Silent laptop advice.

I have a Dell Inspiron. No noise issues on that one. Performance wise I would invest in one with a strong processor (i7 8000+ or better), 16 GB RAM and a large internal SSD (for your other VST's)

Thanks for the reply's.
I'll check them out.

I currently have a laptop a bit old and too noisy (Clevo W350ET with Core i7 alias LDLC Saturn). While waiting to change it very soon for a more powerful PC (and quieter I hope), I tried without no results to add a large external fan under the PC, running at low speed. I obtained much better results by modifying the power options of Windows (advanced options) by favoring passive cooling. to go further, it is also possible to limit the percentage of maximum frequency of the CPU used. Of course, these two options significantly reduce the power of the laptop, which no longer allows to use Pianoteq with a high internal sampling rate. After a quick test (in demo mode) on my daughter's Asus VivoBook pro (i7 8750H) laptop and my wife's macbook pro (intel core i5), even in stress mode, these two laptops are much quieter.

Bruno

MSI GL62 7RD, with a very generous cooling system (it's an entry level gaming laptop) it runs below 40ºC most of the time, and has tuneable cooling fans. The only source of noise is the hard drive which can be turned off in energy options (or removed) since the operating system is on a SSD.

So, I finally decided and bought a a-grade,
Second hand HP EliteBook Folio 1020 g1.
It is fanless, so no noise at all.
I'm still tweaking windows, but I can play with the same latency as my Numa compact 2. Performance index is about 89 - 105.

Tomorrow I will try to install Debian for further testing,
Regarding audio performance it is usually a way better choice than windows.

Okay, I've got a dilemma. On one hand, I want my laptop to be silent.
On the other hand, I want it not only to perform well, but to do so where I plan to take it; which means in one of these houses in Alicante which I think about getting for the holidays. In other words, I'm looking for a silent laptop... in a setting where I'll rely heavily on the fan.
So, is there a way to combine both, or am I condemned to either use it right in the stream of the air conditioner, or to make do with a noisy laptop?

Hmm, I am on Pop!_OS (ubuntu-based). With Debian, are you using the -lowlatency kernel line or the -generic kernel line, I am curious? The -lowlatency line helps me more achieve lower latency (e.g., ~3-5ms) in JACK without xruns, regardless of fine-tuning the rest of the system. In the -generic kernel, I have to keep latencies higher, and still get the occasional xrun (especially if networking services are active).

This is intriguing to me. On my laptop (System76 Darter Pro), changing the governor to "performance" (which just changes the EPP policy to "performance") keeps the CPU running at a maximum sustainable frequency. It downclocks automatically as necessary for thermal reasons. Ceteris paribus, small/thin/light laptops will not do as well as a bulkier laptop for cooling reasons. They just can't sustain higher frequencies (and heat generation) as well as systems with more airflow/movement. Heat builds more quickly and dissipates more slowly. They are fine for normal desktop use, but as soon as you put sustained workloads on them you get more fan noise and/or CPU throttling.

In my own tests, and according to anandtech (https://www.anandtech.com/show/10959/in...i3-7350k/3) CPU frequency scaling still occurs much too slowly (and with little or no concern for DSP load) to be of any use to dynamically boost low latency throughput, so the baseline frequency of the CPU basically determines DSP capacity. They are excited about getting scaling down to 10-30ms, and we are talking about dynamic audio processing in the sub-10ms latency range. This gels with my experience, which sums as follows: if Pianoteq has caused the CPU to increase in frequency, it has likely already caused a DSP overload (even an inaudible one). Likewise, Pianoteq consistently reports a performance index conducive to the lowest frequency I have allowed. So Pianoteq performance potential is tied to whatever is your maximum sustainable baseline frequency. For my laptop, which runs a Core i5 8265U, this seems to be about 2100-2500MHz (1600MHz is "guaranteed" by the TDP of the CPU, and 800 MHz is considered to be most efficient for this particular CPU).

You can get a stronger CPU at the same TDP and get better performance per energy used or heat generated. You can improve cooling system efficiency (bigger case, better cooling). You can undervolt the CPU.

Apart from that, the new intel pstate "powersave" governor (which defaults to the "balance_performance" EPP) acts like a more efficient and responsive ACPI "ondemand" governor. I can either change to the "performance" governor, which tells the computer to run at highest thermally-sustainable frequencies, or increase the minimum (baseline) frequency while staying in the "powersave" governor. The latter gives more fine-grained control over performance vs energy consumption/heat production (and fan noise).

Hypertreading is enabled (System reports 4 cores, but processor has 2 cores)

That gels with my experience. Under "performance" EPP my processor jumps to max or near-max (around 3.8ghz) and stays there as long as there are no cooling issues. For example, playing pianoteq it stays up there (due to the intermittent workload pianoteq demands). However, if I am transcoding (or doing any other sustained max throughput work load), within 10-15 seconds it downclocks to around 2.2-2.5, which is still 6-900mhz above TDP. I am sure it would downclock more in a hot ambient environment.

I have never experienced frequency scaling occurring proactively enough to prevent DSP overloads, which is why I say low latency throughput (and thus pianoteq) performance is tied to the processor's current minimum frequency, whether default or set by the user (via EPP performance or "min" p-state parameter)

What does your pianoteq performance index read? At around that frequency mine reads around 65-70. At full "performance" mode it reads 100-130.

I have read and now understand how hyperthreading can conflict with latency and low latency throughput, so will try disabling it and also all the performance sucking vulnerability side-chain timing exploit mitigations that now accompany it (spectre, meltdown, zombieload, etc). Your processor simply doesn't have hyperthreading anyway.

This is starting to get off topic, but I am looking forward to comparing Pianoteq and system low latency performance and throughput with and without hyperthreading enabled. Especially since theoretically, disabling hyperthreading means disabling all the performance-sucking protections needed to maintain the safety of the structural flaws inherent in hyperthreading that are breeding all these exploits: https://thehackernews.com/2018/06/openb...ading.html

Even without those mitigations in consideration, hyperthreading massively increases thread latency, and offers marginal (if any) throughput benefit as the number of physical cores increase. For example, a 4 core system with hyperthreading (8 virtual cores) only acheives a theoretical performance boost equivalent to one additional physical core, at a great cost of latency. Real-world performance is less, and much so considering the mitigations in place to avoid Spectre, Meltdown, Zombieload, etc attacks.