By Merle Nicholson, Tampa PC Users Group
merle@merlenicholson.com

The latest kinds of PCs aren’t made by HP and Dell. If you want to really find out what’s going on you can look at Alienware and VoodooPC, smaller computer makers that do direct internet sales. Here’s what’s going on in the PC industry. All I describe is commonly available, but just not necessarily mainstream.

Dual-Core Processors: Both AMD and Intel are producing CPUs with two processors. Intel’s workstation CPUs are not available yet, but will be soon. I’m much more familiar with the AMD product, but suffice it to say that Intel’s claims are just as emphatic as AMD’s that their design is best. Of course, you can go to various - usually gamers’ - websites and read extensive performance comparison tests. The gamers tend toward the AMD products because they are fairly easy to “overclock” - that is, run at higher speeds than they were designed for. The ultimate gamers’ CPUs, however, seem to be the AMD 64 FX single core CPUs. There are two available; the most powerful (FX57) runs at 2.8GHz, has 2x64 and 1x1MB cache at $1,100 or so, including just CPU and fan. Games are rarely multi-tasking just yet, and gamers tend to do only one thing at a time.

Dual-Core processors seem mostly to have an advantage in multi-tasking environments where other things are going on in the background. You know how it is. You get something started on the computer and then just relax and fold your hands because trying to do anything else is useless or will even jeopardize your project. With dual processors, well, you have another processor! Go ahead, do more stuff. AMD’s processor is called Athlon 64 X2 Dual-Core, and ranges in performance ratings from 3800+ to 4800+. The prices range from about $430 to $1100. That’s processor and fan only. Intel’s prices are supposed to be lower when they’re available. Offsetting some of the processor cost for AMD is built-in support for common and cheaper DDR 400 memory. The on-chip memory cache on X2 processors is very large. L1 Cache on some models is 128KB, L2 is 1MB, but then multiply those times two!

PCI-E (PCI Express): This is a new version of PCI device expansion slots that greatly improves on the standard old PCI expansion slots. They come in flavors of PCI-E x1, PCI-E x2, PCI-E x4, PCI-E x8, and PCI-E x16, with bandwidths of 512MB/s to 8GB/s. The PCI-E x16 is used for video cards and is now common enough that the price differential over AGP is negligible. This interface provides 3-1/2 to 4 times the speed of AGP. In addition to that, many boards have SLI support that takes two PCI-E x16 slots for two video cards sharing the graphics load at a ludicrous graphics speed. So this is the latest in fast 3D video. ATI and NVIDIA are the major players here so far. Many of the GPU (Graphics Processors) on these cards are overclocked, so all of them have fans and large fan shrouds on the card. You should research these thoroughly to make sure the fan noise is acceptable. Some expensive notebooks are using PCI Express graphics. There are a few add-on cards, drive controllers, network interface cards and firewire cards using PCI Express.

Serial ATA (SATA I & II): This is a method of connecting hard drives to your system. The past method (now called PATA-Parallel ATA) is limited to 133Gb/s burst rate. This is the highest momentary burst it will take and in reality is not achieved very often. The sustained rate is more important, but the burst rate is an indication of the ultimate transfer rate. SATA-I will do 150Gb/s, comparable to SCSI, and the SATA-II specification provides for 300Gb/s. Many older PATA systems are 33Gb/s and 66 and 100 are most common.

SATA drives are common enough now that there is little or no price differential. They are physically distinguished by the small diameter red data cable that replaces the wide flat cable. Most new ones also use a different power cable. Many SATA drives on the market are just PATA drives with some conversion interface to SATA built in. One indication of that is that many will also have a 4-pin power connector. The best drives are those designed as SATA from the start. The most famous and the oldest is the Western Digital 36GB, 10,000RPM Raptor, but its newer 74GB counterpart is the fastest of all SATA drives. They’re both relatively expensive on a dollar per byte basis.

Maxtor has two new drives designed from the ground up, called MaxLine III, in 250 and 300GB, 7200RPM, but with 16MB cache. They have NCQ (Native Command Queuing) capability. And Hitachi is doing new things using SATA-II.

NCQ: Native Command Queuing is simply the capability of a disk storage system to re-order each request for efficiency, instead of supplying data in the order requested. The idea is to minimize the disk head movement, saving time. In a single-user, single-threaded environment, NCQ doesn’t supply too much advantage because the requests for data are mostly sequential. But with servers – and multiple core processors, there’s a very real advantage. The hard drive controller must be designed for NCQ as well, and so far that’s harder to find. Look for NVIDIA motherboard chipsets using RAID that may have support for NCQ. Also some hard drive controller cards have it.

RAID: RAID has been around in the SCSI world forever, nothing new here. But RAID has finally made it mainstream because of the availability of SATA on common motherboards. So you can expect many new motherboards to have SATA RAID support in RAID configurations of 0-Striping, 1-Mirroring, 0+1-Striping+Mirroring and JOBD (Just a bunch of disks). Some new motherboards provide RAID support for PATA.

For pure speed, RAID 0 is used with two new identical hard drives. The two drives become one logical drive, and the files (and the load) are divided between the two. It’s very fast. RAID 1 mirrors all the data on one to the other also, providing redundancy. And that’s not fast – or slow for that matter. But for pure speed, RAID 0 uses two separate cables and doubles the bandwidth to the drives.

Dual Channel Architecture Memory: This speeds up data flow in and out of memory. The motherboard will have two independent channels to memory in hardware. This is the format of NVIDIA’s nForce chipsets for AMD Athlon XP systems and Intel’s i850E, where each memory bank has its own memory channel and an arbiter distributes the load between them and plays traffic cop for incoming data. You must have two compatible (identical) memory sticks. It’s best to buy them packaged as a pair. The memory controller recognizes them on boot-up.

Motherboards: There are a surprising number of motherboard manufacturers that are trying to supply the needs of hobbyists and of computer companies like Alienware and VoodooPC. Many appeal to gamers, who apparently are willing to spend the extra to have leading-edge fast computers. There are four or five chipset makers for these; ATI and NVIDIA seem to have gotten the jump on everyone. VIA and SIS are there too. ATI and NVIDIA are of interest to gamers because they’re familiar with the brands through their leading-edge video cards, and the hope that some compatibility will exist between the motherboard and video card. So the things to look for are the subjects in this article: Dual-Core, SATA, NCQ, RAID, PCI-E and SLI.

Cooling Systems: The new larger processors do generate more heat. AMD and Intel both are developing manufacturing methods to reduce the thickness of the wafer (the new ones are 90 nanometers), which reduces heat requirements. And you may have noticed that notebooks get special attention with slower processors (less computing power) and sophisticated CPU speed controls.

The idea in cooling systems is three-fold. One, it’s a good goal to reduce the fan noise as much as possible; two, some people overclock their systems to the point where ordinary fans can’t keep up with the increased heat and risk burnout; three, it’s neat and clever.

The high capacity cooling systems center around liquid cooling, where, in the places you would have fans and/or heat sinks, you mount a “water block” - just a little box the right size clamped down with two pipes to attach plumbing. They don’t really use water but instead use some kind of coolant whose heat retention properties exceed that of water. In any case, these water blocks are interconnected with plastic pipes and finally wind up connected to some radiator device that also contains a pump.

The radiators are mostly external to the PC. Not all though. One mounts inside, where the normal air exhaust fan would be. Most radiators have a large diameter fan that moves slowly with as little noise as possible. And, of course, it can get more sophisticated, with temperature controls, readouts and variable fan speeds, depending on how much cooling is actually required.

There are other things in the cooling arena. The most interesting is CPU cooling, where the radiator can be of immense size, accommodating a large slower-moving fan. Also of interest is the use of “heat pipes,” which are liquid-filled metal piping that moves the heat out to larger-finned radiator devices. These work well and are in the neighborhood of $50-$70. There are case fans designed to be quieter.

Building your own system: You need to do quite a bit of research. Start with the processor and then the motherboard to support it, then look at the features you want on the motherboard. Then go to the motherboard website and research the compatibility lists. You’ll want to select the video cards and memory from their approved compatibility lists. They’re available on the manufacturers’ websites. Mostly. If you deviate, there’s some risk that it won’t work, or worse yet, have problems where you can’t identify the source.

The major manufacturers maintain extensive documentation, FAQs, and even model-specific forums for users to exchange information. u