Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The nMP seems to have impressive performance with FCP, but so far we haven't seen much written about its performance in other fields, in particular software development.
To be honest I don't think many people need high performance machines for software development anymore; I bought my PowerMac G5 and later my current Mac Pro because I did a lot of C/C++ development with complex projects that needed a machine with a lot of clout to be able to compile them quickly.

However, compilers have improved a great deal to the point that I'm not sure there are many software development projects out there anymore that really need to compile the whole thing from scratch anymore. I suppose in theory OpenCL could be used for that purpose, but I'm not sure it's an area that needs the best possible performance anymore. Really the main hardware requirements are dependent on what your end product will be; for example if you're building a game then you may need a good graphics card, or if you're working on professional video editing software then you'll need the kind of machine that might actually run that, but even then you may be better throwing together a cheap test PC for that kind of thing (assuming your software will run on it).

For example, in my own case what I do is mostly programming, so my next computer purchase is more likely to be a Mac Mini, even though I'd love a new Mac Pro, as the Mac Mini is plenty powerful enough for programming work (even so I'll wait for the next update first).

I've been trying to compare it to other desktops, workstations & servers for running Java application servers, databases, assorted VMs and Java-based development environments, but haven't arrived at a conclusion as to whether the nMP is a good fit.
Probably not; I don't believe any of these uses will use OpenCL to run commands, in fact most server farms running this kind of software are now more likely to use machines with tons of relatively low powered ARM processors instead and/or lots of machines. So IMO there isn't really a desktop that's optimal for that kind of thing anymore, plus if you're just mostly doing testing it's something that a Mac Mini or iMac should be more than capable of handling as they have enough cores/threads to show how well your parallel components run.

Hell, I have several Java programs plus a database and DNS server running on a Synology DS212j (NAS) and it has, iirc, a 1.6ghz single-core ARM processor and 256mb of RAM, and when I was still using a G3 iMac I had a web-server and MySQL database running on it for testing; unless you're doing something that requires a lot of parallelism and or brute power then pretty much any modern machine should be plenty.


So yeah… if it weren't for the fact that I enjoy also do quite a bit of video transcoding, and like to do gaming from time to time as well, then I'm not sure I'd even seriously consider a new Mac Pro at all, other than from the "wow I want one" perspective. Right now I'm waiting to see what the next Mac Mini update looks like as that'll determine whether I get a current Mac Mini cheaply, or a new Mac Mini, though there's still the possibility I'll buy a new Mac Pro anyway because I'm terrible with my money :)
 
These are weird comparisons. After Effects has some features that are still entirely CUDA dependent.


Not for long they are going All OpenCl. Not sure if that means Cuda is to be depreciated... probably not. Adobe is big enough to support any technology needed.
 
Maya/Cinema 4D is probably enough reason to need a Mac Pro with all the bells and whistles alone. Seriously those apps scare me on anything less than a beast computer.

Heck, I run Maya on my Mac Mini now (though primarily just for character animation).

It could certainly be faster, but it's more than usable. Doesn't mean I wouldn't want to replace it with a Mac Pro, however :)
 
Heck, I run Maya on my Mac Mini now (though primarily just for character animation).

It could certainly be faster, but it's more than usable. Doesn't mean I wouldn't want to replace it with a Mac Pro, however :)

Yeah..."animation" is such a broad space that even within Maya it really depends on what you're doing, whether or not its multithreaded or OpenCL aware.

FX animation in Maya...well sometimes multithreaded. Sometimes not. Depends.

Character animation in Maya? Virtually never multithreaded. Does not use OpenCL.
 
Yeah..."animation" is such a broad space that even within Maya it really depends on what you're doing, whether or not its multithreaded or OpenCL aware.

FX animation in Maya...well sometimes multithreaded. Sometimes not. Depends.

Character animation in Maya? Virtually never multithreaded. Does not use OpenCL.

Yup, the characterization of "animation" is this thread is more than a bit vague here and your needs really do vary depending on your particular application.

I'm curious to see benchmarks, but I'd wager that most any Mac user of a high end 3d package is going to see at least some performance increases from the nMP.
 
That sounds like a problem specific to your machine. I've been building PCs all my life (in addition to buying Macs - people can like both!) and haven't had that problem since the move to Vista of all things. I think it's because the 64-bit version of Windows dropped a lot of legacy hardware support and required all new drivers and validation. Oh, and I only run WHQL certified drivers. Anything else and you are potentially compromising the stability of the system.

That said, I don't know what OEMs like Dell do with their systems. Any OEM PC I've ever used had a ton of crapware installed out of the box.

At least with Windows 8, there is an option built in to reset the OS to a completely stock version without any of that. (at least, that was my understanding from a friend that bought an XPS12 earlier in the year)

Honestly, I love the internal design of the Mac Pro (externally, I'm not a fan) but if you plan on running Windows on it, I'd build a PC for less.
The Mac Pro basically has no expansion options unless you are willing to invest a lot of money in external Thunderbolt or USB3 hardware.

With a PC I can just buy an enterprise-level 4TB disk and put it in the case if I need extra storage.
Things like that are a major hassle with the new Mac Pro unless you buy something like those ridiculously expensive Promise Thunderbolt enclosures.
I can buy any number of off-the-shelf PCIe cards, and I don't need to worry about whether they're compatible with OS X or not.

The Mac Pro is basically for people that work in Final Cut Pro X, or people that are hitting the ceiling of what a quad-core i7 can do and need more CPU cores in OS X. (assuming their application can take proper advantage of the additional cores)
Don't forget that you are moving back a generation in performance with the Mac Pro CPU (it's using Ivy Bridge-E) so each individual core is slower than a Haswell i7 at the same clockspeed.

I suppose it's possible that some people have a use for that GPU power for things on OS X besides OpenCL, but I'm having a difficult time coming up with anything which requires OS X.

Thunderbolt 2 is basically Thunderbolt 1, but they now send data over the second channel instead of it being dedicated to video. (so video now shares bandwidth with other devices)

If latency matters, you're still probably best with dedicated PCIe interfaces. (or hardware which bypasses the PC for monitoring)

OpenCL has been around for a number of years at this point, and very little has taken advantage of it.
It's a nice dream, but buy the machine for what it can do now, not what it might (or might not) be able to do in the future.

If that future arrives, the current GPUs may not cut it for OpenCL performance any more.

I know someone that does video mastering for films, and FCPX just doesn't cut it for his needs - none of the GPU accelerated encoding tools do.
He needs as much CPU performance as he can get. (which means a PC with two or more CPUs)

If Apple offered a system with a single GPU and dual CPUs, it's something he would be considering - but the ridiculous mark-up on the CPUs would probably just make him build a PC anyway.


I absolutely agree with your points there, beta. Thank you for them. I am very heartened to hear about stable PC builds: that's just the thing I want to hear about as I consider building my own system. I mean, people talk about cost effectiveness, speed, and fun-factor in building something yourself but long-term stability just doesn't get mentioned too often.

Yes, I'm finally coming round to accepting your conclusion that the few stability issues I've had may be due to quirks in this particular machine. I read another user of this machine exclaim in exasperation that Dell seems to be using its customers as beta testers for their top-of-the-line workstation. Go figure.

I think I will indeed give a PC build a go with good recommended parts. Thank you very much again for your considered points.
 
I absolutely agree with your points there, beta. Thank you for them. I am very heartened to hear about stable PC builds: that's just the thing I want to hear about as I consider building my own system. I mean, people talk about cost effectiveness, speed, and fun-factor in building something yourself but long-term stability just doesn't get mentioned too often.

Yes, I'm finally coming round to accepting your conclusion that the few stability issues I've had may be due to quirks in this particular machine. I read another user of this machine exclaim in exasperation that Dell seems to be using its customers as beta testers for their top-of-the-line workstation. Go figure.

I think I will indeed give a PC build a go with good recommended parts. Thank you very much again for your considered points.
Two helpful links if you go down the build route.
www.reddit.com/r/buildapc

http://pcpartpicker.com/
 
Of course it is. How is this a question? It's the leanest, snappiest, most powerful machine there is, but more importantly it's awesome and a killer status symbol.

It's like asking if a high end Mercedes is worth it if you "don't go to fancy parties or hollywood premieres." No such thing as a girl too pretty etc etc.
 
This article from last year about the impact of OpenCL on photo editing is pretty interesting, particularly the interview with the Photoshop architect...
http://www.tomshardware.com/reviews/photoshop-cs6-gimp-aftershot-pro,3208.html

(sorry if this was already posted here -- if so, I missed it)

It definitely gives the impression that OpenCL is the future for Photoshop etc. So the GPU power of a nMP will eventually be very useful for photography. The question is when...
 
Not for long they are going All OpenCl. Not sure if that means Cuda is to be depreciated... probably not. Adobe is big enough to support any technology needed.

Their niche will probably remain in HPC work. The primary advantage to OpenCL is ubiquity. That isn't as big a deal if you're working with clusters of nodes running Linux. Apple is using AMD on the Mac Pro for this generation. Their biggest sellers are notebooks, which may all migrate to integrated graphics if intel doesn't stumble over the next couple cycles. Interestingly some of the iOS hardware could technically support OpenCL, but they don't currently support that.
 
....
Thunderbolt 2 is basically Thunderbolt 1, but they now send data over the second channel instead of it being dedicated to video. (so video now shares bandwidth with other devices)

If latency matters, you're still probably best with dedicated PCIe interfaces. (or hardware which bypasses the PC for monitoring)

There is little so far to indicate that latency is an issue with TB v2. Clearly if Intel implemented their isochronous and congestion control badly there will be a problem but desgregating the traffic doesn't mean it is necessarily bad and lacks QoS mechanisms.

Intel could have screwed up but the fact that they initially left it out ( probably because didn't have big enough transitor budget to deal with management overhead at high speeds) is actually a good sign. (they are at least aware of the issues involved).



OpenCL has been around for a number of years at this point, and very little has taken advantage of it.
It's a nice dream, but buy the machine for what it can do now, not what it might (or might not) be able to do in the future.

Can throw the same generalizations around about x86 cores. 16-20 cores machines have beeen around for years and there isn't much software that leverages them.

There is a limit to how quickly most established commericial code bases evolved. Apple kept around Carbon ( as scaffolding to get off OS 9 quicker) for years before dropping it (and making new Apps with access to newest APIs go Cocoa only).

It wasn't until 2013 that the entire Mac line up from top to bottom could do OpenCL. At this point the infrastructure is pervasive across the entire line up. That means developers don't have to put in "If happen to have OpenCL then ... otherwise ... " . It can be "if this a modern Mac then when OpenCL do ... otherwise what used to do ... " That makes a significant difference if there is a widespread adoption by the software developers. It isn't just code fragments for a small subset of Mac... Eventually as the the older Macs get retired it is going to be all Macs can.





I know someone that does video mastering for films, and FCPX just doesn't cut it for his needs - none of the GPU accelerated encoding tools do.
He needs as much CPU performance as he can get. (which means a PC with two or more CPUs)

Because he has already evaluated the newest FCPX on the newest Mac Pro can came to that conclusion? Or he has been in the FCPX can't possibly work camp for a couple of years?

It is more than awfully odd when folks swear up and down that graphics work can't be done well on the Graphics Processing Unit (GPU) and needs to be done on the non graphics oriented computational device. Either there is something deeply lacking in the GPU or this is really about legacy software with the limitation... not the CPU/GPU capabilities.

I can see "my database query processing engine doesn't work better on the GPU". But "my graphics problem needs to be pulled out of the GPU" should beg a question. As GPUs gain more general purpose capabilities it even more so begs a question. They are beyond a very limited subset of graphics primitive operations at this point.
 
.... Interestingly some of the iOS hardware could technically support OpenCL, but they don't currently support that.

1. Apple has had a stick up their butt about dynamically compiled code before (OpenCL can be static, but would need specific static per hardware). At one point they tried to ban everyone from using everything but their Objective C and using their javascript implementation.

[ nevermind apps at the time having embedded Lua and other facilities to get work down without causing problems. ]

Apps need to confirm to a farily rigid model so Apple and pull their strings ( terminate , manage , etc.) with the resource management of iOS.
Also there are probably manageable power consumption issues which again is running into Apple "lock down" as opposed to considering tradeoffs.


2. Some missing pieces that are falling into place. SPIR and OpenCL v2's increased support for shared memory contexts. All the iOS devices have integrated graphics where sharing "VRAM" out of the system's RAM.


To a smaller extent Android partially suppressing OpenCL with Renderscript [**] basically makes it is non high priorty issue also since competing platforms aren't leveraging the PowerVR OpenCL compatibilities either.


[**] as opposed to extending OpenCL to address the subset of issues they had about portability and optimizations.
 
A better question is what apps are you using and will they exploit the new Mac Mini Pro.

As for Lenevo, I wouldn't give them one cent. They are a Chinese company that has made it abundantly clear they want to take over the entire desktop market and have used tactics that are shades of old Microsoft. While Apple products may be built in China, it is still an American company.

IBM originally owned the Thinkpad and had a Chinese company make it for them along with their desktops. Later, IBM got rid of its desktop/laptop division and thus Lenevo took it over entirely and all of the profit is Chinese so to speak with a pissy little amount going to the American middle man.
 
1. Apple has had a stick up their butt about dynamically compiled code before (OpenCL can be static, but would need specific static per hardware). At one point they tried to ban everyone from using everything but their Objective C and using their javascript implementation.

Yeah..I only brought it up because I can think of ways to use it. Many iOS apps aren't exactly ports of OSX counterparts even if they do implement similar features or algorithms from older code bases. I would have thought encouragement to use OpenCL in the development of new apps might help bring more productivity software to iOS. As you point out OpenCL has finally gained proliferation through the entire Mac lineup. It makes things more interesting if it's available wherever possible.
 
Right now I'm waiting to see what the next Mac Mini update looks like as that'll determine whether I get a current Mac Mini cheaply, or a new Mac Mini, though there's still the possibility I'll buy a new Mac Pro anyway because I'm terrible with my money :)

I think I'm probably in the same boat as you! I think I'll wait to see what the next generation of Mac Mini looks like (PCIe SSD please!), and will read the nMP reviews once enough people are using them.
 
The Mac Pro is basically for people that work in Final Cut Pro X, or people that are hitting the ceiling of what a quad-core i7 can do and need more CPU cores in OS X. (assuming their application can take proper advantage of the additional cores)
Don't forget that you are moving back a generation in performance with the Mac Pro CPU (it's using Ivy Bridge-E) so each individual core is slower than a Haswell i7 at the same clockspeed.

So lets set aside the high end video argument, and get back to the OP question.. is the nMP worth it if you aren't into high-end video/animation?

What you are really saying is that the nMP is actually ideal for those who aren't in those areas. Folks who spend their days in FCX or Adobe offerings would seem to have the most to gain, especially when there's the big dropoff to an iMac or rMBP.

And is the slightly better clockspeeds on a i7 in an iMac really actually going to give you better performance than the nMP when you consider all the other factors? (SSD speed, memory bus, video, etc?)
 
And is the slightly better clockspeeds on a i7 in an iMac really actually going to give you better performance than the nMP when you consider all the other factors? (SSD speed, memory bus, video, etc?)

If you are talking about a 4 core nMP, then yes, definitely -- the iMac will be faster for the non GPU accelerated tasks.

I went in the Apple store yesterday and did a quick comparison opening 11 Nikon D800 RAW images (36 megapixels) from Adobe Bridge to Photoshop CS6, and then merging them to a panorama on a MBP with a 2.3 ghz processor and the 4 core nMP.

The MBP took 47 seconds to convert the images, and 183 seconds to create the panorama. The nMP was barely faster, at 42 seconds and 176 seconds.

What I really wanted to compare was the fastest MBP with the 2.6 ghz processor, vs. the six core nMP, but they did not have either of those on the floor. And I also wanted to compare Lightroom and Photoshop CC, but they would not let me install those (Photoshop 6 was installed).

Based on these results I assume the 2.6 ghz MBP would have edged a bit ahead of the nMP, and the fastest iMac would be distinctly ahead. My conclusion is that the 4 core definitely would not be worth getting for photography work.

What I still don't know is how much difference the extra cores in the six core model would make. Converting the images does use multiple cores, but merge to panorama is single threaded. And Lightroom (where I spend most of my time) generally is pretty good about using multiple cores.

And another issue is RAM -- the nMP I tested was the stock configuration with 12 gigs of RAM, and the MBP had 16 -- Photoshop runs out of RAM doing that panorama (even with 16 gigs) so that might have hobbled the nMP in the test, and conversely, a 32 or 64 gig nMP might be noticeably better even in single threaded tasks if they need enough RAM (while the MBP of course tops out at 16).

I need to buy a new machine (I am still using a first gen Mac Pro), but it will only be for photography. I still don't know if I would be better off with a top-end MBP or a six core nMP.
 
I need to buy a new machine (I am still using a first gen Mac Pro), but it will only be for photography. I still don't know if I would be better off with a top-end MBP or a six core nMP.

Let assume both are of the same speed.

MBP, ram and CPU will be fix and none upgradeable.

nMP, Ram max 64gb (maybe future 128gb) and also self upgrade of CPU when price drop.

Lastly also depend on how long you going to hold on to the machine.
 
This article from last year about the impact of OpenCL on photo editing is pretty interesting, particularly the interview with the Photoshop architect...
http://www.tomshardware.com/reviews/photoshop-cs6-gimp-aftershot-pro,3208.html

(sorry if this was already posted here -- if so, I missed it)

It definitely gives the impression that OpenCL is the future for Photoshop etc. So the GPU power of a nMP will eventually be very useful for photography. The question is when...

Thank you for sharing. The article seems very interesting and good. I'll take time and will read it thoroughly.
 
What I really wanted to compare was the fastest MBP with the 2.6 ghz processor, vs. the six core nMP, but they did not have either of those on the floor. And I also wanted to compare Lightroom and Photoshop CC, but they would not let me install those (Photoshop 6 was installed).

Based on these results I assume the 2.6 ghz MBP would have edged a bit ahead of the nMP, and the fastest iMac would be distinctly ahead. My conclusion is that the 4 core definitely would not be worth getting for photography work.

What I still don't know is how much difference the extra cores in the six core model would make. Converting the images does use multiple cores, but merge to panorama is single threaded. And Lightroom (where I spend most of my time) generally is pretty good about using multiple cores.

Yeah, LR does seem to soak the cores well, and with D800 files, that's a lot of soaking to do.

Of course, the other part of this is, loaded rMBP is $2700, loaded 27 iMac is $2800, and the base nMP is $3k. So the price difference isn't huge there anyway.
 
Of course, the other part of this is, loaded rMBP is $2700, loaded 27 iMac is $2800, and the base nMP is $3k. So the price difference isn't huge there anyway.

For the MBP, I would definitely go for the 2.6 ghz 1 Tb SSD (and Applecare) so I am looking at $3,648. And if I went nMP, I would get 6 core, 1 tb, 32 gig RAM, and Applecare, so $5048 with D300's or $5448 with D500's.

iMac is a non-starter for me -- I have a 30 inch wide gamut matte monitor already. The 27 inch glossy display would be a downgrade, and I don't want both on my desk. On the other hand, having the 15 inch retina as a secondary display is actually pretty appealing, so another point in favor of MBP.

----------

Thank you for sharing. The article seems very interesting and good. I'll take time and will read it thoroughly.

Here is another interesting article -- the open source RAW converter Darktable has already integrated OpenCL, with very good results...
http://www.darktable.org/2012/03/darktable-and-opencl/

So this demonstrates that it is possible to leverage GPU for RAW conversion, which gives hope that Adobe at least theoretically could do it? And given the previous article I posted, I suspect they are working on it.

That being said, if OpenCL does come to Lightroom (and more of Photoshop), the GPU's in a nMP might still be overkill for it.
 
For the MBP, I would definitely go for the 2.6 ghz 1 Tb SSD (and Applecare) so I am looking at $3,648. And if I went nMP, I would get 6 core, 1 tb, 32 gig RAM, and Applecare, so $5048 with D300's or $5448 with D500's.

iMac is a non-starter for me -- I have a 30 inch wide gamut matte monitor already. The 27 inch glossy display would be a downgrade, and I don't want both on my desk. On the other hand, having the 15 inch retina as a secondary display is actually pretty appealing, so another point in favor of MBP.

----------



Here is another interesting article -- the open source RAW converter Darktable has already integrated OpenCL, with very good results...
http://www.darktable.org/2012/03/darktable-and-opencl/

So this demonstrates that it is possible to leverage GPU for RAW conversion, which gives hope that Adobe at least theoretically could do it? And given the previous article I posted, I suspect they are working on it.

That being said, if OpenCL does come to Lightroom (and more of Photoshop), the GPU's in a nMP might still be overkill for it.


The iMac is a non-starter for me as well. My choices are really the hex (which I have on order) or a Hackintosh. The price difference between my current order and a ~$2k hackintosh makes that a compelling discussion. That's why I really want to see some more results from the Hex/D500.

LR/PS are my biggies as well (besides WoW :rolleyes:) so the question of that future OpenCL support just muddies the waters further.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.