Hi Team, Like the subject line says, I found an odd problem. I bought a new HP laptop with Core I5-8250 (newest 8th generation) quad core with 8GB and SSD a couple weeks ago to replace an old 2008 Dell Centrino Core 2 Duo which I have left hooked up for now. WSJT-X running on the new laptop will not decode reliably, or even at all, for long periods of time unless I run HDSDR at the same time. Tried multiple USB sound cards and the internal audio, same results. Old laptop which has the same audio feed from the radio decodes fine. Running HDSDR in parallel to WSJT-X on the new laptop lets the new laptop decode perfectly. This has been quite the head scratcher. With the steps I have taken below, I believe this could be a bug, or perhaps some obscure feature interaction of the newest CPU/GPU products (like the Intel Optane memory feature for example).
Please read on for setup details and perf observations. Also I have a 2nd station, same type radio (K3) with desktop core I-5 2nd gen, no GPU which works fine. Below is a long read because it is complicated with many variables which I hopefully have reduced. As a result of this behavior and investigation I have the following setup working: I have the new laptop (Win 10 Pro) paired with an Elecraft K3 operated remote. Audio goes from the K3 line out/in (internally transformer isolated) jacks to the laptop via the headphone/mic input jack and via a USB audio converter. The audio is also run to the old laptop mic in jack via the USB monitor out jack (so should be buffered and has separate volume control) as a backup and for performance comparisons. Splitting the audio cable between sound cards, or not, makes no change to this behavior. I have found that I need run HDSDR in parallel with WSJT-X in order to reliably decode some or any signals. I spent days chasing this down and found it by accident while trying to look at the audio for interference possibilities. I was also running N1MM+ which at first appeared to directly relate to this, but I think I have since proved it is not the cause but does seem to have some effect when not running HDSDR. I believe HDSDR (or possibly any similar graphic intense program perhaps) is acting as a GPU and/or CPU load and when loaded enough, decoding works. When working right, I crank out 30-40 decodes on a busy 40M band in a second or 2. Turn off HDSDR and decoding usually stops completely, or only decodes a few stations, though at program startup it will often go a cycle or 3 OK. Occasionally it starts working with a few decodes, then get full decodes for a short time during extended receive sessions then back to no decodes. Changing to Fast or Normal do not seem to affect it. Watching Task Manager CPU and GPU numbers, I get iffy decoding perf at low GPU numbers like <10%. GPU numbers 15-40% give the best results. The CPU is usually 8-15%, anything higher is rare and short when just operating WSJT-X. Running HDSDR on any audio source (stereo mix, mic in or USB line in) does not seem to matter. Hit stop, decodes basically stop. Hit Start, decodes resume full blast. When I minimize HDSDR (or hide the app window behind other screens) the GPU is 2-5% and decoding is poor or none. When I have HDSDR open as the top window, it is about 20-30% GPU and decoding is perfect. When the HDSDR window is hidden behind other screens it is about 3-5% decoding is poor but does seem to decode better then when minimized. Changing WSJT-X.exe priority to higher like real-time makes no change either (but does help the old slow laptop decode more stations). I thought maybe HDSDR was touching the audio buffer perhaps somehow preventing buffer overrun or corruption of some sort, but I can choose audio sources that WSJT-X is not connected to (in settings) and get the same behavior, so audio buffer corruption or loss do not seem likely, but maybe not ruled out. I have run 2 instances of WSJT-X on the new laptop, one on the mic in, the other on the USB line in, and see the decodes when working are nearly identical and near identical to the old laptop in parallel. Stop HDSDR, both instances on the new laptop stop decoding. The old laptop continues on fine. This is what leaves me to believe HDSDR is not impacting the audio buffers, but more likely a CPU/GPU load related behavior. Why I would have no clue. The audio levels are all set to be about the same, 30-40dB on the WSJT scale. HDSDR sees that at S9+30 on 40M with lots of activity. I have tried a range of audio levels and that does not change this behavior. I run all my machines via remote desktop client but I found the same results running locally also. I can improve the old laptop by turning off desktop background and other experience options to lighten the load, or increase priority, but these type of changes have no effect on the new laptop. Thanks for reading, Mike K7MDL
_______________________________________________ wsjt-devel mailing list wsjt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wsjt-devel
