Performance differences in Citrix HDX Thinwire Encoders

Performance differences in Citrix HDX Thinwire Encoders

2015-09-04
/ / /

Per my previous post, changing the Citrix HDX Thinwire Encoder on the fly, we can test the performance differences in the different encoder’s Citrix provides.  I have done so by running through a demo of the Uniengine Heaven benchmark.  The demo is exactly 4 minutes and 20 seconds long.  I did a perfmon trace of the CPU %, total bytes sent in MBits/sec and the Thinwire Output in MBit/sec.

Time for some results!

Compatibility Mode (Encoder 0x0)

DeepCompressionV2Encoder (Encoder 0x1)

DeepCompressionEncoder (Encoder 0x2)
(Rollover the mouse on the next images to compare graphs)

CompatibilityMode vs DeepCompressionV2Encoder

CompatibilityMode vs DeepCompressionEncoder

DeepCompressionV2Encoder vs DeepCompressionEncoder

The cumulative totals should help us get an understanding of the differences between the encoders:

   CPU Total ThinWire Total Network Total (Mbytes)
DeepCompressionEncoder 5531.00 3693.28 540.51
DeepCompressionV2Encoder 5621.67 3684.75 539.74
CompatibilityMode 4197.54 3690.58 553.21
   CPU Total ThinWire Total Network Total (Mbytes)
DeepCompressionEncoder 98.4% 100.0% 97.7%
DeepCompressionV2Encoder 100.0% 99.8% 97.6%
CompatibilityMode 74.7% 99.9% 100.0%
Interestingly, CompatibilityMode uses 25% less CPU then either DeepCompression Encoder.  From what I see though the frames per second appears less for CompatibilityMode then the other two.

2 Comments

  1. Rasmus Raun-Nielsen 2016-10-10 3:30 am

    Great article!

    Interesting graphs. Seems like DeepCompressionV2Encoder is behaving very much like DeepCompressionEncoder – must be because of the full screen being in motion. I would suspect a lower utilization of blended images (still- and moving images, for instance browsing) – do you agree?

    Reply
    • trententtye 2016-10-11 8:08 pm

      It’s difficult to say. DeepCompressionV2Encoder is an evolution on DeepCompressionEncoder and from what I’ve read it’s supposed to more CPU heavy (true) and use less bandwidth (also true) but neither were huge differentiating factors between the two codecs (the differences was less than 2% either way). I didn’t test anything other than full screen, full motion graphical 3D programs so it’s possible the advantages of V2 are understated in this test scenario. Why I chose this scenario; large screen redraws were required (client wanted 3D modeling over WAN) and though periods of idle time exist when modeling it was more important to have fast fluid refreshes so I chose to test with something that required it.

      Reply

Post a Comment

Your email address will not be published. Required fields are marked *

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.