Plenty of reasonable ways to multi thread a word processor, actually. But the more basic issue is that a word processor can't even consume one whole CPU.
Anything with a GUI could make decent use of multi-threading (and this is the most common way Cocoa will force multi-threading.) But again, you might not be doing enough work to consume more than one core anyway.
The devil's in the details. I can seriously see recycled code (development process using shortcuts) causing issues with releasing resources, as it won't work with the API's properly. And even knowing this, can the team developing the API's actually close all the doors for illegal operations that cause such issues. It just doesn't seem realistically possible due to things like time constraints and all the potential ways of writing code. I figured you'd see this is what I was getting at with that example.
But as you mention, it's core usage is minimal, and not necessary at all = it shouldn't even be attempted.
OS X does force multi-threading, especially if you are using Cocoa. I once wrote an app that had no explicit threading, and yet when I ran it, it was running with 16 threads.
If you don't mind, can you point me to some proof as to how this actually works?
I can't see it as truly forced from what I recall (currently under the impression that what you're talking about here has to do with out of order processes running simultaneously, such as running the GUI elements simultaneously on separate cores from the actual data process). Not processes that are based on recursive algorithms.
Multi-threading also doesn't require a rewrite. I'm also not sure what these tools are you're referring to. Pthreads was ratified in... 1995? Mac OS X has supported threads since launch, OS 9/Carbon supported threads.... Again, I'm not sure what you mean by tools, but if you have a compiler that doesn't support multi-threading, you're probably working on a 386 running DOS.
Tools that were developed in-house, as there wasn't anything commercially available at the time those tools were needed, and are still in use. This is how I understand is part of what's going on with Adobe for example.
Now as those tools are likely based on previous tools, there's a good chance they don't conform (or fully) to the open specifications you're referring to.
You can do HD and a second display on an iMac (not that the built in display isn't really nice.) Why would a student need PCIe expansion for Photoshop?
Yes, the iMac can use a second display. I was talking about more than that or sizes the GPU can't handle sufficiently as well (i.e. trying to run say 3x or more 30" monitors).
As per Photoshop needing PCIe, it's more to do with drives. The MP, as any system, is limited to how many disks you can place internally. The iMac is worse, as it can only take up to 2x for some models.
Now it's possible they can use USB or FW for expansion (will allow them access to additional capacity), but it can't increase speed (particularly relevant to the iMac). For the MP, they can use RAID (software or hardware based) to increase their speed as well as capacity.
Now I wouldn't have thought that this was necessary for students, but the information I'm getting indicates otherwise (I do assume their information is accurate, not exaggerated).
Apple simply doesn't care to compete in all markets. They've made this widely known over the years... The prosumer tower market is one place they've repeatedly made clear they're not competing. It's been this way for many years.
But in the case of the SP MP, they do chose to compete, as that product exists.
H.264 encoding, however, is a critical part of a workflow. I may need to do multiple encodes to make sure my video quality is good over an entire film. If it takes me X hours to do that (during which I can't really do much work), I'd pay good money to do that in X/2. And honestly, if I'm cutting film, I'm probably getting paid enough to where spending an extra few thousand on that horsepower is not going to bother me.
Most likely, if I'm in pro video editing, my software costs more than the top end Mac Pro anyway.
As I said, it all comes down to specific usage. If yours is that high, it has more relevance as to which machine will better serve your needs.
But for others, this may not be the case (H.264 encoding doesn't exact as much of their time for whatever reason).
This is why when the RAID questions come up, I have to try and get the smallest details out of the user to figure out how to best go about getting them a usable solution, not one that ends up useless, and a waste of funds. Those details are critical.
Which is true, expect in this case, where threading came before consumer machines. True, we didn't get an OS that supported multiprocessing until 2001, but threading is not some new thing. Developers have been doing it for a long time. In this case, the hardware had to catch up to software.
As mentioned before, I'm sticking with the enterprise market, not the consumer side (so I'm skipping the likes of the P4). Multi-threading started there, not for consumers.
So it's still a Top Down/Trickle Down approach.
Not sure where you are seeing this. After Effects and Premiere can both use 8 cores, as an example.
Logic is still limited to 4x cores per CPU, as IIRC, those with SP Hex and DP Dodeca systems noticed that they were only using 4x cores per (4x on the SP Hex, 8x on the Dodeca).
There is absolutely no reason things should be fixed to 4 cores unless a developer put in a hard limit. All the software I write scales on the fly, and I have been able to very easily use all 8 cores.
Exactly my point. The code that set this limit was developed years ago (i.e. single, maybe dual core CPU's out at the time, and more cores per CPU weren't expected any time soon). So they took a short-cut, and fixed the core limit, with the expectation that when more cores finally came, they'd fix it with a re-write at that time.
And what's being seen, are the vendor's that have been dragging their feet at producing the necessary re-write. Again, this is where cost reasons are the most likely culprit, as both the necessary tech and personnel that know how to do it already exist.
After following this conversation for several days now I still don't get why Haswell would end the Mac Pro line. Even if it is a more expensive part, it won't be that much more expensive than current Xeons. I understand the wafer scale and all, and that there will be more bells and whistles on chip, but I believe that this is just the general direction of high end systems for the coming years. A solution on a chip for the high end. Apple will still continue to sell the Mac Pro, and people will continue to buy it. Especially so if Apple's market share continues to grow.
It's not so much Haswell, but Apple's margins applied to a system based on those parts.
As the price keeps going up, the sales volume will shrink if it's too high. This has already happening as a result of the Nehalem based machines, as the enthusiasts are falling by the way-side, and at least a fair few of the independent pros here on MR are indicating they're nearing their breaking point/questioning the viability of the OSX platform at higher financial costs. As users that will by the MP shrink (can't or refuse to pay the MSRP at that time), there's fewer system sales to divide the R&D expenses. Even the component costs go up, as Economy of Scale is going backwards. So the per system costs will increase (direct and indirect costs applied, but pre profit = total system cost to Apple). In order for Apple to retain their desired margins, prices will have to increase again.
So at some point, the price will exceed what pros can actually afford to pay (classic Supply and Demand economics), and will hit a crux point for Apple where they will determine the MP is EOL, and kill it.
Now it's possible Apple could reduce their margin on the MP to keep it a viable, selling product line (I've mentioned this before, but I'm not sure if it's been noticed). But given the recent trends with Apple (consumer goods taking off and has become a significant portion of their total profit margins, the enterprise market shrinking for Apple, and their desire for high margins), I don't see this happening.
One more note, and this one is a very long term outlook. It is entirely possible that we will see Apple (and many PC manufacturers) dump x86 for the ARM platform as it advances over the coming years. If that were the case then the Mac Pro would be an (likely much cheaper to manufacture) ARM based workstation, making all of our hopes and fears in this particular thread moot.
Not going to happen.
ARM was never intended for that, and can't be as it currently exists (designed to be a low power consumption embedded processor, particularly for portable consumer devices). If they were to adapt it for workstation use, it would become another competitor with Intel and AMD (far larger, more complex, and use as much power).
This is why their products are filling the processing role for things like netbooks and the iPad (Apple's A4 = built from an ARM Core 8). These types of products is what it was actually designed for.
It's theoretically possible they could try to create enterprise grade parts, but it's not their specialty, and they'd have to contend with Intel. The ARM processor filled a hole, and why they've been successful with it.
x86 and the whole desktop/laptop full OS is about to get marginalized.
Depends on where you're talking about geographically speaking. Developing nations, particularly Asia, are where the desktop parts are selling like mad (they're after cheap systems). As their income continues to increase, they'll want more portability and move to laptops and netbooks (need something a smartphone can't fill, particularly where screen real estate isn't sufficient enough).
Chip makers such as Intel are well aware of this, and is where they're aiming their marketing strategies at.
For developed countries, they're aware that the consumer market is changing. It's also why they're focused on the enterprise market here (i.e. shift to cloud computing/clusters).
When this will truly happen, is when ISP speeds are sufficient everywhere and cheap enough for the vast majority of consumer users to utilize such implementations. Software vendors are happy with this shift, as they can make more money with the pricing models that will accompany it (i.e. usage based, not only make money when a new version is released). Intel's fine with it (AMD too, whether they like it or not
), as they'll be selling the CPU's that the clouds/clusters will be built from.
My only point was that if Apple moved to a completely consumer gadget oriented product lineup in the future... ie iPhone, iPad, ARM based MacBook Air, ARM based iMac... then the only job the "Pro" machine would really have is development for that ARM gadget line. In that case I don't believe Apple would maintain an x86 Mac Pro or the like just for high end users. In other words they would in effect kill off their present Pro user base.
This is my point.
The difference is, I see the sales volume dropping to the point it's actually a realistic outcome as things are now.
If we lose the x86 Mac Pro then we will once again have to turn to SUN and the like and we will once again be buying machines that cost as much as a new cars. People who complain about the price of the Mac Pro have no idea what the alternative is, cost wise. I remember all too well when a nicely loaded consumer PC was $4000+ and a similar Mac was significantly more.
Intel and other CPU makers are aware of these issues, and are taking pains to keep the value high (cost/performance will be better than with previous Unix based workstations and servers of the past).
The trick, will be what system vendors decide on for acceptable margins, which ultimately will be reflected in the MSRP.
But there will be some price increases. How much is the real question...
As far as the kids go, that has sadly already happened to a huge degree. My first machine was an Atari 800. As I got older I got my hands on all sorts of nice hardware including a couple of SPARCstation's. I learned Basic > Assembly > Fortran. Kids today take "computer science" and learn all about extremely technical stuff like "spread sheets" and "power point presentations." It's a dark time, for certain.
Even the engineering side is affected, as there's no real-world experience while in college, and the mentorship system that existed in the past has de-evolved (intended to be an additional 4 years of on-the-job training, is in the process of disappearing entirely IMO) due to the senior guys being forced to retire or laid off because they're deemed "too expensive". So there's no one there to teach them what they don't learn while getting their degree. What they do get, is inconsistent and has gaps (those they learn from teach what they can, but they can't teach what they don't know, or if they're no longer there = bits and pieces from multiple people - it's messy).
As a result, design and product quality suffers. Parts selection and testing is one notable area from what I've seen (build off of simulation software, but don't test out their parts for real world behavior based on the presumption that the simulation data is comparable with real world, which it typically isn't).
