post #1 of 1
Thread Starter 
Hello guys.

Some time ago i found a product called Bigfoot Killer Xeno Pro.
(If you are not interested how I came to the idea simply scroll to "The experiment" part.)

LInk to review:

This network card and its predecessors were promising better gaming. The card itself was a commercial flop with overhyped marketing which promised "less lag", and most review portals panned this card to be a hoax.

Reading most of the reviews, and looking at the actual management software I realized that reviewers were checking "common" parameters as network ping from gaming machine to some gaming service, or comparing "display to display" online games on Xeno and other cards. Usually the results were same. Onboard network cards had bad results, but any dedicated Network card was on par with Killer Xeno but at lower price.

Really very few (if any) reviewer mentioned that Killer Xeno is providing a custom QOS packet scheduler. EG you could start a download, while playing an online game, and the software assured that your game was getting higher priority than the download, so your game experience was more optimized compared to other network cards.

A) Windows QOS vs Killer Xeno QOS
Main feature which this card provides is a custom QOS packet prioritization, where online game (identified by DirectX dependency) gets high priority, while other processes get "Default" priority. Each single reviewer kept Windows QOS enabled, thus network communication has passed through 2 different QOS packet schedulers and results were really not telling how the card is working.

B) "Better latency" hoax
For this is definitely responsible directly Bigfoot networks corporation. Early trailers for this card its predecessors contained such information.

So ... In the end I had a product where nor reviewers were able to measure real impact, and the developer himself had provided really NO methodology if, and how this card can improve gaming. Then I realized that "Actions per minute" is a way which players measure how good are they at ... Starcraft II, but how to measure this in environment of a online game?


bulb.gifThe Experiment

I was playing Ultima online for a while (a fan project). I found a spot with a door, and a zone which does not allow me to enter. After communication with the GMs I set up a macro which:

a) Opened the door (they remained opened for exactly 30 seconds)
b) Moved me towards the zone.
c) Entering the zone created a server-generated message (if it was repeating too often it was not spamming whole screen, rather adding a number at the end), and ported me 1 field back.
d) After 30 seconds the door closed and I had in my log specific number of "entrance attempts"


bulb.gifThe results

Onboard Network: 600 APM (10 actions every second)
Killer Xeno: 660 APM (11 actions every second)

Both network devices were providing same Ping from my computer to gaming server, both are average results from 5 measures.

Later I repeated the test with a dedicated Intel Pro 1000 Nic card, and it was 660 apm again.


The outcome
Method I use could be considered to be a spamming the game server, and I did so on low populated server, and with agreement of GMs. In the end I have only results which are true for this specific game, and specific to my location, yet can be important as well.

Usually the "quality of communication" is measured by bandwidth (for download/upload purposes) or latency (yes for gaming). Lets looks at example values:

Upload: 0,25 mbit/s
Download 4,62 mbit/s
Ping (UDP): 60 ms
Ping (TCP): 120 ms

Which values really arent speaking much about online game experience, like the ping can be better. What you dont see that ping (UPD) may be consistent, or very inconsistend and from time to time get over 250 ms. Its simply because those values are : "momentary", measured during quite short timeframe. But add values like:

APM: 600 (actions per minute)
APS: 10 (actions per seconds)
UDP Latency range: 30-45 ms

And we have something that say much more about communication quality specifically for online gaming.


Is there anyone able to code a benchmark like that? smile.gif I dont believe it should not be that hard (i have not passed over the phase using macroing and scripts, rather than real programming language).