Mac Pro Could adopt Intel's Newly Announced 22nm Xeon

[dexum175]

SOURCE: http://www.patentlyapple.com/patent...d-adopt-intels-newly-announced-22nm-xeon.html

"Earlier today, Intel announced a number of new systems coming to market later this year. They touched on Silvermont, Bay Trail, Haswell and Xeon processors shifting to 22nm technology. The latter will provide Apple's Mac Pro users with a nice power boost upgrade if Apple so chooses to stay with this processor this fall. Their all-new Silvermont is unlikely a processor that Apple will use but it could always slip into a lower priced Mac mini if Apple so chose to entertain expanding the OS X market. Last week Intel illustrated that next generation Ultrabooks, premium notebooks like the MacBook Pro and high-end desktops like the iMac with Haswell could experience a huge leap in graphics power, as in 100 to 200% depending on the configuration chosen." - Patently Apple

 

[MyMac1976]

SOURCE: http://www.patentlyapple.com/patent...d-adopt-intels-newly-announced-22nm-xeon.html

"Earlier today, Intel announced a number of new systems coming to market later this year. They touched on Silvermont, Bay Trail, Haswell and Xeon processors shifting to 22nm technology. The latter will provide Apple's Mac Pro users with a nice power boost upgrade if Apple so chooses to stay with this processor this fall. Their all-new Silvermont is unlikely a processor that Apple will use but it could always slip into a lower priced Mac mini if Apple so chose to entertain expanding the OS X market. Last week Intel illustrated that next generation Ultrabooks, premium notebooks like the MacBook Pro and high-end desktops like the iMac with Haswell could experience a huge leap in graphics power, as in 100 to 200% depending on the configuration chosen." - Patently Apple

I don't think Atom is on the radar at Apple. iOS is built around ARM and MM's and 13 MBP's are just fine running normal mobile processors.

 

[deconstruct60]

SOURCE: http://www.patentlyapple.com/patent...d-adopt-intels-newly-announced-22nm-xeon.html

"Earlier today, Intel announced a number of new systems coming to market later this year. They touched on Silvermont, Bay Trail, Haswell and Xeon processors shifting to 22nm technology. The latter will provide Apple's Mac Pro users with a nice power boost upgrade if Apple so chooses to stay with this processor this fall.

There already are Xeon processors on 22nm technology; Xeon E3 v2 . ( Ivy Bridge is on 22nm technology. It just isn't optimized for it). There are Xeon E5 v2 ( "Ivy Bridge") coming later in Q3 but frankly the current Xeon E5 options would have provided a nice power boast upgrade months ago. There is nothing particularly power optimized coming with E5 v2. About exactly the same TDP constraints. In some cases increased core count and in most cases some relatively minor 0.2-0.3 GHz clock speed bumps.

Their all-new Silvermont is unlikely a processor that Apple will use but it could always slip into a lower priced Mac mini if Apple so chose to entertain expanding the OS X market.

Silvermount is too weak for a Mac Mini. An Airport Express or Time Capsule perhaps but not a Mini. It is doubtful that Apple would move off of the Marvell ARM SoC in the those two though since the integration with Wifi radio stack is better.

If Apple did a better home NAS solutions ( e.g., private cloud ) than the Time Capsule then one of the Silvermont variants would make alot of sense. I don't think Apple really cares about redundantly using bandwidth to do syncs on the local LAN rather than dragging all of that up their boxes to swap data between two boxes in your house.

high-end desktops like the iMac with Haswell could experience a huge leap in graphics power, as in 100 to 200% depending on the configuration chosen." - Patently Apple

Extremely unlikely except in some very narrow corner cases of recompiled code. There is some integer/bit manipulation vector (SIMD) processing that may see some 100-200% increases because the old systems used software to approximate what is now fixed function hardware. But general computations? No. Not even close to those kinds of jumps. Try 10-20%. That is closer to reality.

Similarly may see some jumps when comparing 4 core older systems to 8 core newer ones to see 100% increases on apps that scale perfectly by core count. But if do "apples to apples" comparison with same core counts not going to see anything like that.

 

[xav8tor]

No news here. They really are reaching for Mac Pro hopeful news. One sentence. All the original PR says is that Intel plans to "refresh" the Xeon @ 22nm to deliver better performance per Watt and "other features." I don't care about less power anything...at the wall or in the box. I still need GHz and bandwidth in addition to core count...MORE power please! "Other features" probably means an integrated GPU or some other feature that is useless in a pro setting.

 

[deconstruct60]

I don't care about less power anything...at the wall or in the box. I still need GHz and bandwidth in addition to core count...MORE power please!

You'd be in the small lunatic fringe. Even enterprise server shops are interested in more power efficient servers. That doesn't mean the computation "horsepower" goes down. Practically nobody wants chips that consume more electrical power just for giggles. IF that were true the old Pentium 4 designs would still be viable on the market. They aren't. Leaks current like a sieve doesn't really deliver much bang for the buck computationally.

"Other features" probably means an integrated GPU or some other feature that is useless in a pro setting.

Faster memory , better blended turbo speeds ,

In the case of the Xeon E3 v3 better SIMD, better instructructions per cylce, better load store bandwidth , better instructure trace cache, etc.

this has exceedingly little to do with bring integrated graphics to Xeon E5 solutions this year. But frankly, a OpenCL 1.2 capable iGPU wouldn't be useless in a "pro context".

 

[xav8tor]

You'd be in the small lunatic fringe. Even enterprise server shops are interested in more power efficient servers. That doesn't mean the computation "horsepower" goes down. Practically nobody wants chips that consume more electrical power just for giggles. IF that were true the old Pentium 4 designs would still be viable on the market. They aren't. Leaks current like a sieve doesn't really deliver much bang for the buck computationally.

Faster memory , better blended turbo speeds ,

In the case of the Xeon E3 v3 better SIMD, better instructructions per cylce, better load store bandwidth , better instructure trace cache, etc.

this has exceedingly little to do with bring integrated graphics to Xeon E5 solutions this year. But frankly, a OpenCL 1.2 capable iGPU wouldn't be useless in a "pro context".

Lunatic, huh? Is the same true of all overclockers, and all users of frequency driven professional applications, true of anyone who has found themselves CPU bound due to the GHz and not the core/thread count? While I have many apps that benefit enormously from multiple cores and other new gen architecture improvements, there are still apps out there that are substantially clock speed dependent. I use one...heavily. If my workstation is running it well, it means I am making money. If it runs slowly, then I am losing money.

My monthly power bill comprises a very small portion of my overhead costs, easily offset by a faster running machine. So, no, I don't care one bit about having a CPU that saves me ten bucks on next month's bill if it doesn't offer a measurable, noticeable gain in performance. Look at single threaded benchmarks comparing the 3770 to the 4770. They aren't exactly impressive gains. Actually, they aren't all that impressive overall. The fact that a W3690 can't even keep a GTX 670 busy running a frequency driven application is the only thing stopping me from going to the Titan in my Mac Pro. Why bother?

 

[CaptainChunk]

Lunatic, huh? Is the same true of all overclockers, and all users of frequency driven professional applications, true of anyone who has found themselves CPU bound due to the GHz and not the core/thread count? While I have many apps that benefit enormously from multiple cores and other new gen architecture improvements, there are still apps out there that are substantially clock speed dependent. I use one...heavily. If my workstation is running it well, it means I am making money. If it runs slowly, then I am losing money.

My monthly power bill comprises a very small portion of my overhead costs, easily offset by a faster running machine. So, no, I don't care one bit about having a CPU that saves me ten bucks on next month's bill if it doesn't offer a measurable, noticeable gain in performance. Look at single threaded benchmarks comparing the 3770 to the 4770. They aren't exactly impressive gains. Actually, they aren't all that impressive overall. The fact that a W3690 can't even keep a GTX 670 busy running a frequency driven application is the only thing stopping me from going to the Titan in my Mac Pro. Why bother?

But also remember that E5s are mostly targeted to enterprise/server customers. Workstation customers are a very small piece of that pie. In server environments where there are dozens upon dozens of servers installed in racks, power consumption DOES matter. And parallelism wins the day over clock speeds in those environments, as well. So, who is Intel going to listen to first?

But then again, I probably wouldn't go as far as calling the high-end workstation crowd "lunatic fringe," either. More than anything, software developers need to start embracing changes in CPU technology by making software more multithread-aware. In some ways, the hardware is moving at a much faster pace than the software...