2000 mainboards

Epox MVP3C2

Very similar to the older Epox EP-58 from the previous year, and equally good. The only obvious changes were to a switch block rather than jumpers for the voltage setting, one less ISA slot, and ATA-66 hard drive support. None of these were particularly important, for good or ill. Once again, we wished these had more cache, but they were nevertheless a firm favourite here, second only to the VA-503, and we always kept a few handy. As ever with Epox boards, they were reasonably priced, easy to work with, and very reliable.

The driver CD was impressive: it is routine stuff these days, but Epox were one of the first to offer a really good software bundle: Norton Anti-Virus and Ghost (both worth having) and one of those neat-looking integrated setup routines that is supposed to install the drivers for you. In practice, as is almost always the case with these things, you got better results by doing it manually.

We used to make a point of using the official VIA DMA hard drive controller drivers (which were good) rather than the Hi-Point ones which installed by default — the Hi-Points gave better benchmark figures but were no faster in real life and could be troublesome. (The same goes for the 503, by the way, or anything else with an MVP3 chipset: if you are going to load IDE drivers, use the VIA ones, they are the best. In reality though, the best of all for Windows users was to use the default Microsoft-supplier IDE drivers — these were fuss-free and fastest of all. It was only from about 2002 that the VIA drivers became clearly the best ones to use.)

• CPU support: 6x86, 6x86MX, C6, K5, K6, K6-2 to 500MHz, K6-III, P54C, P55C.
• Speed: 66, 75, 83 and 100MHz.
• Slots: 4 PCI, 2 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 384MB.
• Cache: Surface mount, 512k pipeline burst.
• Chipset: VIA Apollo MVP3, Award BIOS.
• Best With: K6-2
• Status: Legacy.

FIC VA-503+

Well over four years old, and still the best all-round Super 7 motherboard. With many hundreds, probably thousands, of 503+ systems under our belt by this time, and a good few still to come by way of repairs and trade-ins and upgrades, we had grown used to the complex and confusing jumper layout — indeed, we had learned it off by heart long since. That aside, we have only a couple of quibbles about the 503, both age related: it would have been nice to have three SDRAM slots instead of two SDRAM and four EDO (which was rarely used by this time), and some people wanted ATA-66 hard drive support. In reality, this was a trivial matter: ATA-66 was very little if any faster than ATA-33 with the then-current generation of hard drives, and we missed it not at all.

Where the 503 did shine was in its support for recent hardware: so far as we know, it is the only board apart from the Gigabyte GA5AA that supported both the K6-2/550 and the K6-III+. (For this last, you need an unpublicised beta BIOS — the K6-III+ runs at a most unusual 2.0 Volts and is not a drop-in replacement for a standard K6-III.)

The 503+ had proved to be everything we could want in a main board: very fast, quite inexpensive, and decently reliable. The VIA MVP3 chipset was still the clear Super 7 winner: SiS and ALI made competent competitors but the MVP3 was both faster and more compatible. And that wonderful 1MB 503+ cache took a lot of beating: with a K6-2 a 503 system creamed the average Celeron, and with a K6-III the 503 was still the best thing you could buy this side of an Athlon in the 900MHz class.

Postscript: February 2002. We've bought our last batch of VA-503s. When our current hoard runs out, that's it. After what is quite possibly the longest main board production run ever, the 503 is no more. We will miss this cheap, flexible, wonderful old faithful the way we missed the 386DX-40. Vale 503.

• CPU support: All Super 7.
• Speed: 66, 75, 83, 95, 100, 112 and 124MHz.
• Slots: 3 PCI, 3 ISA, AGP
• RAM: 4 72-pin FPM, EDO or BEDO and 2 168-pin PC-100 SDRAM, up to 512MB.
• Cache: Surface mount, 1MB pipeline burst. (512k optional.)
• Chipset: VIA Apollo MVP3, Award BIOS.
• Best With: K6-2, K6-III
• Status: Legacy.

FIC PA-2013

The ATX brother of the VA-503+ was similar in most respects. Like the 503, it was around for a very long time and was still the best of breed when it was discontinued sometime in 2001. It had no 72-pin RAM support — not really an issue by this time — but a welcome third SDRAM slot; an extra PCI slot; and the jumper layout was equally confusing but at least easier to reach. Like the 503, the PA-2013 had no peer in its category: if there was a better Super 7 board we'd like to have met it.

• CPU support: 6x86, 6x86MX, C6, K5, K6, K6-2, K6-3, P54C, P55C.
• Speed: 66, 75, 83 and 100MHz.
• Slots: 4 PCI, 2 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB.
• Cache: Surface mount, 1MB pipeline burst.
• Chipset: VIA Apollo MVP3, Award BIOS.
• Best With: K6-2, K6-III
• Status: Legacy.

FIC SD-11

These were described on the 1999 page. They remained a steady seller even after the arrival of the second generation Athlon boards, right up until the new breed of Socket A CPUs arrived. Over time we became quite fond of the SD-11. They had not got any smaller, so they only just squeezed into the cases, the layout was weird, and the manual remained a bit odd, but the markings on the board itself were perfectly clear if you could read the tiny things, and most of the time they just performed without fuss or bother. Reliability was on a par with most decent boards; not outstanding but certainly adequate, and in reality we suspect that most of the problems we had with these could be traced down to the ridiculous CPU mounting method. Slots were a bad idea when Intel introduced them with the Pentium II in 1997, and did not improve over time.

The price gradually crept down as the months went by too: the SD-11s wound up considerably cheaper than a BX board.

Postscript: February 2002. These had a renewed lease of life for us in late 2001 and early 2002 when a regular supplier of ours offered us a number of them complete with Athlon Thunderbird 700 CPUs (yes, slotted Thunderbirds, not Athlon Classics). The price was such that they made a very attractive low-cost upgrade alternative to a Duron or Celeron. 700MHz was off the pace in 2002, but if you were on a budget and upgrading from a 6x86-333 or a Pentium MMX-200, a Thunderbird 700 was a huge advance.

One thing to watch with these: they are very demanding about power supplies. Even many of the AMD certified units cannot run an SD-11 with a slotted Thunderbird reliably, you need a unit that is certified for the particular combination of main board and CPU. Our otherwise excellent A-Open PSUs, which powered Athlon XPs and the massively demanding Thunderbird 1400s without the slightest bother, were not sufficient for this pairing. If you are getting hard drive read errors with an SD-11, odds-on you need to upgrade your power supply. And why should such a modest seeming rig require so much power? We have no idea. In the end, we come back to our favourite motto: we don't care why it works, just so long as it does work.

• CPU support: Athlon.
• Speed: 200MHz.
• Slots: 5 PCI, 1 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB.
• Cache: N/A.
• Chipset: AMD 751/VIA VIA 686, AMI BIOS.
• Best With: Athlon (what else?)
• Status: Legacy.

Microstar MS-5187

Microstar have been around for a long time. We remember using their 386 boards sometimes. The 5187s came to us by mistake: we wanted Baby AT boards, but these were ATX. The main feature of interest was their VIA MVP4 chipset; essentially an MVP3 with SiS style built-in video. They were beautifully packaged and came with what was then a truly impressive set of drivers and utility software: from anti-virus stuff right up to dial-in remote control software.

Full marks so far. The glossy manual seemed nicely done too — until you realise that the only important part, the CPU jumper settings, are all back-to-front! Still, you could work it out with a little careful thought. Less easy to fathom was the silly hardware monitoring feature: the boards report "system hardware abnormal" and beep loudly on boot until you either use an on-board fan connector or disable the CPU fan sensor. Not so much a case of poor design but rather just poor documentation and badly chosen BIOS defaults. (These days, when CPUs run so much hotter and it would make sense to do this, all the main board manufacturers do the opposite. Go figure.)

The boards themselves were fairly standard ATX fare, reasonably laid out, quite small and yet still nice and roomy. Performance of the MVP4 was noticeably slower than that of an ALI-5 or an MVP3 with a stand-alone video card though. Apart from anything else, you lost 8MB of main RAM to the video adaptor. (Switching this back to 4MB in the BIOS used to help — 64MB was a lot back then, remember.) The chipset-integrated sound worked just fine, and was very configurable via the BIOS, but the driver install program could be fussy, particularly if you were upgrading from an older soundcard, so take care. Once it was up and running, it was fine.

All in all, a competent board once you got used to its oddities (and stopped relying on the manual!), but not one that was ever going to set your eyes alight. SiS have been making integrated chipsets a lot longer than VIA, and the SiS530 in the ASUS P5S-B blew its socks off.

• CPU support: K6-2, K6-III, M-II.
• Speed: 66, 75, 83, 95 and 100MHz.
• Slots: 3 PCI, 1 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 512MB.
• Cache: Surface mount, 512k pipeline burst.
• Chipset: VIA Apollo MVP4 (with sound and UMA video), Award BIOS.
• Best With: M-II.
• Date: 25th January 2000.

ASUS P5S-B

These filled a major gap in our motherboard line-up, one that had troubled us for some months: the entry-level integrated board. In the past we had several very successful main boards with built-in video and/or sound: originally the excellent ASUS SP97-V, and then several other more-or-less similar products, all based on the little-known and under-rated SiS 5581 chipset. Although the performance of integrated sound and video is never as good as that of proper stand-alone components, for many purposes ultimate performance is not the point: for a student system, or a network workstation, fuss-free low-cost reliability is the aim — and an integrated board is typically about $100 cheaper than a standard board with separate sound and video cards. Although it served us faithfully for several years, the 5581 had seen its day: limited in theory to 83MHz, we never really trusted it much past 66MHz, which in practice meant that a 6x86MX-300 or K6-2/300/66 was the upper limit.

The P5S-B used the long-promised SiS 530 chipset, the 100MHz development of the 5581, and supported the then-current generation of entry-level CPUs, notably the K6-2 450 and 500. As always with ASUS products, our first impressions revolved around a reasonably sensible layout (bar the placement of the switches at the front right — just where they are impossible to reach once installed in a system), excellent build quality, and superb documentation.

On closer inspection, minor flaws became apparent. The prime issue was detail layout. The two sets of dip switches covering a host of options were scattered every which way. Despite the excellent manual, these were tricky to set up. The optional on-board ESS Solo soundcard worked just fine, though we still wished it had been a Soundblaster. Alas, ASUS stopped making the P5S-B too early and we had to go back to FIC and Epox for our Super Sevens. But perhaps it was just as well: over the next few months the return rate proved to be a little higher than we would have liked.

• CPU support: K6-2, K6-III, most other Socket 7.
• Speed: 66 to 100MHz.
• Slots: 4 PCI, 2 ISA
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB.
• Cache: Surface mount, 512k pipeline burst.
• Chipset: SiS 530, Award BIOS, integrated AGP video, ESS sound chip.
• Best With: K6-2/450.
• Status: Legacy.

ASUS P5A-B

We looked at these as a possible replacement or supplement for our FIC and Epox Super 7 boards. On the whole we saw the VIA chipset as superior to the Ali Aladdin 5 set these were based on. On the other hand, ASUS had the habit of converting us to their point of view, so we were interested in trying these out. They cost about the same as an Epox or a 503, perhaps a few dollars more, so it was purely down to which one we would rather use.

Three things determined our preference for the MVP3 over the Aladdin 5: our benchmarks rated the MVP3 a fraction faster, the Ali was never really happy with TNT video cards (though the more recent AGP drivers were certainly much improved), and the Ali UDMA drivers, like the Intel ones, conflict with real mode CD-ROM drivers. In normal use you should not run both chipset UDMA and real mode (DOS type) drivers at the same time, but in the workshop it is often needed for troubleshooting. Could the ASUS implementation of the Aladdin persuade us?

As so often with ASUS products, the presentation was excellent and the driver CD as good as any we had seen. Jumper settings were clear enough, though still inferior to the outstanding Epox system. The board itself was rather small but neatly laid out. Most things were easy to reach, though the power supply connector was too close to the edge of the board and tended to foul on other objects in the case. We didn't like the way that one of the mounting screw holes was actually inside the CPU socket (you can see it at the top right of the socket in the illustration) — this is cumbersome and creates extra work in a production environment, and becomes a real pain if you have to troubleshoot. Just the same, these were an impressive board and we'd have been happy to use more of them, but in the end they were still not as good as the 503+.

• CPU support: All Super 7.
• Speed: 66, 75, 83, 95 and 100MHz. Unofficial 105, 110, 115 and 120MHz support.
• Slots: 3 PCI, 2 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB. (256MB cachable.)
• Cache: Surface mount, 512k pipeline burst.
• Chipset: ALI Aladdin 5, Award BIOS.
• Best With: K6-2
• Status: Legacy.

A-Trend ATC5200

Yet another MVP3 board, on the face of things as typical as any other. They came in two versions, quite different but sharing the same model number. Setup was easy enough once you worked out a few odd little things: sometimes on the first version (not illustrated) the serial port cables only seemed to work properly when you put them in a particular order — an odd fault that one — and the settings in the manual for the K6-2/500 were wrong. The boards came ex-factory with some weird jumper settings that make no sense, but that's common too. Main board detail design and documentation errors are, alas, standard fare these days. There are very few motherboard makers who get it right all the time. In fact, if pressed, we couldn't name even one. Epox are close.

The ATC5200's Award BIOS was a little odd, but you quickly get used to these things. Initially we rather liked these, but in the longer term reliability was not as good as we'd expect. Several of ours went back to be replaced more than once. Worse, quite a few of our ATC5200s showed the type of minor but significant problem that makes a main board very time-consuming in the workshop: notably AGP-related issues that don't show up until a board is installed and running all the proper drivers. It's much nicer, when a board fails, if it just stops completely. That way you can spot the trouble right away and not have to waste time troubleshooting it. We won't be buying A-Trend boards again in a hurry.

• CPU support: All Super 7 except K6-2/550, K6-III+.
• Speed: 66, 75, 83, 95 and 100MHz.
• Slots: 3 PCI, 2 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB.)
• Cache: Surface mount, 512k pipeline burst.
• Chipset: VIA MVP3, Award BIOS.
• Best With: K6-2/450, K6-2/500
• Status: Legacy.

ASUS K7M

ASUS added a couple of small touches that set these apart from our usual first-generation Athlon board, the FIC SD-11. It was much smaller, about half the size of the SD-11, had the usual excellent ASUS presentation, and an extra pair of USB ports. Best of all, where the SD-11 had a miserly single serial port, the K7M had the regulation two. Sure, if you use a PS/2 mouse you only need one for a modem, but what if you want a second modem, a data logger, a machinery controller, or any of the hundreds of other serial devices?

But although we liked the K7-M, it was almost $100 more than an SD-11 or any of several other similar products, and $100 for a second serial port seemed excessive. We would have paid a little more for K7M, but not that much. On the other hand, if price was no object, this was probably the best of the first-generation Athlon boards.

• CPU support: Athlon
• Speed: 200MHz.
• Slots: 5 PCI, 1 ISA, AGP
• RAM: 3 168-pin PC-100 SDRAM, up to 768MB.
• Cache: N/A.
• Chipset: AMD 751/VIA 686, AMI BIOS.
• Best With: Athlon.
• Status: Legacy.

Epox EP-7KXA

Epox sat out of the first round of Slot A boards for the Athlon. This was their first Athlon board and one of the first of the second generation Slot A boards with the VIA KX133 chipset.

For some reason beyond our ken, these copped a bath at Tom's Hardware — slow unstable, an utterly useless board, according to Tom — but according to Ace's Hardware (a less well known but much more reliable site) they worked perfectly. As usual, Ace was right: we sold a good few of them, with perfect satisfaction and a zero return rate. Another excellent Epox board.

All the standard KX133 features applied, including AGP 4X and 133MHz SDRAM, and Epox didn't skimp on connectivity — plenty of PCI slots, an ISA slot for older hardware, and twin serial ports. There was also built-in VIA sound which was adequate for occasional business use but not a patch on a real stand-alone Sound Blaster card for music or games.

The new-style Award BIOS offered lots of options but the interface was clumsy and a clear step backwards. Alas, the new Award BIOS is everywhere now, Epox were just the first to use it. (Why is it that the BIOS makers can't just stick with a good thing when they get it right? The previous Award BIOS, like the wonderful old AMI Color BIOS of 386 days, was near-perfect. Nobody minds if you tinker with a product to improve it, but Rule One is "when you get it right, don't touch it!")

All in all, a very competent Epox board at an excellent price, which saw us through nicely until the socketed Athlons took over in August or September of 2000.

• CPU support: Athlon
• Speed: 200MHz.
• Slots: 5 PCI, 1 ISA, AGP
• RAM: 3 168-pin PC-133 SDRAM, up to 768MB.
• Cache: N/A.
• Chipset: VIA KX133, Award BIOS.
• Best With: Athlon.
• Date: 11th July 2000.

ASUS K7V

This second ASUS Athlon board was based on the then-new VIA KX133 chipset, rather than the original AMD and AMD/VIA combination chipsets that the first generation Athlon boards used. On paper they looked impressive. The KX-133 allowed for AGP 4X and (more significantly) 133MHz RAM. Sensibly, the K7V retained full port connectivity, but we wondered if it wasn't too soon to abandon ISA completely — particularly as the place where the ISA slot would have gone was filled by one of those utterly useless AMR slots. The peace of mind that the extra flexibility of an ISA slot or two brings was then and still is well worth having. ISA may be slow and old-fashioned, but there were still plenty of jobs which only ISA could do (and if there is any performance penalty to it, no one has ever managed to demonstrate it).

In practice, we soon found that the lack of an ISA slot caused difficulties. With new machines these could sometimes be worked around at the cost of some expense or inconvenience, but for upgrades ISA was essential. The time for ISA to disappear would come, but not in '00, and not in '01 either. There were just too many peripherals that still needed it.

The K7V quickly developed a huge and in our view quite unjustified reputation around the web. ASUS had gone to a lot of trouble to make overclocking possible, but they hadn't worked hard enough on the nuts-and-bolts practicality side of the K7V. All modern motherboards are prone to having plug and play resource assignment problems from time to time — that's an inevitable part of the moronic plug and play design philosophy of not giving the human enough power to override the automatic system when it makes mistakes — but the A7V was worse by far than most competing boards. Not one of our A7Vs failed to work correctly in the end — but not one of them worked first time up.

• CPU support: Athlon
• Speed: 200MHz.
• Slots: 5 PCI, AGP
• RAM: 3 168-pin PC-133 SDRAM, up to 1.5GB SDRAM or VCM.
• Cache: N/A.
• Chipset: VIA KX133, Award BIOS.
• Best With: Athlon.
• Status: Legacy.

Transcend TS-AVD3

With our most popular CPUs, we spend quite a bit of time looking at the available range of motherboards and choosing the best two or perhaps three. These make up the vast bulk of our mainboards. Even where there seems little to choose between a number of boards, it is still best to pick one and standardise on it, as familiarity is a very important part of being able to work reliably and efficiently with a main board. It is obviously important to learn of any particular oddities it has (such as compatibility issues — which are still more common than you might imagine). Less obviously, knowing the minor details of a board off by heart (like jumper layout, BIOS settings, and I/O cable connections) is a significant aid to both productivity (because it saves time) and reliability (because you are less likely to make mistakes).

Socket 370 CPUs, however, made up a very small proportion of our sales in 2000: perhaps five percent or so. By the time you subtracted upgrades and repairs to Baby AT form systems (which required an older Baby AT main board of course, perhaps a KA-6100 with a converter card) we only used a handful of Socket 370 motherboards for new-build systems, and our choice of the AVD3 for a number of them had more to do with random chance and who we happened to be ordering something else from on the same day than any particular merit of the board itself. In reality, we are not qualified to comment in any detail on Socket 370 boards: we just didn't see enough of them.

With that said, the Transcends did all that we expected, and were reassuringly familiar because the business end of the VIA Apollo Pro chipset, the south bridge, was very similar to the south bridge chip on the ubiquitous MVP3 and MVP4 Super 7 chipsets, and exactly the same as that on the VIA Athlon chipsets. In fact, you can even use the same drivers.

• CPU support: Celeron, Pentium III
• Speed: 66 to 133MHz.
• Slots: 5 PCI, AGP, ISA, on-chip sound.
• RAM: 3 168-pin PC-133 SDRAM, up to 1.5GB SDRAM or VCM.
• Cache: N/A.
• Chipset: VIA Apollo Pro 133, Award BIOS.
• Best With: Celeron.
• Status: Legacy.

FIC AZ-11

FIC's first Socket A board for the Duron and Thunderbird. We had mixed feelings about these boards. On the one hand they arrived promptly: we had AZ-11s available within a week or two of being able to buy Socket A CPU chips for them, when most of the other manufacturers were dragging their feet. In their favour also was their simplicity: plug them in, add some RAM and a CPU, and away you go. FIC went to some trouble to make them overclocker-friendly, with small bus speed steppings in software, and safe emergency recovery built into the BIOS.

But the AZ-11 did not delight the true-blue techie: FIC had joined the recent cheapskate rush to leave out all the ISA slots (then still an essential for decent flexibility) and provided only a niggardly single serial port. The new BIOS, for all its practicality, had the feel of condescension to it and was not as fast and practical in the workshop as the traditional type (pioneered by AMI in 386 days, and since then perfected by Award themselves). There were no jumper settings to speak of (which is just as well, given FIC's rather horrible layouts in recent years!) but they continued to use illogical and difficult placements for the connector block. Worse than the SD11, in fact.

When they first came out (August 2000) the AZ-11 was rather expensive but this gradually changed as more and more competing boards arrived. The FIC tradition is to leap out of the blocks with a new-technology board, which starts out quite expensive, and keep on making it for a long, long time: the PA-2005, PA-2007, and VA-503+ were all very long-running boards, and all became very reasonably priced after the first few months. On the whole, we looked forward to a long run with these, and expected to mostly use them in entry-level systems running Durons (though we used them for Thunderbirds as well from time to time). However, the lack of connectivity crippled them as a mainline product for a full service dealer like Red Hill. The cost to us of having to replace a three month old motherboard because our customer needs a second printer port, a stock market ticker card, or perhaps a second modem is significant. Given that we soon had two excellent alternative boards available at equal or lower cost, before too long we relegated the AZ-11 to emergency use only, to cover out-of-stocks.

In August 2000, when these first came out, we wrote: "we will be having a good hard look at competing Socket A boards too ..... we'd like to see a nicer BIOS, better LED block layout, two serial ports and an ISA slot, and all that at a price competitive with the AZ-11, and with reliability as good or better." It seemed like a big ask, but both Epox and Soltek soon had boards which came very close to our ideal. So we neglected the AZ-11 for most of its market life, selling just a handful of them after the initial flush. Yet, in its own way, it remained indispensable. It was one of the very few early Socket A boards that was AMD certified to run the highest speed and hottest-running Athlons, and could be trusted to keep the power up to a Thunderbird 1100, which a lot of other boards could not. And for all its lack of features and general clunkiness, the AZ-11 was astonishingly fuss-free and reliable. It was discontinued in the winter of 2001. Then, with good stocks of this unlovely but solid and reliable board available at very low prices, we snapped them up again to use for entry-level systems running Duron 750s and the like.

• CPU support: Thunderbird, Duron
• Speed: 200MHz.
• Slots: 5 PCI, AGP, on-chip sound.
• RAM: 3 168-pin PC-133 SDRAM, up to 1.5GB SDRAM or VCM.
• Cache: N/A.
• Chipset: VIA KZ133, Award BIOS.
• Best With: Duron.
• Status: Legacy.

Epox 8KTA

Close to the ideal Socket A board. They were a fair bit smaller than the FIC AZ-11, but they offered quite a bit more: twin serial ports, an ISA slot, no less than six PCI slots. Notice that Epox didn't waste any space on those useless AMR (Audio Modem Riser) slots. (See the three ASUS illustrations above to look at these — they are the very short brown slot that looks like a one-third size AGP connector.)

In typical Epox style, the LED connectors were laid out so that they were easy to see and difficult to get wrong — a small thing perhaps but something of a fetish for us here at Red Hill. This is something that is so easy to get right that we can never understand why most manufacturers get it wrong. The only limitation of the 8KTA was that despite having the same chipset and the same number of DIMM sockets as the AZ-11 the 8KTA could take only 768MB of RAM. (Or so the manual claimed — in practice, we didn't know anyone who could afford to try it out back then!)

The 8KTA came in at around the same price as an AZ-11: in other words, very reasonable. Hands-on, they were a pleasure to work with and one of our most popular Socket A motherboard for the six month period spanning the end of 2000 and the start of 2001. There were several minor variations: the 8KTA+ added a locking AGP slot and extra voltage options for overclocking, while the 8KTA2 had ATA-100 instead of ATA/66 controllers. The only really significant difference was in CPU voltage provision: the 8KTA supported up to Thunderbird 1100, the 8KTA2 and later versions of the 8KTA+ added the power-hungry Thunderbird 1200 as well.

• CPU support: Thunderbird, Duron
• Speed: 200MHz.
• Slots: ISA, 6 PCI, AGP, on-chip sound.
• RAM: 3 168-pin PC-133 SDRAM, up to 1.5GB SDRAM or VCM.
• Cache: N/A.
• Chipset: VIA KZ133, Award BIOS.
• Best With: Duron, Thunderbird.
• Status: Legacy.

Soltek 75KV

Our other favourite first generation Socket A board. Larger than the Epox 8KTA and a little more expensive, they offered the same excellent features: twin serial ports, an ISA slot, plenty of PCI slots, extra fan connectors if you wanted them. The extra room, however, did not make them easier to work on — for some incomprehensible reason, the LED connectors were crammed together in one of the all-too-common modern senseless arrangements. Still, that's the worst thing we have to say about the 75KV, so we can't really complain.

It was the Soltek software bundle that was the only real point of difference they had: like FIC and Epox, Soltek bundled Norton Anti-Virus (not our favourite anti-virus but at the price — essentially zero — a great boon), Norton Ghost and a virtual CD drive package that only works properly if your CD-ROM drive is D: or you can be bothered trying to figure the documentation out. (Which we can't. Software is supposed to work right out of the box. If it doesn't, we buy better software.) Soltek didn't stop there though, they threw in WinFax as well — at the price, that was an amazing bargain.

The really weird thing though is that the first time you install Windows on one of these, there is an extra pop-up on your desktop, offering "bargain" software downloads. Suspicious as ever, we checked it out. Essentially, it offers nothing at all of value, mostly just advertising. At first we thought a free WinZip would be nice, but it turns out to be just the same unregistered shareware one you can download from anyware on the Net and probably already did. But where does it come from? Load the exact same copy of Windows on another computer and the Phoenix.net stuff doesn't show up at all! It turns out that the download code actually sits in the Soltek motherboard's BIOS! An extraordinary idea, and quite a technical achievement, but utterly useless. Still, it disables easily enough, so we put up with it. They were a nice motherboard just the same.

• CPU support: Thunderbird, Duron
• Speed: 200MHz.
• Slots: ISA, 5 PCI, AGP Pro, 4 USB ports, on-chip sound.
• RAM: 3 168-pin PC-133 SDRAM, up to 768MB SDRAM or VCM.
• Cache: N/A.
• Chipset: VIA KZ133, Award BIOS.
• Best With: Duron, Thunderbird.
• Status: Legacy.