Announcement

**Dr Mordrid** · 1 May 2002, 17:58

1. make sure that write verification is turned OFF on Maxtor drives of ATA100 or ATA133 design. This is on by default on new drives and is supposed to turn itself off after 10 power-ups. Unfortunately it doesn't always work that way. Maxtor has a utility to manually turn it off on their download page. Look for WVSET.EXE. Turning off WV ups performance considerably.

2. Make sure you install the FastCheck tool that came with your Fasttrak and choose the option for it to run at every bootup. Once installed go to the Options page and turn OFF the S.M.A.R.T. feature (which is ON by default). S.M.A.R.T. slows down the array by constantly checking the drives status. Speeds should increase significantly by turning it off.

3. On Fasttraks other than the TX4; if you are running a two drive array make sure they are NOT on the same cable. Otherwise the drive on the master connection (end of the cable) will prevent simultaneous interleaved writes to its partner, thus slowing down the array as a whole.

4. Speaking of cables, note that I mentioned that the end cable connector is the MASTER. This is a mistake a lot of people make, especially with ATA100 and ATA133 drives. IT DOES MATTER which connector you use for which drive. While the drive will work on the opposite connector, it may not work at peak performance.

In the case of a single drive connected to a cable it should be configured as a MASTER and put on the end connecttor.

Dr. Mordrid

**Sciascia** · 2 May 2002, 05:48

In the case of a single drive connected to a cable it should be configured as a MASTER and put on the end connecttor.

OK, here is a question for you. In an effort to clean up my case I cut the master connectors off of the cables (all I have is the slave connectors) to shorten them. I only needed about 6-8" of length and the excess was just a mess. I have noticed that my throughput does not seem to be as fast as I thought it should be.

What is the difference between the 2 connectors? I always thought that they were the same pin out.

**Brian Ellis** · 2 May 2002, 06:37

Aha! One of the mysteries of electronics, called characteristic impedance and ground loops. Have you ever wondered why a 40 pin connector should have an 80 conductor cable?

The answer is because high-speed switching requires clean signals. In order to achieve this, it is first of all necessary that the inductance/capacitance ratio is constant. This is done by inserting a ground wire between adjacent signal wires. This produces a cable with a finite characteristic impedance. If the far end is correctly terminated with a resistor of the same value, the signal travels up the cable, close to the speed of light, and that is that. If the termination does not have the same resistance, the signal will be reflected back and this can have two effects: the returned signal, typically a nanosecond or so later, may cause the state of the switch at the originating end to change, depending on the amplitude of the reflection. Worse, electromagnetic radiation may result and cause neighbouring conductors to pick up crosstalk which would be interpreted as a different state: result? Chaos.

Now, if the intermediate ground wires were connected at both ends of the cable, they would form closed loops, which would have inductive signals induced in them from the inductance of the signal wires, and a current would flow round the loop. This would upset the characteristic impedance and the crosstalk suppression. Result? Chaos.

I have not actually taken the connectors of these cables apart to verify what I'm about to say, but I believe that there is at least a 98% chance it is correct: The logical grounding for the intermediate wires is inside the blue plug. Only one single ground wire will be connected to the black plugs, so that no loop can be formed. The single ground wire connection is probably one of the two outside wires.

Therefore, if you cut off the blue plug, the 39 intermediate wires will be floating. Result: impedance mismatching and probable chaos. At the best, the system will assume it is not an ATA connection.

I would guess that if you cut off the end black connector as close to the middle connector as possible (flush with the connector body), making sure there were no short-circuits, it would work OK. As Doc says, if you have the end black connector there, but unconnected, you will have an unterminated stub and this will also cause reflections, albeit more attenuated. You must plug the end connector into your master drive, but I'm reasonably confident that it does not matter whether it is the original end one or the cut-down end one. If my hypothesis is correct, then the most important thing is that the blue connector be plugged into the m/b or RAID board.

Sorry if I've gone too techie.

**Dr Mordrid** · 2 May 2002, 06:41

In addition to what Brian has noted each connector has a pin absent on the other. These are mainly used for cable select, but there are other signaling considerations as well.

Another thing you want to avoid is using the old 40 conductor cables witih dives rated at ATA66 and above. These drives require 80 conductor cables, which have a ground for each signal wire (as noted by Brian) as well as additional signaling absent in the 40 conductor cables.

40 conductor cables should only be used with ATA33 and ATA16 drives.

Dr. Mordrid

**Sciascia** · 2 May 2002, 07:59

Brian,

Good explanation, a little in depth, just how I like them. When I did cut the cables, it was with a razor, right at the base of the connector and I did make sure there were no shorts. Being a little on the anal side in most cases, I wouldn't have used it if I thought there was a achance of a short. I am not rich enough for that game

.

I am using 80 pin cables here, just wondering how much of a difference there is between the actual master and slave connector. I can probablyu scrounge up a few cables to do a little comparison testing. It would give me something to do this evening.

**KRSESQ** · 2 May 2002, 20:13

Update:

No luck so far. Stats remain more or less unchanged. I should note that the buffered read/write speed for the 17 gig is 21 MB/sec, and on the RAID the buffered read time is 33 MB/sec and the buffered write time is 20 MB/sec. It still doesn't prevent dropped frames using anything other than Vid Tools for capture. Suppose I should count my blessings.

Kevin

**Sciascia** · 5 May 2002, 03:44

HELP...

I replaced the cables and still have no luck. throughput seems to be about half of what it should be. I have 2 WD400 (40G) HD s and made sure that ATA100 is enabled. I also have a single WD600 (60G) for my system drive on the MB IDE controller. IT is actually faster than the RAID system.

Does anybody have any ideas? I tried swapping slots and rebuilding the array in different configs, played with the PCI utilization slider, made sure that SMART is disabled and played with the write buffer. I also disabled my firewire card and removed all of my USB devices. No improvement.

I am at a loss of ideas now. Anyone have some suggestions?

**Brian Ellis** · 5 May 2002, 04:13

Sciasia

No comfort for you ... but join the club! I have exactly the same problems. Maybe we should rename Fasttrak as Snailtrak :-( In my case, I've a TX-2 striped array with 2 Maxtor 60 Gb 7200 rpm platters: it is barely faster than the 30 Gb 5400 rpm C: disk and does cause drops. I found a marginal improvement by reducing the PCI usage to about 60%, which makes me thing that the bugger is a bus hog, far worse than Live! I even replaced the Live! out of the machine and put in an SB16, thinking that two hogs may be worse than one, but that made no difference, whatsoever. There is no IRQ conflict (or any other) on the machine.

I conclude that, at least as far as my machine is concerned, Snailtrak simply does not live up to its Promise

Your machine is totally dissimilar to mine (PIII/450, Intel chipset, 384 Mb 100 MHz SDRAM, Bravo Baby m/b, SB16, 3Com network, Marvel G-200, Mitsumi CD drive, Pioneer AO-3 DVD/CD burner), to the extent that we have nothing in common other than a Promise card.

**Sciascia** · 5 May 2002, 06:01

Brian

What OS are you using these days? I wonder if it could in fact be a WinXP issue as was mentioned earlier. If that is the case, I wonder why there aren't more reports on it. As far as I know others are getting great performance from their Promise RAID controllers, but these days the majority are using the MB's with built in controllers. I doubt that there could be much difference between the two, I imagine it is pretty much a Promise controller just hardwired to the PCI bus.

Is anybody using WinXP with a Fasttrak 100 successfully?

**Brian Ellis** · 5 May 2002, 08:07

AH! Sorry to disappoint you, mine is on WIN98SE (I use Marvel G-200 in hardware mode).

Doesn't XP require Dynamic Disk to get it to work properly?

**Dr Mordrid** · 6 May 2002, 10:42

I know you don't want to hear this, but I have both FT100 and TX4's running on my XP betaing systems. They're slightly slower in XP, but certainly are not slouches.

The 4 drive Maxtor D740X array is benching at 98 mb/s sequential writes in both the SANDRA and the RT2K benchmark and about 94 MB/s sequential reads (Win2K).

WinXP shows about 90 mb/s writes and 88 mb/s reads.

Dr. Mordrid

**Sciascia** · 6 May 2002, 11:41

Well with 4 drive arrays, you are about double of what I am benchmarking with my piddly 2 drive array. So that means that it should be operating properly now.

Sequential read 55mb/s

Sequential write: 41 mb/s

About right?

**Dr Mordrid** · 6 May 2002, 12:39

Probably as fast as those poor l'il WD's can go

Dr. Mordrid

**Brian Ellis** · 7 May 2002, 01:27

Huh! Sandra gives me best scores of
Seq. Read 19 Mb/s (which is bloody awful)
Seq. Write 43 Mb/s (which is acceptable)
Av. Access 6 ms (which ain't bad)
Index 18384
The buffered read is 149 Mb/s but the buffered write is 42 Mb/s.

Promise RAID 0 striped on 2 Maxtor 60 Mb 7200, no partitioning. The daft thing is I'm getting drops on capture, not on playback. Does this mean that it is using the buffers?

Win98SE, PIII/450: 384 Mb RAM. Marvel G-200, SB16 ATA.

Announcement

Mediocre Fasttrak Performance?

Mediocre Fasttrak Performance?

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment