ASUS P2-99 motherboard
Review date: 5th May 1999.Last modified 03-Dec-2011.
High speed, high quality, low cost. Any questions?
The only problem with ASUS motherboards is that they're not cheap enough. Their base model motherboards, like the popular P2B series, certainly aren't very expensive, but at $250-odd versus $200 or so for an Abit BH6, they lose some sales. Abit got a bad name in the early days of the BX6 motherboard, with various horror stories being told of dead-on-arrival boards. Opinions differ on Abit's current quality levels, with some vendors still reporting considerably higher return rates on Abit boards and others not, but regardless of reliability their boards still sell well in the overclocking community. Abit was the first major manufacturer to make Pentium 2 motherboards that allowed you to tweak CPU core voltage setting to push recalcitrant processors into higher speed brackets, and they're still the first name that comes to mind, despite all of the major Abit features now being available in the products of more "staid" manufacturers, like ASUS.
ASUS' new P2-99 uses Intel's 440ZX chipset, a slightly less capable but also slightly cheaper version of the enormously popular BX chipset. The features the P2-99 lacks, though, aren't important to most users, even rabid gamers. And the board's got almost as much overclocking and expansion potential as serious gamers' favourites like the current P2B and BH6 boards. The P2-99 sells for $199 (Australian dollars), versus $259 for the current version of the P2B. A current model BH6 will cost you about the same.
ZX limitations
The 440ZX chipset (see Intel's page about it here) is just like the 440BX except for two things - the amount of memory and the number of bus mastering PCI slots supported.
The ZX chipset supports only up to four "sides" of DIMM memory; most DIMM modules are double-sided, and so two normal modules are as much as the chipset can handle. Accordingly, the P2-99 is equipped with only two DIMM sockets, and supports, according to the manual, up to 512Mb of RAM.
If 512Mb isn't enough for you, I strongly doubt saving $50 on a motherboard is high on your list of priorities. If you've already got RAM that you want to use on a new board, though, the two slots could be annoying. If your RAM's only 66MHz rated, though, you're probably going to have to scrap it if you want to run even the standard 100MHz Front Side Bus (FSB), let alone higher speeds, so for most users the loss is minimal.
The ZX chipset also supports only four bus mastering PCI slots. Some ZX boards have the more usual five PCI slots, but only four of them can be used for cards that need a bus mastering slot. The P2-99 simplifies the matter, and only has four PCI slots, along with the usual single AGP slot and three ISA slots. One ISA slot shares its back panel hole with one PCI slot, so you can only use one of them at a time. Since few users these days have more than a couple of ISA cards, if any at all, this is no big deal.
There's also a lower-specified "ZX-66" version of the ZX chipset, which only supports 66MHz FSB, and is designed for use only with Celerons, which at present all use 66MHz FSB as standard (although they pretty much all work fine at higher speeds - see below...). ZX-66 is likely to be seen only in real bargain basement motherboards, and is of no interest to overclockers.
What you get
The P2-99's packaging is as good as every other ASUS board's; motherboard, not very useful software CD, excellent manual, IDE and floppy cable, little bag of extra jumper blocks. The board itself is, as is normal with ASUS products, immaculately constructed, with high-grade componentry. And, unlike early P2B revisions, the manual spells the word "Pentium" correctly on the cover!
Unlike higher-specced motherboards, the P2-99 doesn't have onboard temperature monitoring, but it does have headers on the motherboard for both CPU and power supply temperature probes. Motherboard temperature sensors are close to useless, as they don't tell you how hot the really important parts of the computer are.
Setting up
All ASUS boards still use the tried and tested, and highly popular, jumper setup method. Jumpers are sets of pins that stick out of the motherboard, invariably cryptically annotated in one point Myopia, but actually pretty easy to find with the help of a good manual, such as comes with the P2-99. To change the jumper settings, you slide little plastic dinguses with contacts inside over the jumpers according to the diagrams in the manual. The eight-pin FSB jumper block on the P2-99 offers a plethora of possible settings:
|
FSB speed |
PCI speed |
| 66.8 | 33.4 |
| 75 | 37.5 |
| 83.3 | 41.65 |
| 100.3 | 33.43 |
| 103 | 34.3 |
| 105 | 35 |
| 110 | 36.67 |
| 112 | 37.33 |
| 115 | 38.33 |
| 120 | 40 |
| 124 | 31 |
| 124 | 41.33 |
| 133 | 33.3 |
| 133 | 44.33 |
| 140 | 35 |
| 150 | 37.5 |
The many small steps mean that you're likely to be able to wring some extra speed out of a large number of processors, without losing reliability. And, somewhat startlingly, the really high settings seem to have at least a chance of working!
Fancy features
A standard feature for serious overclockers these days is CPU core voltage (Vcore) adjustment. Many processors won't run at a particular speed at their recommended core voltage, but will run if the core voltage is increased. Increasing the voltage makes the processor run considerably hotter, so you need a well cooled case and/or a huge heatsink on your CPU (see my own solution to the case-cooling problem here).
A more recent overclocking feature is cranking up the "VIO", or input/output voltage. The primary effect of this on many systems is to allow RAM to run faster; in many cases it permits regular PC100 RAM to operate properly at 133MHz FSB. The downside, again, is more heat and possibly reduced lifespan.
The P2-99 supports both Vcore and VIO tweaking, but not in a very advanced form - there's a jumper for each setting on the motherboard, with a standard and "TEST" setting. I don't know how much the TEST setting increases the VIO, but I'm informed that it boosts Vcore from the standard 2.0 volts to 2.15. In my tests, both changes did indeed make a difference.
Testing
I first tried the P2-99 out with a brand new retail boxed Intel SL37F Pentium II 350. Like all current Pentium IIs and all Celerons ever made, SL37F is "multiplier locked"; it can only run at 3.5 times the FSB, regardless of what multiplier you set on the motherboard. If you want to run it faster than its stock speed, you can only do it by tweaking the FSB.
I wasn't anticipating very exciting overclocked results from the P-II. Pentium IIs have separate cache RAM that runs at half of the processor speed, and historically has proved to be rather touchy about running faster. Tweaking cache latency with special utilities (see here) has helped some people wring more speed out of P-IIs, but the current model Celerons with their smaller but full speed cache built right into the processor chip are a better bet. The older SL2W8 and SL2YK P-II 300s and SL2WY and SL2TV P-II 333s are the usual P-II overclockers' favourites; the more recent multiplier-locked versions are much less fashionable.
Nonetheless, I was pleasantly surprised at how much faster than stock the P-II was willing to run. In a stripped down machine (hard drive, CD-ROM drive, two plain 64Mb PC100 SDRAM modules, ASUS AGP-V3400TNT 8Mb TNT video card and nothing else - no sound card, no network card) the P-II 350 powered up perfectly happily, as you'd expect, at 100MHz FSB.
But it also worked at 105, 112, 120 and even 133MHz FSB, for a CPU speed of 466MHz! When I bumped up the Vcore and VIO settings via the jumpers provided, it went on to run stably at 140MHz FSB, or 490MHz. Startlingly, Windows loaded even with the motherboard wound up to its honking 150MHz top FSB speed (processor speed 525MHz!), but it was flaky, and programs crashed at random.
I installed Quake 2 and left the computer chugging away at 490MHz with the standard Q2 "attract mode" demos running; it was still perfectly happy two hours later.
At 140MHz FSB the PCI bus is clocked at only 35MHz, which is barely higher than the 33MHz standard setting and should work with pretty much all PCI devices. This doesn't necessarily go all of the FSB/PCI speed combinations; It's all very well to crank a computer up to ludicrous speed when it's got not much in it, but this might not say much for the possible performance in an upgrade situation where all of the old hardware can't just be traded in on newer gear.
The V3400 TNT card was happy as a clam at all of the speeds I tried, despite the AGP clock speed also being locked to the FSB (you can set it to be two thirds of FSB, or the same as it; the first setting is the default, and the second is what you choose for 66MHz FSB). Again, this trouble-free result might not be duplicated by other hardware.
Nonetheless, this was darned impressive performance from a nondescript 350MHz processor. If you can manage 466 or 490MHz, you're beating out a 450MHz P-II, which at present costs more than twice as much as the 350. Bargain!
Next, I reset the voltage-goosing jumpers on the P2-99 and yanked the faithful Celeron 300A out of my usual computer. This processor, like most of the old Slot 1 Celerons, is happy to overclock; it's been running at 450MHz for months on an old P2B that has no voltage adjustment of any kind.
On the P2-99, the C-300A ran perfectly at 450MHz, as you'd expect, and also at 105MHz FSB (466MHz) without voltage adjustment. 112MHz FSB (504MHz) was very flaky indeed until I set the Vcore jumper to the TEST position; presto, stability. Higher than that, though, was impossible; Vcore and VIO increasing together wouldn't make Windows load at 115MHz FSB (517MHz), although the computer could get to the BIOS setup screen OK. I tried 120MHz FSB and lo, the computer didn't boot any more.
No matter. The P2-99 seemed perfectly happy to run the 300A at 504MHz all day, which is 68% faster than the processor's sticker speed and 12% faster than the speed I normally run the thing. 12% is not a serious increase; canonically, any performance improvement of less than 10% isn't noticeable at all. I wouldn't buy a new motherboard for 12% more speed. But if you're upgrading from, say, an old Pentium machine, you can pick up a PPGA (Socket 370) Celeron 333 for less than $200 (Australian dollars), which includes the price of a Slot 1 adapter card (like this one) to let you use the socketed chip on a Slot 1 board. Even if you only manage to get 415MHz out of it (5 x 83MHz), you've got more than 90% of the performance of a P-II that costs four times as much!
Overall
The motherboard is something you shouldn't skimp on when building a PC. But if you can find a cheap board with high-grade features, like the P2-99, why not? With everything that's needed by all but the looniest overclockers, this is the board to buy if you want a quality product at a bargain price. It's not quite as good a choice for overclocking as boards that allow several quantified steps of voltage adjustment, but it should work just as well for most users, and the price is definitely right.
The ASUS web site
At the time of writing, the ASUS site contains no information on
the P2-99 motherboard.
There's a reason for this.
If you're in Europe or the USA, you probably can't buy the P2-99 as a retail product. It's apparently available there, but only in an OEM (Original Equipment Manufacturer) version. There's another board, the P2Z, that's very similar, and occupies the P2-99's place in the ASUS retail lineup in the larger markets. The P2Z has the same 440ZX chipset and the same slots, but also has the LM78 voltage monitoring chip which the P2-99 lacks. The P2Z is therefore somewhat closer to full P2B specification than the P2-99, and would probably be priced a little higher if both boards were sold through the same channel in one place. For most users the difference is trivial.
