Over the next several days, you'll be hearing a lot about Intel's significant upgrade to the Pentium 4 platform. Soon enough, that brand new Canterwood board you have will be yesterday's news as two new words will be on the lips of all enthusiasts... Grantsdale and Alderwood.
Well, after that dramatic intro, I suppose the real question is what exactly is the Grantsdale and Alderwood all about, and more specifically, why should you the reader care? Despite the drama, the D915P/G and D925X (Grantsdale and Alderwood respectively) really are large jumps in the Pentium 4 platform's history, and offer some serious features that will be impossible to ignore. Since this is a new technology, we'll break this article down in a FAQ format so that it (hopefully) answers any questions you may have.
What are Grantsdale and Alderwood?
Both the D915P/G and D925X are new Express Chipsets designed for the Pentium 4 platform. The chipsets support Intel's HyperThreading technology, Dual Channel DDR2, and PCI Express. They will support wireless access point (WAP) onboard, which Intel claims that along with a wireless card, you can setup a wireless network in four simple steps. I personally doubt this will be secure out of the box, so expect a few more steps than that in order to secure that wireless network.
Intel High Definition Audio, which was something mentioned in our interview with them last month, is a new technology that improves greatly on the onboard audio we were used to before. Of course, RAID is supported by the ICH6R, making the need for 3rd party controllers unnecessary. However, you can expect some of the high end boards to still provide this as an option.
How are they different?
The Grantsdale is geared towards the mainstream/professional user, while the Alderwood is targeted more towards the enthusiast. Though both boards share similar features, there are a few differences that distinguish them from one another. In terms of memory support, the Alderwood only supports DDR2, whereas the Grantsdale will support DDR and DDR2. The Grantsdale will also offer onboard graphics as an option in the form of the 915G.
As you've probably figured out now, this two tier release is similar to the last one we've seen from Intel. Basically, the Springdale market is moving to Grantsdale, and Canterwood to Alderwood.
What is this PCI Express?
The story of the day is going to be PCI Express (PCI-E). Before continuing, let's have a look back at a bit of history.
ISA - Back in the days of single MHz speed PCs, the Industry Standard Architecture (ISA) was the main connection within your PC. Everything from sound cards to video cards went in here. It began as an 8-bit and later expanded to 16-bit, running at whopping speeds between 8 and 10MHz.
Considering the speed of the PCs back then, and their purpose, the ISA standard was fine. Forgotten from this time was the Extended Industry Standard Architecture (EISA) which allowed for 32-bit transfers, but it wasn't any faster than ISA and it was expensive to implement.
VLB - As PC gaming began to take shape, the Video Electronics Standards Association (VESA) introduced the VESA Local Bus (VLB). In simple terms, the VLB was a 32-bit version of the ISA connection and was required for the newer graphics cards at the time. The VLB ran at speeds between 25 and 50MHz.
The VLB ran synchronously with the processor (the 486 at the time), and shared the bus between the VLB and CPU. The problem of course is something similar to what we experience today when the AGP and PCI slots get overclocked. As CPU speeds increased, so did the device plugged into the VLB.
PCI - Version 1 of the Peripheral Component Interconnect (PCI) was first proposed by Intel back in 1991, followed by version 2 a couple years later. It is a 32-bit bus and supports speeds of 33 MHz and 133MB/sec, with later specs allowing 64-bit transfers at 66 MHz. It was the first plug-and-play connection that would automatically configure IRQs and other settings, making the addition of new devices much easier.
Using bus mastering, PCI adapters are able to perform tasks at the same time with the main processor. PCI has the ability to run asynchronously from the processor, so faster CPUs will not put the PCI slots out of spec. Although the PCI bus was separated from the CPU, it did share the bus with other PCI devices. With PC games becoming ever more popular, a change was needed in order to maximize the graphics card's potential.
AGP - The Accelerated Graphics Port (AGP) is exactly what the name implies. Designed by Intel, this connection was based on the 66 MHz PCI specification and runs at 266MB/sec, though modern cards now run 2.1GB/sec. It is a dedicated point-to-point channel, and instead of fighting it out with other PCI slots, the AGP slot has it's own direct link to system memory.
One of the early advantages of AGP was that game textures can be retrieved from system memory, and in theory, video cards would require less memory (making them cheaper) as they did not need to store these textures anymore. I think we know how all that went, but nonetheless, the dedicated high speed connection was required eventually, and the faster connection to system memory does come into play when your video card buffer fills out.
Why the need for PCI Express?
PCI Express is designed to replace PCI and AGP as we know it. There are several advantages to PCI Express but the main thing is to get rid of all the bus traffic we have right now. As it stands, moving graphics to its own AGP slot alleviated the graphics traffic that overloaded the PCI bus. However, hard drive data, networking traffic, sound data, as well as any peripherals plugged into your PC all eat away at the same 133MB/sec the PCI bus allows for. A simple analogy would be as follows:
Imagine the above chart as back in the days of ISA and PCI. You have four lanes of traffic merging into one, hence the bottleneck will occur at the merge. By moving to AGP:
By moving graphics data to it's own AGP, we take away one of the contributors to the PCI bottleneck. However, we can still see there's still a potential traffic jam once data begins to flow.
The above image illustrates the type of flow we can expect from PCI Express. I've simplified it, but basically data will no longer share one path, but rather be split into multiple paths. The standard calls for something similar to dual channel, but in the case of PCI-E, there is a transmit pair and a receive pair. PCI-E speeds will run up to 2.5GB/s with a transfer rate of 200MB/s. For those of you keeping score, that is up to four times faster than PCI.
Costs should be on par, if not lower to replace PCI slots. Since PCI Express is point-to-point, the number of trace routes are drastically reduced. This does not mean PCI will be gone by next week, as there is still an adoption period, but both PCI and PCI Express are capable of coexisting together. At the software level, nothing has really changed, so you should have no problems running a PCI Express system with the current Operating System you're using now. For Intel's launch, expect to see the review boards equipped 16x PCI-E slots instead of 8x AGP slots. Later on, there may be some motherboards where you will have an AGP slot, but this isn't native to the Grantsdale and Alderwood, but rather something like a bridge over the PCI bus.