The PCI standard has more ins and outs than you might be aware of. We attempt to sort them out, as well as highlight a problem that can cause software authorisations to expire before their time.
A PCI card is a PCI card, right? Well, I recently reviewed ESI's new Julia soundcard, and among its list of features is compatibility with both +3.3-Volt and +5-Volt PCI slots. Since many of you may not be aware that there are differing standards out there to trap the unwary, here's a bit of background information that may save you buying a soundcard that possibly won't fit into your PC.
The PCI specification defines two basic types of expansion connectors that may be found on a motherboard — one for systems with 5-Volt signalling levels, and the other for systems using 3.3-Volt signalling levels. This specifies the voltage level of the various clock and timing signals, but not necessarily the power supply voltage. A particular card may require both 5-Volt and 3.3-Volt power supplies irrespective of its signalling-level voltage.
The most common PCI cards are expansion cards whose edge connectors have a notch at pin positions 50 and 51 (about 2.5 inches from the end of the connector nearest the backplate — see photograph), which allows them to be plugged into 5V slots. The 3.3V expansion cards have a notch at pin positions 12 and 13 (about half an inch from the end of the edge connector nearest the backplate). These keying notches prevent you from plugging in an incompatible expansion card and causing any damage. Most early PCI slots were 5V only, and didn't even provide power at the 3.3V position. However, 3.3V slots are now becoming more widespread, so with both notches on the Julia and, correspondingly, dual 3.3V/5V operation, ESI are claiming wider compatibility than the norm. Cards, like the Julia, offering dual-voltage operation are generally referred to as 'Universal Cards'.
Voltage apart, there are quite a few other PCI-slot variations. The normal PCI buss (whether operating at 3.3V or 5V) is 32 bits wide and runs at 33.3MHz (normally quoted as 33MHz), offering a maximum bandwidth of 133MB/second. One way to increase this maximum bandwidth is to use a faster clock speed, of 66.6MHz, but to operate at this increased frequency both the slot and card must be designed to support it, as outlined in the PCI version 2.3 standard. The other way is to increase the buss width to 64 bits, as defined in the PCI 64 standard, resulting in a much longer slot. The first part of a 64-bit slot is identical to the 32-bit one, with a 5V key or a 3.3V key, or both, but has a 64-bit extension.
PCI-X (eXtended) is a new, higher-speed version of the PCI buss jointly designed by IBM, HP, and Compaq. A new PCI specification was submitted way back in 1998 that proposed a further increase in PCI buss speed, to 133MHz, so PCI-X consequently ups the clock speed from 66MHz to 133MHz. Version 1.0 of PCI-X specifies a 64-bit connection running at speeds of 66, 100, or 133MHz, while PCI-X 2.0 adds PCI-X 266 and PCI-X 533 speeds, offering transfer rates of up to 4.3GB/second — 32 times the speed of the original PCI buss.
PCI-X also features enhanced reliability through error detection, and it is backwards compatible, so you can install a PCI-X card in a standard PCI slot, although its performance will be throttled down to 33MHz. You can also install both PCI and PCI-X cards on the same buss, but the buss will then run at the speed of the slowest card. Believe it or not, you can even install a 64-bit card in the shorter 32-bit slot, as long as its notches are in the right place on the edge connector, although (once again) its performance is likely to plummet as it runs in 32-bit mode.
Over the years I've probably installed as many different audio interfaces as anyone anywhere, but I never assume I know better than the manufacturer when it comes to installation. Unfortunately, a few other musicians think they do, which is why forums are cluttered with requests for help when a new interface isn't recognised properly by Windows or doesn't work once the drivers are in place.
The simple rule is this: always carefully read the manual and any 'read me' documents for installation instructions before doing anything else. Many interfaces (particularly those of the Firewire or USB variety) require their drivers to be installed before the hardware is plugged in, for correct operation. This is often because Windows may partially recognise the new hardware as something already supported by its bundled generic drivers and attempt to install them automatically. If you have the most suitable and up-to-date drivers already in place on your hard drive you're far less likely to run into problems when your new hardware is recognised.
Even where an audio interface manual advises you to install the hardware first, it may still instruct you (after rebooting your PC) to ignore Windows' own search for compatible drivers and instead run a specific EXE (executable) file on the bundled CD-ROM. This will not only install several required drivers but also various other utilities needed by the interface.
By now, many of you must be shaking your heads in disbelief, but don't worry — I've never seen any 64-bit slots or expansion cards either, and none of the motherboards in any of the PCs I've had during the last 13 years have featured any PCI slots other than standard-width 32-bit ones running at 5V, with a 33MHz clock and 133MB/second maximum throughput.
Initially, PCI-X expansion sockets were reserved for high-end servers and workstations featuring dual Opteron and Xeon processors, rarely appearing on motherboards for the masses. However, AMD users will have seen the longer 64-bit 66MHz slots alongside 32-bit 33MHz ones on more recent motherboards (since 2002) featuring chipsets such as AMD's own dual-processor 760MPX. More recently, Intel Pentium 4 users could get in on the act with expensive motherboards such as Supermicro's P4SCT range, which (as well as providing three PCI-X slots) uses PCI-X to connect its Serial ATA/150 hard-drive controller. Such slots offer four times the normal PCI bandwidth (533MB/second) and benefit expansion cards using speedy technologies such as Ultra2 SCSI and Fast Ethernet.
Many Mac users will also already know a lot more about slot types than we PC musicians do, since some configurations of the G5 provide three PCI-X slots, one running at up to 133MHz and the other two at up to 100MHz. These slots support universal and 3.3V signalling, but not 5V signalling, which is why the latest soundcards intended for both Mac and PC use (such as ESI's Julia) must provide both notches in their edge connectors. Unfortunately, if any of your existing 32-bit PCI cards are 5V only (like my Echo Mia, shown in the photo) they can't be plugged into these 64-bit slots.
Why haven't we PC Musicians come across more about PCI-X in our travels to date? Well, soundcard manufacturers are generally quite happy with the 133MB/second bandwidth of the 32-bit 33MHz PCI buss, which can already support a vast number of audio tracks. A single mono, 16-bit, 44.1kHz track only requires 88200 bytes per second, so you could theoretically run over 1500 simultaneous tracks before saturating this standard PCI buss. Mind you, all the other devices on the PCI buss have to share the same bandwidth, and there's plenty of potential for PCI latency timing conflicts between these various devices, as I discussed in PC Notes October 2004. In addition, those running multiple DSP cards (such as the Powercore and UAD1) alongside their multi-channel soundcards often seem to push the boundaries.
PCI-X may be of more interest to musicians running video alongside audio, but I suspect that the real reason we've seen so few PCI-X peripherals is the final arrival of PCI Express, which I first mentioned in PC Notes way back in November 2002. This serial version of PCI has a tiny slot just one inch long (so no, you can't just plug any PCI card into it). As I discussed in my PC Notes column in the September 2004 issue, PCI Express slots are already appearing alongside the old 32-bit, 33MHz PCI slots on motherboards featuring Intel's new 925X/915 chipset family.
I suspect that these new PCI Express slots will eventually take over from both 32-bit and 64-bit slots, although during the changeover 32-bit PCI slots are provided so that you can still plug in your existing expansion cards. PCI Express provides the huge advantage that each of its slots has access to an entire 'lane' of the PCI Express system, and won't have to share the 512MB/second bandwidth with other devices on the buss, which is excellent news for musicians. In the meantime, if you're thinking of buying a PC with PCI-X or PCI Express slots, make sure it also has enough standard 32-bit slots to house your collection of 5V-only PCI cards.
This month I suffered a really annoying PC problem that had a simple cause and solution but far-reaching ramifications. After routinely rebooting from my Internet-enabled partition into my special 'XP Review' partition to audition a new soundcard, I discovered that all of my Waves plug-ins refused to run because their six-month authorisation had expired.
Since I'd only installed them a couple of weeks previously I knew that something had gone wrong, and I initially suspected that installing the Windows XP Service Pack 2 (SP2) had somehow interfered with Waves' Pace copy-protection system, which I've never had problems with in the past. To check, I rebooted into my 'XP Music' partition that still had Windows XP SP1 installed, to find exactly the same problem. However, during further investigation I discovered a file on my PC stamped with the correct day, month, and time but dated 2005. Using Windows' Find function I discovered quite a few more, and determined that at 2.30pm that afternoon my system clock had jumped ahead by one year — and this was why my time-sensitive plug-ins had expired.
I certainly hadn't altered the clock myself, so my next suspicion was that the PC had contracted a virus (among many other possible effects, your system time or date may be affected). However, a complete trawl of every drive using my virus checker with its latest database update showed no infection. Although it was only a couple of hours before I noticed the system year alteration and changed it back from 2005 to 2004, by then I'd lost time-sensitive authorisations to a lot of expensive software. I'm still no wiser as to the cause, but fortunately Waves were sympathetic to my situation and renewed my licence. If something similar ever happens to you, it may be possible to restore an image of the affected Windows partition saved before the date change. However, Pace copy-protection remains impervious to such chicanery.
Occasionally, you may find disk utilities such as Scandisk reporting invalid time-date stamps and offering to change them for you, but again you should be cautious — some other copy-protection systems rely on post-dated files (I've got one dated 24/01/10 on my PC), and may permanently refuse to launch their applications after any change.