|
SX and Nuendo both support the Quad core cpu you left a link to but, at this time there is some problems.
This was written to explain it a bit better:
I posted this over at N.com, but I'll repeat it here..
There were 2 very distinct and separate areas brought to light navigating the issues presented by the Dual Quad adventure.
1. Nuendo places a a primary resource loading- both VST and CPU - when enabling large numbers of I/O, I believe its a buffering/ resource allocation mechanism. I don't see that as bug. It has an accumulative effect as the core count increases. This I see as a bug :-)
I suspect that has always been the state of play , the move to Quadcores has simply highlighted the fact to a greater extent. This is witnessed in the disproportional delta between VST/CPU meters. Again, we need to remember that the VST meter is not purely a CPU meter, it is a VST resource meter - what exactly that entails is the question . Also we need to note that the issue is consistent whether the I/O's are physical or virtual. Nuendo still reserves resources in the same way.
2. There is a distinct loss of scalability at lower latencies , proportionally more substantial than being witnessed from 2-4 Cores on the Single CPU Platform. Also , switching MP off resulted in substantially higher VST readings / TM readings , severely hampering low latency performance-that in combination with the first portion of the bug in turn translated to the system being hobbled below 1024 samples with the high number of I/O. With lower numbers of I/O , the behaviour is not as severe until below 128 samples, but having said that, whereas on a Single Socket system, moving from a 2-4 Core configuration resulted in a respective scaling of 52%-34% for the latencies from 256-032, on the Dual Socket systems it resulted in a scaling of 27% - 05% on the latencies from 256-064 , and a 13% decrease at 032 samples.
Other multi threading audio apps do not suffer from the loss of scaling being witnessed - Sonar/Reaper. Why the behaviour is more pronounced on the Dual Quad configuration is something that only Steinberg can answer. It is directly related to how they implement the X-Scaling - meaning scalablity per core. If it is behaving differently on Quad Duals compared to Dual Quads, there is obviously more involved than only the X number of cores. This is no different to when the spectre first raised its head with the Dual Dualcore Opti's, when it was reportedly working on Quad Singlecore configurations, it shouldn't have mattered that the behaviour wasn't exactly the same. Thats when some lost sight of the forest.
|
