Will having an external graphics card help displaying large waves?

Currently I have an intel i3 dual core processor @ 2.13 GHz with integrated graphics and 4 GB RAM. I have no external graphics. When I try to display large waves (0.5 million points) it take a lot of time. I expect that having a medium-range external graphics card (not a high end gaming one) will help in this. I am asking since I am considering buying new laptop and if the graphics card will help then I'll take that into account.

Can anyone confirm this? (I know that a lot of factors came into consideration, but a general statement is fine).
The i3 is old so you will probably see a significant improvement by upgrading to i5 or i7 which will more than likely come with improved onboard intel graphics. A separate graphics card would give you a substantial boost in Gizmo plots and some help with memory but I see no indication that it is relevant in your application. Outside Gizmo most of the calculations are preformed by the CPU so I do not think that you will see much speed improvement by getting a system with a dedicated graphics card. I also think that upgrading from 4G to 8G RAM would be very effective.

A.G.
WaveMetrics, Inc.
Thanks. But I am focusing more on 2D plots, does these also get benefit? In another words, I know that Igor (through the Gizmo function / tool) utilizes OpenGL for rendering. Does it also use it for 2D?
ali8 wrote:
Thanks. But I am focusing more on 2D plots, does these also get benefit? In another words, I know that Igor (through the Gizmo function / tool) utilizes OpenGL for rendering. Does it also use it for 2D?


IGOR does not directly use OpenGL except in Gizmo. This however does not mean that your operating system does not make OpenGL calls for display purposes.

Going back to what I said earlier: an upgrade of RAM would be the obvious first choice. Then consider going to i5/i7 with newer graphics hardware.

A.G.
WaveMetrics, Inc.
For the sake of completeness, I got rid of the slowness and everything is now fine with the same PC...but there is a catch.

When viewing the waves as 'markers', everything is pretty slow as described above. The only way to get rid of that is by changing the view to 'line between the points', so that there are only lines and no points (markers). It seems that this reduces the computational requirements somehow.
When you use markers, each marker is drawn individually, so there are simply a lot more drawing operations. AG's advice to upgrade the processor is a good one, and possibly an upgrade to your graphics system would help, but I don't know how much. It seems to me that drawing 500,000 markers isn't very useful- don't they overlap to the point of being useless?

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
Do the recommendations here still apply for IP7?

- When resizing a graph (with 56 2D plots) it is taking my admittedly slow computer a few seconds.
Yes, I'd say that everything above is still valid for IP7.

One thing to watch out for in IP7 is that if you're using a High-DPI display (sometimes called 4K or 5K), drawing certain graphs can be several orders of magnitude slower (compared to a standard DPI display). This is true on both Macintosh and Windows, though possibly to different degrees.

Pretty much all of Igor's drawing code executes in the main thread, so for drawing you want the fastest processor clock speed you can get. On the other hand, many analysis routines can be threaded, so for analysis having multiple cores will likely net you better performance. Outside of Gizmo, the graphics card doesn't really matter.
aclight wrote:
Yes, I'd say that everything above is still valid for IP7.

One thing to watch out for in IP7 is that if you're using a High-DPI display (sometimes called 4K or 5K), drawing certain graphs can be several orders of magnitude slower (compared to a standard DPI display). This is true on both Macintosh and Windows, though possibly to different degrees.

Pretty much all of Igor's drawing code executes in the main thread, so for drawing you want the fastest processor clock speed you can get. On the other hand, many analysis routines can be threaded, so for analysis having multiple cores will likely net you better performance. Outside of Gizmo, the graphics card doesn't really matter.


Thanks Adam - now I get why resizing took 40+ seconds on a H-DPI display and 3s on a standard display. Is there any plans to move drawing (does drawing = GUI?) out of the main thread for IP8?
Okapi wrote:

Thanks Adam - now I get why resizing took 40+ seconds on a H-DPI display and 3s on a standard display. Is there any plans to move drawing (does drawing = GUI?) out of the main thread for IP8?

If you haven't done so already, take a look at:
DisplayHelpTopic "High-Resolution Displays"


Practically, there's not much we could do to move drawing outside of the main thread. Qt supports this to some extent but this would require changing an enormous amount of our code and probably isn't something we'll do in the near future.

Practically, there's not much we could do to move drawing outside of the main thread. Qt supports this to some extent but this would require changing an enormous amount of our code and probably isn't something we'll do in the near future.


I'm not the uber-expert with OpenGL but maybe adding a 2D plot mode based on gizmo would be easier than having multi threaded plotting code?
Qwt has an experimental QwtPlotGLCanvas, http://qwt.sourceforge.net/class_qwt_plot_g_l_canvas.html, which allows to use OpenGL for 2D plots.