3

We are Rendering 16 live Streams at our client machine through one of our applications and the resolution of the video streams are as 4CIF/MPEG4/25FPS/4000Kbits.

The configuration of the client machine is below.

HP Desktop Machine:

Microsoft Windows XP Intel (R) Core2 Duo CPU E8400 @ 3.00 GHz 2.99 GHz, 1.94 GB of RAM Intel (R) Q45/Q43 Series Express Chipset (Inbuild)

The CPU usage of the machine peaks 99% for 16 streams.

After some discussion, we had decided to install external graphics card to reduce the CPU usage.

So that, we have tried following graphics cards.

  1. NVIDIA Quadro NVS 440 - 128 MB
  2. Radeon HD 4350 - 512 MB GDDR2
  3. Redeon HD 4350 - 1GB DDR2
  4. ASUS EAH 4350 Silent 1GB DDR2

But the performance wise there has been no difference - even a drop in performance.

So, what is the purpose of these external graphics cards?

Really it will reduce the CPU usage? What parameters have to check, if we want to reduce the CPU usage?

1
  • 4
    An external GPU will only reduce CPU load if the software is designed to utilise an external GPU in some way. If the software isn't designed as such, then it will never touch the GPU and there will be no improvement.
    – DMA57361
    Jan 13, 2011 at 9:11

3 Answers 3

8

The common purpose of graphics cards it to produce immediately visible graphics, whether it is rendered 3D games, producing a visual graphical user interface, or more lately decoding video to be viewed on a monitor.

If you are encoding 16 data streams to be sent to other machines to be rendered to screens by those machines then this is different to how the graphics cards you have are intended to be used but, with some extra software, is not beyond what they are capable of.

A decent graphics card, with some extra software written to make use of it, can encode video faster than a CPU can but the encoding software has to be written to make use of the GPU or nothing special will happen.

See http://www.nvidia.co.uk/object/IO_62559_uk.html for an example in Adobe Premiere

PowerDirector will use the graphics card to encode video for you

In short you can't just plug in a graphics card and expect your software to start using it, you need to use software that actually knows about GPU accelerated encoding in the first place.

What software packages are you using to get your video from wherever your source is to your endpoints?

-=EDIT=-

As a side note, all of the graphics cards you have listed are very low-end budget cards that are very limited in terms of processing power, granted two of the 4350s have 1GB of RAM, but in all honesty that amount of ram is so common on graphics cards these days that it's almost like saying "my computer has a processor."

Those cards may be able to play Full HD video, and possibly play a couple of 3d games with the effects turned down (or even off), but don't expect any ground-breaking performance boost in any games or for video encoding, those 4350's really are the bargain bucket line of video cards.

1

i have a dual processor pc thats using a 64 bit microsoft operating system. i bought the radeon hd6670 graphics card in the hope that i could watch movies in 1080p but the cpu usage is at 98+. after reading your interesting article i was wondering what software i should use to get my server pc (xeon intel 2 single core processors) to share the work load?

1
  • Welcome to Super User! Unfortunately, this post is not an answer. Please form your own question. Might not hurt to take a look at the FAQ (superuser.com/faq) and format your question a little better either.
    – rtf
    Oct 29, 2012 at 17:03
0

You have a low-end CPU that's trying to do heavy computational work. It's getting overloaded, so you add a low-end GPU, and you wonder why the system can't do heavy computational work. Well, it's a low-end system.

If the application you're running supports computing on the GPU, then adding a high-end GPU will make a massive difference. If it doesn't, adding a GPU will make no difference at all. GPUs only accelerate graphics automatically, not computation.

Most likely, the best solution is to upgrade the CPU. For example, a Q9450 will roughly triple your CPU's computing ability.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .