Microsoft's latest operating system, although primarily designed for business and network systems, seems to offer many features that are attractive to musicians, including greater stability and support for dual‑processor machines. But is there enough software support to make upgrading worthwhile? Martin Walker investigates.
People differ greatly in the way they react to a new operating system release. Some buy it the day it comes out because they simply must have the latest version of everything. Others look back on the time it took them to iron out operational problems with their current operating system, and how many tweaks they've needed to optimise it over the years, and leave well alone. This is commonly referred to as the 'If it ain't broke, don't fix it' approach, and is the reason why many musicians are still happily running Windows 95. Most foolhardy are those who get hold of pre‑release beta versions of the very newest operating systems, either legitimately or via the back door, and try to use them not as intended — for in‑the‑field testing and feedback — but for serious music work.
The most common industry advice with a new operating system release is to wait for at least a few months to let the lemmings discover the pitfalls before taking the plunge yourself. So, since it's now been almost a year since Microsoft's Windows 2000 operating system was released, it's time to take a more considered look at its advantages and disadvantages for the PC musician.
Before I wade into the main issues that affect musicians, it's worth quickly going over the main features of Windows 2000. There are three versions — Professional, Server, and Advanced Server — and which one you need depends on how many PCs you need to support. The cheapest Professional version should suffice for all musicians, which is lucky, since the Server version starts at about £800, while the Advanced Server is more like £3000. You can buy the full Professional version at a street price of about £250, as an upgrade from Windows 95/98 for about £170, and from Windows NT for about £100.
It's a mistake to think of Windows 2000 as an update of Windows 98, since all three versions of it are based on Microsoft's operating system for business users, Windows NT 4.0, and as such have rather different aims. For most musicians the attraction is not so much the feature set, although this does offer the tempting prospect of multi‑processor support, but rather its claims for much better reliability and stability. These qualities are crucial for those running 24‑hour‑a‑day applications such as office networks and web servers, and even though running a PC 24 hours a day for months probably isn't high on most musicians' lists of priorities, trusting it not to crash during a vital session is. Like Windows NT before it, Windows 2000 runs each application in its own section of 'protected' memory, so that even if one does crash, the system will carry on regardless without requiring a reboot.
However, this stability does come at a slight price: you have to have Windows 2000‑compliant setup programs to install software. If your choice of music software doesn't specifically state its Windows 2000‑compliance, then you may not be able to install and run it successfully, even though it runs beautifully on Windows 95 and 98. Some manufacturers, like Steinberg, have embraced Windows 2000 wholeheartedly, and any version of Cubase from 3.71 onwards can be installed under Windows 2000. Philippe Goutier's Wavelab, also marketed by Steinberg, makes good use of dual processing, and he claims that it gives slightly better performance under Windows 2000 than with Windows 98. Cakewalk have been similarly enthusiastic, and all their current products can be run successfully on Windows 2000. However, although Emagic's Sound Diver synth librarian/editor is compatible, they don't recommend that their Logic users upgrade to Windows 2000 until a new compatible version is released in Spring 2001.
There is a sneak workaround for those who desperately want to use a particular application that doesn't officially offer Windows 2000 support, but of course I can't officially recommend this, or be held responsible for any consequences. If you are about to upgrade from Windows 95/98 to Windows 2000, and you leave an existing non‑compliant application on your hard drive, there's a good chance that it will still run on Windows 2000 after the upgrade, although MIDI timing might well be seriously affected.
Another aspect of Windows 2000 that may not sound immediately useful to musicians is increased data security. However, while it's vital to those running Internet servers and office networks that malicious hackers can't gain access to cause damage, many of us now spend a lot of time surfing the Net at home, and Windows 95/98 has no security measures built in to stop this. Both Windows 2000 and Windows ME (the latest in the long line of Windows 95/98 releases) also have file protection, which ensures that badly behaved applications can't change a protected system file. Both these aspects will help to keep your PC running smoothly and reliably.
Graphically speaking, Windows 2000 looks fairly similar to the Windows 95/98 family, and Windows ME has already had a graphic makeover to make it look almost exactly the same as Windows 2000, so you won't really notice much difference if you make a sideways upgrade. The main differences are under the surface, and in the area of third‑party compatibility and support.
Some of the 'new' features that were added in the development of Windows 2000 from Windows NT 4.0 have already been available to Windows 95/98 users for some time. These include Plug and Play for easier hardware detection, and support for USB and FireWire. Some people have also claimed that Windows 2000 USB performance with music peripherals is better than with Windows 98, but I've yet to hear conclusive evidence. DirectX 7.0 is also included in Windows 2000, once again to bring it into line with Windows 98. In NT 3.51 there was no DirectX support at all, while DirectX 3.0 support was added in NT 4.0. However, in this early version DirectSound was only supported by an emulation mode through the MME (MultiMedia Extensions) drivers, and most of us already know how much soundcard performance drops with software synths when using emulated drivers. The reason for this was due to NT's security features, which prevented multiple applications accessing the same piece of hardware. Under Windows NT, the version of DirectX couldn't be upgraded by the user either, so those applications that checked for the required version and even offered to install a later version of DirectX because they needed one of its new features could only be run on Windows 95/98.
One new Windows 2000 feature that may sound useful to musicians is Hibernation — the ability to save the current state of your PC including open applications, window positions, and desktop, so that when you turn it back on everything is exactly as you left it. If you're in the middle of working on a song this could be a godsend. If, however, like most musicians, you have lots of memory installed, it may take longer to use Hibernation than shutting down and booting up normally, since the entire contents of RAM are first saved to disk, and then reloaded afterwards. In addition, your PC needs to have a fully ACPI‑compliant BIOS to fully support this feature. While most machines sold in the last 18 months or so will have this, others won't — see the Installation section for more details.
Given that buying a faster processor is the most popular upgrade among PC users, it seems a waste to simply throw the old one away or stick it in a cupboard to gather dust — and perhaps the most attractive feature of both Windows NT and 2000 for musicians is that, unlike the Windows 95/98/ME family, they can support multiple processors. Windows 2000 Professional can use one or two processors, while the Server version can support up to four, and the Advanced Server up to eight! For musicians using software plug‑ins to run all their EQ and effects, the thought of being able to add more processing power simply by plugging in a second processor alongside their existing one is enticing.
However, there are restrictions. Not only does your motherboard have to be specially designed to take two or more processors, but you can only use one of a few suitable processor models. The main one at the moment is the Intel Pentium III, although at least one motherboard manufacturer makes a model that can use two Intel Celeron processors, even though Intel does not support this. Most importantly, however, to gain any significant benefit the software also has to be specially written or adapted with multiple processors in mind.
The way multi‑processing works is that applications are divided into multiple 'threads', semi‑independent processes that can be run in parallel. Even with a single processor there can be advantages in this programming approach. Many applications use multiple threads to enable multi‑tasking, so that one task can carry on while another is started; and when multiple processors are available, different threads can be allocated to each CPU.
With some processor‑intensive programs, such as 3D graphics and CAD software, it's comparatively easy to split off different functions to each processor. However, the situation is somewhat more complicated when it comes to an application like a MIDI + Audio sequencer, since all the different elements are being streamed in real time, and must remain in sync.
Music applications can be split up so that audio mixing and effects are handled in one thread, MIDI processing in another, and user interface responses in yet another, and while it's possible to specifically assign each task to a separate processor, you can also let Windows handle its CPU resources dynamically across a single processor, by giving them different priorities. The lowest priority is nearly always given to the user interface, which is why your screen updates get sluggish when you run lots of real‑time software plug‑ins.
This approach is taken by quite a few different applications such as Steinberg's Cubase VST, Wavelab, and Nuendo, Emagic's Logic Audio, and all of Cakewalk's multitrack products. If you want to see whether your own applications are written using multiple threads, you can use a utility like TaskInfo2000, which I discussed in PC Notes August 2000 (you can download this at www.iarsn.com/downloads). I've shown it running alongside Cubase VST in the screenshot.
When this sort of application is used with two identical processors, the entire audio processing workload is normally handled by one processor, with any remaining tasks left to the other one. Since audio processing is by far the most significant overhead for any music application, this approach results in a typical overall performance improvement of just 20 to 30 percent — not that much considering the extra cost of setting up a dual‑processor system. To get further improvement you need to split the audio processing in some way between the two CPUs, so that it can be processed in parallel. This means added code and complexity, and until far more of us are running dual‑processor systems I suspect that few PC applications will be rewritten to run like this. However, with the introduction of the new multi‑processor Mac G4 series, several music developers have devoted more time to dual‑processor support in their Mac versions. Emagic have optimised version 4.5.1 of Logic Audio Gold and Platinum for dual‑processor Macs in such a way that the second CPU is used for all the main audio processing including plug‑ins and mixing, while the first CPU is used for the Mac OS, screen redraws, MIDI processing, audio data loads and saves, live soft synths and their plug‑ins, and live inputs and their plug‑ins. By distributing the audio processing in this way, performance improvements vary between 20 and 60 percent, depending on what proportion of your music is being processed 'live'. Sadly, at the time of writing these optimisations aren't yet available for the PC version, but Emagic are currently working on it.
Another developer working in this area is Steinberg, whose Advanced Multiple Processing Support was introduced on the Mac version 5.0 of Cubase VST. This Advanced MP code will also put in an appearance in the imminent PC version 5.0 revision 2. Once again the audio processing is split between the two processors, giving much larger performance boosts of 50 to 60 percent. Steinberg's Product Manager Dave Nicholson has written a very informative article about the differences between a conventional single‑processor system, a conventional dual‑processor system, and their Advanced Multi‑Processor system in the Knowledge Base section of their web site. You can find this at service.steinberg.de/knowledge.nsf/show/mac_multiprocessor.
Neither approach will give a dual 800MHz processor system the same performance as one with a single 1.6GHz processor, but then buying two 800MHz models is considerably cheaper. Looking at a December PC catalogue, 800MHz processors were the slowest advertised on the Pentium III list, and cost only £188 including VAT. So, assuming a 50 percent improvement running the updated Logic Audio or Cubase VST when they become available for PC, you could have the equivalent of a 1.2GHz Pentium III for just £380, though this would probably equal only a 1GHz Pentium III for less optimised software.
Ultimately, if you want to run the maximum number of real‑time effects, buying the single fastest processor you can afford is the best bet when running music software. However, there are other advantages of running a dual‑processor system that many people overlook. By splitting the graphics, file I/O, MIDI, and main audio processing into different threads you gain a more responsive user interface, and if these threads are then handled by two processors, the situation improves noticeably. Every developer I spoke to independently stressed this: when a single‑processor system is pushed to the limit it becomes sluggish and unresponsive, whereas a dual‑processor system remains far more comfortable to use. In fact, the second CPU can be loaded to 99 percent while the first will still happily respond to key presses and open new windows; if you tried this on a single CPU system it would appear to have crashed.
Using dual processors can also increase the maximum number of simultaneous audio tracks beyond what is possible with a single CPU, although ultimately this is still limited by the speed of your hard drive. It can also benefit MIDI timing — Emagic claim that by splitting audio and MIDI processing to separate processors on the new Mac version of Logic they achieve "a new level of USB MIDI timing precision", which must be good news for all musicians. The only disadvantage is that CPU performance meters tend to be far less accurate.
Sadly, the biggest stumbling block to widespread use of Windows 2000 by musicians is a scarcity of suitable soundcard and MIDI drivers. The current situation will be far clearer if I briefly explain the different types of Window driver in use, starting with Windows 95. This used so‑called VXD‑style drivers, and each soundcard developer had to add into these the appropriate routines to support DirectSound for lower latency. As most of us remember, without DirectSound support Windows used an 'emulation layer' that pushed latency back up to the level of the more basic MME drivers.
Meanwhile, Windows NT 4.0 required a completely different design of driver from the consumer Windows 95/98 platform. Since it was primarily a business OS, few soundcard manufacturers bothered to devote much effort to the extra development required. While the few NT soundcard drivers that were released were fairly successful for audio recording and playback, as I mentioned earlier, DirectSound was supported in NT only through a high‑latency emulated version, which made NT rather unsuitable for many software synths. There were also various problems with MIDI timing under NT, which is the reason why neither Steinberg nor Emagic ever released versions of Cubase and Logic specifically for NT systems.
Like Windows 95, Windows 98 could also use VXD‑style drivers, but on this platform Microsoft also introduced a completely new design: the WDM (Win32 Driver Model) format. This is intended to simplify future driver development, by providing a unified design suitable for both consumer and business operating systems, such that one driver will be totally compatible across all future platforms. Another advantage of WDM drivers is that they are better‑suited for busses like USB and FireWire. WDM also incorporates a standard low‑latency wave interface into Windows itself, so that a WDM driver automatically gets both MME and DirectSound support from Windows without needing special code added by each soundcard manufacturer. Unfortunately, WDM drivers are far more similar to the NT ones than to the older VXD model used by both Windows 95 and 98. Given that so few soundcard manufacturers had thus far developed NT drivers, it's perhaps hardly surprising that even fewer took up the option of WDM when Windows 98 was launched, especially since their existing Windows 95 VXD ones still worked well under Windows 98.
In the case of Windows 2000, two types of driver can be used — Windows NT 4.0 and WDM — but you must use one or the other for each hardware expansion device in your PC, including your graphics card, MIDI interface, soundcard, and so on. You cannot, therefore, assume that Windows 95/98 drivers will run under Windows 2000, since most are still in VSX format. Moreover, although existing NT4 drivers may work well with some peripherals, NT4 soundcard drivers will still suffer from the same limitations as they did when running under NT4. Obviously, the best solution is to use WDM drivers, but here we return to the fact that few soundcard manufacturers have developed them, so we face a chicken‑and‑egg situation: until Windows 2000 is taken up by huge numbers of musicians, few manufacturers are likely to write suitable soundcard drivers for it.
Until recently, there was a further complication for WDM audio driver developers. WDM provides a single driver model for all future Microsoft operating systems, as well as multi‑client access to audio hardware, so WDM drivers should, for instance, allow you to run a soft synth and audio sequencer on the same stereo output or mix together loads of audio streams, using WDM's own audio mixer. Unfortunately, however, this built‑in audio mixer has some 30mS of latency.
At the NAMM show back in February 2000, Cakewalk managed to get representatives from 24 music hardware and software companies to discuss support for a WDM driver extension to get around this limitation. The initial response to their detailed proposal was 'overwhelmingly positive', and Cakewalk have made great progress since then by working closely with Microsoft themselves to resolve this latency problem. The result is the publishing of a new WDM Kernel mode streaming interface that bypasses the internal mixer, and which will work with any WDM audio driver running on Windows 98, 98SE, or 2000. These interfaces are now available to every audio manufacturer who expresses an interest, and will also be included in the Whistler development kit (of which more later).
Cakewalk have already managed to achieve 10mS latency with an unmodified Soundblaster Live! card, and 1.5mS with M Audio's Delta series, and have used this technology in their next‑generation SONAR multitrack digital recording system, to be launched at the Winter NAMM show in January. SONAR is optimised for both Windows 2000 and WDM, and provides "ultra low‑latency mixing as well as real‑time effects processing on input".
If you decide to take the plunge, there are various things you should check before parting with any money (see the Windows 2000 Checklist box, left). Microsoft also have a Windows 2000 Compatibility web site (see screenshot on page 154) which offers more pre‑installation information. Bear in mind that installing a new operating system is the most fundamental change you can make to your PC, and that should anything should go wrong during the procedure you might be left with a PC that needs its hard drive completely reformatting and every application reinstalled from scratch. Windows 2000 performs plenty of checks on existing applications and drivers to check for compatibility, but the chance of many music applications being included in this checklist is unlikely.
For this reason, it's well worth considering a clean install on a different partition, or better still another hard drive, since this will allow you to start with a clean slate and lose all the unnecessary dross that always accumulates on every hard drive given half a chance. Your Registry will also start its new life in a fresh and compact state, and generally you'll give your PC the best chance of optimum performance in its new incarnation. The only time you should consider upgrading over the top of your current operating system is if you've got loads of applications already installed that you don't want to (or can't) reinstall, or as mentioned earlier, as a way of running an application that can't be installed under Windows 2000, but which runs quite well if you smuggle it in under cover of the upgrade.
However, the Windows 2000 installer can also automatically configure a multiple‑booting system with either Windows 95/98 or NT 4.0, so that each time you boot your PC you can choose which version of Windows you want to run. If you decide to do this, or to upgrade over the top of an existing operating system, you can start the Windows 2000 Setup.exe file from inside Windows. Otherwise, you can boot from the Windows 2000 CD‑ROM if your PC allows this, or failing that use the supplied boot floppy disks. Before you complete the installation you'll have to decide which format to use for your operating system hard drive (see Formatting Hard Drives box on page 150).
It goes without saying that you should back up everything before you start. You should make sure you have a copy not only of all your songs, accounts, and word processing files, but also things like plug‑in presets, INI files that contain all your favourite settings, VST Instrument patches, and all those update files that you downloaded to bring your applications up to their latest versions, and which you'll probably now need once again. This is another good reason to keep your data either on a separate partition, or on a different hard drive altogether, safely out of harm's way.
On my Internet travels I came across a user poll asking whether people would use a consumer version of Windows 2000 if one were available, and an overwhelming 78 percent said 'yes'. This is hardly surprising, since if you combine the stability and security of the current Windows 2000 Professional version with the multimedia and consumer support of Windows 98 you would have a winner. Well, Microsoft have long wanted to take this path, so that they can concentrate on further development of one rather than two major Windows platforms. The problem in the past has been that the hardware requirements of NT have always been significantly higher than those of 95/98.
However, such is the march of technology that when the next version of Windows 2000 (codenamed Whistler) is launched, its requirements will be supported by most new consumer PCs, and it will incorporate many of the remaining bells and whistles from the consumer platform. It will be available as a minor upgrade for existing Windows 2000 users, and a major upgrade for those still running Windows 98/ME. Two versions are expected — Professional and Personal — with the more expensive Professional version having more security features and support for Intel's new 64‑bit Itanium processor. Microsoft are hoping to launch both in the latter half of 2001. A beta 1 version has recently been released for industry testing, and even at this early stage initial feedback is reasonably positive. One of the biggest issues being addressed is software and hardware compatibility. To be widely accepted by the consumer market, Whistler needs to be able to run a much wider range of soundcards, graphics cards, modems, and applications than Windows 2000 currently can. Microsoft have already confirmed that many more applications like games will run on Whistler than on Windows 2000, and the former also has a Compatibility Mode that makes an application think it's being run on Windows NT 4.0 or Windows 95. This makes things a lot easier, although you should be wary of using this to run low‑level utilities.
Ultimately, the Personal version of Whistler should be the perfect operating system for home and small business users, and because it's mainly intended for consumers it will also be significantly less expensive than the current £250 Windows 2000 Professional. If things turn out as Microsoft plan, Whistler could become the most popular as well as the most reliable version of Windows for the musician, as well as knocking the twin problems of hardware and software compatibility on the head once and for all. We live in hope...
Although Microsoft suggest a minimum of a Pentium 133MHz processor and 64Mb of RAM to run Windows 2000, most industry experts think a 200MHz Pentium and 128Mb a more sensible recommendation. You'll also need a hard drive at least 2Gb in size, with 1Gb of free disk space. The full version can obviously be installed on a blank hard drive, but you can also upgrade from Windows NT 3.51 or 4.0, or Windows 95/98. You can't upgrade from Windows 3.1.
One decision that needs to be made before installing Windows 2000 is how to format your hard drives. Windows NT used a hard‑drive file system named NTFS (New Technology File System), which was different from that used by Windows 95/98, although it could also recognise the FAT16 formatted drives used by Windows 95. NTFS provides more security for a system that will get used by various people, by keeping multiple copies of its master file table, to protect against corruption and data loss. FAT32 was introduced in Windows 95 OSR2 and wastes far less space than FAT16 by having a variable cluster size, and this is what most people now use with Windows 98.
Windows 2000 recognises both NTFS and FAT32, and during an upgrade from Windows 95/98, the installer will ask you if you want to convert the partition to NTFS. If you're going to run it alongside Windows 95/98 as a multi‑booting system you will have to use FAT32, or your common data be invisible when running the Windows 95/98 partition. Conversely, if you already have an NT 4.0 installation and want a multi‑booting system you'll have to choose NTFS, although you should be aware that Windows 2000 upgrades the drive to NTFS version 5. This can still be used from NT4, but its CHKDSK drive error checking utility won't work any more.
If you have need to retain compatibility with Windows 95/98, most experts agree that NTFS is a better file system than either FAT16 or FAT32, although a few people have found its hard‑drive read/write performance under Windows 2000 worse than either NT or 98, especially with older motherboard chipsets. If you're in any doubt about changing the partition type during the install, choose the 'No Changes' option, since you can always change your hard‑drive format afterwards using the Convert utility, or with a third‑party application like PowerQuest's Partition Magic.
If you are intending to upgrade, here are the things to check before you get out your credit card.
- Make sure that your motherboard BIOS is Windows 2000‑compatible, and if not, whether a flash update is available. This should ensure that features like Power Management work correctly.
- Make sure all your hardware expansion cards have Windows 2000‑compatible WDM drivers by looking on the manufacturers' web sites. You can install these before you upgrade.
- Failing this you will need Windows NT 3.51 or 4.0 drivers, although these will only provide emulated DirectSound support.
- Check that all your software is Windows 2000‑compatible.
There are some specific areas to watch out for when attempting to run existing software under Windows 2000. Because it handles CD‑ROMs differently, CD‑R software that worked in Windows NT 4.0 or 98 has to be updated, so you should double‑check to see if a new version has been released. You should be very wary of any utility program that changes things at a low level, such as a hard‑drive defragmenter. Those written for NT or 98 won't work, although Microsoft do bundle their own. By their very nature, virus‑checking utilities are specific to each operating system, so you will definitely need a new version.
If you want to take advantage of Windows 2000 multiple processing you will also need:
- A motherboard with dual‑processor support.
- Software applications that are multi‑threaded, and preferably optimised to give optimum performance with multiple processors.
As mentioned in the main text, Windows NT was generally avoided by music developers because of its poor MIDI timing, and was especially bad when users attempted to run applications that used code intended for Windows 95/98. The situation with Windows 2000 is much better, since its design ensures much more stable timing, but once again only if the application has been written using Windows 2000 APIs (Application Program Interfaces), and not 95/98 ones. Splitting tasks between two processors doesn't seem to make MIDI timing any worse either. The most important thing to ensure is that your MIDI interface has Windows 2000‑compatible drivers, and there are still not too many of these around!