Here's another little tidbit to put cores in practical terms: I sometimes run
a torture test called Prime 95. which uses FFT calculations to push the AMD CPU
to 100% utilization on all four cores. Even with all four cores heating up at
100%, I can still run PSDR 2.01.16 with no audio dropouts, clicks, or rasps.
And this is with PSDR set to medium frame rate (25 FPS), with NR engaged, and
at 192k. Oh yeah- HRD log and HRD DM780 are also running on a second HD
monitor. And I'm composing this email. I can, however, cook an egg on my CPU
at 54C unless all the fans kick to high gear.
On the other hand, if I start just one program with a high DPC driver, PSDR
freaks out, gives me quivering panadapter trace, and won't transmit until I
restart.
Jim
> From: pete...@osr.com
> To: telli...@itsco.com; vk...@bigpond.com; jsqu...@msn.com;
> flexradio@flex-radio.biz
> Date: Wed, 1 Dec 2010 10:51:12 -0500
> Subject: RE: [Flexradio] Maximized Window & CPU Use
>
> I agree with all of this, it's very well thought-out and said. Very well
> indeed. The only thing on which I beg to differ is the claim that multiple
> cores add to context switching overhead. Actually, they don't... Windows is
> VERY smart about context switching now, and takes into account some very
> sophisticated parameters about the system's physical configuration: NUMA,
> Hyper-Thread, MultiCore, etc.
>
> The one thing I'd add is that it's VERY important to differentiate between
> Hyper-Thread systems and MultiCore systems. They are VERY different animals.
> A system with 2 physical processors (sockets) where each of those two
> processors is capable of 2 HyperThreads will be significantly inferior in
> performance to a system with 2 physical processors (sockets) where each of
> those processors has 2 MultiCores. So, in a word, MultiCore GOOD, HyperThread
> NOT AS GOOD.
>
> Finally, while Tim didn't allude to this in his excellent summary it's "all
> the rage" on the intertubes... in terms of speed of memory access and NUMA
> architectures: Whether a system has one shared memory bus or one memory bus
> per physical CPU isn't particularly important for running PowerSDR. The NUMA
> architecture machines will have less memory contention every time, and that's
> fine, but it's not an important factor in choosing a system for PowerSDR.
>
> If the above is too technical for some readers, the best (admittedly biased)
> simple summary I can provide is "at this point, for the current generation of
> chipsets, systems using AMD chipsets tend to be better than systems using
> Intel chipsets."
>
> P
>
> -----Original Message-----
> From: flexradio-boun...@flex-radio.biz
> [mailto:flexradio-boun...@flex-radio.biz] On Behalf Of Tim Ellison
> Sent: Wednesday, December 01, 2010 8:51 AM
> To: David Walker; jsqu...@msn.com; flexradio@flex-radio.biz
> Subject: Re: [Flexradio] Maximized Window & CPU Use
>
> First off, I'll add my disclaimer that anything I say here may be trumped by
> Neal and you should take what he says as the absolute truth. He lives in this
> trench daily. I stand on the outside and look in.
>
> A CPU /mobo upgrade will make a big difference if you take a big enough leap.
>
> For as long as I can remember, I have used exclusively Asus mobos. I have
> been very pleased with both the Intel and AMD versions.
>
> In my last "technology upgrade cycle", I replaced just about all of my PCs
> with AMD Phenom CPUs. I personally think that they have lower "ambient" DPC
> latency and faster PCI-E bus I/O. I have no hard data to prove this point.
> Intel seem to do this hokey thing with IRQ stacking that makes tweaking
> peripheral bus performance difficult.
>
> PowerSDR is a mathematical application. Lots of floating point math. The most
> important aspect of designing a real-time audio processing application is
> getting data off of the peripheral bus, be it Firewire, USB or gigabit
> Ethernet, through the memory controller and to a CPU that has sufficient
> L1/2/3 cache to quickly pass the data into the CPU for number crunching. Add
> multiple cores to the mix and you have context switching to deal with that
> can be a hindrance, but not as much now as say 3-4 years ago due to updated
> Windows operating systems.
>
> So I look at CPU cache, the speed of the front side bus and the speed of the
> RAM when choosing a CPU/mobo. One reason I like the AMDs is the memory
> controller is on the CPU and that speeds things up.
>
> I also never get a motherboard with on-board graphics. They are convenient,
> but can be slow and steal RAM for buffering.
>
> If you are looking for a tried and true bill of materials for building a high
> performing PC for SDR applications, get one from Neal. He is the man in this
> regard and is current with the latest "toys".
>
>
> -Tim
>
>
> -----Original Message-----
> From: David Walker [mailto:vk...@bigpond.com]
> Sent: Wednesday, December 01, 2010 7:56 AM
> To: Tim Ellison; jsqu...@msn.com
> Subject: Re: [Flexradio] Maximized Window & CPU Use
>
> okay Tim no worries. I understood not a word of that :-)
>
> Do you think a motherboard and CPU upgrade would make any difference as
> Jim suggested? They radio works fine with my application screening or
> minimisation methodology but I guess I am keen to get a the system working as
> well as I can. Plus I've never been happy with this Gigabyte board I have the
> bios features are fake.
>
> I can upgrade to a i5 650 CPU and an Asus P7H55-MUSB3 motherboard quite
> easily. Would that be sufficient or would you recommend a Quad Core AMD
> system?
>
> cheers,
>
> Dave VK2NA
>
> On 01/12/10 23:28, Tim Ellison wrote:
> > This behavior has been around since the PowerSDR 1.x days. I suspect it has
> > something to do with the GDI+, the class-based graphics API for C/C++
> > applications in VS.
> >
> >
> > -Tim
> >
> >
> > -----Original Message-----
> > From: David Walker [mailto:vk...@bigpond.com]
> > Sent: Wednesday, December 01, 2010 1:35 AM
> > To: Tim Ellison
> > Cc: flexradio@flex-radio.biz
> > Subject: Re: [Flexradio] Maximized Window& CPU Use
> >
> > Hi Tim,
> >
> > When I run PowerSDR minimised or have say a web browser cover up the
> > application I notice I get much less CPU loading which is great for CW and
> > especially with N1MM contest logger running. I get no lags or crackles in
> > audios etc..
> >
> > This is why I am wondering about the graphics performance of PC with
> > PowerSDR. I notice that if I cover the band buttons with an application
> > say, Notebook the latency and CPU use is reduced, so the application
> > switching button colours in TX mode slows my system down.
> >
> > I'm just experimenting here trying to get setup for the next contest.
> >
> > Dave.
> >
> >
> > On 1/12/2010 12:02 PM, Tim Ellison wrote:
> >
> >> "Would getting a better graphics improve the situation?"
> >>
> >> Probably not. The rendering of the PowerSDR Panadapter does not use and
> >> graphics hardware acceleration. Most GPUs are optimized for 3D graphics,
> >> not 2D rendering.
> >>
> >> I would make sure you have the latest Win7 drivers installed.
> >>
> >>
> >> -Tim
> >>
> >>
> >> -----Original Message-----
> >> From: flexradio-boun...@flex-radio.biz
> >> [mailto:flexradio-boun...@flex-radio.biz] On Behalf Of David Walker
> >> Sent: Tuesday, November 30, 2010 5:38 PM
> >> To: flexradio@flex-radio.biz
> >> Subject: [Flexradio] Maximised Window& CPU Use
> >>
> >> Hi,
> >>
> >> I tend to use the smallest PSDR window size possible to minimise my CPU
> >> load which is normally between 8 and 20%. On the odd occasion I wish to
> >> maximise the screen the CPU load increases to around 20% to 48%.
> >>
> >> I only have an ATI Raedon X300 PCI-E graphics card with 520MB onboard RAM;
> >> latest drivers of course.
> >>
> >> Would getting a better graphics improve the situation?
> >>
> >> Cheers,
> >>
> >> Dave
> >>
> >> _______________________________________________
> >> FlexRadio Systems Mailing List
> >> FlexRadio@flex-radio.biz
> >> http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
> >> Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/
> >> Knowledge Base: http://kc.flexradio.com/ Homepage:
> >> http://www.flexradio.com/
> >>
> >>
> >
> >
>
>
> _______________________________________________
> FlexRadio Systems Mailing List
> FlexRadio@flex-radio.biz
> http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
> Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/
> Knowledge Base: http://kc.flexradio.com/ Homepage: http://www.flexradio.com/
_______________________________________________
FlexRadio Systems Mailing List
FlexRadio@flex-radio.biz
http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/
Knowledge Base: http://kc.flexradio.com/ Homepage: http://www.flexradio.com/
