Mac Pro: It could be worse

[FluJunkie]

So, was browsing the Apple site, and musing about the occasional "new" Mac Pro language and the like, and Pro users feeling abandoned.

The good news is, it could be worse. The Mac Pro site talks about current graphics cards, modern-ish chips, SSDs...

It could be worse. You could be a scientist.

The Apple Science section? Has gems like the following:

And for those times when you want to boot directly in Windows, there’s Boot Camp, which is included with every new Mac running Mac OS X Leopard.

With up to 10.5TB of storage and a quick Fibre Channel interface, Xserve RAID storage allows you to access and archive your data quickly — at less than $2 per gigabyte.

The Apple Workgroup Cluster is an all-inclusive, high-performance cluster perfect for sequence analysis, data mining, and other embarrassingly parallel operations.

The 64-bit Mac Pro with up to 32 GB of RAM and up to 4 TB of local storage...

This whole page: http://www.apple.com/science/solutions/workgroupcluster.html

Tutorials in Sharing Information with Mac OS X Leopard Server Wikis

A "cluster computing" page that's just a link to Lion Server...

An image slideshow on the banner that heavily features Xserve...

---

So whenever you're feeling blue, head over to the Science section of Apple's site. Then when you look back at the Mac Pro, you'll feel the sudden rush of euphoria that comes from knowing at least the site is talking about products that Apple still sells.

 

[derbothaus]

I hope the class action dude sees this. New reason for existence.

 

[sjinsjca]

Not to mention some of their case studies, such as http://www.apple.com/science/profiles/vatech2/ ...which describes Virginia Tech's supercomputer cluster, based on the long-obsoleted Xserve rack-mounted Mac.

Apple needs to get serious about the scientific market. But as a market segment it's rarely if ever mentioned on Apple's job site. (I've looked.)

 

[Zarkizon]

If you find errors on the Apple website, you can fill out this form here concerning issues (like outdated content) on the site.

A few years ago I noticed that on the MacBook Air page they still were advertising OS X Leopard as the OS when Snow Leopard had been out for a substantially long time, and it was fixed not long after.

 

[deconstruct60]

So, was browsing the Apple site, and musing about the occasional "new" Mac Pro language and the like, and Pro users feeling abandoned.
....

The good news is, it could be worse. The Mac Pro site talks about current graphics cards, modern-ish chips, SSDs...
.....

This whole page: http://www.apple.com/science/solutions/workgroupcluster.html

An image slideshow on the banner that heavily features Xserve...
....

If those pictures do date from the end of XRaid era that picture of a portable "half rack"

with 9 XServes is probably petty close to being replaceable by a new Mac Pro with dual E5 2670s. If can leverage one of the newest GP-GPU cards and a dual E5 2670s, even more so.

----------

Apple needs to get serious about the scientific market. But as a market segment it's rarely if ever mentioned on Apple's job site. (I've looked.)

Apple should hire biologists , theoretical physicist, and chemists to sell packaged Mac solutions? Not.

Apple doesn't need very many basic science folks because they don't do basic science. There is no "Bell labs" or "Watson labs" equivalent at Apple.
There are probably more than few working in Apple stores though (not particularly doing science).

 

[Ruahrc]

Apple should hire biologists , theoretical physicist, and chemists to sell packaged Mac solutions? Not.

Apple doesn't need very many basic science folks because they don't do basic science. There is no "Bell labs" or "Watson labs" equivalent at Apple.
There are probably more than few working in Apple stores though (not particularly doing science).

I don't think he is suggesting Apple hire biologists and conduct cancer research, or even to sell Macs to other biologists. What he is suggesting is that Apple start marketing more towards scientific markets. And start developing products for scientific applications.

However, honestly I think they have the scientific area fairly well covered now. Aside from rackmount servers, high-end Mac Pros are pretty well equipped for scientific work, and there is something to be said about the whole OSX/Linux/Unix synergy.

Look around in most science/engineering departments and you see a lot of macs sitting on desks. High-end equipment are almost always connected to PCs though, I think that's because the Windows OS is more open for development especially when it comes to highly customized hardware addons.

 

[itsmrjon]

Look around in most science/engineering departments and you see a lot of macs sitting on desks. High-end equipment are almost always connected to PCs though, I think that's because the Windows OS is more open for development especially when it comes to highly customized hardware addons.

I can vouch for this. I work in a lab that develops CFD (computational fluid dynamics) codes and every desk has a mac on it. However, our cluster is all tyan hardware running redhat.

By no means would I ever use a mac pro as a 'super computer,' however I currently use it for testing code and visualizing data. It's really hard to test an MPI application without multiple processors on the machine, and it's not worth wasting computational units on supercomputers to fix bugs.

Hence... need a new mac pro

 

[sjinsjca]

Apple should hire biologists , theoretical physicist, and chemists to sell packaged Mac solutions? Not.

Agreed, certainly. At no point did I advocate for packaged Mac solutions in my post above. My point is: there is untapped potential in university and industrial R&D labs, and Apple is not talking effectively to those possible Mac (and iOS) customers.

 

[goMac]

If those pictures do date from the end of XRaid era that picture of a portable "half rack"
Image

with 9 XServes is probably petty close to being replaceable by a new Mac Pro with dual E5 2670s. If can leverage one of the newest GP-GPU cards and a dual E5 2670s, even more so.

You can tell it's from the end of the XServe era as they're featuring that third party RAID box they pushed instead of the XServe RAID.

 

[deconstruct60]

What he is suggesting is that Apple start marketing more towards scientific markets. And start developing products for scientific applications.
.....
High-end equipment are almost always connected to PCs though, I think that's because the Windows OS is more open for development especially when it comes to highly customized hardware addons.

The "developing products" and open to development don't match up. Microsoft provides the OS and development tools. Apple needs to provide the platform and something nurturing for 3rd parties to do the apps. If Apple does core OS X and drivers, keeps moving XCode moving forward , and keeps the window open for Intel and other high end compiler vendors (e.g., PGI) , then it shouldn't be a problem.

It's really hard to test an MPI application without multiple processors on the machine, and it's not worth wasting computational units on supercomputers to fix bugs.

It wouldn't hurt for for Apple to contribute some porting help ("free" hardware and/or people ) to some projects like OpenMPI & OpenMP , but as long as they incrementally improve OS X's Posix and Unix foundation that should not be a hard problem for partners/contributors/etc that don't work for Apple to make work well.

There are some gaps on OS X like Infniband, but that isn't a show stopper.

My point is: there is untapped potential in university and industrial R&D labs, and Apple is not talking effectively to those possible Mac (and iOS) customers.

I think Apple does far more talking in showrooms and user deployments than on some web pages. The web pages are a "nice to have", they are not really the core driver of sales. Most people in labs can build their own solutions if given the basic building blocks to work with. Just about all places where the "data center priests" or "Uniformity IT priests" don't run the show the Macs have a chance.

It is more important for Apple to find multiple areas for the Mac Pro to expand into than to supersaturate one or two ( e.g., media and R&D labs).

It doesn't help to have 3-4 year old stale case studies up. However, I doubt it is significantly holding them back. If pressed I wouldn't be surprised to see them take down what is there.

 

[Bubba Satori]

Really sad.

 

[goMac]

The rumors of Apple looking for a replacement to fill the hole left by the XServe is good though. It reminds me of the death of the Newton. Apple said they were looking at replacements for the eMate and Messagepad, and it led to the iBook and (many years later) the iPhone.

With Apple pushing FCPX and pro features hard, one can only hope that means pressure from the top down to keep and advance pro Mac hardware as well.

For all the fuss when it came out, FCPX is shaping up to be really strong pro software. I'd hate to be one of the people that freaked out and switched away when it was first released.

 

[spock2112]

I actually run a scientific lab, largely off of Mac Pros and 27" quad-core i7 iMacs. XGrid, while a bit awkward at times, works well to get a grid setup in place quickly and easily. We can have a single box there that people can use for office apps, programming, etc. and that also donates a good number of cores to the common grid. It's been a very cost-effective solution given the dual-purpose each machine serves.

 

[chrono1081]

I can vouch for this. I work in a lab that develops CFD (computational fluid dynamics) codes and every desk has a mac on it. However, our cluster is all tyan hardware running redhat.

By no means would I ever use a mac pro as a 'super computer,' however I currently use it for testing code and visualizing data. It's really hard to test an MPI application without multiple processors on the machine, and it's not worth wasting computational units on supercomputers to fix bugs.

Hence... need a new mac pro

Your job sounds cool...just sayin...(I'm not being sarcastic either).

 

[throAU]

To be fair, the xserve kinda outlived its usefulness.

Apple is not a high-end storage specialist, and will never compete well in that space (the big boys like EMC and Netapp do that as their core business and are both very good and very far ahead of apple starting from nothing). If you want high speed storage, you buy NetApp, EMC or similar (or roll your own using ZFS on FreeBSD or Solaris).

The xserve software is still available if you want to share out a filesystem from even a mac mini or mac pro connected to a decent SAN.

Yes they could do with a rack mount box. fit some rails to the next mac pro and it would be good enough - but seriously, if you just want raw compute, you are better off buying a server from someone who specializes in that space. Even apple's own cluster doesn't run apple hardware.

It's not their focus, and trying to use Mac hardware in that environment (bulk compute, as opposed to visualization of the computed data as per the dude above) is very much a case of square peg in round hole...

 

[goMac]

To be fair, the xserve kinda outlived its usefulness.

Apple is not a high-end storage specialist, and will never compete well in that space. If you want high speed storage, you buy NetApp, EMC or similar (or roll your own using ZFS on FreeBSD or Solaris).

The xserve software is still available if you want to share out a filesystem from even a mac mini or mac pro connected to a decent SAN.

I think you mean XRaid, not XServe.

 

[InuNacho]

Curious what your Apple Workgroup Cluster might look like? Configure and price your system now.

You forgot the link to nowhere http://wgcc.apple.com/.

If it's any consolation the Pro section of the site still has ads for FCS3 and the Mac Pro still being advertised as new.
http://www.apple.com/pro/

 

[itsmrjon]

Your job sounds cool...just sayin...(I'm not being sarcastic either).

Sometimes it's cool. I'm technically a Mechanical Engineer/Applied Mathematician but spend most of my time doing computer science. Think of it as atypical programming. Some of the codes we build are pretty awesome, some are really boring.

Heres a video of a code I just started writing this past weekend.

http://www.youtube.com/watch?v=qi3vWgtTBzw

It's a lid driven cavity flow with a Reynolds number of 10,000. There are 2,000 particles in the domain which is [0,0] to [1,1] square.

Solution is done using the MAC Method and the particles are Montecarlo (massless) particles which use linear interpolation and an Euler scheme for integration.

This code is currently setup to run on 8 processors (shared memory at the moment), written in FORTRAN 2003, and that simulation took about 16 hours to run on the Mac.

 

[chrono1081]

Sometimes it's cool. I'm technically a Mechanical Engineer/Applied Mathematician but spend most of my time doing computer science. Think of it as atypical programming. Some of the codes we build are pretty awesome, some are really boring.

Heres a video of a code I just started writing this past weekend.

http://www.youtube.com/watch?v=qi3vWgtTBzw

It's a lid driven cavity flow with a Reynolds number of 10,000. There are 2,000 particles in the domain which is [0,0] to [1,1] square.

Solution is done using the MAC Method and the particles are Montecarlo (massless) particles which use linear interpolation and an Euler scheme for integration.

This code is currently setup to run on 8 processors (shared memory at the moment), written in FORTRAN 2003, and that simulation took about 16 hours to run on the Mac.

Thats so cool! I couldn't even know where to begin with something like that.

 

[xVeinx]

The problem with the XServe was that it wasn't as practical a solution overall. Most of the scientific computing these days is geared towards high performance, high memory, etc. In order for Apple to compete in this space, they would have to put more money and time into R&D for these types of applications, increasing the flexibility of their servers and be able to include more CPU/Memory configurations, GPGPU, interconnects, et. al.. Apple can't be everything in the computer world, and when the market branched into either highly-specialized hardware or low-margin, high volume machines, the interest died.

 

[itsmrjon]

Thats so cool! I couldn't even know where to begin with something like that.

This is the paper I'm referencing for developing the code. http://diyhpl.us/~bryan/papers2/frey/divers/McKee S., The MAC method.pdf

It's just a matter of following along and building the program to solve the provided equations. Complications only occur when you try to do something new.

If you're interested in trying to write a CFD code, there are many simple cases/solvers you can get started with. Look up the heat equation, that one is rather popular for learning numerical methods.
http://www.cfd-online.com/Wiki/Finite_difference

CFD is honestly one of those fields where you kill for speed. Going 2 years without updating a workstation is pretty unheard of around these parts. Sadly I'm getting tired of waiting for a new Mac pro, I've been eyeballing a Cray psc lately... Apple has no love for us scientists anymore.

 

[itsmrjon]

It wouldn't hurt for for Apple to contribute some porting help ("free" hardware and/or people ) to some projects like OpenMPI & OpenMP , but as long as they incrementally improve OS X's Posix and Unix foundation that should not be a hard problem for partners/contributors/etc that don't work for Apple to make work well.

There are some gaps on OS X like Infniband, but that isn't a show stopper.

I think the third-party market fills the software gap rather well for apple. PGI offers some awesome compilers that allow for using shared/distributed memory programming on OSX. I've never had much of an issue running my codes on my mac with the exception that it simply does not have enough memory/cores. (I mention PGI rather than intel because PGI supports CUDA on C and FORTRAN. They also have the PGI CDK which is pretty awesome for building your own cluster.)

 

[chrono1081]

This is the paper I'm referencing for developing the code. http://diyhpl.us/~bryan/papers2/frey/divers/McKee S., The MAC method.pdf

It's just a matter of following along and building the program to solve the provided equations. Complications only occur when you try to do something new.

If you're interested in trying to write a CFD code, there are many simple cases/solvers you can get started with. Look up the heat equation, that one is rather popular for learning numerical methods.
http://www.cfd-online.com/Wiki/Finite_difference

CFD is honestly one of those fields where you kill for speed. Going 2 years without updating a workstation is pretty unheard of around these parts. Sadly I'm getting tired of waiting for a new Mac pro, I've been eyeballing a Cray psc lately... Apple has no love for us scientists anymore.

Thanks for the info! I'll give these a look later today.

 

[FluJunkie]

This is the paper I'm referencing for developing the code. http://diyhpl.us/~bryan/papers2/frey/divers/McKee S., The MAC method.pdf

It's just a matter of following along and building the program to solve the provided equations. Complications only occur when you try to do something new.

If you're interested in trying to write a CFD code, there are many simple cases/solvers you can get started with. Look up the heat equation, that one is rather popular for learning numerical methods.
http://www.cfd-online.com/Wiki/Finite_difference

CFD is honestly one of those fields where you kill for speed. Going 2 years without updating a workstation is pretty unheard of around these parts. Sadly I'm getting tired of waiting for a new Mac pro, I've been eyeballing a Cray psc lately... Apple has no love for us scientists anymore.

Yay computational science people. I'm an Epidemiologist, who works mostly with outbreak simulations.

My fondness for Apple and science is...mixed. They make great client machines, but their server offerings are non-starters, and if the Mac Pro doesn't get updated soon, their workstation offerings are pretty sub-standard. It's hard to justify "But I like my laptop's OS..."

 

[goMac]

It wouldn't hurt for for Apple to contribute some porting help ("free" hardware and/or people ) to some projects like OpenMPI & OpenMP , but as long as they incrementally improve OS X's Posix and Unix foundation that should not be a hard problem for partners/contributors/etc that don't work for Apple to make work well.

Eh, I think OpenMP is kind of done. I mean, it's easy to code for, but stuff like OpenCL is much more powerful. Yeah, OpenCL is more difficult to code, but it can scale way way better, especially for scientific uses.

My feeling from Apple is if you don't want to go all the way to OpenCL, they really want you to use libDispatch instead.