NVIDIA nForce Chipset Preview
IGP - Integrated Graphics Processor
This content was originally featured on Amdmb.com and has been converted to PC Perspective's website. Some color changes and flaws may appear.The IGP (Integrated Graphics Processor) of the nForce chipset is most similar to what is commonly known to PC enthusiasts as the North Bridge of the chipset. This part of the nForce chipset holds the components that run the memory, video, and processor functions.
TwinBank Memory Architecture
The memory controller that NVIDIA has built into their AMD chipset is called the TwinBank Memory Architecture. Using this revolutionary memory controller, NVIDIA can offer the manufacturers the option of a standard 64-bit memory bus or an advanced (and a bit more expensive) 128-bit DDR Intelligent Memory Interface. This will essentially allow the motherboards that feature this option to boast a 4.2 GB/sec maximum memory bandwidth, higher than any current available system for any home platform.
There were many questions initially on how NVIDIA can provide a 128-bit memory interface without introducing a new RAM standard. Obviously, NVIDIA is not going to do that, so their claim to have 128-bit memory performance is really just a dual 64-bit memory interface. There are two, completely separate memory controllers in the IGP-218 (version with this option) and that of course requires the installation of at least two sticks of memory.
While at first guess you may think this is making us return to the days of SIMMs, EDO and matched pairs, you’d be wrong. In fact, the nForce IGP-128 only supports 3 DIMM slots, making it impossible to have matched pairs, actually. However, in order to have the dual memory architecture to work you need to have one DIMM in the first bank and at least one in the second. Interestingly and excitedly, they do not have to be the same size or speed. This enables most users to convert into the NVIDIA nForce motherboards without too much of a problem. Of course, you should expect to get the best system performance from a system that uses the same memory in each bank.
DASP – Dynamic Adaptive Speculative Pre-processor
The acronym DASP is NVIDIA terminology for their pre-processing feature that is built into the IGP north bridge. Simply put, the DASP on the nForce acts much in the same way as the prefetch features do on the Palomino and Morgan core AMD processors. NVIDIA is the first chipset manufacturer to offer this option on a chipset and that means that unless NVIDIA licenses it out (they have a patent pending on the technology), we may not be seeing it on any future ones either.
The benefits of the DASP feature will be most easily seen in programs that allow the DASP to accurately predict the continuous memory calls. Streaming media will see the most actual rise in performance here, anywhere from 5-15% depending on the applications. Synthetic benchmarks that test memory capabilities will also show good increases, perhaps in excess of 20%, over the current leaders.
For those techies who want the word straight from the horse’s mouth, here is the official tech doc clarification on the DASP feature:
DASP is an intelligent agent that monitors CPU requests and looks for access patterns that it can successfully predict. When it recognizes such access patterns, it exploits unused memory bandwidth to load its cache with data the CPU is expected to request later. When the CPU requests the data, it is returned to the CPU immediately rather than after waiting for the memory access. For such requests, latency is reduced from 40% to 60%. DASP is carefully engineered to make efficient use of memory bandwidth while minimizing overall latency.
Also, below you will see a diagram and two graphs supplied by NVIDIA to demonstrate the power of the IGP’s DASP feature.