|[Home] [Credit Search] [Category Browser] [Staff Roll Call]||The LINUX.COM Article Archive|
|Originally Published: Wednesday, 17 May 2000||Author: Brian Richardson|
|Published to: enhance_articles_hardware/Hardware Reviews||Page: 1/1 - [Std View]|
The Great Camino Benchmark: Part Two
Part Two of Brian's review of the "Camino" (Intel i820) chipset. This reviews LINUX installation on an i820 motherboard, and introduces the benchmarks used for testing.
It's a Slot 1 motherboard, taking any Pentium II/III CPU. The two newest features that the average user will love are AGP 4X and UltraDMA 66. AGP 4X increases the performance using 3D video cards (like the RIVA TNT2, GeForce and S3 S540 chipsets). The UltraDMA 66 controller, when used with an 80-pin IDE cable, allows a maximum transfer rate of 66 MB/s to the hard disk drive. Both features appeal to gamers, but UltraDMA 66 by itself can be nice for file server applications.
The board has two USB slots, 5 PCI slots, 1 AGP slot, an AMR connector for AC97 modems, and on-board AC97 audio (note: AC97 audio is unlikely to work in LINUX).
In the blue corner, wearing the mauve trunks ... Our reference platform is the AMI S781 motherboard. It uses the Intel 440BX chipset, providing AGP 2X and four standard SDRAM DIMM slots. The Intel PIIX4 southbridge provides an UltraDMA 33 hard drive controller, lacking the speed of the Intel i820's controller. The 440BX is one of Intel's best chipsets, and allows the same CPU technology to be compared between both boards.
What's in the box? To make things fair, each test will be run using the exact same peripherals on each system:
Installing Corel LINUX 1.0s My first installation was Corel ... simple, painless, brainless. Corel's installer had no problems running on either test system. Both systems were up & running KDE in a matter of minutes. From the looks of things, both systems ran equally well. Corel, like most LINUX distributors, doesn't include the UDMA/66 patch in the 2.2.x kernel because it doesn't work well with all ATAPI devices ... so testing that will require a kernel patch.
After playing with Corel, I decided to start benchmarking. I used the BYTEmark LINUX/UNIX utility first (version 2, 10/95). It comes with a precompiled LINUX binary, which makes things easy for a relative newbie like myself. However, the remaining benchmarks I dowlnoaded were source-only ... and Corel's default installation neglected to install any C compilers (foo!). Rather than mess with installing extra packages, I moved on to a different distribution.
Installing Mandrake 7.0 On to a larger distro ... Mandrake 7.0 is one of my personal favorites. I installed "everything" using the custom/development settings (about 1395MB). It took longer than Corel to install, but gave me similar ease-of-use results. GNOME was up without issue on the AMI S781, and installation went the same on the SuperMicro board.
Mandrake doesn't include UDMA/66 support in 7.0 (good news, it's standard in the 2.4 kernel). But, there is a patched kernel, which allows the unmotivated slob (like me) to use faster hard drive access without the hassle of building a new kernel ... you snickering LINUX gurus should remember I'm still a newbie in many respects.
Once GNOME came to life, I began compiling the sources of the various benchmarks I had at my disposal. After weeding out the ones that wouldn't compile, didn't have good documentation, or didn't seem to apply to my style of benchmarking (many were network benchmarks, something I don't want to get into today) I had my toolbox for the task at hand.
My major benchmarks are BYTEmark v2 and UNIX Bench 4.0.1. ... text based benchmarks that have been ported to several UNIX platforms (BYTEmark even runs in DOS, for those who care). Both run in a reasonable amount of time, and put their emphasis on overall hardware performance. UNIX Bench has a file system test, which I can utilize for UDMA/66 testing. I'd like to use DOOM's "timedemo" feature, but I can't find any of my original WAD files to make the game run ... where are those pesky floppy disks? DOOM's an old game, but it's easy to setup without the need for heavy game hardware (I'm trying to avoid any benchmark that requires lots of specific hardware configurations).
Now What? In the next installment, I'll show my benchmark results. That will lead into an interpretation of what they mean for folks who want to use RAMBUS or the i820 chipset for their LINUX computing needs. This will also lead to what I hope to be a good discussion on benchmarking LINUX in general, lots of posts covering what I may have done wrong, and general negative comments from folks who regard me as "somebody who sucks".
You will find I don't do everything per the directions in the LINUX Benchmarking HOW-TO. The HOW-TO refers to X-engine as an outdated program, so I left it alone. I wanted to do the kernel build times as suggested, but I ran out of time (I have a meeting with some computer company next week, begins with a "C"). Since Whetstone is so similar to BYTEmark, I stuck with BYTEmark. I also didn't discard categories of UNIX Bench or calculate means other than those provided by the benchmark itself ... I'm sure there are valid reasons for those suggestions, but I decided to use the results as-is.
Anyhow, stay tuned for the last installment of "Brian's Excellent Benchmark Adventure" ...
Brian Richardson toils in a software labor camp just north of Atlanta, Georgia. After a twenty hour day writing useless software that does unimportant tasks (such as booting the computer and testing all of its major functionality before the OS loads), Brian panhandles on street-corners wearing a sign that says "Will Install LINUX For Food".