The main criterion for testing was at least 5 PCI slots, that's why we excluded such boards of micro ATX format as Gigabyte GA-7ZM, Pc-Chips M805LR and Microstar 6340. We gave preference to those boards which had an external ATA100 controller. That's why the boards from Abit and Iwill without ATA100 didn't take part in testing.
So, in the end we chose 8 boards: Abit KT7-RAID, ASUS A7V, MSI 6330 (K7T Pro), Iwill KV200-R, Gigabyte GA-7IXE, Gigabyte GA-7ZX-1, PC Partner K133ASA and Soltek SL-75KV+. The test was carried out on the base of 2 configuration testbeds.
So, here they are:
AMD Duron 600
128 MBytes SDRAM CAS3 (SPD; really CAS2 up to ~160 MHz) PC133 (Apacer/Infineon) or 128 MBytes VCRAM CAS2 PC133 (Kingston/NEC)
Seagate Barracuda ATA II ST315320A (15 GBytes 7200 rpm UDMA66 2 MBytes Cache Buffer)
ASUS AGP V7100 (nVidia GeForce2 MX 32 MBytes SDRAM), driver 5.32 beta1 (worked more stable than 5.33 from the complete set with the card)
Guillemot Maxi Sound Fortissimo (Yamaha YMF744), driver 4.07.1240 (WHQL Certified)
Lite-On 40x UDMA33
CNet Pro 120
Genius GM56PCI-L(Lucent Winmodem), driver 5.66 (the later 5.75 often break the connection in my line)
Sony 1.44" The test was carried out under the control of Windows 98 SE Rus. Additionally we installed VIA 4 in 1 Service Pack 4.24 and MS DirectX 7.0a Rus.
Because of the fact that we had some difficulty in installing the cooler Golden Orb on Abit Iwill boards, we used a cooler nearly hand made (6800 rpm). When at 20 degrees Celsius, we used coolers from Thermal Take, the temperature of the processor at the regular frequency didn't exceed 36 degrees Celsius, and with the new cooler we hadn't more than 39 degrees Celsius, what is not so bad.
Description of the boards
Some manufacturers offer the following variants for the mentioned boards:
Besides, Gigabyte, MSI and PC Partner offer the boards in micro ATX format (GA-7ZM, MSI 6340 and K133MSA correspondingly). Besides, we expect the all boards to be based on KT133 with integrated ATA100 controller in the South Bridge (VIA 686B).
Because of the big size you can face some problems when installing Abit and ASUS into many cheap ATX-blocks. The memory slots on Abit board are located not very convenient (set firmly against AGP card), and the capacitor which is located behind the first PCI slot hinders in installation some boards. As for Iwill, ATX power supply connector is in the center of the board, what causes some difficulty as well. The capacitors located in front of the forth PCI slot on Microstar board don't allow installing some other daughter boards. The all boards (except PC Partner) have coolers on the North bridge of the chipset, and Abit has it even with a fan.
Compatibility and performance
The most difficult to install was Abit. All other boards hadn't any problems with equipment installation. However, all boards with standard AWARD 6.0 (except Soltek) had wrong image of jpg files in the Explorer window, refused to play video on the media player and had some crack in D3D-games in the system with SCSI and IDE disabled, though it took place when using AMI BIOS or AWARD Medallion. We managed to get the video adapter and SCSI-controller work good on one IRQ on the both Gigabyte boards. The performance tests were carried out in the less hard system, although some boards still had some difficulty in installing. First of all to check the compatibility of the boards, we installed the modem on one IRQ with video (into the first slot for the boards with AWARD, into the third one for the boards with AMI). When installing drivers for the modem, Microstar and Abit had some problems. ASUS and Abit had problems with VCRAM work together with the external ATA100 controller (Iwill performed good), that is the system buzzed when initializing Windows graphics interface. But it happened from time to time and in general the systems worked stable.
Unfortunately, the most test programs and utilities give quite different results when at repeated testing (each test was made 5-7 times). That's why we used the programs which had minimum result difference:
Testing the processor and the memory:
Testing the disk system:
Each board was tested with those BIOS versions and drivers, which we received in the complete set:
All test were carried out under Win98SE RUS v 4.10.2222A, all programs in autoload and tray were deleted. When changing the boards were reinstalled the OS entirely. The memory speed was set to maximum performance for all boards in all settings, AGP 4x and Fast Write mode were enabled (where it was possible). We used video mode 1024*768, HiColor, 85 Hz, if it's not indicated apart.
Testing the processor and the memory
You can see, that the performance of the operations with floating point depends on the real processor frequency, the performance of the integer operations depends also on the memory working speed. Abit works with the memory quite good, the board has many additional settings. Gigabyte on AMD falls behind because of the low memory working speed (CAS2 PC100). Unfortunately, Gigabyte GA7-ZX-1 and PC Partner don't have a possibility of a compulsory setting of CAS2 at 133 MHz of memory bus, and we couldn't find PC133 module with CAS2 with SPD.
Testing the disc system
Unfortunately, we excluded Ziff-Davis Winbench 99 from the programs of testing HDD performance because of too different results.
You see, that there is no an unquestioning leader - the disk benchmarks have different concepts for performance defining. Although in general, the external controllers are not quicker than the South bridge of VIA chipset, in the real applications the result difference constitutes less than 1%, that's why we carried out the performance tests on the "basic" controllers.
Neither memory speed nor HDD performance influence the working speed of business applications. As for games, you can see there are both evident leaders and outsiders, but the performance is not the only criterion to choose the board.
Performance dependence on the memory settings
In the description to ASUS motherboard you can read that VCM SDRAM from NEC is compatible with SDRAM industrial standard and it ensures more than 50% performance increase with 30% power supply reduction. Unfortunately, we didn't notice it, moreover, in Quake III the performance falls regarding CAS2 PC133 memory. VCRAM refused to work on AMD750, however, when installed together with SDRAM, it was found by SPD reading program (from the complete set SiSoft Sandra Millennium Edition). As for other results, I want to say that the engineers of VIA shortened the gap between different memory types.
Only 3 boards allow changing the multiplier on the processor (ASUS allows only in jumper mode; unfortunately this mode doesn't suit to all processors), and only 4 of them allow changing the core voltage. The ceiling on the IDE-system with overclocking with the bus constituted 117 MHz for all boards (except both Gigabyte boards and PC Partner), on the SCSI-system it was 120 MHz. When we soldered up the known configuration bridges on Abit and Soltek, we reached as much as 914 MHz (101.5*9) at 1.8V on the core. The maximum performance was achieved in 112*8 mode (896 MHz) on the IDE-system and in 120*7.5 mode (900 MHz) on the SCSI-system for the both boards. We show the overclocking results for the definite Duron 600 with the definite hardware, so don't generalize them for all processors of this series.
It's a leader in general performance. Rich expand possibilities. The biggest number of memory settings. And very bad compatibility. Other lows include a big size. The lighter version, ABIT KT7, is a traditional choice of a gamer-overclocker.
The reliable and fast board from the respectable manufacturer. We didn't face any problems, but for rare buzzing when working with VCRAM together with Promise. Similar to Abit, the board has a huge size. We like the board, but we'd prefer to wait for the variant without Promise with VIA686B.
The cheapest board on the obsolescent AMD750. It possesses not very good performance and scant additional possibilities. No problems occurred during the work. It may be recommended for users with PC100 memory.
With the real CAS2 PC133 memory it could become the leader in our review (the board was tested in CAS3 PC133 mode). The high speed and stability come together with traditional minimum overclockability. A good board for a traditional user. The CD contains a dream for a lazy user - flash-writer under Win9x, though beta-version;)
The most expensive board in the review, but it's the only one that possesses a full sound controller. It's of a relatively small size (as compared with other boards with ATA100). Low performance, though high reliability which allows to be recommended for complex systems on the base of IDE.
Average performance, good reliability, moderate price - just a dream for a usual user:) The highs are a small size with good location of connectors. It's the first competitor for Gigabyte GA7-ZX-1. Well, it's for you to decide, what to choose.
High performance, stability, promising overclockability, and low price (alike Gigabyte) make this board an optimal alternative for the majority.
PC Partner K133ASA
Similar to Gigabyte, it was tested in CAS3 PC133 mode and the performance wasn't mind-blowing. A bad choice for overclockers. Though it has quite low price (at the level of GA-7IXE4). In general, the products of this company and its clones (Aska, Supergrace, Vtech and others) are worth to pay a close attention to.
External ATA100 controllers
With one HDD on IDE cable they don't give any advantage over an integrated ATA66 (and they won't until there are HDDs with a bandwidth more than 66 MBytes/s). That's why the external ATA100 controller without RAID support on ASUS A7V board is unnecessary. The controllers on Abit and Iwill boards support RAID, but we would recommend the latter for those who want to build a high-performance and fault-tolerant disc system based on IDE. As for HPT integrated in the Abit, it has too many complains from users. Besides, this board hasn't a brilliant compatibility.
"The memory of a new generation" didn't impress me much. Though bearing in mind the price for brand name modules CAS2 PC133 SDRAM and CAS2 PC133 VCRAM, the latter may turn out the most successful purchase in the nearest future. However, today it makes no sense, perhaps because of poor support for VCRAM in BIOS of the motherboards. Though we should notice that the memory has the following highs:
Well, VCM is quite a good choice for those who don't want to get involved in memory selection, or those who possesses a board without a compulsory setting of the working mode passing by SPD.
Write a comment below. No registration needed!