy when using a fast core..

the preview renders can happen on low resolution versions so while a user may still have to wait on a low core count system, it can be kept under control (a minute or so for a judge_able preview)

the 4 days of rendering.. yeah, adding more cores can cut that time drastically.. but, you're not actually shortening your personal workload at all.. you're shortening the computer's workload (work time).. if my renders come back in 2days instead of 4, i've still worked 30hrs on the project..

whereas if i did the same thing with clock speed, say 2.5ghz vs 4ghz.. a 30hr project on 2.5ghz may be done in 26hrs with 4ghz.. and i'll have a less frustrating overall vibe going on..

to me, shaving off 4 hours of my personal time > overall project time..

if all you're doing is rendering other people's models then yeah, more cores is definitely the right path.. but the people i know of who are in that business have hundreds of cpu cores at their disposable.. and i highly (highly!) doubt anyone is going to buy 5- 24core macpros to build a renderfarm with.. they're getting their cpu elsewhere even if it's a mac pro driving the farm..

but back to the individual modeler/renderer.. that's where the GPU for rendering comes into play.. the things have thousands of cores in them.. i did a quick example in one of these threads around here using my 5770.. one image with openCL enabled then the same image without.. the render time sped up more than 2x just using a 5770.. more powerful cards and two of them at that? it's pretty much a no brainer where someone like me needs to be looking towards for enhanced render times.. all the while, keeping my fast clock speed and saving some $ as well..

add(i couldn't even afford/justify buying a 24core macpro.. i can definitely afford/justify spending $5000 on a computer.. and i really do believe a fast 6core with dual GPU and openCL applications will complete my renders faster than a 24core computer can.. and at the same time, the much cheaper computer will also be faster/smoother during the times I'm actually sitting at the desk)

 

[VirtualRain]

i'm saying that 12cores are actually worse because they are slower.. more specifically, they are slower during the times a user is actually interacting with the computer.. generally, when a process is maxing out all available cores, it's a point where the user is no longer sitting at the desk working.

Great point. You're on a roll here

 

[haravikk]

i'm saying that 12cores are actually worse because they are slower.. more specifically, they are slower during the times a user is actually interacting with the computer.. generally, when a process is maxing out all available cores, it's a point where the user is no longer sitting at the desk working.

If you're trying to use a computer when all cores are being maxxed out then you're going to suffer the same responsiveness issues regardless of the CPU's basic clock-speed. However, in that case the 12-core processor still has far more head-room for demanding processes which means it's either more likely to have spare cores for you to continue working, or if fully utilised it will finish the job faster, either way it's going to remain the faster processor.

It's usually up to you the user anyway, as most demanding applications will let you limit the number of threads they use, so on the 12-core (24 thread) processor you might set a limit of 20, leaving you with four hardware threads left for whatever else you want to do. You could even run two demanding 12-thread processes and it'd be almost like having two slower six-core processors at your disposal, you can pretty much run any combination you like to get the right balance of performance and responsiveness.

But it's also worth remembering that Turbo Boost isn't just about single-core performance; it has various different stages it can operate at. I believe in the case of the 12-core processor that this means it can effectively operate like an 8-core, 6-core, 4-core or single core processor depending upon the actual workload. So if you're only running a workload that uses 4-cores then it should be about as fast as if you were running that workload on the 4-core processor, except that unlike the 4-core if you need more cores, you've got them.

 

[Nugget]

But it's also worth remembering that Turbo Boost isn't just about single-core performance; it has various different stages it can operate at. I believe in the case of the 12-core processor that this means it can effectively operate like an 8-core, 6-core, 4-core or single core processor depending upon the actual workload.

This is largely true for for the 4, 6, and 8 core processors. In turbo boost conditions they all perform remarkably similarly. The 12 core does take about a 15% hit in peak performance for single-core tasks due to its lower turbo boost limit. This is based on geekbench scores for the CPUs (measured using other platforms currently, but the numbers should translate directly to OS X).

I think that even at the "money is no object" level there's still a decision to be made between an 8 or 12 core nMP, driven by the nature of your tasks. You do give up a little single-core potential if you want those extra cores.

One question we won't know the answer to until the nMP is available is if there are turbo boost duration issues at play. It may be that we see turbo boost sustainability differences between the 4, 6, and 8 core CPUs for longer-duration tasks. Those differences will depend on the cooling and power limitations of the nMP specifically so cannot be inferred from other devices like the geekbench numbers are currently measured.