|
||
|
|
|
              |
|
 |
 |
||||
 |
|
|||
by William Van Winkle |
||||
| The trouble with working in a technical industry like ours is that only geeks understand our everyday jargon. Still, you can't hid from computing. Understood or not, these technologies permeate everyday life, and every once in a while people are forced to grapple with them. This is about the time when confusion sets in. With real geeks, you're safe. They're the ones who hang out on Web hardware forums and understand all the essentials, even though they delight in debating the merits of driver 102.84.20a versus 102.84.20b. The weekend geek is more dangerous. This is the guy who subscribes to two computer magazines and thinks he knows everything but in fact has developed an unhealthy passion for the technology itself and not how it can benefit his life. Weekend geeks will drop $4,000 on a PC, packed with the most five-star components they can research, and then hold it against you, the seller, when it either doesn't work or becomes obsolete in six months. And then there are the masses of lay people, the users who can surf and save but not configure or troubleshoot. These are our parents, spouses, teachers, preachers, stylists, grocers, neighbors, and friends. This is also the bulk of those who make up consumer PC sales. There's no need to worry about real geeks. In a technological nuclear winter, they'll be the ones sipping contentedly on microbrew stockpiles and playing Xbox in their underground bunkers. But the weekend geeks and lay people are the ones prone to confusion. The only difference is that the lay people will admit their fear and ignorance, which makes your job as a technological advisor a lot easier. Nobody wants to make a bad purchase. Nobody wants to be stuck with a flat and no spare on the info highway. This is why the rumors and rumblings that trickle down from the heights of geekdom can sow a lot of uncertainty and worry. People just don't know what to think. In a sense, you're a counselor as well as a system builder. Part of your job is to unearth this confusion, drag it into the open, and settle these issues for the benefit of your clients. To illustrate, I thought we'd address a few of today's leading lip-biters. Preparing for Longhorn This is a big one. I see so many articles discussing all of the cool translucent 3D effects Longhorn's UI is going to have, not to mention the multimedia functionality, collaboration tools, next-generation searching, and so on. As the Longhorn (now officially called Vista) hype builds, one might think that it will take a supercomputer stuffed into an ATX case to run the new OS, and such a presumption wouldn't be far off if you'd seen the original minimum required hardware list that emerged from WinHEC 2004: dual-core 4GHz CPU, 2GB RAM, 1TB hard disk, 1 Gbps Ethernet/802.11g wireless, and a graphics processor that runs three times faster than today's GPUs. (Mind you, even now I still have PC hardware manufacturers telling me that no consumer needs Gigabit Ethernet. To all you thick-headed product managers: wake up already!) You read this and think, "Good Lord, Longhorn is due out next year, and dual-core 4 GHz CPUs don't even exist yet! Whatever shall I do?" Wait, it gets better. So at WinHEC 2005, a new set of minimum requirements emerged: a modern CPU, 512MB of RAM, and a GPU capable of supporting the Longhorn Display Driver Model (LDDM), which should be covered by any DirectX 9-compatible ATI or NVIDIA card with the appropriate drivers. Now, remember when the original Windows XP system requirements arrived. Can you imagine running XP on a 300 MHz CPU and 128MB of RAM? Microsoft said you could—and those were the recommended specs. The minimum specs were 233 MHz and 64MB. I mean, I had an Evo notebook with a 700 MHz Pentium III and 256MB of RAM than was buckling under XP. I finally took pity on the poor thing and turned it into a Xandros Linux terminal. But if today's mainstream configs are the minimum recommended for Longhorn, should we make the old assumption that you'll need 2X to 3X the machine to get anything resembling decent performance? This is where the fear and confusion comes in. The reality is that Microsoft will do things a bit differently with Longhorn. The WinHEC 2005 specs actually refer to the "Longhorn-Ready" requirements. There will also be "Longhorn-Capable" requirements that aim at older CPUs, older GPUs, and only 128MB to 256MB of memory. The functionality and UI will be scaled back on "Capable" boxes and will probably look more like Windows 2000 than the glossy, glassy, translucent Aero interface of Longhorn proper, but both will be similar under the hood. In fact, Longhorn alpha testers are already reporting that the OS runs just fine even on CPUs such as AMD's Athlon XP 2000+. So this time it looks like the recommended hardware list may be pretty close to accurate. For now, assure your customers that the don't need to paralyze their buying process and wait for Longhorn, which will be both 32- and 64-bit out of the box. Especially if you sell a 64-bit CPU with your new systems, there's nothing to worry about. Two Cards, Two Slots, Go-Go-Go! I recently did a hands-on review roundup of about a dozen fairly high-end PC power supplies. The lowest stated wattage in the bunch was 480W. My goal was to see if these units were up to the task of powering today's and tomorrow's dual graphics platforms. NVIDIA's SLI is already out, ATI's CrossFire is starting to ship, and both Intel and VIA have two PCI Express slots on some of their latest designs. I thought about how dual graphics might get used in the future, and running a graphics-heavy application in the foreground with some moderately stressful CPU crunching in the background (say, for multimedia encoding or whatever) seemed plausible. So I built a system with a Pentium 4 3.8 GHz , four DDR2 modules, an nForce4 SLI motherboard, and two 6800 GT cards. Six months from now, these will not be impressive specs, but the systems going out now with such hardware will be called on to perform next year's work loads. Now, I did some digging and ascertained that my configuration had only about a 330W overall draw but pulled a peak amperage of 40A from the +12V rail. That's primarily from the CPU and PCIe cards. My test was simply to run Prime95 in the background, which put about a 50% load on my single-core CPU, and then run just one game test from 3DMark05 in the foreground to hammer the 6800 cards for a minute. Of the 12 power supplies I tested, all of which were from reputable manufacturers, two would not power up my system and another four locked up during my test. Incredibly, I even had a 650W power supply with four +12V sub-rails (13A+18A+16A+8A) fail this straightforward trial. Presumably, the seasoned geeks buying SLI systems today know enough to consult NVIDIA's SLI-certified hardware list (www.slizone.com/object/slizone_build.html) and pick a model from that, and if they don't then you should. (Note: OCZ's impressive 600W PowerStream was the only non-SLI-certified PSU to pass my testing, and I have it from OCZ that a cabling update will soon earn the model its certification.) But it won't be long until dual graphics slots migrate down from the enthusiast level, and that's when we'll start to see people trying to cut corners and pinch pennies. Confusion will set in as most buyers and many sellers won't understand the power-related underpinnings of the problem, and systems will start failing under what seems like ordinary usage conditions. So consider this a preemptive strike against confusion. Bone up on your power supply knowledge and, if necessary, run tests similar to mine to make sure your dual graphics systems really will perform as expected down the road. Performance or Platform? We all know the old cliche about no one ever getting fired for buying IBM. Now, it's about playing things safe by buying Intel. Intel has "the platform." That's what emerged from the RAM cover story I did recently on workstations. I asked the Intel PR guy to help me make a case for Xeon when AMD's Opteron was kicking all kinds of booty across the benchmark board. "You can read what you like into this or that score," replied Intel, "but we have a platform and they don't." AMD, of course, was like, "Whatever. We're kicking your Xeon booty. And your little Pentium doggy, too." This bickering rhetoric is all well and good, but the question that gets left in many buyers' minds is: Who's right? You can't tell a customer something stupid like "In the real world, that little performance edge doesn't matter." That would be like a car salesman telling you that three miles per gallon and a few cup holders really isn't a big deal. You'd walk and find another salesman more sympathetic to your priorities. I test a lot of motherboards, and this gives me a chance to try out the whole platform argument first-hand. If Intel is right, if having the CPU, chipset, motherboard, and whatever else all produced under one roof does bestow a benefit, I should see it when comparing against competitors that don't have that unity. And ultimately, I would say that Intel's argument has some merit, although it's not the great deciding force Intel likes to tout. In situations where I'm testing an Intel CPU and chipset on an Intel motherboard, the platform is rock solid, period. This is particularly true when it comes to the launching of new chipsets. In comparison, take an Intel CPU, pair it with ATI chipset (for example only; I'm singling out ATI for any specific reason), and plant it on a cheap, second-tier motherboard with a young BIOS, and you're just asking for nightmares. I've seen it time and time again. But this isn't playing fair. The weak link here is the motherboard manufacturer, not necessarily the compatibility between the CPU and chipset. For instance, I've never had a problem with an AMD processor and an nForce4 Ultra chipset backed by a first-tier motherboard. NVIDIA is justifiably manic about quality control across its board partners, and you could argue that the Athlon/nForce "platform" is every bit as good and solid as the Pentium/945 "platform"—better, in fact, if you figure in all of those killer nForce bonus features. Customers used to ask me whether Intel was better than AMD, and the question remains as inscrutable as weighing performance against platform. There are too many variables and priorities involved to arrive at a meaningful answer. For the minority of customers buying solely on the basis of performance in a handful of applications, a few benchmark tests should tell you what you need to know. Otherwise, you must push the customer toward discussing features and flexibility rather than performance and platform. Performance and platform are ambiguous, often subjective terms, and this is where the confusion creeps in. Once you identify tangibles such as the customer's budget, usage scenarios, and skill level, then you can mesh these with product tangibles such as features and upgrading flexibility to arrive at a meaningful answer. Doing this is cumbersome. The customer wants you to fill in the blank and instead you're launching into an essay. But if you can pull the client through the process and have him arrive at his own conclusion rather than one you've spoon-fed, his confusion will vanish and he'll feel safe, both in his product choice and in you as the one to supply it. |
||||
 Back to top |
||||
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form. |
||||
