Intel's Hottest Quad Core Ever: Build A QX6850 PC Without Busting Your Budget
Here's how to roll your own computer equipped with Intel's top-of-the-line quad-core processor, along with CPU benchmarks and some lower-cost project alternatives.
Intel hasn't yet posted full documentation for the QX6850. However, the QX6800 datasheet will get you close. Most of the specs are the same, but the QX6850 is faster. It increases the clock speed of each of the cores from 2.93 GHz to 3.0 GHz.
Most importantly, the speed of the front-side bus has been boosted from 1033 MHz to 1333 MHz. The bus is used for interprocessor communication among the four cores and also as the link between the cores and memory.
Cranking up the bus is crucial for getting maximum performance out of a multicore design. With the front-side bus approach used by Intel, each core has to "ask" for access to the bus before it can communicate with memory. That process is called "contention," for obvious reasons -- the cores must "contend" for access to the bus, and sometimes they don't get it right away. When access is delayed, it takes longer to get required instructions or data out of RAM, a situation that's called increased memory latency.
This spotlights one area where AMD has been ahead of Intel, and which the smaller chipmaker has justifiably crowed about. Namely, AMD has used integrated memory controllers in its processors since 2003, when it launched its Opteron server and Athlon desktop devices. An integrated controller connects all of a processor's cores directly to the memory, thus cutting latency to the bone and holding it there, regardless of how many cores a chip has.
Interestingly, Intel's emphasis on cranking up the speed of the bus in the QX6850 hints at a future design direction. As cores proliferate, the load on a front-side becomes more of a stumbling block against maximum performance. Accordingly, Intel is expected to go to an integrated memory controller in Nehalem, a completely new microarchitecture that will succeed the QX6850's Core architecture in late 2008.
(click image for larger view)
Graphics comprise the most stringent portion of the benchmark tests.
No company makes cooler-looking mobos than Asus. They make liberal use of bypass capacitors -- a practice not always adhered to by Taiwanese motherboard makers in the early days. The caps direct any stray voltage to ground and make sure your processor stays within operating specs.
The P5K Premium board used in this project is also a harbinger of a new trend among high-end boards, in that it features built-in wireless. Personally, I prefer the assured high-speed connection of a wired link. Plus, living in an urban area, I envision wireless thieves lurking behind every lamppost. However, the feature certainly makes initial wireless set-up easier and will be welcomed by most suburban users.
Like all Asus mobos, this one is replete with cooper heat pipes. These help direct heat away from the processor and chipsets. Is this stuff really necessary? Well, a year into its life cycle, my previous build, a dual-core Athlon PC with Asus M2N32 motherboard is running just fine. That's all the more comforting, given that its processor runs hot, and I've got the box stuck in a location with poor external airflow.
The Achilles' heel of this project turned out to be its $60 graphics card, an NVidia GeForce FX 5200. It's a perfectly adequate card -- Nvidia's own site characterizes it as "mainstream." However, because graphics are key factor in benchmarks, it drags down the ratings of the overall PC.
This isn't fair; if you're not using the machine for gaming, hot graphics aren't a must-have. However, it's consistent with the way ratings work these days. The canonical case is the Windows Experience Index in Vista, where your score pretty much hinges on your graphics card.
2014 Next-Gen WAN SurveyWhile 68% say demand for WAN bandwidth will increase, just 15% are in the process of bringing new services or more capacity online now. For 26%, cost is the problem. Enter vendors from Aryaka to Cisco to Pertino, all looking to use cloud to transform how IT delivers wide-area connectivity.
The UC Infrastructure TrapWorries about subpar networks tanking unified communications programs could be valid: Thirty-one percent of respondents have rolled capabilities out to less than 10% of users vs. 21% delivering UC to 76% or more. Is low uptake a result of strained infrastructures delivering poor performance?