We noticed when we first starting using the ABIT board that it overclocks the front-side bus by 4MHz by default. This isn't necessarily a bad thing but we wanted to point out that we manually set the front-side bus back to default for benchmarking. We'll take a more in-depth look at this board in a future article and explore more of the features that it has to offer.
Support
So what did it take to get these two platforms going this time?
Well, one platform was relatively painless while the other took some doing. First, we'll look at the more difficult platform.
The Intel platform contained a lot of new technology and so some pain was expected.
In the beginning we were waiting on a new NVIDIA driver release in order to support our new PCI-Express NVIDIA GeForce 6800GT.
This driver was finally released and then we went on to the other components.
The next driver we were looking for was the Ethernet driver for the on-board chip.
The on-board Gigabit controller is not Intel's usual chip but a chip made my Marvell.
The stock kernel drivers did not work, so we contacted Intel.
Intel had apparently been working on Linux support because they quickly supplied us with a kernel patch that worked very well and was easy to install. Finally, we wanted to get the on-board audio chip working for testing.
When we contacted Intel for this, we were told that it was in the works but not ready yet.
Several weeks later, we received an e-mail with the audio drivers and went back to work.
The Intel audio drivers we received weren't quite complete, stable, or easy to install.
We did get them working though and we thank Intel for the effort.
I'm sure they will mature over time. We were able to get audio and as long as it was an ALSA native application all seemed to work well. The real difficulty was with anything that needed OSS emulation. This produced very poor sound quality with quite a bit of distortion. Also, if we did not load the emulation kernel modules before the actual driver module, the kernel would usually panic. This was a problem but Intel was kind enough to supply the work-around of loading the module first. At this point all components work and only the audio driver has some improving to do. You can download the kernel config file that we used here: config-D925XCV You can find Intel's drivers for the D925XCV HERE
Since there isn't really any new technology that would need new drivers, the ABIT AV8 was very easy to install a 64-bit or 32-bit Linux on. The only driver that really gave us problems was for the VIA Velocity Gigabit Ethernet chip on-board. The driver is only available in the latest kernels and it's misplaced under the 10/100 drivers in the kernel configuration. Once built-in to the kernel, we found that our kernel panicked as soon as we brought up the device on boot. In order to bring the Ethernet up without a panic, we had to load the driver manually as a module and then bring the device up. I'm sure this will be fixed soon and it was only a minor annoyance. If it would have been really bad, we could have simply popped in a PCI Ethernet card. All other devices worked as expected and even the on-board sound produced good-quality audio. You can download our 32-bit kernel config: config-32bit-ABITAV8 and 64-bit kernel config: config-64bit-ABITAV8
For all of our testing, kernel 2.6.8 was used. Please note that earlier kernel versions may not have all of the drivers mentioned above. If something appears to be unavailable, please upgrade to the latest kernel.
Stability We remember a day where this section of the article used to contain a lot of information about which hardware wasn't stable and which components you shouldn't use with other components. Well those days seem to be over as both Intel and AMD seem to have gotten over any stability problems that may have once plagued them. Both of the platforms tested below were completely stable in our testing. Actually, the only thing we have to report here is that we did experience some initial stability problems on the ASUS A8V that AMD sent over with the Athlon 64 3800+ processor. This stability seems to have stemmed from an early board revision and not from AMD's processor or VIA's chipset. Once we replaced it with ABIT's AV8, all was smooth sailing. Intel's board was as stable as they usually are. Since their boards configure themselves to stable settings by default, there was little to mess with and it all just worked as expected
Below on the left, we have Cosair's latest DDR memory, the 3200XL series. These modules are designed to provide absolutely the lowest memory latency timings possible. We're talking timings of 2-2-2-5 here! This should give any system an edge over typical memory and should seriously be considered if you'd like your memory vendor to guarantee those timings. In our testing we used this memory in both our Athlon 64 system and our Intel 875P system
Above on the right, you'll see the drive we used in all of our setups. The new Maxtor MaXLine III drive is one of the first SATA drives which supports a technology called Native Command Queuing (NCQ). NCQ is a technology which allows the drive to reorder data in a more efficient order for reading and writing. This is a technology seen in SCSI drives and it's finally making it to the desktop market. For this technology to work, both the controller and drive must support NCQ. The 925X supports NCQ while the VIA K8T800 Pro does not. Unfortunately though, not even Intel will benefit from this technology under Linux until the kernel driver supports it. NCQ's Linux driver components are being developed by Jeff Garzik. We contacted him about this and were told that something should be available in 4-6 weeks. That was about 2 weeks ago. He directed us to http://linux-ata.org/driver-status.html for future information. We'll let you know when this support becomes available.