Slashdot Mirror
Linux Intel Chipset Comparison
Diabolus writes: "AnandTech have done a comparison of Intel-CPU chipsets on Linux here. It talks about performance, stability and support issues for the various chipsets; apparently an Athlon chipset comparison is due shortly. Nice to see that it's not just 3D performance now -- Linux is becoming more mainstream among the h/w enthusiast crowd." This is a cool followup to see to AnandTech's October comparison of video cards under Linux.
I've been running a dual boot system for the last week. Working between the two on the same system has proven to me that Linux runs better on the setup that I have. Far more stable. The processing seems to happen much faster. I have a piii 500 with a 7200 rpm 27 gig drive (fat32/win98 single partition), and a 5400 rpm 3.2 gig(3 partition, 100mb swap). I had been running on an ATA66 but that's been removed for the Red Hat 7 Install. So I think it might be at 33 onboard.
To make this long, not very technical, story short, Linux has proven it's speed and stability with the same hardware. What I find interesting i that Linux is running on the slower drive but it still runs faster.
a/s/l here. Sorry, adding domain tags to your s
It was my understanding that the "big" benchmark Spec (the one that supposedly everyone relies on) is platform-independent. That it doesn't really use a "OS" to run, per se.
- I don't care if they globalize against free speech. All my best free thoughts are done in my head.
Probobly because when you build your system, you bought parts that you wanted.
You said you bought Dell, well a lot of people recomend that brand. But I found them sleeping in bed with Intel. No matter how slow/expensive Intel chips are, they still buy them.
A lot of things can determine the speed of the system not just CPU. For instance you could get RAID with 2HD, this will double your HD performace(if you use striping 0, or 0+1) HD performance is where you gonna see the difference(you always buy the fastes HD available) because it's the slowest link in the system, Memory is also important I found 128 is not enought if you use more then one application.
How about a new symbol to go alongside the ubiquitous 'Designed for Windows {95/98/ME/NT}' and 'plug & pray' ones on new hardware? Something like 'Designed for Linux - designed to work', or even just a pretty picture so you know it's tested and includes Linux drivers.
Then I might actually be able to tell if my new kit works on a decent OS before forking out the hardearned or waiting for someone else to try it first!
JJ
I find it quite curious that there is no explaniation of why the BX chipset gave the best results in the real-world test (kernel compile) while it lagged behind (albeit only slightly) in every feature-specific test.
--I assume full responsibility for my actions, except the ones that are someone else's fault.
By definition, what a benchmark is is just a measurement of the speed of a particular piece of hardware at running a particular piece of software. That includes the OS. It makes sense to compare different pieces of software on the same hardware to compare how fast they are at solving a particular problem, or the same software on different hardware to see how fast the hardware is at a particular task, or even different pieces of hardware and software performing a similar task. The win98 benchmarks are (barring bad benchmarking techniques, etc) as accurate at measuring the win98 solution to a problem as the linux ones are at measuring the linux solution to the same problem. They may be slower, or faster, but they fundamentaly are no more accurate. They both tell how fast the hardware/software configuration is at a task.
What really matters is what task you care about, and I'm guessing you care about linux doing something, so linux benchmarks would have more meaning to you. For various reasons, I run windows (and can't switch easily enough to linux, but that's a different discussion), so the windows ones matter more to me. However, I am quite glad to see both being run.
Not only that, but all the JavaScript locked up my browser.
They completely ignored the higher-end chipsets like Intel's 840 and the ServerWorks ServerSet III; it would have been nice to see those mentioned.
...the 82443BX is still the best one that Intel has released so far. The 820, 815, and 840 are all so plagued with problems due to the Rambus incident (the entire existence of Rambus can be classified as an incident; when Rambus dies, progress continues). Not until Intel releases their DDR chipset will another viable Intel offering be seen.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
The benchmark results varied significantly, particularly for one RAID conroller which gave utterly appalling results. It then dawned on me why this was the case. The magazine had performed all their benchmarks under Windows95, yet the RAID controller had no Windows95 drivers available. For this controller they had run their benchmarks using the 16-bit emulation mode.
This is not the first time I've seen such benchmarks run. I've lost count of the number of IDE based systems I've seen not running in optimum UDMA mode.
Coincidence
I hope they do come out with a review for linux but that just might take time. SO what have your experiences been under linux ? Anything people like me should be keeping an eye out for ? Or something you would definitely recommend ?