It is currently Fri Mar 29, 2024 9:30 am


CPU for large sample sets, e.g. NANCY - only ONE solution

Existing and forthcoming Hauptwerk instruments, recommendations, ...
  • Author
  • Message
Offline
User avatar

vpo-organist

Member

  • Posts: 306
  • Joined: Wed Apr 29, 2020 6:49 am

CPU for large sample sets, e.g. NANCY - only ONE solution

PostFri Aug 27, 2021 12:01 pm

As discussed in this thread there are big problems with the CPU performance.

On the manufacturer's page of NANCY only the approximate RAM requirement is mentioned. There is no mention of CPU. So, as an uninitiated user, you can assume that a current PC with good CPU performance and enough RAM will not cause any problems.

Now buyers are reporting the problems with CPU performance. Piotr is reading this thread for sure. But no information comes from him what CPU monster you need for the NANCY sample set. Apparently, no one really knows.

This is comparatively like buying a machine that you plug into a wall socket. If you only use a little power, everything works fine. If you use the full power, the fuse blows.

A sample set is produced of which nobody knows which PC is needed for it. Even if they did know, pretty much ALL NANCY customers would have to buy a new PC. Such a thing is called "developed past the needs of the customer" (not designed for most customers). It would be fair of Piotr to explicitly point out this fact.

The sample set should have been developed for Convolution Reverb from the beginning. It has been reported that a sample set with IR does not sound as good as a multi-channel set. This is logical. If you only use one IR, then you only have one perspective. If the manufacturers would use multiple IR's then you have an absolutely high quality solution.

I think it is high time to align Hauptwerk and sample sets to Convolution Reverb in an optimal way. Then there will be no more big performance problems with a current PC. And then you don't need more than 64 GB. This is pure madness, what a Hauptwerk customer is expected to do.
Offline
User avatar

mdyde

Moderator

  • Posts: 15446
  • Joined: Fri Mar 14, 2003 1:19 pm
  • Location: UK

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Aug 27, 2021 12:18 pm

[Topic moved here.]
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Aug 27, 2021 12:58 pm

I use 16 stereo buses with IR reverbs on dry or slightly wet samplesets to turn them into surround using IRs taken at multiple mic positions. It never sounds as good as a good surround sampleset, but saves some organs that are too dry from being uninstalled.

Unfortunately, even 16 concurrent IRs adds a significant amount of CPU overhead. On my i9-9900, I get 2 - 4 more bars on the HW CPU meter depending on registration -- about the same overhead as going to 96k sample rate, enabling higher definition pitch shifting, or adding another Nancy mic position.

Running multiple IRs in place of multiple mic perspectives and multiple releases certainly saves memory, but whether it saves CPU load at the same level of realism isn't so easy to prove or disprove without a lot of experimentation.

In any case, samplesets with 2, 3, or 4 perspectives never say that you have to run all the available mic positions. They're there so you can choose what works for you. I'm not sure documenting producer-tested CPU loads on their specific machine for specific settings at tutti would be translatable by most people to their own hardware, but I agree it would give at least some indication of unusual CPU requirements before they buy.
Offline
User avatar

vpo-organist

Member

  • Posts: 306
  • Joined: Wed Apr 29, 2020 6:49 am

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Aug 27, 2021 3:34 pm

I don't know your setup. How does Convolution Reverb sound with headphones?
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Aug 27, 2021 5:02 pm

I don't know. I disabled my headphone connection last year.
Offline
User avatar

vpo-organist

Member

  • Posts: 306
  • Joined: Wed Apr 29, 2020 6:49 am

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostSat Aug 28, 2021 3:23 am

Too bad. Using IR's is not so easy for sample set manufacturers and users. The first requirement are several IR's (perspectives) from the recording location, because the organ is tuned to the room. Otherwise it will usually already sound different or worse. There are exceptions where a dry sample set sounds good even with other IR's. I don't have time to make recordings of Caen Dry and Rotterdam Dry at the moment. But it sounds fantastic.
Have you ever presented your setup here on the forum? I would be very interested in that.
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostSat Aug 28, 2021 5:23 am

It's a typical multichannel audio system with 12 front and 8 rear speakers and a subwoofer. Speakers are grouped by model and ranks are assigned to groups more or less by front/rear and pitch range. Default algorithm for speaker selection within group.

4 KRK RP10-3 for rear 32' - 8'
4 Adam T7V for rear 8' and up
4 Adam A8X for front 32' - 8'
8 Adam A3X for front 8' and up
1 SVS PC12+ mixdown
Offline
User avatar

vpo-organist

Member

  • Posts: 306
  • Joined: Wed Apr 29, 2020 6:49 am

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostSun Aug 29, 2021 9:08 am

mnailor wrote:I use 16 stereo buses with IR reverbs on dry or slightly wet samplesets to turn them into surround using IRs taken at multiple mic positions. It never sounds as good as a good surround sampleset, but saves some organs that are too dry from being uninstalled.

Which sample set with which IR's doesn't sound as good as a surround set? Unfortunately, there are not many dry sample sets for which there are also IR's from the same room.

If any IR is used, then a complete re-intonation is often necessary to achieve a good result.
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostSun Aug 29, 2021 12:07 pm

No, I can't compare the same sampleset dry with IR vs wet surround. What I meant to say was that I haven't gotten IR on any dry or semidry sampleset to sound nearly as good as good wet surround samplesets like Caen, Nancy, Oloron, Rotterdam, Brasov, Armley, or Hereford.

I'm still experimenting with IR, but it I haven't gotten it anywhere close on my perception of realism or overall effect.

I've never actually been able to hear any problems with 3 or more releases. Maybe I'm just not tuned into whatever you hear that's inferior to calculated IR reverb.
Offline

adri

Member

  • Posts: 1545
  • Joined: Tue Aug 31, 2004 6:33 am
  • Location: Belgrade, Serbia

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostThu Sep 02, 2021 7:12 am

Some small instruments in not too wet acoustics are now being sampled with 6 or even 8 microphones. Isn’t this overkill?
When I attend a concert, my ears just do stereo and that’s just fine if you pick the right listening spot.

Now large organs are also sampled with a plethora of microphones in wet acoustics. The near channel never mixes well with the far channel. Yet these sample sets take forever to install, load and then you spend a lot of time trying to figure out what listening position works best for you.

In all cases, this devours both RAM and CPU.

Is it overkill ? Many think so. Others love it.

If your computer is not the latest performance monster, there is a problem.

Software and hardware have been in a constant cat and mouse game.

For me personally, four channels is enough. In most cases, I find myself setting an organ to the wettest setting or slightly below it.

Convolution reverb works well on some too dry sample sets but kills the organ in other cases.

There is room for both improvement, simplification as well as not running too far ahead of the average computing power of the end user.

My 2 cents for the moment.
Offline
User avatar

vpo-organist

Member

  • Posts: 306
  • Joined: Wed Apr 29, 2020 6:49 am

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostThu Sep 02, 2021 2:20 pm

Maybe it is overkill. They are just different listening perspectives. You can choose one according to your taste and not load the other channels. 8-Channel is also marketing and you pay for the channels even if you don't use them.

The problem with the cat and mouse game is that no one knows exactly what requirements the cat must meet so that you can really catch every mouse.

If someone says this works and that doesn't work for me, then over time a collection of information forms in the forum that is difficult or impossible to evaluate. Too bad. This is important information from the Hauptwerk customers that should be channeled to get value out of it.
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostThu Sep 02, 2021 3:50 pm

vpo-organist wrote: snip...
If someone says this works and that doesn't work for me, then over time a collection of information forms in the forum that is difficult or impossible to evaluate. Too bad. This is important information from the Hauptwerk customers that should be channeled to get value out of it.


True, anecdotes spread across hundreds of topics are hard to draw any useful conclusions from. But how would you suggest collecting and organizing this?

To me, what is needed is a Hauptwerk capacity planning methodology with a limited number of explicit benchmark tests, a central collection point for users to submit their test results with computer specs and settings, a built-in way to compare test results on similar hardware to identify unreplicable results for human review, auto-updating capacity models to forecast how the benchmarks should do on as yet untested hardware, and a way to compare any sampleset's maximum compute requirements to the benchmarks to help users identify hardware that supports a benchmark that is known to exceed the sampleset's requirements. That's a huge effort.
Offline
User avatar

JulianMoney-Kyrle

Member

  • Posts: 386
  • Joined: Tue Jun 19, 2007 6:23 pm
  • Location: Calne, Wiltshire, UK

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Sep 03, 2021 4:34 am

I am inclined to agree. When trying to decide whether to buy a particular sample set, or which one out of several of the same type, it is very difficult to find useful information on which to base a decision. The forum is all very well, but such information as is available isn't very well organised.

I am lucky enough to be in a position where I have a computer that will run just about all of what is available (I haven't tried Nancy) and a system that can make good use of different listening perspectives (six stereo channels iplus sub-woofer n a large room with a high ceiling). I also have a lot of different sample sets, but the ones I keep going back to are the ones which are recorded in a way that works best on my system, with a sense of presence that best gives the illusion of playing a real organ. Sometimes these are smaller instruments such as Noordbroek. It is also helpful to be able to adjust the perspective, particularly when I am learning or practising a piece.

However, what works best for me may not be best for somebody with a more modest system with limited processing power or RAM, where compromises are required when loading the more demanding instruments, or somebody using fewer channels or headphones, or just in a different room. There is also the matter of expectation, as invididual users will be familiar with different organs in different acoustics depending on where they live and what instruments are available for them to play (leaving aside the question of taste).

I think it would be enormously helpful if there were a central repository of detailed information listing not only what samplesets are available, together with cost and system requirements, but also what type of organ they are and what makes them interesting, and how those buying them have reacted to them, as well as user reviews. It has taken me a long time and quite a number of purchases, for instance, to come to understand the difference between a North German / Dutch Baroque organ and a Thuringian / "Bach" organ, and modern instruments inspired by them, though on the way I have discovered some real gems, such as Zutphen from Sonus Paradisi which hardly seems to have been mentioned on the forum at all. I do think that anyone posting such a review should also give their credentials, as an experienced Dutch organist growing up in Groningen is likely to have a different perspective from a keen amateur living in the UK (such as myself), and somebody familiar with playing the original from which the samples have been created is unlikely to have the same views as somebody who only knows the instrument from recordings.
Offline

larason2

Member

  • Posts: 752
  • Joined: Thu Feb 04, 2016 9:32 pm

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Sep 03, 2021 11:16 am

In previous threads it was mentioned that the vast majority of Hauptwerk users still listen in Stereo. This is the case for me. After early experimentation with IR's, I basically abandoned them, because I found it was hard to get the voicing to match for most IR's and sample sets, even dry ones. When a sample set has more than one perspective, I pick and only load the one perspective I like, ignoring the rest. This minimizes RAM and CPU use. I think for most users this is the best solution for now.

Is having a dry perspective with the room's own IR a solution? As I've mentioned before, I don't think its possible to get a sample set dry enough for my taste. I do have some dry sets with the room's own IR, and I still think the recorded perspective is superior to the dry sample with the IR (it still sounds a bit artificial to my ears). As mentioned before, convolution reverb uses CPU resources, so it's not a perfect solution. Others clearly feel differently than I do though!

Now if I had more than two channels in my sound system, I would first increase the number of speakers in each channel, and route each note to the different speakers using the advanced routing options (to cut down even more on intermodulation distortion). Once I have 4 speakers per stereo channel, then I may consider two speakers in the back for ambiance. Still, even then, I would only load 4 channels probably. It depends on how things sound though!
Offline

mnailor

Member

  • Posts: 1602
  • Joined: Sat Nov 30, 2013 5:57 pm
  • Location: Atlanta, GA

Re: CPU for large sample sets, e.g. NANCY - only ONE solutio

PostFri Sep 03, 2021 11:52 am

Yeah, I have 12 front and 8 rear speakers and 1 sub, and I only load 2 perspectives (4 channels) on 6 - 8 channel samplesets. That means always I omit the closest, dryest perspective -- not a fan of close mics -- I wouldn't put my ear against the organ case and listen for an hour... :D

On a few 4 channel surround samplesets I only load ambient and leave out the closer perspective because it's too close.

For Nancy, after rearranging my audio groups for better front/rear/left/right clarity, I'm back to only loading Front and Middle. I had been mixing a bit of Close before to firm up attacks.

For stereo samplesets, I recently started adding an artificial rear perspective using Rank Perspective routing with release truncation as needed and IR. Stole that idea from a few people here and on the fb group. Of course it was in the user guide, but I missed the significance at the time.

My bus groups are now organized by 4 speaker types, 2 groups in front and 2 in rear, each duplicated both Dry and Wet (IR) in the same mixer preset, to send wet samples to Dry groups and dry or truncated samples to wet groups. I'll use different presets for different IRs and keep the group numbers the same to easily switch out the IR for a given organ by just changing preset number.

That scheme works for samplesets that are wet stereo, dry stereo, wet surround, or surround but too dry for me to enjoy.
Next

Return to Hauptwerk instruments

Who is online

Users browsing this forum: No registered users and 6 guests