Bug in HW8: Free RAM misreported [edit: not a bug]
- JulianMoney-Kyrle
- Member
- Posts: 421
- Joined: Tue Jun 19, 2007 6:23 pm
Bug in HW8: Free RAM misreported [edit: not a bug]
I downloaded and installed HW8 last night. So far it all seems to be working as expected, and I am in the process of rebuilding my 60-odd caches. However, the amount of free RAM is misreported, both the bar graph and the digital readout. I have 128GB, and regardless of which organ I load, it reports 43.8 GB free when it is loading and 45.3 GB once the loading is complete and hte organ is ready to play.
It does load cached organs much more quickly than HW7, and unloads them almost instantaneously.
It does load cached organs much more quickly than HW7, and unloads them almost instantaneously.
Re: Bug in HW8: Free RAM misreported
No bug. Check out the release notes, page 11, for memory management changes. There's a new option to retain part of memory (60% by default) when unloading an organ to speed up loading the next one. You can turn it off if it bothers you.
- JulianMoney-Kyrle
- Member
- Posts: 421
- Joined: Tue Jun 19, 2007 6:23 pm
Re: Bug in HW8: Free RAM misreported
I don't think I spotted this. So rather than release the memory back to Windows, HW hangs onto it and overwrites it as it loads the next sample set? It certainly seems to save some time.
Presumably you can still load two sample sets in succession even when each one is large enough to require nearly all the installed RAM.
Presumably you can still load two sample sets in succession even when each one is large enough to require nearly all the installed RAM.
Re: Bug in HW8: Free RAM misreported
Yes, and yes, it will still fill up available memory if needed and release back the excess over the retention % upon unload.
If you wouldn't mind, perhaps editing the topic a bit would avoid attracting trolls?
If you wouldn't mind, perhaps editing the topic a bit would avoid attracting trolls?
Re: Bug in HW8: Free RAM misreported
Hello Julian,
Thanks very much for upgrading, and glad to hear loading and unloading are aster.
To add to Mark's comprehensive replies, for reference, here is an excerpt from another thread that also covers it, in case it helps:
Thanks very much for upgrading, and glad to hear loading and unloading are aster.
To add to Mark's comprehensive replies, for reference, here is an excerpt from another thread that also covers it, in case it helps:
[I've edited the title of the topic to indicate that it isn't actually a bug.]mdyde wrote:That's working as intended with the default values for the new settings on the "General settings | General preferences | Advanced ..." screen tab, i.e.:
- "Reserve RAM in chunks for fastest organ unloading/loading [may use a little more memory]" = ticked.
- "Max proportion of computer's RAM to keep reserved when unloading an organ" = 60%
You can reduce the latter setting if you like to free up memory whenever organs unload. Even if you set it to zero (keeping the first setting ticked) you should still get significantly faster loading and unloading than earlier versions, but higher settings (e.g. 60%) will give the highest possible loading/unloading speeds.
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Hauptwerk software designer/developer, Milan Digital Audio.
- JulianMoney-Kyrle
- Member
- Posts: 421
- Joined: Tue Jun 19, 2007 6:23 pm
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Dear Martin (and Mark),
Thanks for the explanation; I'm not sure how to edit the titles of posts, so thank-you for doing this for me.
Previously I have used the RAM meter to check that an organ is able to load as configured, though these days I have so much RAM that it isn't really an issue.
I quite often load organs uncompressed as my CPU will handle it and it supposedly improves the polyphony. I am now thinking, however, that as well as taking up more RAM, an uncompressed sample will use more disk space and take longer to load. Load times have been a great problem for me of late as I have an SSD that seems to be running much slower than its stated performance and I haven't been able to get to the bottom of why this is. Since I am now having to rebuild the caches anyway for HW8, I have never overloaded the CPU as regards polyphony (Windows housekeeping is a different matter) and HW8 is supposed to have better polyphony than HW7, perhaps I should change my policy and load everything with lossless compression. Any thoughts?
Thanks for the explanation; I'm not sure how to edit the titles of posts, so thank-you for doing this for me.
Previously I have used the RAM meter to check that an organ is able to load as configured, though these days I have so much RAM that it isn't really an issue.
I quite often load organs uncompressed as my CPU will handle it and it supposedly improves the polyphony. I am now thinking, however, that as well as taking up more RAM, an uncompressed sample will use more disk space and take longer to load. Load times have been a great problem for me of late as I have an SSD that seems to be running much slower than its stated performance and I haven't been able to get to the bottom of why this is. Since I am now having to rebuild the caches anyway for HW8, I have never overloaded the CPU as regards polyphony (Windows housekeeping is a different matter) and HW8 is supposed to have better polyphony than HW7, perhaps I should change my policy and load everything with lossless compression. Any thoughts?
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Thanks, Julian.
- All else being equal, Hauptwerk v8 will only give more polyphony than v7 if you have a recent CPU that has cores of differing performances relative to each other (e.g. the 'performance' and 'efficiency' cores on 12th+ generation Intel i9/i7 CPUs). On older CPUs, or those with all-equal cores, polyphony will be the same as v7.
- All else being equal, v8 should load any given organ faster than any previous version.
- As for previous versions: with rank memory compression turned, any given organ may load somewhat faster than with memory compression on, and polyphony should be slightly higher (but more memory will be needed by the organ, of course).
- As for previous versions: to minimise drive space usage, the cache file is always compressed regardless of whether memory compression is enabled for ranks, so enabling/disabling rank memory compression doesn't affect drive space.
Hence if you have lots of spare RAM and want the fastest possible loading (and highest possible polyphony) then you could turn rank memory compression off.
- All else being equal, Hauptwerk v8 will only give more polyphony than v7 if you have a recent CPU that has cores of differing performances relative to each other (e.g. the 'performance' and 'efficiency' cores on 12th+ generation Intel i9/i7 CPUs). On older CPUs, or those with all-equal cores, polyphony will be the same as v7.
- All else being equal, v8 should load any given organ faster than any previous version.
- As for previous versions: with rank memory compression turned, any given organ may load somewhat faster than with memory compression on, and polyphony should be slightly higher (but more memory will be needed by the organ, of course).
- As for previous versions: to minimise drive space usage, the cache file is always compressed regardless of whether memory compression is enabled for ranks, so enabling/disabling rank memory compression doesn't affect drive space.
Hence if you have lots of spare RAM and want the fastest possible loading (and highest possible polyphony) then you could turn rank memory compression off.
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Hauptwerk software designer/developer, Milan Digital Audio.
- JulianMoney-Kyrle
- Member
- Posts: 421
- Joined: Tue Jun 19, 2007 6:23 pm
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Thank-you for the explanation, Martin. I will keep them uncompressed.
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Thanks, Julian. You're very welcome.
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Hauptwerk software designer/developer, Milan Digital Audio.
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Julian, thanks for bringing up loading uncompressed. I hadn't tried that in a long while. I've now set all my organs to load uncompressed, except San Francisco, Billerbeck, and Nancy, which don't fit in 128 GB without compression.
Loading times are faster with HW 8, with my slowest load at 90 seconds. Previously, loading was more CPU-bound, but now my cache disk is 85% busy at about 950 MB/s.
Loading times are faster with HW 8, with my slowest load at 90 seconds. Previously, loading was more CPU-bound, but now my cache disk is 85% busy at about 950 MB/s.
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
If I read the discussion correctly, there are benefits to loading sample sets uncompressed, rather than using "lossless" compression.
Is that correct? And correct for both HW 7 and HW 8?
Is that correct? And correct for both HW 7 and HW 8?
Bruce
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Hello Bruce,
The (only) benefits of loading ranks uncompressed are:
1. The organ will load faster. (However, with Hauptwerk v8+ the difference in loading times between compressed and uncompressed will generally be less than for earlier versions, since v8+ loads organs more quickly even with with compression enabled.)
2. The computer will be able to manage bit more polyphony (typically about 10-15% higher).
The (only) benefits of loading ranks uncompressed are:
1. The organ will load faster. (However, with Hauptwerk v8+ the difference in loading times between compressed and uncompressed will generally be less than for earlier versions, since v8+ loads organs more quickly even with with compression enabled.)
2. The computer will be able to manage bit more polyphony (typically about 10-15% higher).
Best regards, Martin.
Hauptwerk software designer/developer, Milan Digital Audio.
Hauptwerk software designer/developer, Milan Digital Audio.
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
Hello Mark,
You write ....Loading times are faster with HW 8, with my slowest load at 90 seconds.
Wow! This seems very fast to me.
Could you give a little more detail. I note you have Billerbeck.
How long does Billerbeck take to load in HW8?
How long did it take in HW7?
Stephen
You write ....Loading times are faster with HW 8, with my slowest load at 90 seconds.
Wow! This seems very fast to me.
Could you give a little more detail. I note you have Billerbeck.
How long does Billerbeck take to load in HW8?
How long did it take in HW7?
Stephen
Re: Bug in HW8: Free RAM misreported [edit: not a bug]
I'm away from home for a couple weeks, but from memory: Caen 2.67 Surround loads in 30 seconds, was about 1 minute on HW7. Nancy is now about 1 minute, was 2 minutes. Billerbeck is around 90 seconds, was 3 minutes. My cache directory is on an internal SSD. All samples, 24 bits, full releases.StephenM wrote:Hello Mark,
You write ....Loading times are faster with HW 8, with my slowest load at 90 seconds.
Wow! This seems very fast to me.
Could you give a little more detail. I note you have Billerbeck.
How long does Billerbeck take to load in HW8?
How long did it take in HW7?
Stephen
Overall, my load times are approximately halved by HW8. That does depend on whether it's the first organ loaded after starting HW. It's somewhat faster to load subsequent organs since HW is able to retain memory allocated for the previous organ.
I'm on Windows 11, and freeing memory was very slow. Unloading an organ before HW8 could take over a minute, now it's a few seconds.