Subsystem Testing

The first thing we'll check is the audio. We downloaded and installed to test its CPU utilization.

Like the other Realtek solutions we've used in the past, CPU utilization was fairly high thoughout the DirectSound3D tests. CPU utilization never got to 9%, but it did average in the 4% - 6% range, which is a lot higher than the <2% averages we've seen with the nForce 2.

It's not the end of the world, given that the slowest CPU you would be using with the K8T is an Athlon 64 3200+, but we would have liked to have seen lower numbers. This is a synthetic benchmark though, and since I know all of you enjoy a game or two, let's see how the sound will affect UT 2003 performance.

UT 2003 Antalus Sound Tests

For our UT2003 tests, we ran dm-Antalus benchmarks at 640x480, minimum detail with sound on and off. This was repeated at 1024x768, but with maximum detail. The reasoning is at low detail and resolution, the work will fall on the CPU and motherboard subsystem. Higher resolution is more representive of actual gameplay for most users

With sound enabled, the MSI K8T Neo-FIS2R takes an 8 frames per second hit at 640x480. These results are a little better than what we've seen in the past as 15 - 20 frames per second used to be the average drop.

At a higher resolution and detail levels, the onboard sound's CPU utilization will not be a factor at all. There is less than 0.05% loss in speed when using the onboard sound at this resolution, therefore, I wouldn't hesitate in using the onboard sound for general use and gaming.

In terms of sound quality, I found gaming to be very acceptable, as was the case with movie and MP3 playback. There was some slight distortion when running disk intensive tasks though, as slight static came from my Logitech Z-560s while playing a MP3 off the hard drive.

For recording tests, I used a small microphone that came with my Audigy Platinum, and recorded a few sentences. The recordings sounded as they should, with no crackling or distortion, though sound quality was slightly muffled. Compared to the quality on the MSI nForce 2, there was a difference, so some work needs to be done here.

Hard Drive Performance

We used HD Tach to gauge read and write performance with our Western Digital SE 120GB HDD, and read performance with our Seagate 120GB SATA drive. Unfortunently, I don't have an identical drives to perform RAID testing, but we'll try to follow this up when I can acquire a second drive.

Tests were done using freshly imaged disks, split into partitions of 40GB, 50GB, 30GB (roughly), and benched on both the VIA and Promise controllers.

PATA Performance

I've seen reports of the Promise controller outperforming the VIA controller, but that wasn't the case here (mostly). The Promise controller does edge out the VIA in CPU usage, eating 18.7%, compared to VIA's 19.7%.

SATA Performance

On the VIA controller, the maximum read speed with the Seagate SATA drive was about 60MB/sec, with 45MB/sec being the average. CPU usage was about 20%, simialr to PATA performance. The Promise averaged a slightly lower read speed, but the CPU usage was also lower at 18%.

Network Performance

We used to test the networking speed, and Windows Task Manager for CPU usage. We copied a variety of install files, totalling 758 MB, varying in sizes of 300kb to as much as 60MB per file from the K8T Neo-FIS2R machine, to our ABIT IC7-MAX3 box, which uses an Intel Gigabit CSA controller. We also performed the same test with an ISO image, totalling 761MB.

Both systems were connected via a CAT-5E crossover cable, which should prevent any bottlenecks that would arise with our standard 10/100 router.

Small Files Test - 758MB Total

The Realtek 8110S did a good job with the file transfers, and was about seven seconds faster at uploading the files than it was downloading. However, CPU usage was quite high while uploading... about 55%, whereas downloading averaged 35% according to Task Manager.

Large File Test - 761MB Total

With a large, single file, times were improved in both the upload and download tests when compared to the smaller files. CPU usage mirrored the small file tests, but on average, about 5% less. Overall, I do find these CPU usage results to be fairly high, as Intel's and nVidia's (nVidia is a 10/100 part) solutions are about 30% for the same files.