Playback Speed in windows 7

Post by **robertisaar** » Wed May 05, 2010 6:33 pm

wondering what anyone else running win7 is getting for max playback speeds? i believe i am running the same TP version as when i was running vista, but instead of the ~400 samples/sec i was able to get with an item list consisting of nothing selected, i can't seem to top ~65 samples/sec with the same list. i set the speed multipliers to anywhere from 100 - 10,000, but same speed. CPU usage peaks at about 10% during this process...

Post by **Mangus** » Fri Jun 25, 2010 4:45 pm

This doesn't really belong in bugs, so moving to discussion.

This would be very PC dependent, both in terms of CPU and disk speed (logs are opened as memory mapped files and streamed into memory). I'm able to get ~600hz for some applications, but I'm on a pretty beefy machine.

Post by **robertisaar** » Fri Jun 25, 2010 5:56 pm

i should have updated this a bit when i came across it:

TP SEEMS to only limit speed in certain tabs... if it's the capabilities of my laptop, it would make sense, but it's not that old of a machine...

what i've done was make a "item list" with nothing in it. so i just select that, set the speed multiplier to some number above 1000, and the log BLAZES through at roughly 800-900 Hz when my multiplier is set to 9999 with a full A1 ADX. with the full item list, it's ~100-120, histograms are ~50, a monitor screen with nothing in it is ~45-50, and a full dash is ~45...

and i rechecked processor usage at full speed... now it roughly maxes out core 1, with a little activity on core 0, where before, they averaged 10%...

not sure what to think of that, other than it does look like my processor is the limiting factor here...

i'll have to try and load everything onto my wife's quad-core 8GB beast and see what kind of numbers i can crank out... even though it's likely that 3 of the cores wil be useless for it, i'm sure the 1 Phenom X4 core is a bit tougher than 1 Pentium D core.

Post by **Mangus** » Fri Jun 25, 2010 6:47 pm

TP isn't multithreaded in a way that multi-core machines will show benefit with, at least not really. Data is handled in a separate thread from UI, but in some cases they're synchronized becuase, if they weren't, the UI wouldn't keep up with the data thread and packets would be missed.

You'd notice similar behavior in previous operating systems, although perhaps higher numbers (as OS architecture and hardware targeting changes).

Post by **robertisaar** » Fri Jun 25, 2010 6:52 pm

that would explain the slight activity on 1 core and almost maxing another...

i assumed it wasn't multi-threaded either... very few things are yet.

i imagine it's a pain to code as well...

Post by **Mangus** » Fri Jun 25, 2010 6:57 pm

robertisaar wrote: i assumed it wasn't multi-threaded either... very few things are yet.

It is indeed multi-threaded. Not all applications do work that lends itself well to multi-threading. TunerPro, unfortunately, is one of those applications. Work is spread across threads (and, though Windows, cores) where it makes sense.

In the case of real-time data I/O, the first packet is read and passed to the UI thread. While the UI thread is rendering, the data thread is retrieving and processing (doing claculations on) the next packet. If the UI thread isn't don't yet, the data thread waits for it to finish before passing the new data to the UI thread. Whichever thread is done first is the bottleneck.

Sometimes disk I/O is the bottleneck. Other times graphics capabilities are the I/O (high-end 2D graphics processors, like those designed for workstations, do well with the GDI drawing code as an example).

TunerPro isn't all that hefty of an application. And there's always room for optimization.