NodeXL workstation recommendations?

May 4, 2009 at 6:48 PM
NodeXL prompted me to upgrade to Office 2007, and now it's driving me to upgrade my hardware, too.  I'd appreciate any recommendations on the kind of machine I should be looking to buy.

At this point, NodeXL is putting the heaviest demands on my computing, so if there are trade-offs to be made (I'm looking to buy a decent workstation, but nothing outrageous) I'd like to make the right ones. 

For example - and I'm making this up, since I'm no expert - should I favor more cores, or more MIPS?  How important is RAM vs. a specialized graphics processor? And so on...

thank you
May 5, 2009 at 1:56 AM
(This was answered offline.)

-- Tony
May 15, 2009 at 7:50 PM
Edited May 15, 2009 at 7:58 PM

To summarize some of our discussions about hardware:

Hardware can make a big difference in the performance of NodeXL.

CPU is clearly a bottleneck, but in many cases RAM is the first place to improve the performance of NodeXL.

The desire for lots of RAM leads to a related issue: 64 bit operating systems.  Most desktop operating systems (and laptops) are 32 bit applications with limits on the amount of RAM they can address (4GB theoretically, 3.5GB in practice).  64 bit operating systems can address huge volumes of RAM in contrast.  That said, the current version of Excel 2007 is a 32 bit application, no matter the type of operating system under which it runs.  Until the next version of Office arrives with a 64 bit implementation of Excel, NodeXL is limited to a 32 bit address space.

To maximize the RAM available to a 32 bit implementation of Excel, however, it can be useful to run in a 64 bit environment, which allows the OS to give each 32 bit application its own 4GB memory space (assuming the physical RAM is present).  

Therefore, I believe the current maximal hardware configuration for NodeXL (for those shopping for hardware) would be a fast multicore CPU system with 8GB of RAM running a 64 bit Vista OS.  More RAM is likely to be a waste for now until Excel can address a larger space.

I do not think NodeXL can take advantage of lots of cores or CPUs beyond the cpu on which its thread executes.  A dual cpu or dual core cpu may be as much hardware NodeXL can leverage for now.  Tony may be able to correct me on this!

If you are a developer, NodeXL code can be extracted and compiled to run in a 64 bit environment outside of the constraints of Excel.

One caveat: the 64 bit Vista may not be that friendly for consumer use.  If the main machine you intend to use is the same as the NodeXL machine, it may not be worth the driver hassles and related issues from running a non-standard OS.  

A fast video card is also a nice performance booster, although any mainstream video card should now more than ably support NodeXL (until we add 3D!).

The biggest performance boost is likely to come from adding two or more monitors as large as possible to the NodeXL analysis workstation.  The ability to devote a screen to the network graph alone while another screen hosts the spreadsheet view of graph data is the best way to work effectively in the NodeXL environment.