First of all
It took longer than I thought, but finally the new socket 7 guide has been completed. After long nights of testing it’s good to know that all candidates passed the ‘basic’ benchmarks successfully and fulfil decent quality expectations. Not one board ran instable at 66 MHz clock; the only great hurdle to take is a bus clock beyond 66 MHz, and in spite of Intel’s restrictive specs, most manufacturers have more or less tried to integrate the new settings. If you take into account that the hardware itself has not to be changed, the performance gains at 75 or even 83 MHz bus speed are tremendous. Particularly the L2 Cache benefits a lot when clocking it this high. Two boards came with the VIA Apollo VP2 chipset and 1 MB Cache. Take a look at the charts at the end of this page to compare performance. I think it’s important to mention that all manuals were so far complete and usually good to understand. Multi language manuals would be desirable, since motherboard shipments are lacking this feature. Very important to know is the fact that the 430TX chipset can only use 26 bit tag RAM for the administration of the level 2 cache, 18 bit are chipset internal, externally there’s an 8-bit tag RAM. This limits the cacheable area to 64 MB. If you anyway use more memory, you will lose about 5% overall performance! Microsoft contributes to this by having their operating systems access the highest available memory address first. This means practically that more memory merely makes sense unless you have memory intensive applications benefiting from it. For example if you have 128 MB RAM and if you then only use Word, you will be always working outside the cacheable area. It has the same impact on your performance as if you took some other TX chipset board with e.g. 32 MB and disabled the L2 cache! Please try to avoid breaking the 64 MB barrier, else choose a HX chipset board or one with VIA Apollo chipset.
Since 430TX chipset motherboards work with 3.3 Volts on the ISA bus instead of the classic 5 Volts, some voltage critical cards (e.g. AVM Fritz! ISDN card) may be malfunctioning. Often these troubles can be avoided by ensuring you get a new revision; I’m sorry I have no list of affected hardware. Nevertheless, if you already had some nights fighting with your computer, think about this. Please don’t run immediately your messenger to mail us your problems, we can’t possibly deal with more than a few. Be sure your troubles are not based on configuration mistakes (check IRQs, DMAs, I/O ports, Windows 95 config and conflicts first), then please contact the manufacturer of the affected part and ask for possible reasons. It would be very helpful if you informed us in case ISA voltage is the problem.
All new BIOS revisions support the LS-120 drive; most the IDE ZIP drive as well. Since the C6 is about to be available now, a wave of BIOS updates will be ready to download pretty soon. Due to its aggressive low price, this CPU can be interesting for upgraders who have been using a CPU around 100 MHz. So if you want to have more performance without paying too much and you if you can live without a fast FPU, take the C6 into consideration.
Some results
Intel’s 430TX chipset could have controlled the market, but Intel decided to stunt it for emphasizing the performance and push up the sales of Pentium II CPUs. The Socket 7 market is all but dead though. VIA successfully topped the 430HX’s performance with the Apollo VP2 (or AMD-640) and is going to attack Intel’s AGP advantage with the Apollo VP3 chipset. Please take a look at benchmark results from the final revision of the FIC PA-2012 Rev. 1.2.
It’s no secret: The Apollo VP2 chipset (also AMD-640) is at the moment the fastest for Windows NT, particularly with 1 MB L2 Cache. Important as well: Don’t forget the VIA Busmaster IDE drivers, they push up your performance again. As often, the overall performance depends highly on fine tuning by the manufacturer. A good 430HX and EDO memory can be as fast as an average 430TX board with SDRAM. Once again, the field is too close together to justifiably talking about slow or fast. The performance differences between two boards are usually under 5%, that’s why I would buy a motherboard first by its features (external clock, expandability, stability) and second by its performance. Product cycles are too fast to spend a lot of money on one single component, but this is why everybody should make his own decisions.
Always on the top and the best allrounder at all clock speeds is ABit’s AB-TX5 (430TX), also very stable and comfortable to configure by ABit’s CPU Soft Menu. If a Cyrix/IBM CPU is your choice, take a closer look at the two boards with Apollo VP2 chipset and 1 MB Cache. Taking advantage of the linear burst mode, the FIC PA-2007 and the Shuttle HOT-603 are the fastest boards for those CPUs. The fastest board for the Pentium MMX is currently the Asus TX97-XE. It’s a pity that the 75 and 83 MHz modes don’t run reliably. It’s more difficult making your mind up in case you want to use a AMD K6 CPU. Very fast boards are the FIC PA-2007 or the AOpen AX5T, but others are pretty close. To determine your favorite motherboard, please refer to the benchmark charts and the technical informations of each board.
Testing environment
Of course the motherboards were tested under the same conditions:
CPUs: Intel Pentium MMX 233 MHz (66 x 3.5), AMD K6 at 233 MHz (66 x 3.5) and if possible at 250 MHz (83 x 3) as well as Cyrix/IBM 6x86MX PR233 (187 MHz, 75 x 2.5).
Memory: If supported 64 MB SDRAM (2x 32 MB Toshiba 10ns or Samsung 12ns), else 64 MB EDO DRAM (2x 32 MB TI, 60ns)
Harddisk: Quantum Fireball ST 3.2 EIDE (if supported using Ultra-DMA/2 interface, else DMA/2)
VGA: Matrox MGA Millennium 4 MB, 1024×768 High Color
OS: MS Windows 95 OSR 2.1 (USB supplement), MS Windows NT 4.0 Server (+ Service Pack 3)
Benchmarks: Business Winstone 97
I did not try mixing EDO and SDRAMs. It should work on many boards, but I don’t recommend you to do it. You won’t get more performance if you add some new SDRAM to your EDOs. If you want to use 75 or 83 MHz bus frequency, change over to the faster SDRAM, but keep your EDO memory in case you’re going to stay at 66 MHz. The performance advantage of SDRAM at this bus speed is not worth talking about. Another issue is memory expansion, which has become quite elegant since DIMM modules can be installed seperately; you don’t need to obtain an identical pair and you won’t get trouble due to different contacts or even chip manufacturers. In some boards I tried the mixture of Samsung, Toshiba, Hitachi and LGS SDRAM memory, without any problems.
The Quantum Fireball ST obviously is faster than the Fireball TM used in the former tests. But the performance difference between EIDE hard drives is only visible under Windows NT. Windows 95 does not profit that much from a faster hard drive. This is also a reason why I did without the well known performance champ Seagate Cheetah.
You might ask why I did not use the faster Millennium II. This video card is certainly faster, but not fast enough to clearly push up the Winstone results. The 2D improvements are fine, but 3D quality is, I’m sorry to say, a huge disappointment. For the same money you can get one of the new generation 3D accelerators like Diamond Viper 330 or nVIDIA RIVA128, you can get some information about the new cards at the graphics card review.
For all overclockers
If you plan to overclock your CPU (especially K6 or M2), be aware of a good air circulation inside your case. The best cooler/fan is useless if the heat remains around the processor. Please check SDRAM timings and the item ‘specultative read’ in your BIOS as well. Try the slower settings if you have trouble, not all boards are optimized to the two faster bus frequencies. Furthermore there always is the possibility that your board is unstable running at 75 or 83 MHz bus clock. This could happen with the CPU as well, some just can’t be overclocked too much. Another phenomenon is that all three CPUs don’t get as hot running Windows NT as in Windows 95 or DOS. NT executes the CPU instruction ‘HALT’ whilst idle, something that’s obviously asked too much from a mouse driver collection like Windows 95. To push up the CPU clock another bit, you can also try a higher voltage. For example I ran the Pentium 233 MMX at 83 MHz x 3.5 (291 MHz) to check system stability at much higher processing speed. As you already expect this, was quite a hot affair,I could have cooled the CPU with a hair drier as well. Therefore I don’t recommend to do this permanently, it’s amusing to see the performance of future CPUs, at least for me.
ABit AB-AX5
At 83 MHz, this board is one of the fastest, at normal bus speeds it is better than average. This board features ‘Soft Menu‘, it’s an extension to the Award BIOS 4.51PG, which allows to set the CPU and bus frequencies in the system BIOS. This seems to be made for playing with the different possibilities. Let run your favorite application or OS with 75 or in some cases 83 MHz system bus speed, and write your recordable CDs at safe 66 MHz. Without opening your case, isn’t it great? As in the prior socket 7 review already mentioned, this board has some difficulties with certain types of SDRAM, e.g. the older Corsair modules won’t work properly. If your are looking for a an ATX board to run at the highest possible external clock speed, think about the AX5.
ABit AB-AX5, Rev. 2.2
ATX format; ATX power supply; Intel 430TX; 4x ISA, 4x PCI; 4x SIMM, 3x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz; BF2 function not available, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.7V, 2.5V.
ABit AB-PX5
This board seems to support 75 and 83 MHz bus clock, but I had some trouble with it, even with our latest revision 1.14c. This expressed in hangups and protection errors under Windows 95 as well as blue screens under Windows NT. Particularly the old Corsair memory didn’t want to work properly over 66 MHz. In spite of these problems, performance is not bad at all and the overall results at 66 MHz are satisfactory. Except with the 6x86MX, the board is slower under Windows NT. Maybe a new BIOS can bring the estimated stability at 83 MHz bus.
ABit AB-PX5, Rev. 1.14c
Baby AT format; AT power supply, ATX power connector not equipped; Intel 430TX; 4x ISA, 4x PCI; 2x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz; BF2 function not available, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (with connector); 1x USB (without connector); IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.5V.
ABit AB-TX5
ABit’s newest TX chipset board, the AB-TX5, seems to inherit the crown of the AT boards most suited for overclockers and speed fans: It’s fast, faster and the fastest. Featuring ‘Soft Menu’, it’s very easy to adjust CPU clock ratio. External frequencies are supported up to 83 MHz, very stable by the way. Even the Pentium 233 MMX ran fine at 291 MHz and 3.2 Volts! It’s good to see that the performance of the AB-PX5 was improved much more as well. This is my personal favorite TX board by the way. A new board must first be reliable and stable for me under all circumstances. And if it’s fast, too, I got my candidate. ABit’s TX5 board fits exactly my personal expectations. Please don’t forget that there are more fine boards as well; before you make your choice, go ahead and read some other reviews as well. Maybe you will find some fact or feature you will regard even more important than I do.
ABit AB-TX5, Rev. 1.0
Baby AT format; AT & ATX power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz; BF2 function not available, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (with connector); 1x USB (without connector); IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.7V, 2.5V.
ABit AB-IT5H
This of all boards months ago was the first running perfectly even at 83 MHz bus speed. By the way, the memory timings were all switched to ‘maximum warp’ without errors. This means quite good performance under Windows 95; Windows NT is a bit slower, also the usage of a 6x86MX CPU. But its performance is comparable to some boards with the 430TX chipset! You will get 4 ISA / 4 PCI slots, 4 PS/2 slots and 2 DIMM slots (merely EDO-DIMMs, no SDRAM). The latest revision 2.0 now has switching regulators; if you should own a IT5H, don’t think about upgrading to a TX chipset board, you probably won’t get much more performance, especially in case you’re using over 64 MB RAM.
ABit IT5H, Rev. 2.0
Baby AT format; AT power supply; Intel 430HX; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM (EDO!); Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz; BF2 function not available, I/O chip: ALi; busmaster DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (with connector); 1x USB (without connector); IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.7V; two TAGRAMs for cacheable size of 512 MB
AOpen AP5T
This 430TX chipset board features CPU temp monitoring, 512 KB L2 Cache, 2x DIMM, 4x PS/2 sockets, 3x ISA / 4x PCI slots and outstanding quality. This boards runs stable up to 83 MHz, but strangely I had to switch the K6-233 down to 2.9 Volts (!) to run reliably at 250 MHz with the Toshiba SDRAM. The Shuttle HOT-565 has similar features but no switching voltage regulator. AOpen seems to become a reliable manufacturer for 83 MHz boards just the way ABit is doing with the AB-TX5. Both boards support Ultra DMA/2, 64 MB cacheable area (SDRAM or EDO), no ECC support. If you want to use the 233 MHz K6 processor, be aware you get rev. 3 or newer, because you will need 3.2 V support for this CPU. Revision 3.1 also has jumper settings for 2.2 Volts, which the K6 266 most likely will tolerate. The performance of the AOpen TX chipset boards is fine once again, particularly with the K6.
AOpen AP5T, Rev. 3.1
AT format; AT power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66, 75, 83 MHz; BF2 jumper missing, I/O chip: SMC; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without connector); 1x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.2V. NCR SDMS support ver. 3.07.
AOpen AX5T
Revision 3 has some voltage selection jumpers that are probably reserved for 2.1 or 2.2 V core voltage. Thanks to the 430TX chipset it features the same limitations (64 MB cacheable, no ECC) and advantages (UDMA/2, SDRAM support). Its performance is good, as with the AP5T. Other manufacturers should take a look at some of the AOpen features, e.g. the admirable stability at 75 and 83 MHz bus speed; you can say it’s more stable than fast at 83 MHz. The only comparable boards are the ABit AB-TX5 and partially the Shuttle HOT-565/569.
AOpen AX5T, Rev. 3.0
ATX format; ATX power supply; Intel 430TX; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66, 75, 83 MHz; BF2 jumper equipped, I/O chip: SMC; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 1x USB (not on placed with the other connectors, cable missing as well); IrDA support; core voltages: 3.2V, 2.9V, 2.8V, one reserved jumper, possibly 2.2V or 2.1V. SDMS support ver. 3.07.
AsusTek P/I P65 UP5 w/ CPU card C-P55 T2D
A very special board ist the Asus P65 UP5. You have to buy a CPU card for either two Pentium CPUs or another card for two Pentium Pro CPUs. Since AMD or Cyrix/IBM CPUs do not support multi processor architecture, the only suitable CPUs are the Intel Pentium or Pentium Pro. In spite of the various processor configuration possibilities, Winstone does not benefit of a second CPU, so I did not benchmark the dual CPU configuration, of course. Unfortunately this board does not support more than 66 MHz bus speed; I would have liked benchmarking a Pentium Pro at 75 x 3 (= 225 MHz)! This lack limits the qualification for some CPUs running at 75 MHz (6×86-PR200 and 6x86MX-PR233). Due to this restriction, you will just find results for the K6 at 233 MHz and Pentium MMX at 233 MHz (both 66 MHz bus). I think it has to be mentioned that I could tune up all memory settings to maximum (r/w bursts: x-2-2-2, turbo leadoff etc.) without any problems. If you’re thinking about obtaining this board, please take into account the size and position of the CPU card. In most cases the HDD bays use exactly the space the CPU card would occupy, so please plan on a large scale and get a server case.
Asus P/I P65 UP5 baseboard (Rev. 1.41)
Baby AT format, AT power supply, chipset on CPU card; 3x ISA, 5x PCI, 1x Asus Media Bus extension; 8x SIMM; Award flash BIOS; cache on CPU card (socket 7) or on CPU (P6); external frequencies: 60, 66 MHz; I/O chip: Winbond; busmaster DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without connector); 1x USB (without connector); IrDA support; SDMS support ver. 3.07.
Asus P55T2D CPU card
Dual processor socket 7 CPU card; Intel 430HX chipset; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66 MHz; BF2 jumper not equipped; core voltages: 2.9V, 2.8V, 2.7V, 2.5V.
AsusTek P/I P55 TXP4
Asus still is able to learn, that’s fine. For some months now every baby AT board comes with PS/2 mouse cable as well. Better late than never, I think. The PC 97 compliant boards also have an excellent fan on the item list. The new TXP4 board seems to be a low end version of the TX97 series. It is coming without CPU temp surveillance, without Asus Media Bus and has a minimum size. Nevertheless, 4x PCI / 3x ISA slots, 4x PS/2 and 2x DIMM sockets should be sufficent for most users. Good to see that Asus spent some more money for the PS/2 memory sockets with metal clip. 83 MHz bus clock is selectable, nevertheless running anything but stable. Cyrix 6x86MX PR233 (187 MHz) only ran fine with the slower SDRAM Ras-to-Cas latency of 3/3 clocks. I could make use of the faster 2/2 settings only for the 66 MHz CPUs. Quality is Asus-like, but its performance is nothing what has to be praised. Future BIOSes could surely improve this.
No picture, sorry. Rev. 1.x.
AT format; AT power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz (83 unstable); BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.7V, 2.0V.
AsusTek TX97-E
As the name says, this board meets the Microsoft PC 97 standard. It’s also available as ‘TX97’ without the ‘E’, this means one additional ISA slot, three instead of two DIMM sockets and no PS/2 sockets. If you plan to use future socket 7 CPUs, be aware of the BF2 jumper being equipped (like on my revision 1.11), it enables clock multipliers up to x5.5. Performance and quality is excellent as always and the price is too high as always. The board features external clock support up to 75 MHz. What you should know are the jumper settings for 83 MHz: (FS0, FS1, FS2) 2-3, 1-2, 1-2. Asus does not mention this clock anywhere, so they don’t guarantee its reliability. Nevertheless, everything worked fine and fast as well. You can enable all performance enhancing options in the BIOS, just SDRAM speculative read lead to instability under Windows NT. Due to the PC 97 compilance, additional features are fan monitoring, temperature and voltage monitoring and modem remote on.
No picture again, so sorry. Rev. 1.11
AT format; AT power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66, 75, 83 MHz (83 undocumented); BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.5V, 2.0V. This board meets PC 97 specifications; SDMS support ver. 4.0.
AsusTek TX97-XE
This board has exactly the same specifications as the TX97-E, but comes in ATX format. I will add the benchmark results from the TX97 and TX97-E as soon as I get them. Strange to say why Asus does not implant support for bus clocks beyond 66 MHz into their ATX format boards, the same is the XP55T2P4. As a matter of fact, higher clock rates definitely are selectable (75 MHz: FS0, FS1, FS2 at 2-3, 1-2, 1-2), but are unbearably instable, what a pity. Everything else is Asus-like, nothing to complain about. 4x ISA as well as 4x PCI are standard for ATX boards, 4x PS/2 and 2x DIMM should also be enough. This board is also available with on-board soundblaster compatible sound system. As written above, this board is currently the fastest for the Pentium MMX.
Picture is missing, I know.
ATX format; ATX power supply; Intel 430TX; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66, 75, 83 MHz (75 and 83 MHz unstable); BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.1V. This board meets PC 97 specifications; SDMS support ver. 4.0.
Chaintech 5TDM2
This is Chaintech’s actual AT board based on Intel’s 82430TX chipset. The overall performance is good, nothing to complain so far. It’s the first board with only one cache memory chip of 512 KB. Chaintech has also inplemented switching regulators and an ATX power supply connector in case you want to use this AT board in a common ATX case (PC97). Chaintech now uses a BIOS extension called SeePU where you can adjust CPU speed and voltage in your system BIOS. Chaintech has also added a utility disk (besides the Busmaster IDE drivers) to update the system’s INF files allowing Windows 95 to name the PIIX4 PCI-to-ISA bridge as well as the 430TX/440LX correctly. Very fine!
AT format; AT & ATXpower supply; Intel 430TX chipset; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 55, 60, 66, 75, 83 MHz; BF2 jumper integrated, I/O chip: ITE; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without connector); 1x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V and more depending from BIOS revision.
FIC PA-2007
It seems as it everybody wants to know something about this board. The facts: VIA Apollo VP2 chipset, 1MB L2 Cache (2 banks), 2 DIMM/ 4 SIMM slots, 3 ISA / 4 PCI slots, good jumper descriptions, up to 75 MHz bus speed (no 83!), Award BIOS. Performance is excellent, especially under Windows NT although system sometimes doesn’t boot if cache timing is set at 2 clock leadoff cycle (2-1-1-1); strangely using the 6x86MX at 75 MHz bus speed, it worked pefectly and amazingly fast thanks to the linear burst feature. You may have to select 3 clocks (3-1-1-1) to undergo the booting sequence with the K6 and the Pentium MMX with some BIOS revisions. I regret I was not able to get the 512 KB version in time for some comparison. Its performance should be around the same in this case (3-1-1-1). If you are a speed freak and need a cacheable area over 64 MB, take into consideration the PA-2007, it’s one of the fastest boards at the moment for all configurations.
FIC PA-2007, Rev. 1.0
AT format; AT power supply; VIA Apollo VP2 chipset; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 1024 KB PB-Cache (6 ns); external frequencies: 55, 60, 66, 75 MHz; BF2 jumper integrated, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without connector); 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V.
FIC PT-2007
This board has precisely the same layout as the PA-2007, the difference is the chipset. The PT-2007 is not suited for systems with more than 64 MB RAM due to the limitation of the cacheable area thanks to Intel’s 430TX chipset. The BF2 jumper function has been integrated as well, but bus clocks over 66 MHz are not supported, FIC is strictly keeping the Intel specifications. I’m sorry I haven’t had the time to try the remaining jumper setting possibilities, maybe 75 or even 83 MHz can be found. This will be updated as soon as I’ll have time again to do this. Very hot: It’s the second fastest board for the combination of Pentium MMX and Windows 95 and very fast with the K6 as well.
FIC PT-2007, Rev. ?
AT format; AT power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 55, 60, 66 MHz (also 68 MHz with ICW48C67 clock generator); BF2 jumper integrated, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V. This board meets PC 97 specifications.
FIC PA-2010+
Here we have a VIA Apollo VPX chipset board in ATX format. The K6 at 250 MHz was not logged in correctly, none the less it worked fine. I got real difficulties with the Pentium MMX 233. Sometimes it was logged in as Pentium 133 MMX, but performance was good enough that I assume this were 233 MHz. If somebody wants to operate at 83 MHz, he should be aware he gets the version with ICW W48C67-1 or ICS9147-01 clock generator, all other support 75 MHz at the highest. At 83 MHz bus speed, SDRAM bank interleave had to be disabled, otherwise you would get heavy errors. As all VIA chipset based boards, the PA-2010+ also supports the linear burst for Cyrix/IBM CPUs. You can get performance gains of around 2-3% when using this alternative burst mode, so don’t forget to check this. If you are looking for an ATX board and VIA chipset, please consider the new PA-2012 (1 MB) or the PA-2011 (with 512 KB), they are much faster.
FIC PA-2010+, Rev. 4.0
ATX format; ATX & AT power supply; VIA Apollo VPX chipset; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 55, 60, 66, 75, 83 MHz (83 MHz only with clock generator as described above); BF2 jumper equipped, I/O chip: Winbond; busmaster DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V; SDMS support ver. 3.07.
FIC PA-2011
This is the ATX version of the PA-2007, but just with 512 KB Cache. Technical facts are all the same. Just the most important (performance) is not at the same high level as the PA-2007. The board has an additional connector for AT power supply. Our revision 2.2 does not have PCI special cycle support. By the way, the PA-2010+ does, I don’t know why. The user gets into real trouble if he tries setting the SDRAM timing to 2/2 clocks. The result is that the board didn’t want to talk with me until I made multiple use of the reset swich. I still haven’t found out the reasons for this; it happened with every CPU and with EDOs instead of SDRAM, too. I’m confident a newer BIOS could improve this. I will also update this page as soon as I have some extra time to deal with this. Enable linear CPU burst when using Cyrix or IBM CPUs, you will be rewarded by a few per cents more performance. Be aware of using the Pentium MMX: This board is quite slow with it.
FIC PA-2011, Rev. 2.2
ATX format; ATX & AT power supply; VIA Apollo VP2 chipset; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; AMI flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75 MHz; BF2 jumper equipped, I/O chip: SMC; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V.
Gigabyte GA-586TX3
Gigabyte has three TX chipset based socket 7 boards: GA-586 TX, 586 TX2 and 586 TX3 (3 DIMM sockets). Performance does not diverge very much from each other, so I took the TX3 board for benchmarking our complete ‘suite’. Performance is okay, but I can’t understand why Gigabyte does not back bus clock speeds beyond 66 MHz. You will never be able to run the 6x86MX PR233, which runs at 187,5 MHz (75 MHz bus x 2.5) nor the 6×86 P200+ (75 MHz x 2). The few tests that were left showed nothing abnormal, everything ran satisfactory and at very good performance. Worth mentioning is the board design and quality, as well as the usage of dip switches for clock and voltage selection. The latest Gigabyte boards now feature voltage auto detection, what makes the manual setting absolete.
GA-586 TX3, Rev. 1.x
AT format; AT power supply; Intel 430TX; 4x ISA, 4x PCI; 2x SIMM, 3x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66 MHz; BF2 function integrated, I/O chip: ITE; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.5V, 3.4V, 3.3V, 3.2V, 3.1V. 3.0V, 2.9V, 2.8V, 2.7V, 2.6V, 2.5V, 2.4V, 2.3V, 2.2V, 2.1V, 2.0V. This board meets PC 97 specifications; SDMS support ver. 3.07.
Gigabyte GA-586TX2
The TX2 board has the same features as the TX3. A bad difference is that performance of the TX3 is always better. The lacks of the 430TX chipset: 64 MB cacheable area, no ECC support. Also the lacks of Gigabyte: No more than 66 MHz. Configuration is easy thanks CPU voltage auto detection.
GA-586 TX2, Rev. 1.x
AT format; AT power supply; Intel 430TX; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66 MHz; BF2 function integrated, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.5V, 3.4V, 3.3V, 3.2V, 3.1V. 3.0V, 2.9V, 2.8V, 2.7V, 2.6V, 2.5V, 2.4V, 2.3V, 2.2V, 2.1V, 2.0V. This board meets PC 97 specifications; SDMS support ver. 3.07.
Gigabyte GA-586ATX03
The ATX version of the TX2 is quite a fast performer. As always, Gigabyte keeps Intel conservative specs, 66 MHz at the fastest. Configuration is easy again thanks CPU voltage auto detection.
GA-586 ATX03, Rev. 1.x
ATX format; ATX power supply; Intel 430TX; 4x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66 MHz; BF2 function integrated, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.5V, 3.4V, 3.3V, 3.2V, 3.1V. 3.0V, 2.9V, 2.8V, 2.7V, 2.6V, 2.5V, 2.4V, 2.3V, 2.2V, 2.1V, 2.0V. This board meets PC 97 specifications; SDMS support ver. 3.07.
Gigabyte GA-586 HX2
Gigabyte’s HX chipset board can be equipped with an additional TAGRAM to provide a cacheable area of over 64 MB. For a 430HX chipset board, performance is okay. Again we meet Gigabytes dip switches for CPU selection and for manual voltage selection in case you don’t trust the auto detection system. As all Gigabyte boards the HX2 is als very easy to handle, no problems at all.
GA-586 HX2, Rev. 2.x
AT format; AT power supply; Intel 430TX; 4x ISA, 4x PCI; 6x SIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66 MHz; BF2 function integrated, I/O chip: Winbond; PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 1x USB; IrDA support; core voltages: 3.5V, 3.4V, 3.3V, 3.2V, 3.1V. 3.0V, 2.9V, 2.8V, 2.7V, 2.6V, 2.5V, 2.4V, 2.3V, 2.2V, 2.1V, 2.0V. This board meets PC 97 specifications; SDMS support ver. 3.07.
Mercury W586 VXL
The only Triton VX chipset board is Mercury’s W586 VXL. It’s equipped with 512 KB L2 Cache, but strangely has no DIMM sockets. Nevertheless, performance is not too bad (taking into account that we are dealing with a 430VX board). I would really like to get some board with a faster chipset from Mercury, because the package even includes a PS/2 mouse connector, and this is not very common for less known manufacturers. Who knows what surprise the better boards do offer… I don’t know the price for this board, but equipped with the VX chipset it should be very cheap. Here we seem to have a company which successfully tries to compete with other manufacturers by selling a product based on low-end components and achieving niet performance. One reason is the prosper support of 75 MHz bus speed. Not enough: jumper settings for 83 MHz are JP1, JP2, JP3 to 2-3, 2-3, 2-3. For many readers of this homepage this kind of motherboard is certainly unsuited for their computer at home, but if I had to assemble cheap Windows workstations, this board could become a candidate. The lacks: No ECC support, no SDRAM support, no Ultra DMA, max. 64 MB cacheable area.
Mercury W586 VXL, Rev. 4.1
AT format; AT power supply; Intel 430VX; 3x ISA, 4x PCI; 4x SIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 75, 83 MHz (83 MHz settings as shown above, be prepared for instabilities); BF2 jumper missing, I/O chip: ITE; busmaster DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (with connector); 1x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.7V, 2.5V.
Shuttle HOT-553
This was one of the rare boards which need to be equipped with a VRM module to provide the lower CPU core frequencies. There are a lot of newer boards available and performance feeling is not very exciting. Furthermore there is no support for 75 or 83 MHz (but down to 40 MHz, I was really surprised), due to its age I think. For the same reason, the K6 233 was logged in as K5 at 233 MHz, I had no time to look for a newer BIOS. The TAGRAM (8 Bit) can be replaced by a wider one to be able to ‘cache’ more than 64 MB. This board is no longer up to date. If you want a HX chipset board, select the ABit IT5H or the Asus P55T2P4, they already have 512 KB pipelined burst Cache and are much faster.
Shuttle HOT-553 Rev. 1.1
HOT-106 VRM module
AT format; AT power supply; Intel 430HX; 4x ISA, 4x PCI; 4x SIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66 MHz; BF2 jumper not equipped, I/O chip: UMC; busmaster DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; core voltages: 3.2, 2.9V, 2.8V.
Shuttle HOT-565
Here we have a very hot board. It comes with Intel’ 82430TX chipset and gustily benchmark results. Shuttle always uses larger jumpers, furthermore they are coloured. The 6x86MX produced hangups under Windows NT (entirely at 75 MHz bus), but 83 MHz bus clock ran fine under Windows NT, in this configuration Windows 95 performance seems to be very good as well. The slow motion BIOS settings brought no improvement. The board does support 83 MHz bus clock without a word in the manual, you only have to know that JP36 has to be closed at 5-6.
Shuttle HOT-565, Rev. 1.5
AT format; AT power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66, 75 MHz; BF2 jumper missing, I/O chip: ITE; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without connector); IrDA support; core voltages: 3.3V, 3.2V, 2.9V, 2.8V.
Shuttle HOT-569
The benchmark results are very satisfying. This latest TX chipset board now officially supports up to 83 MHz bus speed. If this board matures some time, it can be counted to the very best, since it is the successor of the 565. The performance at 83 MHz was really fast under Windows NT. It features a real variety of external frequencies. Starting at 50 MHz up to the already said 83 MHz. The BF2 jumper function was integrated as well, so you will be able to run future CPU at a maximum multiplier of x5.5. The passive voltage regulators were all changed to switching regulators without any coolers; this looks like good work.
Shuttle HOT-569, Rev. 1.2
AT format; AT & ATX power supply; Intel 430TX; 3x ISA, 4x PCI; 4x SIMM, 3x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 68, 75, 83 MHz; BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 3.1V, 3.0V, 2.9V, 2.8V.
Shuttle HOT-603
This board is the only one equipped with AMD’s 640 chipset, which corresponds to VIA’s Apollo VP2. The board features 1 MB L2 Cache (2 banks), external frequencies up to 83 MHz, 3 ISA / 4 PCI slots, integrated PS/2 mouse connector (just few cases have the corresponding hole right beside the keyboard connector, the final revision does not have it any longer) and ATX power connector as well. You may use up to 3 DIMMs or 4 PS/2 modules. Dip switches would have been nice, nevertheless the jumpers are well documented. Performance is outstanding, except with the Intel CPU.
Shuttle HOT-603, Rev. 1.2
AT format; AT & ATX power supply; AMD 640 chipset; 3x ISA, 4x PCI; 4x SIMM, 3x DIMM; Award flash BIOS; 1024 KB PB-Cache (6 ns); external frequencies: 50, 55, 60, 66, 68, 75, 83 MHz; BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse on board; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.1V.
Soyo 5XA5
Intel 430TX chipset, 512 KB L2 Cache, 4 ISA as well as 4 PCI slots and support for 75 and 83 MHz are the main characteristics for this little ATX board. You may insert up to 3 DIMM modules and if necessary 2 PS/2 modules. The position of the BIOS may prevent you from using some slots with cards which use the usually free margin between connector strip and slot. The clock configuration has to be made on a ‘mice piano’ or dip switches, as some people say. Problems or errors could be solved by disabling SDRAM speculative read and switching the RAS-to-CAS delay to 3/3 clks. This shows how much space for improvement is left. As the Asus TX97-XE, Soyo’s TX board meets PC 97 standard, too.
Soyo SY-5XA5, Rev. ?
ATX format; ATX power supply; Intel 430TX; 4x ISA, 4x PCI; 2x SIMM, 3x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66, 75, 83 MHz; BF2 jumper equipped, I/O chip: Winbond; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse; 2x USB; IrDA support; core voltages: 3.2V, 2.9V, 2.8V, 2.1V.
Tyan S1564D ‘Tomcat IV’
As it was with the Tomcat III, the newest dual Pentium board from Tyan has similar features: Intel 430HX, 512 KB PB-Cache, four memory banks, four PCI / five ISA slots and switching regulators. But no support for 75 or 83 MHz bus clock. If you take into account that such a board is particularly made for server systems you can understand why Tyan did not implement higher bus clocking. Tyan is known for excellent reliability and obviously they decided to remain on the ‘safe shore’ of 66 MHz. All tests ran without any difficulties, I was surprised that my ‘memory mix’ consisting of 2x 32 MB EDO, 4x 16 MB EDO (2x 2k and 2x 4k refresh) as well as 2x 4 MB FPM (total of 136 MB) worked as well. Please consider that this mix contains of EDO and FPM, 16-chip, 12-chip, 8-chip and 2-chip modules just as symmetric and asymmetric equipped types! You will rarely find such a fine motherboard. I would have loved testing larger memory modules but had none available here, sorry. Design and quality are excellent, you are able to use all four PCI slots in full length. Winbench results don’t differ by using two processors, because a second CPU will just push up applications running simultaneously, but not the primary. You will need to obtain multiprocessor software in this case.
Tyan S1564D, Rev. ?
Dual socket 7, baby AT format; AT power supply; Intel 430HX; 5x ISA, 4x PCI; 8x SIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 50, 60, 66 MHz; BF2 jumper missing, I/O chip: SMC; busmaster DMA or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without cable); 1x USB (without cable); IrDA support; ECC/parity support; two TAGRAMs allow cacheable area of 512 MB; core voltages: 3.2V, 2.9V, 2.8V.
Tyan S1571 ‘Titan Turbo’
To my knowledge this is the first board from Tyan supporting 75 and 83 MHz. Equipped with Intel’s 430TX chipset and 512 KB of L2 Cache, the performance of all CPUs running at 66 MHz bus clock is quite bad, there’s a lot to be improved. I hope it’s due to the early revision and BIOS that the modes of 75 and 83 MHz are extremly unstable as well. Neither the M2 at 187 MHz nor AMDs K6 at 250 MHz were reliable. Even the slow motion memory timings were as shaky, what a pity. The quality of this Baby AT board is very good as always; SDRAM sockets are good as well, they help to eject the module when opening the locks. The jumper descriptions are okay, but why still jumpers? Other companies have more elegant methods like dip switches or even the possibility of setting clock frequencies in the system BIOS (ABit). Very cool is the amount of memory slots. 6x PS/2 and 2x DIMM are waiting to be equipped. But in most cases it is not very clever to equip the 430TX chipset with more than 64 MB, since more can’t be ‘cached’. There should be some time for working on the higher external frequencies. The various memory banks as well as the 4 ISA and 5 PCI slots are typical for Tyan and a buying argument for many people. I will update this page as soon as I get a newer revision of this board. The board’s layout is excellent, I think it could be worth waiting for the final revision when performance will be much better – let us hope so.
Tyan S1571S, pre-release board
Baby AT format; AT power supply; Intel 430TX; 4x ISA, 5x PCI; 6x SIMM, 2x DIMM; Award flash BIOS; 512 KB PB-Cache (6 ns); external frequencies: 60, 66, 75, 83 MHz (unstable over 66 MHz); BF2 jumper equipped; I/O chip: SMC; busmaster Ultra DMA 2 or PIO 4 onboard; FDC onboard; 2x serial 16550; 1x EPP/ECP parallel; 1x PS/2 mouse (without cable); 1x USB (without cable); IrDA support; core voltages: 3.5V, 3.4V, 3.3V, 3.2V, 3.1V, 3.0V, 2.9V, 2.8V, 2.7V, 2.6V, 2.5V, 2.4V, 2.3V, 2.2V, 2.1V, 2.0V.
Benchmarks Windows NT
Testing environment
Of course the motherboards were tested under the same conditions:
CPUs: Intel Pentium MMX 233 MHz (66 x 3.5), AMD K6 at 233 MHz (66 x 3.5) and if possible at 250 MHz (83 x 3) as well as Cyrix/IBM 6x86MX PR233 (187 MHz, 75 x 2.5).
Memory: If supported 64 MB SDRAM (2x 32 MB Toshiba 10ns or Samsung 12ns), else 64 MB EDO DRAM (2x 32 MB TI, 60ns)
Harddisk: Quantum Fireball ST 3.2 EIDE (if supported using Ultra-DMA/2 interface, else DMA/2)
VGA: Matrox MGA Millennium 4 MB, 1024×768 High Colour
OS: MS Windows 95 OSR 2.1 (USB supplement), MS Windows NT 4.0 Server (+ Service Pack 3)
Benchmarks: Business Winstone 97
I did not try mixing EDO and SDRAMs. It should work on many boards, but I don’t recommend you to do it. You won’t get more performance if you add some new SDRAM to your EDOs. If you want to use 75 or 83 MHz bus frequency, change over to the faster SDRAM, but keep your EDO memory in case you’re going to stay at 66 MHz. The performance advantage at this bus speed is not worth talking about it. Another issue is memory expansion, which has become quite elegant since DIMM modules can be installed seperately; you don’t need to obtain an identical pair and you won’t get trouble thanks to different contacts or even chip manufacturer. In some boards I tried the mixture of Samsung, Toshiba, Hitachi and LGS SDRAM memory, without any problems.
The Quantum Fireball ST obviously is faster than the Fireball TM used in the former tests. But the performance difference between EIDE hard drives is only visible under Windows NT. Windows 95 does not profit that much from a faster hard drive. This is also a reason why I did without the well known performance champ Seagate Cheetah.
You might ask why I did not use the faster Millennium II. This video card is certainly faster, but not fast enough to clearly push up the Winstone results. The 2D improvements are fine, but 3D quality is, I’m sorry to say, a huge disappointment. For the same money you can get one of the new generation 3D accelerators like Diamond Viper 330 or nVIDIA RIVA128, you can get some information about the new cards at the graphics card review.
Some boards do not always appear in the charts; the reason usually is instability or the missing support. Only if the particular configuration was able to complete the benchmark did I quote the result. The performance relation between the boards does not necessarily keep the hierarchy from the 66 MHz charts (K6 and Intel MMX); some boards have to be run at slower memory timings to work stable, so a board being average at 66 MHz could be very fast at 75 or 83 MHz thanks to the optimizations of its manufacturer.
Benchmarks with the Cyrix 6x86MX PR233 (187 MHz, 75 x 2.5):
Benchmarks with the AMD K6 at 233 MHz (66 x 3.5):
Benchmarks with the AMD K6 at 250MHz (83 x3):
Benchmarks with the Intel Pentium MMX at 233 MHz (66 x3.5):
This was it for now. Every new board that arrives will be benchmarked as soon as possible and of course added to this review. If you want us to test some other motherboard, please contact the motherboard manufacturer, not us. He will send us a sample for testing.
Benchmarks Windows 95
Testing environment
Of course the motherboards were tested under the same conditions:
CPUs: Intel Pentium MMX 233 MHz (66 x 3.5), AMD K6 at 233 MHz (66 x 3.5) and if possible at 250 MHz (83 x 3) as well as Cyrix/IBM 6x86MX PR233 (187 MHz, 75 x 2.5).
Memory: If supported 64 MB SDRAM (2x 32 MB Toshiba 10ns or Samsung 12ns), else 64 MB EDO DRAM (2x 32 MB TI, 60ns)
Harddisk: Quantum Fireball ST 3.2 EIDE (if supported using Ultra-DMA/2 interface, else DMA/2)
VGA: Matrox MGA Millennium 4 MB, 1024×768 High Colour
OS: MS Windows 95 OSR 2.1 (USB supplement), MS Windows NT 4.0 Server (+ Service Pack 3)
Benchmarks: Business Winstone 97
I did not try mixing EDO and SDRAMs. It should work on many boards, but I don’t recommend you to do it. You won’t get more performance if you add some new SDRAM to your EDOs. If you want to use 75 or 83 MHz bus frequency, change over to the faster SDRAM, but keep your EDO memory in case you’re going to stay at 66 MHz. The performance advantage at this bus speed is not worth talking about it. Another issue is memory expansion, which has become quite elegant since DIMM modules can be installed seperately; you don’t need to obtain an identical pair and you won’t get trouble thanks to different contacts or even chip manufacturer. In some boards I tried the mixture of Samsung, Toshiba, Hitachi and LGS SDRAM memory, without any problems.
The Quantum Fireball ST obviously is faster than the Fireball TM used in the former tests. But the performance difference between EIDE hard drives is only visible under Windows NT. Windows 95 does not profit that much from a faster hard drive. This is also a reason why I did without the well known performance champ Seagate Cheetah.
You might ask why I did not use the faster Millennium II. This video card is certainly faster, but not fast enough to clearly push up the Winstone results. The 2D improvements are fine, but 3D quality is, I’m sorry to say, a huge disappointment. For the same money you can get one of the new generation 3D accelerators like Diamond Viper 330 or nVIDIA RIVA128, you can get some information about the new cards at the graphics card review.
Some boards do not always appear in the charts; the reason usually is instability or the missing support. Only if the particular configuration was able to complete the benchmark did I quote the result. The performance relation between the boards does not necessarily keep the hierarchy from the 66 MHz charts (K6 and Intel MMX); some boards have to be run at slower memory timings to work stable, so a board being average at 66 MHz could be very fast at 75 or 83 MHz thanks to the optimizations of its manufacturer.
Benchmarks with the Cyrix 6x86MX PR233 (187 MHz, 75 x 2.5):
Two boards have an advantage supporting the linear burst mode for IBM or Cyrix CPUs, surprisingly the ABit TX5 nevertheless outperforms them under Windows 95.
Benchmarks with the AMD K6 at 233 MHz (66 x 3.5):
Benchmarks with the AMD K6 at 250MHz (83 x3):
Benchmarks with the Intel Pentium MMX at 233 MHz (66 x3.5):
This was it for now. Every new board that arrives will be benchmarked as soon as possible and of course added to this review. If you want us to test some other motherboard, please contact the motherboard manufacturer, not us. He will send us a sample for testing.