The FIC KW-15 is an interesting mix of features with an emphasis on flexibility. The most immediate thing that sticks out about this board is it’s dual slot architecture, which means support for both Slot 1 and Socket 370 processors. Having the capability to stuff either processor might make a lot of sense in companies that tend to hand down processors and do a lot of component sharing over the life of a computer. I could foresee great benefit in standardizing on a board, like the KW15, that could take a wide variety of different processors depending on the needs of the end user. Perhaps a cheap Celeron for the basic clerical type work and maybe a faster PIII for someone involved in content creation.
The next thing to notice is the slot complement. There are 5 PCI slots and an AMR slot. There is an OEM option to stuff a single ISA slot, but this will probably not be a common occurrence and wasn’t stuffed on my review sample. The integrated video take the place of an AGP slot, which means only a PCI video card could be used in place of the integrated video.
The AMR (audio, modem riser) slot is something integrators use to save money and customers disdain for performance reasons. For those users who don’t know, an AMR slot takes a plug in module that gives software based modem or audio capability to the system by having the processor emulate the hardware functions in software. This method is usually a cheap means of implementing this capability, but is inferior in terms of system performance. While emulating the modem, your processor is being robbed of CPU cycles, cutting into your ability to multi-task with other applications efficiently.
Only two DIMM slots are available on the KW15, which is the usual configuration on i810e boards. These DIMM slots can each handle up to 256MB modules for a maximum of 512MB. I don’t have a problem with 512MB being the maximum since this is enough for all but the most power hungry users. However, two DIMM slots usually don’t start out with large capacity DIMMs, but get upgraded over time and the most likely scenario is to want to add in more lower capacity DIMMs over the lifetime of the board, which makes having only two slots somewhat limiting. I’d prefer to see at least three on most boards just to increase upgrade possibilities.
The integrated nature of the i810e chipset means that it supports the audio codec standard AC’97. Intel’s support of AC’97 comes by way of the 810ICH chip onboard. The audio portion supports 3D positional audio, A3D, Q3D, wavetable and 3D surround sound. The audio support would be fine in business settings where you might want to listen to an occasional multimedia CD or other audio sources in a work environment. Home users that are big on playing MP3 or mastering music might want greater HW based capability, but most users that just need basic audio support will be happy. The audio can be disabled to use either an AMR audio solution or a PCI plug in card.
The integrated video consists of a 230MHz RAMDAC capable of running 8 bit color depths up to 1600×1200 resolutions at a vertical refresh rate of 85Hz. There is 4MB of SDRAM display cache installed on the mainboard and a technology called DVMT (Dynamic Video Memory Technology) which the manual claims provides the equivalent of 8MB video memory performance. No one is going to win many Quake tournaments using the integrated video on this board. While it can run 3D games, it vastly inferior to high-end AGP boards like the Nvidia Geforce. But while running office apps, it plugs right along and would be a reasonable solution for users of 2D applications on 17 or 19" monitors. This wouldn’t be my choice for running larger 21" monitors. The video can be disabled and an add-on PCI video card can be used as an option.
The KW15 employs an ICS9248BF-87 clock chip for the generation of the various CPU and Bus frequencies. This gives the board the ability to support the standard 66, 100 and 133MHz CPU frequencies as well as a few others that fill in the cracks. This board is not meant to be an overclockers dream board and will function best at the standard recommended bus speeds. I did have some limited success overclocking a few of my processors, but not under all conditions. I will cover this in the diagnostics section.
The KW15 is a standard ATX form factor. The ATX power connector is well placed at the very top of the board making it easy to keep the main power harness out of the way of the fan/heatsink for the S370 part. The side location of the floppy and IDE connectors are easily accessible and will pose no interference problems. I did find a few of the jumpers uncomfortably placed for easy changing. There are two jumpers that need adjustment depending on which processor type you use, J2 and J10. J2 was crammed up against the Slot 1 socket so close that it was tough to change even with needle nosed pliers.
When running a FCPGA Coppermine PIII in the S370 socket, it’s necessary to place an R/C based terminator board in the Slot 1 slot in order to properly terminate the traces that are shared by both sockets. Once installed and the memory is in, it’s very difficult to place a fan on the S370 socket. The terminator board may need to be removed in order to get the fan placed properly. Other than those little caveats, the board should fit well in most ATX cases.
The included manual is adequate in its explanation of the hardwired jumpers and interface cable placement. I would like to have seen more effort placed on the bus speed selection section, since this was not clear and needed to be viewed in the BIOS in order to know what was really available. The board could have also benefited from better silkscreened text next to the jumpers and headers to explain their functions. I found myself having to look at the manual a lot when clear silk screening would have improved the useability.
Following the RWT tradition of thoroughly testing the board, I ran numerous iterations of benchmark software under both Windows98SE and NT4 with SP6. I ran three processors during my testing: Intel Celeron 433MHz Slot 1, Intel Celeron 466MHz S370 and an FCPGA PIII 550MHz processor by itself and on an IWILL Slocket II adapter. Each processor was run under each OS at length and showed few problems.
I started out with Windows98SE and ran the batch mode tests of the ZD Winstone 99 1.2 and Content Creation Winstone 2000 over 8 hours each while rebooting between each run without seeing any system hangs or problems. I also employed Passmark’s Burn-In for over 24 hours without any hiccups. With both DIMM sockets stuffed with two 64MB SDRAM modules, I also ran the BCM Memory Stress tests for a full day without any data errors being flagged.
I then installed Windows NT 4.0 with SP6 and reran all the benchmarks with the various test processors and did not see a failure. The batch mode of Winstone99 and CCWinstone2000 is a pain because of the need to log into NT during a reboot, which requires attended operation. I therefore executed only a half dozen runs with each processor, rather than an 8 hour straight thrashing.
Going back and forth between a Slot1 and Socket370 processor obviously took a few jumper changes and occasionally required going in and clearing the CMOS settings to reinitialize bus speeds. This confusion can be common to many boards and isn’t something I worry about. I initially had trouble with any Socket370 chip until I realized the R/C terminator board included in the box had a real purpose and helped the situation tremendously.
R/C Terminator Kit
The manual states that the terminator board is required for all FCPGA processors, but considering it’s function, I would recommend it’s use any time you’re using an S370 part. I was disappointed when attempting to overclock the PIII 550MHz FCPGA and found that FIC had not implemented a ¼ divisor for the CPU speed. At 133MHz, the PCI slots were at 44MHz (implying a 1/3 divisor), but not the desired PCI speed of 33MHz. This is most likely just a BIOS tweak because the data sheet on the ICS 9248-78 clock generator chip employed is capable of a ¼ PCI divisor. Hopefully FIC will eventually release a BIOS fix for this specific problem.
Interestingly when using an IWILL slocket II jumpered for 133MHz bus speed, the BIOS gave the indication that the PCI bus was at the desired 33MHz. When using the FCPGA chip at either a default setting in the IWILL slocket II or in the S370 socket, the 133MHz bus speed yielded a 44MHz PCI clock speed. This increased PCI clock could potentially cause some PCI cards to exhibit stability problems and probably should be avoided in business situations. I was also able to run the 466MHz Celeron (normally using a 66MHz FSB) at 75MHz for a total of 525MHz without any stability problems as well
There were some differences in the two processor slots that showed up under overclocking situations. I was able to overclock the PIII 550MHz FCPGA to 733MHz (a 133MHz FSB speed and PCI bus speed of 44MHz) using an IWILL slocket II in the Slot 1 position without any signs of stability problems. Attempting to do the same thing with the FCPGA in the S370 yielded disaster with no video at boot. I have a feeling the terminator board was included to help out this S370 stability situation, but having longer trace lengths going to multiple slots from the chipset is the downside to dual slot architectures. There are some overclocking options for this board, but most business will run this at default speeds for data integrity reasons.
A small observation was the soft ATX power switch employed during the testing didn’t always work as expected. During some testing, the power was on immediately after the supply was turned on, even when the soft power switch had not been touched. If the power was properly shut down using this switch it subsequently worked as expected, but if power were inadvertently shut off it would not function properly. Maybe a BIOS fix could solve this.
Be the first to discuss this article!