Ok, so I've got a spare SCSI controller lying around, and a spare LVD-80 cable to go with it. I've been wondering if this could be done, and a few other people have recently asked about it in this forum, so I figured what the hell, why not?
What I'm talking about of course is rounding LVD cables. Splitting the wires apart is decidedly safer than with IDE cables, because all you've got to do is get a small rip going in the clear (saran wrap-esque) plastic wrapping. No knives involved, no chance of exposing the wires. The issue of capacitance (or interference) is already taken care of because the wires are twisted off into pairs.
I started off seperating each pair of wires, but besides being way more work than rounding a cable should be, I noticed that the twists in the pairs were becoming much looser. So I decided to seperate them into pairs of 5 or so, and then to tear from there.
Due to the fact that it's really late (or early), I decided to only round one segment of the cable. I decided to round the middle section, and to stick my test drive at the very end of the cable (well, technically the terminator is at the very end), so that if there were any signaling or latency or what-have-you problems introduced by this rounding, they'd have plenty of time to show up.
blah blah blah, you get the idea. on to testing.
First off, the drive was detected in the SCSI bios, which was a great sign. Windows detected it and I was able to access it, browse the contents, and play a few MP3s from the drive. Everything great so far. For my main testing I decided to use Adaptec SCSIBench. Plusses for this were: allows testing to loop until I stop it, and it allows me to test random, sequential, and same sector I/O tests. Since I wanted to know if rounding reduced the performance or maximum bandwidth, the same sector I/O tests are important, as they simply measure how fast data can be transferred from the drive's cache.
Using a transfer size of 128KBytes, same sector I/O, this test has been running for about 10 minutes now with no errors. Transfer rate is hovering around 70MBytes/sec, very close to the theoretical limit of 80MBytes/sec, especially taking overhead into account. This is as well as this drive has ever performed on that test, so I can conclude that there was no loss in performance, at least not in the area of maximum bandwidth.
Sequential I/O drops down to about 20MBytes/sec, which again is typical for this drive, no loss in performance here.
Random I/O drops the furthest to about 6.5MBytes/sec, which is still typical.
I can even play MP3s from the hard drive while performing these tests, no problem.
My preliminary conclusion: rounding resulted in no degredation in performance or stability.
I will eventually round the entire cable and perform more extensive tests on it. I would be happy to test this on an LVD-160 cable as well if anyone has one lying around that they'd like to donate to the cause of greater knowledge for anandtech users.
What I'm talking about of course is rounding LVD cables. Splitting the wires apart is decidedly safer than with IDE cables, because all you've got to do is get a small rip going in the clear (saran wrap-esque) plastic wrapping. No knives involved, no chance of exposing the wires. The issue of capacitance (or interference) is already taken care of because the wires are twisted off into pairs.
I started off seperating each pair of wires, but besides being way more work than rounding a cable should be, I noticed that the twists in the pairs were becoming much looser. So I decided to seperate them into pairs of 5 or so, and then to tear from there.
Due to the fact that it's really late (or early), I decided to only round one segment of the cable. I decided to round the middle section, and to stick my test drive at the very end of the cable (well, technically the terminator is at the very end), so that if there were any signaling or latency or what-have-you problems introduced by this rounding, they'd have plenty of time to show up.
blah blah blah, you get the idea. on to testing.
First off, the drive was detected in the SCSI bios, which was a great sign. Windows detected it and I was able to access it, browse the contents, and play a few MP3s from the drive. Everything great so far. For my main testing I decided to use Adaptec SCSIBench. Plusses for this were: allows testing to loop until I stop it, and it allows me to test random, sequential, and same sector I/O tests. Since I wanted to know if rounding reduced the performance or maximum bandwidth, the same sector I/O tests are important, as they simply measure how fast data can be transferred from the drive's cache.
Using a transfer size of 128KBytes, same sector I/O, this test has been running for about 10 minutes now with no errors. Transfer rate is hovering around 70MBytes/sec, very close to the theoretical limit of 80MBytes/sec, especially taking overhead into account. This is as well as this drive has ever performed on that test, so I can conclude that there was no loss in performance, at least not in the area of maximum bandwidth.
Sequential I/O drops down to about 20MBytes/sec, which again is typical for this drive, no loss in performance here.
Random I/O drops the furthest to about 6.5MBytes/sec, which is still typical.
I can even play MP3s from the hard drive while performing these tests, no problem.
My preliminary conclusion: rounding resulted in no degredation in performance or stability.
I will eventually round the entire cable and perform more extensive tests on it. I would be happy to test this on an LVD-160 cable as well if anyone has one lying around that they'd like to donate to the cause of greater knowledge for anandtech users.
Facebook
Twitter