Laptop for PI Processing (Dream Machine?)

[jerryyyyy]

I have been processing using a 32-bit Dell M2400.  This series is the so-called portable workstation series, which I have used in research for many years.  They have higher quality componets that the regular Dells like the Latitude series and separate graphics cards.  Anyway, looks like they have come out with my dream machine to replace the current model now 4-5 years old. 

This one is 64-bit.  The processor is a i7 and look like 4 core with 8 threads.... the devil is in the details with the i7 processors.

http://www.dell.com/learn/us/en/04/campaigns/precision-m3800-workstation

Comments appreciated regarding running PI.  Note the resolution of the Retina screen....

Here are the leaked specs:

The M3800 is powered by a fourth-generation Intel Core i7-4702HQ eight-threaded quad-core processor with a clock speed of 2.2 GHz and a maximum turbo-boost frequency of 3.2 GHz, which allows the CPU to run more efficiently under undemanding operating conditions, then speed up when resource-hungry software demands more processing power. The base configuration has 8 GB of 1600 MHz memory on board, expandable to 16 GB, and comes with a 500 GB hybrid drive, including an 8 GB cache, which is expandable to 1.5 TB of HDD, SSD, or hybrid storage.

The graphics hardware is Nvidia's Quadro K1100M with 2 GB of DDR5 VRAM. Dell Precision Senior Product Marketing Manager Mano Gialusis said it provides "true workstation-class" performance, optimized for software such as Adobe Creative Cloud, Autodesk Entertainment Creation Suite, and Avid Media Composer. He also noted that Dell's Optimus system monitors the graphics hardware and helps boost battery life, a critical consideration in a lightweight workstation. The K1100M can drive multiple external monitors via an optional D3000 USB 3.0 docking station, and also supports Dell's 31.5-inch Ultra HD UltraSharp 32 monitor, slated to ship later this fall.

The built-in screen is a 15.6-inch five-finger multitouch UltraSharp LED-backlit display available in native HD (1920×1080) or a Retina-beating QHD+ (3200×1800) versions. Asked about OS support for the high-resolution IGZO display, Gialusis told StudioDaily it has improved with the latest version of Windows. "We're launching the M3800 with Windows 7 and Windows 8.1," he said. "Windows 8.1 scales beautifully with the high-resolution panel. People who want to load with Windows 7 can go in and make adjustments, with different scaling percentages available on the system as well as the ability to adjust the resolution, as needed, to one of many settings, including 1920×1080.

[Phil Leigh]

Having just built a desktop machine SOLELY for Pi...
1) the graphics card makes little difference... Pi is not graphics-intensive
2) 16Gb RAM - as fast as you can get for the motherboard - is a good idea.
3) my new 8-core/4Ghz AMD machine runs Pi tasks >twice as fast as my old 4-core machine 2.4Ghz AMD machine (Win 7 64-bit in both cases)

Pi uses all the cores you can throw at it and anything that stops paging to disk (i.e. lots of RAM) is a good idea.

[georg.viehoever]

AMD is good when you want to have a cost effective machine. If you want performace, go for one of the latest Intel CPUs (haswell). Thje more memory channels the better - image processing is often memory bandwidth bound.

[Phil Leigh]

Sadly for me cost has to come before performance 

[jerryyyyy]

This is of course for work, and C++ programming, but I travel a lot and need a good road machine.  What is interesting is that the Haswell processors come in many many variants.  Initially I was hoping I could get one similar to a Mac AIR but the mobile versions do not have as many threads.  So, one way or another the machine is going to be heavilier but will have the 8 threaded processor. 

[EorEquis]

I went and had a look at Anand Tech's image manipulation benchmark when I built my desktop machine a year ago...like Phil, I built it specifically to process astro images.

At the time, the i5 3570K was 3rd on that list, with the other 2 options being considerably (many times) more expensive.  I adore it.  Even now, at a year out, I'm very pleased.

Phil is right : Graphics card is of no value to PI.  A high end graphics card does not help do complicated tasks...it helps do simpler tasks over and over again.  Great for rendering an object a zillion times as it moves across the screen in a game...not so much for calculating alignment points in a 30 frame stack.

RAM is king.  16 at least, 32 is even better.  The less swapping PI has to do, the happier it'll be.

Jury's still out on PI using the 4 "virtual" cores, imo.  I've run it on an i7 with them, and on my i5 without, and seen little difference.  Admittedly, an older i7, but it has the "8" cores, and I've not seen PI really benefit from them.  There's a bit of black magic there I don't understand though, so i'm sure the experts and/or PTeam can offer better insight than I in that regard.

[pfile]

Phil is right : Graphics card is of no value to PI.  A high end graphics card does not help do complicated tasks...it helps do simpler tasks over and over again.  Great for rendering an object a zillion times as it moves across the screen in a game...not so much for calculating alignment points in a 30 frame stack.

for now this is very true, but if PI ever grows OpenCL capabilities a high end video card will speed up some tasks in PI by orders of magnitude... many of these processes are already "embarrassingly parallel" - that's why PI is so good at saturating all your cpu cores.

not sure if the lack of native f64 support and/or incomplete IEEE f32 compliance on some GPUs is a permanent barrier to entry for PI though.

rob

[EorEquis]

Phil is right : Graphics card is of no value to PI.  A high end graphics card does not help do complicated tasks...it helps do simpler tasks over and over again.  Great for rendering an object a zillion times as it moves across the screen in a game...not so much for calculating alignment points in a 30 frame stack.

for now this is very true, but if PI ever grows OpenCL capabilities a high end video card will speed up some tasks in PI by orders of magnitude... many of these processes are already "embarrassingly parallel" - that's why PI is so good at saturating all your cpu cores.

not sure if the lack of native f64 support and/or incomplete IEEE f32 compliance on some GPUs is a permanent barrier to entry for PI though.

rob

An excellent point, and thanks for the reminder.

[jerryyyyy]

One place it seems to use the cores appears to be in the BatchPreprocessing Script, where with my dual core it seems to process two images at a time.  Eight might be better.