<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
	<title type="html"><![CDATA[Modartt user forum - Performance Index - large fluctuations warning]]></title>
	<link rel="self" href="https://forum.modartt.com/extern.php?action=feed&amp;tid=7938&amp;type=atom"/>
	<updated>2020-11-30T16:43:25Z</updated>
	<generator>PunBB</generator>
	<id>https://forum.modartt.com/viewtopic.php?id=7938</id>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971746#p971746"/>
			<content type="html"><![CDATA[<p>Thanks Ethanay!</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-30T16:43:25Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971746#p971746</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971711#p971711"/>
			<content type="html"><![CDATA[<div class="quotebox"><cite>BTMASS wrote:</cite><blockquote><p>Thanks Stephen.&nbsp; I originally did this when I got my Surface Pro back in June.&nbsp; It&#039;s a useful hack but IMO needs to be used with some caution in that forcing 100% processor usage all the time takes its toll on the hardware.&nbsp; In the i5 SurfacePro&#039;s case the little guy gets very hot.&nbsp; It has no fan, so is totally passively cooled.</p><p>As I say I think I&#039;ll experiment with seeing if I can lock in the CPU to a lower amount.&nbsp; Maybe I can find a happy medium between acceptable performance and heat.</p></blockquote></div><p>Regardless of the (higher) frequency you specify, the laptop may quickly downthrottle to a lower or even minimum frequency under sustained loads for thermal reasons. I would run a CPU benchmark test for 10min or so and see what the &quot;steady sustainable&quot; frequency is that the device can maintain under load and use that as the target frequency.</p><p>For me, when i was using a laptop, i just set it to max frequency (3.9ghz) and that translated to 2.2ghz sustained speed. so that is the understanding that i used to tune pianoteq performance for maximum possible realtime quality while avoiding xruns</p>]]></content>
			<author>
				<name><![CDATA[ethanay]]></name>
				<uri>https://forum.modartt.com/profile.php?id=497</uri>
			</author>
			<updated>2020-11-29T18:25:26Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971711#p971711</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971691#p971691"/>
			<content type="html"><![CDATA[<p>Thanks Stephen.&nbsp; I originally did this when I got my Surface Pro back in June.&nbsp; It&#039;s a useful hack but IMO needs to be used with some caution in that forcing 100% processor usage all the time takes its toll on the hardware.&nbsp; In the i5 SurfacePro&#039;s case the little guy gets very hot.&nbsp; It has no fan, so is totally passively cooled.</p><p>As I say I think I&#039;ll experiment with seeing if I can lock in the CPU to a lower amount.&nbsp; Maybe I can find a happy medium between acceptable performance and heat.</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-29T12:04:34Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971691#p971691</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971675#p971675"/>
			<content type="html"><![CDATA[<div class="quotebox"><cite>BTMASS wrote:</cite><blockquote><p>I can change the power profile on the laptop to minimum 5% and maximum 5% processor power and that locks the frequency to 1400 MHz.&nbsp; This gives me a steady CPI of 65.&nbsp; This works for a bit and then rather annoyingly the power profile automatically changes itself.&nbsp; My current take will be to return to the Surface Pro 7 i5 ...</p></blockquote></div><p>This webpage might have some relevant information--</p><p><a href="https://www.minitool.com/news/cpu-throttling-issue-how-to-fix.html">https://www.minitool.com/news/cpu-throt...o-fix.html</a></p><p>--</p>]]></content>
			<author>
				<name><![CDATA[Stephen_Doonan]]></name>
				<uri>https://forum.modartt.com/profile.php?id=4838</uri>
			</author>
			<updated>2020-11-28T15:01:58Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971675#p971675</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971674#p971674"/>
			<content type="html"><![CDATA[<p>This has been an expensive (time wise) but elucidating experience for me.&nbsp; I come from the viewpoint that more computing power always means better performance and that&#039;s clearly not the case with Pianoteq (and indeed perhaps other music creation software - I&#039;m a newbie at all this frankly).&nbsp; </p><p>I can change the power profile on the laptop to minimum 5% and maximum 5% processor power and that locks the frequency to 1400 MHz.&nbsp; This gives me a steady CPI of 65.&nbsp; This works for a bit and then rather annoyingly the power profile automatically changes itself.&nbsp; My current take will be to return to the Surface Pro 7 i5 and see if I can make that work nicely - I only really replaced it as I thought it was having an issue due to lack of processor power and that does not seem to be the case at all.</p><p>As I say, very educational for me!</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-28T14:56:27Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971674#p971674</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971607#p971607"/>
			<content type="html"><![CDATA[<p>Many thanks guys!&nbsp; Yes, much food for thought.&nbsp; Ulp....</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-26T22:44:10Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971607#p971607</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971604#p971604"/>
			<content type="html"><![CDATA[<p>yes, this issue on the Surface sounds to me like a combination of CPU frequency scaling enabled and also passive cooling.</p><p>The issue on the new laptop is common with frequency scaling alone. As I said, modern CPUs to my knowledge just can&#039;t ramp up their speed fast enough for low latency applications, and the overhead creates a delay that causes buffer overruns.</p><p>Re: the CPI, I mean in terms of setting your performance parameters, if you you set Pianoteq&#039;s quality settings (sampling rate, polyphony, # of microphones, etc) for real-time use so that they do not create the risk of Pianoteq&#039;s processing demand exceeding the minimum indicated on the CPI, you should be safe. This is only relevant for real-time rendering (e.g., while you are playing; not for rendering pre-recorded MIDI files into audio, which can happen outside of real-time and so there is no risk of buffer issues, so you should almost ALWAYS use highest quality settings unless you are going for a specific effect as a mixing engineer or producer...)</p><p>If you want to stabilize the processor speed (which is a good thing for real-time or low latency performance because it eliminates the overhead of the CPU takign extra cycles to scale up and down and allows it to devote those cycles to processing user input as well as giving you a more predictable and reliable low latency performance level), you need to lock the frequency at a certain speed. For quiet operation (if fan cooled) and/or battery life, choose a lower frequency. For max performance, choose a higher frequency. Can&#039;t have it all...it&#039;s a tradeoff.</p><p>I am not sure how to accomplish this on Windows 10 but I bet even if there aren&#039;t built-in tools to do this sort of performance tuning (e.g., custom power profiles that allow you to tweak the CPU performance settings) there are probably tons of 3rd party (and free) programs or apps that will help you do it. You should be able to do it dynamically in user space (e.g., after bootup and login). Any app or widget that gives you fine control over ideally both CPU min/max frequency and multi-threading (which is ideal to disable in low latency work) should suffice.</p><p>I agree with Stephen, laptops very commonly depend on cpu frequency scaling to keep the laptop cool and save battery life because the common user does not notice the milliseconds of latency when browsing/typing/listening/watching that matter to those of us using laptops for real-time or low latency processing work.</p><p>They also commonly depend on thermal throttling (which is thermally-reactive downscaling of CPU frequency) to prevent overheating. What this means is: don&#039;t expect the CPU to perform at or even near it&#039;s &quot;max frequency&quot; setting. For example, I had an i5 quad core ultrabook that was actively cooled via both fan and heatpipe (so, relatively very effective cooling for a small, thin laptop) that had a max frequency of 3.9ghz,&nbsp; but it could only sustain ~2.2ghz for more than a few seconds at a time. Still, that locked the CPI at about 125 for me.</p><p>Part of this is another engineering tradeoff: a powerful processor running at full-tilt needs excellent cooling to run sustainably at those higher speeds. Modern thin and light laptops just don&#039;t have the space and airflow to dissipate enough heat to do that. So those &quot;high spec&quot; ultrabooks look good on paper, until you need them to use that high performance over an extended period of time (e.g., transcoding or compiling), then you find that they perform significantly more slowly than similarly-spec&#039;d well-cooled desktop computers.</p><p>That should give you plenty to google (or duckduckgo if you prefer) to figure out how you need to configure your system <i class="far fa-smile smiley"></i></p>]]></content>
			<author>
				<name><![CDATA[ethanay]]></name>
				<uri>https://forum.modartt.com/profile.php?id=497</uri>
			</author>
			<updated>2020-11-26T21:59:39Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971604#p971604</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971599#p971599"/>
			<content type="html"><![CDATA[<div class="quotebox"><cite>BTMASS wrote:</cite><blockquote><p>... and the processor running between 1400 and 2900 MHz</p></blockquote></div><p>Frequency scaling. That&#039;s something that is often turned off or on in the computer&#039;s BIOS firmware (called &quot;Intel Speedstep&quot; in the BIOS of my desktop comptuter). In Linux, frequency-scaling can also often be disabled using a kernel parameter specified manually or automatically at bootup, and in some cases the CPU cores&#039; frequencies can be set to a fixed value using a userspace utility app after bootup.</p><p>I suppose laptops often depend more than desktop computers on CPU frequency scaling, for the sake of keeping cool and to conserve the battery, but frequency-scaling does often affect real-time audio use.</p><p>You might try an internet search using &quot;frequency scaling&quot; and your tablet or laptop brand and model, to see if you can turn it off, at least temporarily, and see if that helps.</p>]]></content>
			<author>
				<name><![CDATA[Stephen_Doonan]]></name>
				<uri>https://forum.modartt.com/profile.php?id=4838</uri>
			</author>
			<updated>2020-11-26T20:44:39Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971599#p971599</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971593#p971593"/>
			<content type="html"><![CDATA[<p>Hi ethanay - many thanks for your input and you are of course quite right in your comments.&nbsp; It doesn&#039;t make any sense and that is because it’s not true… How stupid was I thinking that fix made any sense and of course the large fluctuations error came back again on the Surface Pro running in this lesser performance mode as well. I’ve actually tried most of the power profile options now, including the Molten Music profile I set up which really does make the PC work on maximum performance all the time. The problem I have encountered, however, with the Surface Pro that I have (it’s an Intel i5 model) is that it’s passively cooled with no fan and tends to get pretty toasty sitting there at full tilt all the time.</p><p>However, I have never been one to let common sense get in the way and so I purloined another laptop, this one with a fan and a relatively meaty processor (at least as far as mobile computing is concerned). I am getting exactly the same issue with the “warning large fluctuations” reminder. Currently this PC is telling me the performance index is fluctuating between 72…194 and the processor running between 1400 and 2900 MHz.&nbsp; I’ve tried toggling the multicore rendering tick box on and off but that doesn’t seem to make much of a difference at all.</p><p>I’m currently running on 48000/48000/128 with auto (optimistic) and I think I can detect the very occasional little crackle. I know that many people on this board get crackle free, low latency results from quite modest systems so I am quite hoping it’s just something I need to tweak.</p><p>One quick question - from your comment “So moral of the story: if you have an unstable CPI, you should take the LOWEST number as the actual performance to avoid xruns in the buffer and therefore audio glitches or dropouts” - how can you do this? Is there anywhere in Pianoteq to limit this and keep it to the lowest number?</p><p>I guess it’s back to the drawing board for me! All the best and thanks again for your help.</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-26T19:35:23Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971593#p971593</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971583#p971583"/>
			<content type="html"><![CDATA[<p>i should also add i have seen fluctuating CPI on systems that rely heavily on cpu thermal throttling (eg poor engineering on cooling system on many lightweight thin consumer systems that seem powerful according to specs) because they can&#039;t sustain max performance under load.</p>]]></content>
			<author>
				<name><![CDATA[ethanay]]></name>
				<uri>https://forum.modartt.com/profile.php?id=497</uri>
			</author>
			<updated>2020-11-26T15:50:30Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971583#p971583</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971582#p971582"/>
			<content type="html"><![CDATA[<div class="quotebox"><cite>BTMASS wrote:</cite><blockquote><p>Turned out (I think so anyway, at the moment!) that I was hoisted by my own petard.&nbsp; I had set up the SurfacePro from the get go to run at 100% performance all the time and followed Molten Music’s excellent video about how to tweak the SurfacePro and force it to throw all its might at every task, rather than act like a normal tablet and try and conserve power at all costs.&nbsp; As soon as I reverted to Windows 10 Balanced Power plan things started to calm down and the Fluctuation Warning disappeared.</p></blockquote></div><p>In my experience the only times I have gotten such large fluctuations in performance index is with CPU scaling enabled (eg 65-124), so your report seems backward to me (though I am not arguing, I just don&#039;t understand why or how taking it out of performance mode resolved the issue!)</p><p>The warning is not dangerous. CPI just indicates how much overhead you have before CPU load. If it fluctuates overhead is unpredictable. This is a problem because no modern chipset is capable of adjusting performance without adding several (sometimes over a dozen) milliseconds of latency. That is also why multi-threading is also not recommended for latency-sensitive setups (yes, more throughput overall, but at a cost of increased latency overhead).</p><p>So moral of the story: if you have an unstable CPI, you should take the LOWEST number as the actual performance to avoid xruns in the buffer and therefore audio glitches or dropouts. (Maybe this explanation could be helpful for Modartt to include in documentation). If you take a higher number you put yourself at high risk for real-time audio rendering problems.</p>]]></content>
			<author>
				<name><![CDATA[ethanay]]></name>
				<uri>https://forum.modartt.com/profile.php?id=497</uri>
			</author>
			<updated>2020-11-26T15:47:02Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971582#p971582</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971527#p971527"/>
			<content type="html"><![CDATA[<div class="quotebox"><cite>Qexl wrote:</cite><blockquote><p>My thanks also to Stephen_Doonan and aWc, giving such good advice as always!</p></blockquote></div><p>Indeed and thanks to you again Qexl.&nbsp; </p><p>It&#039;s a shame the Mac Mini didn&#039;t work out for me as it&#039;s a very neat little unit in terms of format and size.</p><p>Who knows in the future whether the M1 chip or a variant of it will make it&#039;s way into the iPad Pro line, which would be very attractive.</p><p>Cheers</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-25T08:15:58Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971527#p971527</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971516#p971516"/>
			<content type="html"><![CDATA[<p>Well, super good news there BTMASS.</p><div class="quotebox"><cite>BTMASS wrote:</cite><blockquote><p>The variation seems more wide than my desktop PC but I’m not sure if that makes any difference or not</p></blockquote></div><p>Should be fine - it&#039;s only a real prob if the audio is glitching - but every lappy will have limits. Amazing how such unusual things can gain us the performance we ask from our machines. </p><p>Feeling really happy for you <i class="far fa-smile smiley"></i></p><p>My thanks also to Stephen_Doonan and aWc, giving such good advice as always!</p><p>Cheers.</p>]]></content>
			<author>
				<name><![CDATA[Qexl]]></name>
				<uri>https://forum.modartt.com/profile.php?id=4633</uri>
			</author>
			<updated>2020-11-25T04:14:51Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971516#p971516</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971496#p971496"/>
			<content type="html"><![CDATA[<p>As an aside, I did get an email response from Focusrite.&nbsp; I&#039;m sure they won&#039;t mind me sharing the gist of it here:</p><p>&quot;As discussed on the phone, the M1 processor systems are very new and as we&#039;re not given early access to them, we haven&#039;t had a lot of time to test them. But rest assured, Our QA team are thoroughly testing them! </p><p>We will update the following article when we know more:<br /><a href="https://support.focusrite.com/hc/en-gb/articles/360017407320-Focusrite-product-compatibility-with-macOS-11-Big-Sur-on-Apple-Silicon-based-systems">https://support.focusrite.com/hc/en-gb/...ed-systems</a></p><p>That being said, have you tried the unit in your DAW? I have noticed a few people online saying that their unit does work as expected in a DAW, but Focusrite Control does not recognise it.&quot;</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-24T16:19:34Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971496#p971496</id>
		</entry>
		<entry>
			<title type="html"><![CDATA[Re: Performance Index - large fluctuations warning]]></title>
			<link rel="alternate" href="https://forum.modartt.com/viewtopic.php?pid=971495#p971495"/>
			<content type="html"><![CDATA[<p>Well I seem to have it sorted out now, thank goodness. I first tried attaching my desktop PC to see if it could be a PC issue.&nbsp; That also displayed a range in the performance indicator (171..192) but no fluctuation warnings.</p><p>Turned out (I think so anyway, at the moment!) that I was hoisted by my own petard.&nbsp; I had set up the SurfacePro from the get go to run at 100% performance all the time and followed Molten Music’s excellent video about how to tweak the SurfacePro and force it to throw all its might at every task, rather than act like a normal tablet and try and conserve power at all costs.&nbsp; As soon as I reverted to Windows 10 Balanced Power plan things started to calm down and the Fluctuation Warning disappeared.</p><p>It seems to run well at 48,000/48,000/192 with Auto (Optimistic) polyphony just fine, with a performance indicator ranging 82..122.&nbsp; If there is any latency my old knackered ears can&#039;t detect it.&nbsp; The variation seems more wide than my desktop PC but I’m not sure if that makes any difference or not.</p><p>Many thanks for your help everyone!</p>]]></content>
			<author>
				<name><![CDATA[BTMASS]]></name>
				<uri>https://forum.modartt.com/profile.php?id=7087</uri>
			</author>
			<updated>2020-11-24T16:16:17Z</updated>
			<id>https://forum.modartt.com/viewtopic.php?pid=971495#p971495</id>
		</entry>
</feed>
