Apache Labs Community Forum

Hi all,
I have had the spectral lights (the red and yellow ones on the top left hand corner of the screen) turned on for a while. They flash at me and I've been trying to optimise my setup to get the maximum screen update rate without getting regular flashing from them.
I run a moderately fast HP I 7 machine running win 10 pro. The overall pc loading doesn't exceed 25% with thetis running at 13~17%. I have the latest thetis and P2 soft and firmware running. I am running two screens with an effective pixel screen area of ~ 3500 X 1025 with a 1.5 MHz frequency coverage (2X 27inch curved LG monitors set at 60 hz).
My video card is a quatro 2000 . I tried using two of them with one screen on each card, but there was no change.
What I am finding is that my limits seem to be set by whats coming through from the 7000 mkII. I can run RX1 with with panadaptor at ~ 30 frames / second. Changing to panafall drops that to ~25 frames/second. Add RX2 and I'm down to ~ 20 frames/second.
If I move the display coverage back down to ~ one screen ~ 2000 X 1025 then things improve and I can get back up to ~ 30 frames/second running most things.
The real question for me is what is my bottleneck ? looking at the data rate coming from the 7000 it seem to stay at ~80 Mbytes/second (It is a 1gigabyte link). I'm not sure if I need to go for an even faster machine or I just don't have this machine setup correctly.

Regards Bob

There are a couple of things to note.

1) with ~3500 resolution in X, there will be around 3000 pixels in the spectral area

that is 3000 line segment draws for the panadapter
if filled, another 3000 line vertical line draws
if active peak another 3000 segment draws, or 3000 vertical (if filled)

That is per frame. So that could be ~9000 lines per frame, 60 frames a second, That is ~540,000 lines per second. If those lines are not equal to 1 pixel width, there is additional overhead.

So for two receivers, that would be over a million lines per second. That is a lot of lines !

2) If they are quadro 2000's and not a variant of the 2000 then they have a gpu passmark benchmark of 940, which is quite low, even comparing them to cards that are 8 years old. They are designed as workstation compute cards, and I am not sure how well they cope with this sort of thing. Are you able to try a different card from a friend/work?

3) Screen display width is always the issue for Thetis. Some performance might be gained by me splitting the calculations over a number of cores, and then combining the data to send to the gpu, but I am not sure a whole lot will be gained. When I profile Thetis at around 2500 pixels wide with a single RX and leave it running for about 30 seconds, we can see that ~20% of its total cpu time is spent calling directX2d DrawLine. We can go down the route of multi threading those DrawLine calls, say building the frame up in 4 thread, but man, so much work !
4) Waterfall is an insignificant overhead as it happens

5) there may be millage in adding a decimation or whatever the word is, so that we can for example, ask WDSP.GetPixels to return 1/4 of our screen real estate, and then step 4 everything as we draw it out, so we are iterating through 1/4 of the data and drawing 1/4 of the lines.

6) you could also switch on logical processor view on taskmanager, and check if one is maxed at at nearly 100%. Also use the cpu% view on thetis status bar, but switch into Thetis only mode by right clicking it and check what it is actually doing instead of entire system. You could also turn the GPU column on in taskmaster processes to see what is going on (or use the nvidia overlay if you have that stuff installed for the quadro).

Cheers,

Richie.

Thanks Richie,
Ok it looks like I am pushing my luck here, and not just fighting a windows setting. I will try to lay hands on a serious gaming video card and see how I go with that. But even if I'm out of luck there 20 frames /sec is still not bad. Having 2 27" curved screens showing the whole band is very very nice. I'm lusting after a single screen .....34" curved which will do the same. Until you have experienced it one has no idea. Again thanks.

Bob

I am adding that decimation stuff Bob, so may be a solution. It actually looks really good.

This is showing a decimation of 8. I have not yet finished the waterfall pixel increase so the image looks odd. Notice how the fill doesn't quite fill the outer line, this is due to the vertical lines now being 8 pixels wide. For those wanting high frame rates, the decimation can be increased.
and with a decimation of 2
R.

I am shocked that dual Quadros will not run the OP's desired configuration.

As a benchmark, I have an EVGA GeForce 1070 graphics card feeding triple Dell 30Hz, 4K monitors over three separate Displayport connections.

This configuration is by no means state of the art, but it has absolutely no problem keeping up with a 30Hz screen refresh over a 17+ million pixel area (rough estimate of spectral display area). I suspect it would have no problem even at 60Hz if I had faster 4K monitors. That's all of 80M at 768KHz sampling rate.

Sadly, due to cryptocurrency mining, even my decrepit old GeForce 1070 is going for $400 used on eBay.

Thanks Richie....in anticipation. I took my gauge of the pixels being used by the display in the bottom left.
I'm running double the frequency coverage of the 768kHz..... 1536 kHz.... I suspect that maybe where I'm running into trouble. I went there because I'm also using the 7000 with a 2 and 70 CM convertor to monitor the VHF and UHF bands and I wanted to be able to look at the repeater inputs at the same time as the outputs.
Richie by the way did you see my post on overdeviation on FM ? thats another of my tribulations.

Regards Bob

Hi with my sons help I found the HP monitor software and took this shot of RX! and RX2 on The video cards don't look to be the problem.
also looked at the network transfer rate RX1 only ~80Mbyte/second add RX2 it goes to ~150 Mbyte/sec.

Regards Bob

I'm running an i7-7700k. I pushed my RX1 and RX2 sample rates up to 1536KHz. Still no problem reaching a stable 30Hz update rate at ridiculous window sizes across triple 4K monitors. Using ResMon (built into Windows, just run it from the "run" box), Thetis shows approx. 150 Mbits/s (not bytes) receive, 11 Mbit/s transmit, for a total of 161 Mbit/s network utilization (rounded to nearest Mbit/s).

Something would seem to be bottlenecking your PC, which ought to be doing just fine.

I see you using CPU Affinity. I strongly recommend not assigning any "affinity" to Thetis, let it do it's own thing with respect to how it utilizes the CPU.

What CPU is in your PC, exactly?

Not quite sure these days. I was told I had a good machine......but in relation to what ?
But this is what I have.

Regards Bob

Compared to a i9 (Q4-2021) :

https://ark.intel.com/content/www/us/en ... 4599,64621

or choose another Intel cpu to compare to your Xeon.

regards, Leon

So it's an E5, not an i7. If you compare my i7-7700 to your E5, (just Google it) the E5 you have is not that much less powerful. It should not be a problem.