For some time now, I’ve been futzing with four Via KT266A boards: the Shuttle AK31, the Epox 8KHA+, the Asus A7V266-E and the Abit KR7A.
You’ll see facts and figures during Christmas week, but for those of you a trifle more impatient, a few comments about them now.
While I can’t get really wound up about these boards, I don’t get really wound down, either. The KT266A chipset looks to be a solid performer, no matter what the implementation.
There really isn’t much difference between any of them performance-wise, either. Features and price should be the deciding factors for you here.
Maybe the Abit does a little better, and the Shuttle does a little worse, but the Abit and Asus cost almost twice as much as the Shuttle and Epox.
True, the Asus has ATA100 RAID and onboard sound, and the Abit has ATA133 RAID, and if those are features you’ll actually use, fine, but if you’re not, you’re going to pay an exorbitant sum for the few extra percentage points in performance you’ll get.
A few of those extra dollars do go towards better quality, and both the Epox and Shuttle are bargain boards. The Epox has had some gestation and QC problems. The Shuttles seem to be functionally OK, but barebones. It’s kind of a shame that there isn’t more options besides around $100 or around $200.
Not Much Adventure for the Average
We’re at the end of a generation of systems: .18 micron CPUs/DDR266. There’s not just a lot of headroom to play with unless you’re ready to get hardcore.
Something a lot of people don’t realize is that AMD is treating overclockers like the troops in Afghanistan are treating al-Queda, slowly but surely giving them less and less room to work in. The obvious one was to increase the difficulty in unlocking the CPU. The less obvious one was restricting the range of voltages.
If the voltage regulation on the socket A board works right, 1.85V is only a 6% increase in voltage over default. Even if the voltage regulation went high (as did many earlier socket A boards), it still was usually a bit shy of 10%.
Contrast that to the 15% voltage increase that was often available on BX boards, and this can really cramp an overclocker’s style.
So voltage modifications have become almost a necessity lately for serious, much less extreme overclocking.
I would say about half the current XP overclocking owners haven’t unlocked their CPUs. I’d bet at least 90% haven’t done a voltage mod. If AMD tightens the screws anymore, Intel is going to end up getting back most of the overclockers by default.
The reality is the overclocking population is like a pear, a small hard core inside, with the rest much softer.
Frankly, most overclockers just want fat, easy, pretty safe improvements, and that’s not what you’re not going to get with this equipment. You’re going to have to make hardware modifications and take some real risks to get big improvements.
If you (or the person you’re building for) is the average person who values reliability over raw speed, if you want a pretty fast but safe system, this is pretty good stuff. An XP/KT266A setup is the first Athlon system I’d feel comfortable setting up for the average Joe. I’m not afraid the processor is going to self-immolate after the slightest excuse to do so and that the motherboard stands at least a fighting chance of being reasonably bulletproof.
If you have serious mechnical overclocking skills, and are fed up with every Moe, Dick and Harry doing just about as well as you with about five minutes’ worth of knowledge, this is pretty good stuff for you, too. You’ll leave the wusses behind.
That leaves everybody else in the middle. For those of you in that category, you’ll probably get a lot more jollies down the road with .13 micron CPUs.
No Having Your Cake and Eating It Too
I’m getting a lot of messages asking me about which mobos will work with Thoroughbred. The answer right now is, “We don’t know.” Based on AMD’s track record in obsoleting older mobos with newer processors, the prudent answer is “Assume nothing.” Come on folks, how many times does AMD have to screw you before you notice?
We’re supposed to see the first Thoroughbred in about three months. Supposed to see. While I think AMD will more or less hit that target this go-round, you may not necessarily want it, given that Thoroughbred II (also known as Barton, with SOI construction and likely to be a sizable improvement over T1) will follow about six months after T1, and looks to be the last structural improvement planned for the Athlon.
Early in 2003, we’ll then have Hammer, which will absolutely require a new mobo, and probably a good deal else.
Within the next three months, we’re probably looking at DDR333 memory and boards becoming the standard, and another six months for them to actually get it done right.
On the other side, we think Northwoods at some point in time will become effective competition, though it’s beginning to look like later in 2002 rather than sooner.
2002 could be a pretty tumultuous year: delays in products, new generations of equipment not quite working right; then big leaps towards the end of the year. By the end of it, we’re probably looking at overclocked speeds of 3.5GHz, maybe more, but most of that will probably come later in the year.
What you should do depends on how much you’re hurting now and how often you can or want to upgrade.
If you’re more than a bit off the beaten track and reliability counts more than adventure, an XP/KT266A looks to be a nice safe haven for most of 2002. You can take the year off, and by the end of it, we’ll be looking at Bartons vs. mature Northwoods, with Clawhammers just starting to show up.
If you want more adventure, wait for the initial .13 micron, but with adventure comes uncertainties and problems.
Don’t count on anything you buy today being able to work tomorrow. If you sort of need something today, and sort of can’t really afford replacing it soon, buy less now. If you spend only $80 for something like the Shuttle, $130 for an XP1600, and $50 for a 256Mb stick of DDR, that will hold you pretty decently as is until the smoke clears on .13. If Thoroughbreds work with it, consider that gravy. If they don’t, it’s only $80.
If it has to work with a Thoroughbred, wait until the company says it will work with a Thoroughbred.