Futuremark 3DMark2001 SE build 330
The ASUS A8N32-SLI Deluxe is quite distinctly slower than the A8N-SLI Premium. The difference of 508 points can’t be regarded as a measurement error. So this is the first time the new chipset from NVIDIA meant to build a SLI platform with two full-width PCI Express x16 slots performs considerably slower than the older version of the chipset which only provides 8 PCI Express lanes per each graphical slot.
Futuremark 3DMark05 build 120
The newest version of Futuremark’s benchmarking suite runs at the same speed on both mainboards because the graphics subsystem performance is the main speed-limiting factor here. The main architectural deficiency of the nForce4 SLI X16 is not so conspicuous, even though the ASUS A8N32-SLI Deluxe is formally slower than the A8N-SLI Premium.
The CPU performance test from 3DMark05 makes it clear that the ASUS A8N32-SLI Deluxe is the slower of the two mainboards. In this test the central processor has to process some vertex shaders and to calculate the flight trajectory of the airship. The amount of data transferred between the central processor and the graphics subsystem is probably so large that the narrowness of the HyperTransport bus between the CPU and the chipset begins to affect the performance of the system, even though slightly. This is about the same thing that we can expect to happen in older games that do not use shaders, e.g. in Unreal Tournament 2004 . You’ll see the results of our gaming tests shortly while next go WinRAR and PCI Express Test.
The system bus – HyperTransport plays this role on the AMD64 platform – is loaded fully during a data-compression task, so it is natural that the ASUS A8N32-SLI Deluxe is slower than the A8N-SLI Premium. The gap isn’t bigger than 10%, but it’s a fact that the narrower HyperTransport bus connecting the CPU and the chipset has a negative effect on the performance of the nForce4 SLI X16.