My apologies for this lengthy post, but this is complicated stuff (for me at least) and I would like to be as clear as I possibly can....
I'm a frequent user of VST effects in GS3.04 and have found the following behavior with hyperthreading (HT) enabled or disabled. Is there anyone that can confirm these observations so that I know its not my system that's acting weird?
First, in all cases disabling HT solves the problem. But...
For a typical setup I have one GPulse instance (Aux) and ~8 VSTs inserted (Channels). The type and order of the effects does not matter. The following then occurs:
1) With HT enabled, the VSTs stop processing after a while (time interval is random). They have to be reinserted to get them to work. GPulse instances just keep working throughout. I have never observed this with HT disabled. Also, the moment the VSTs stop is not clearly correlated to a particular event.
2) I noticed that since GS3.04 and with HT enabled, VSTs and GPulses are processed together on one of the two 'virtual' CPUs (HT makes the CPU think that its actually two separate CPUs. Sort of schizophrenic...) Now, the first time I checked this in the task manager the CPU power that is needed to stream the samples and do the mixing (msg32.exe) is cleverly provided by the second 'virtual' CPU. I actually noticed a (small) increase in the max. polyphony and the number of VSTs I can use at the same time. Nice, exactly as Tascam said it would...However, at other times when I checked (eg. after a windows restart), VSTs, GPulses and disk streaming were put on the same virtual CPU leaving the other one doing nothing! Variations are also possible: VSTs on #1, streaming #2 or vice versa, or both on #1 or on #2.
3) With HT disabled a particular set of effects takes up ~30% of CPU power. Opening the same project with HT enabled showed only ~18% CPU activity, while everything was working as with HT disabled (for as long as it lasts [see issue no.1]). However, with HT either disabled or enabled the GS3 CPU indicator consistently reads ~30%. Does HT handle VSTs almost 200% more efficient or is the Task Managers CPU meter wrong?
These issues make it very difficult for me to use VSTs with HT enabled on my system, even though the (potential) advantage of HT is convincingly clear to me.
Any ideas are much appreciated...
Win XP Pro
Pentium IV 3.0 Ghz HT, 800 Mhz FSB, Prescott
ASUS P4C800 Deluxe
2 x Corsair PC3200LL/DDR400 matched 512 Mb pairs (total 2Gb)
Seagate 80 Gb (system disk)
2 x Western Digital RAPTOR 37 Gb in RAID0 (samples)
nVidia Riva TNT64Pro graphics card (PCI)
RME HDSP9632 with AO4S-192