# of cores and hyperthreading

[amagras]

I wanted to update (and inevitably resurrect) this post with information because it may help others in the future.

I've been comparing PT behaviour with and without hyperthreading resulting in: PT works in my PC (listed on the signature) with HT and #CPU-2 better than with no HT and #CPU-1, this is 8CPU-2 in the first case against 4CPU-1.

Thanks to all of you for chime in!

[Darryl Ramm]

It's great to share tips, but comments about CPU scaling are kind of meaningless without knowing what plugins and what workload they are running.

And in some cases you may need to know the versions of plugins.

And even more fundamental, what are you actually meaning when you say "works better" -- is that lower indicated CPU use (measured how)?, a higher throughput number (like done for DVERB tests?), or ability to run certain workload with fewer DAE/AAE CPU errors?

What may be interesting is to develop DVERB like workloads of other plugins.

[amagras]

Good idea! You and me could start by picking two plugs and maybe others will do the same until we have an extensive database. Anyways I'm still not very clear on how to make the Dverb test but I'm on it.

[Darryl Ramm]

Originally Posted by amagras

Good idea! You and me could start by picking two plugs and maybe others will do the same until we have an extensive database. Anyways I'm still not very clear on how to make the Dverb test but I'm on it.

There is lots of discussion on DVerb on DUC, especially around testing Win PC builds.

I'm not going to have time to help build stuff, but the workload that would interest me is VI workload (especially Superior Drummer) with a mixed general EQ/reverb workload, throw in some Eleven etc.. And that is even more interesting to things about because it's not just adding more tracks and throughput that is interesting. Its more likely having some fixed fairly heavy Vi workload and then adding processing plugins to that and other audio tracks. Folks doing full on composition work, not just laying down rhythm tracks, might have hugely complex multi-instrumental VI workloads. I wonder what exists there already for test/demo workloads (from vendors or others).

I've also thought it would be fun in the past to build some sessions that did nothing but demo/test for plugin denormalization bugs. That would be fun...

[amagras]

I must have added "if you have time" of course. You can count on me if you happen to have time for making some comparisons, as I said I'm on it, just lead the way.

[Amack]

amagras,

I have a XPS 8700 too (with 24 GB, msata SSD, and Windows 8.1 though). I have also gone thru all the optimizations and done some testing with different processor Bios settings. Nothing seems to have much of an effect on performance. Nothing seems to approach the kind of performance guitardom claims for similar testing on one of his old computers running HD. (He seems unwilling to test it on "vanilla".) However, I didn't notice any difference in using his Dverb approach vs. mine with a "Tone Generator" - (set to pink noise so its easier to listen to for extended times). You might want to see if you can get better results than I've been able to. In my experience it's usually easiest and best to minimize the number of variables (i.e. plugins) when trying to isolate/solve problems. I (and likely many others) would really appreciate any feedback you have on the process and/or your results if you decide to pursue that approach. See http://duc.avid.com/showthread.php?t=370379 for details.

Amack

Originally Posted by amagras

I must have added "if you have time" of course. You can count on me if you happen to have time for making some comparisons, as I said I'm on it, just lead the way.

[amagras]

Actually the difference was very small in my PC too. Thank you Amack, I'll take a look. I haven't had time to research on how to do the tests. Maybe it's in that thread?

[Amack]

Yup

Originally Posted by amagras

Actually the difference was very small in my PC too. Thank you Amack, I'll take a look. I haven't had time to research on how to do the tests. Maybe it's in that thread?

[guitardom]

Originally Posted by Amack

amagras,

I have a XPS 8700 too (with 24 GB, msata SSD, and Windows 8.1 though). I have also gone thru all the optimizations and done some testing with different processor Bios settings. Nothing seems to have much of an effect on performance. Nothing seems to approach the kind of performance guitardom claims for similar testing on one of his old computers running HD. (He seems unwilling to test it on "vanilla".) However, I didn't notice any difference in using his Dverb approach vs. mine with a "Tone Generator" - (set to pink noise so its easier to listen to for extended times). You might want to see if you can get better results than I've been able to. In my experience it's usually easiest and best to minimize the number of variables (i.e. plugins) when trying to isolate/solve problems. I (and likely many others) would really appreciate any feedback you have on the process and/or your results if you decide to pursue that approach. See http://duc.avid.com/showthread.php?t=370379 for details.

Amack

The 8700 is a cheap desktop and has I believe the slowest 4th gen i7 you can get. It's not exactly going to be a powerhouse. You have also never said if on latency mon if you have any drivers taking over 1ms, if you have the c states and eist disabled in the bios among a number of other things. But expecting a 4790 to have high end performance is stretching it, especially at high sample rates where you seem to have some obsession. Nobody is able to make albums at 96k (or 88.2 or 192) with that processor. A lot of us couldn't do it with that processor at 48k. I know a 6 core is the smallest I can work on successfully due to plugin needs.

Also for the 40th time. There is no difference with HD software! Especially not now that most all the features are in the regular version including disk cache which is the ONLY feature HD contained that could give it better performance in a couple scenarios. It's down to some automation features and surround being unlocked along with it being unlocked to run the higher end hardware.

The dverb test and your tone generator thing are completely different and have no comparison to each other. The tone generator works or it don't. That simple. The dverb test is testing CPU, realtime processor stability, and disk I/O. It also gives users a relative way to compare systems and typically determine if someone has issues or not in there performance. The tone generator tests real time stability only. Once it works, your done. It's terribly limited in scope.

[amagras]

Originally Posted by guitardom

The 8700 is a cheap desktop and has I believe the slowest 4th gen i7 you can get. It's not exactly going to be a powerhouse. You have also never said if on latency mon if you have any drivers taking over 1ms, if you have the c states and eist disabled in the bios among a number of other things. But expecting a 4790 to have high end performance is stretching it, especially at high sample rates where you seem to have some obsession. Nobody is able to make albums at 96k (or 88.2 or 192) with that processor. A lot of us couldn't do it with that processor at 48k. I know a 6 core is the smallest I can work on successfully due to plugin needs.

Also for the 40th time. There is no difference with HD software! Especially not now that most all the features are in the regular version including disk cache which is the ONLY feature HD contained that could give it better performance in a couple scenarios. It's down to some automation features and surround being unlocked along with it being unlocked to run the higher end hardware.

The dverb test and your tone generator thing are completely different and have no comparison to each other. The tone generator works or it don't. That simple. The dverb test is testing CPU, realtime processor stability, and disk I/O. It also gives users a relative way to compare systems and typically determine if someone has issues or not in there performance. The tone generator tests real time stability only. Once it works, your done. It's terribly limited in scope.

My computer wasn't cheap to me. I don't know that you and Amack have going on but I find your post a little angry. Peace