A case for 64 bit Windows

With 64 bit computers becoming mainstream (its hard to find a machine thats not 64 bit capable these days) the question being asked more often is “Should I install a 64 bit operating system or a 32 bit operating system?” While the advantages of a 64 bit OS might be more obvious for those running native 64 bit software, or those who have a need for more than 2GB of memory, its a bit more murky as to whether regular 32 bit applications will also perform well on a 64 bit OS.

A 64 bit capable processor has an increased number of registers, as well as an improved floating point unit (FPU) design with double the number of FPU registers. 64 bit operating systems can take advantages of the expanded register set available to a 64 bit processor. The advantages to this are fewer memory accesses for data that can be stored in registers, leading to faster execution of computationally intensive processes. You can read more about this from this Cakewalk X64 white paper from several years ago.

Here is an interesting benchmark showing that 64 bit windows can perform better even with the same hardware and software. Some of the benchmarks are relevant to digital audio processing as well and so apply to DAW users.

Windows Vista Benchmark: 64-Bit Faster Than 32-Bit

A Closer Look at 32-Bit vs. 64-Bit Windows

6 Replies to “A case for 64 bit Windows”

  1. Great Blog Noel. Is about time you made a statement about this. We few that have been saying that Sonar runs better on a 64 bit OS have had to listen to those that insist that the only advantage is memory size. Those of us that have taken the 64 bit route have found that that is just a very nice advantage but not the only one.

  2. My ears told me that when I use 64bit on a REVERB (and the reverb supports this….) (Sonar 7,8,8.5 does native 64bit (even on 32bit (the audio pipeline uses double accurracy (if selected) (This is available in Sonar 64bit and Sonar 32bit) it sounds more reall. I assume this is because reverbs are numbers crunched with your 24bit audio data wich can lead to long numbers. The longer those numbers the more accurate the “reverbtail”. And because a native 64bit computer probably needs “less resources” (I mean in a 32bit you call it ; “double precision” In a 64bit it’s just …. native resolution?) it stays a snappy OS even when playing back an intensive project….

    Anyway, a 64bit platform wants much more RAM and the coming combination of 4/8/16GB ram in a mainstream DAW together with Solid State Disks will lead to snappier OS’es and the artist will have the ability to monitor from within the DAW because of fast real time processing so the listener/performer does not notice latencies. So prices drop and the customer will smile.

    With the new Lexicon PCM VST plugin I wonder if people will notice “better tails” on a 64bit system compared to a 32bit system…. Let’s wait and see.

    Sonar 7 (and soon 8.5) on my X2 with Vista64 feels snappier then when I was using WINXPPRO.

    1. Hi Frans,

      Yes, using a double precisicion (64 bit) audio stream on a 64 bit capable reverb will generate audibly smoother tails. Thats a common benefit.

      Some clarifications: Its important to understand the distinction between 64 bit math VS 64 bit addressing.
      Any 32 or 62 bit process on both 32-bit and 64-bit operating systems and are capable of doing 64-bit math, but only a 64 bit operating system can do 64-bit addressing, as shown in the table below.

      ————————————— 64-bit-math ——– 64-bit-addressing ——— FPU benefits

      X64, 64bit process ………………… x ………………………… x …………………………….. x
      X64, 32bit process ………………… x
      X86, 32bit process ………………… x

      On a X64 OS and running a native 64 bit application you get the maximum performance benefit when doing 64 bit math due to the extra 64 bit FPU registers. Note however that there is no SONIC benefit by using a 64 bit application. i.e. assuming your computer has the horse power to process all the 64 bit audio streams it should sound bit identical to a 32 bit OS/App doing the same. You can verify this for yourself by rendering audio to a wave from both platforms and then doing a phase inversion test on the results from both.

Leave a Reply to John Cancel reply

Your email address will not be published. Required fields are marked *