Pete, thanks for the comments.
On my AMD 945 (3750 passmark), I could see about 350 decoders max.
On two bands at 96 Khz, there were around 600 signals. If that's
representative, it's fair to assume that the max load is during the
daytime - with 3 active bands (10/15/20). Assuming 300 signals per band,
and adding a bit of pad, to call it 1K signals total.
Assuming that passmark score/decoder was predictive in my case, I
extrapolated that it would take about something with about 11K passmark to
keep up (for the most part) skimming 5 bands @ 96 Khz on a busy RTTY
contest.
It's interesting that your much more capable rig (2-3x higher passmark
score) was running only 25% higher decoder count. I've seen data that the
AMD cpu do not provide as good of a performance as the intel for unknown
reasons with this application (I assume it's due to some CPU or code
optimization specific to the compiler or libraries). However, as a separate
data point, Wes told me his current setup (i7 hasselwel category CPU) was
almost able to keep up in a 5x96 configuration.
Lot of assumptions and variables - so I don't have a lot of confidence in my
conclusions.
But I do think this means the best of the current generation would be up to
the task. But only just so...
73/jeff/ac0c
www.ac0c.com
alpha-charlie-zero-charlie
-----Original Message-----
From: Pete Smith N4ZR
Sent: Wednesday, February 18, 2015 5:44 AM
To: Jeff AC0C
Cc: rbn-ops@yahoogroups.com ; RTTY Contesting ; skimmer@dxwatch.com
Subject: Re: [RTTY] RTTY Skimmer Server computer hardware needed to handle
WPX without throttling
Hi Jeff. For whatever it is worth, here's what I did this weekend. I'm
running an AMD FX8350 (passmark 9000+) with 8 GB of RAM.
1. turned off CW skimming for the weekend.
2. Set up low band and high band rotation at sunrise and sunset,
running 4 bands at a time
3. Limited coverage to 96 KHz
4. Tailored my RTTY segments, starting with the "Contest" default
setting, to make sure I wasn't listening where the RTTY wasn't. The most
important change was to define two RTTY segments for 40M, 7040-7050 and
7080-7120. I think I probably missed some Region 2 and 3 activity in
the gap, but also got rid of a lot of that pesky CW stuff.
The best I was able to get was 440 decoders. All 8 cores were engaged at
the same levels. Whenever the CPU usage would spike over 70 percent,
throttling would occur. The number of decoders would drop fairly
sharply (down to 350 or so), then rebound to at or near the critical level.
A preliminary look at the RBN traffic for the weekend suggests about a
quarter of the CQWWCW daily totals(all modes). I made about 29,000
spots per day, and I don't hear very well - certainly as compared to WZ7I.
73, Pete N4ZR
Check out the Reverse Beacon Network at
http://reversebeacon.net,
blog at reversebeacon.blogspot.com.
For spots, please go to your favorite
ARC V6 or VE7CC DX cluster node.
On 2/17/2015 6:50 PM, Jeff AC0C wrote:
I ran Alex’s RTTY Skimmer Server over the WPX contest weekend and had a
tough time decoding multiple bands without throttling, even though the PC
it was running on is pretty stout.
If you ran RTTY Skimmer Server over WPX and managed to do so on multiple
bands without throttling, I would be very interested to know what your
computer configuration was. I need to put together a new box for this
task and have no idea how to size the hardware.
73/jeff/ac0c
www.ac0c.com
alpha-charlie-zero-charlie
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|