Promise RAID cards

[John Wray]

I'm (slowly) putting together a new analog capture system, and am thinking about how to arrange my storage. I'm planning on having a system disk and a data disk on IDE 0, DVD-R/RAM and CD-R/W on IDE 1, and a capture array on a Promise RAID controller. Is there any reason to prefer the fastTrak TX4 over the TX2? My RAID array will only be used as a RAID 0 capture drive.

Do the 4 channels of the TX4 buy anything other than hot-swap and RAID 0+1? Can the TX2 be used with four drives (in pairs, either configured as one large array, or as two separate arrays)?

There's no reason to consider the new TX2000, is there? I presume that ATA/133 doesn't make any difference for sustained transfer?

John

[IM_Riktar]

Promise TX4 gives you greater speed increases once you put 3 or more HD's on it since the TX4 has 4 (master only) ports, whereas the TX2 handles 3 or more dives by adding the drives on in a master/slave fashion.

So from a pure speed point: 2 drives = no difference (To speak of) between the TX2 and the TX4. Once you add more than that the TX4 starts to pull away in sustained transfer.

[John Wray]

Promise TX4 gives you greater speed increases once you put 3 or more HD's on it since the TX4 has 4 (master only) ports, whereas the TX2 handles 3 or more dives by adding the drives on in a master/slave fashion.

Why is that? For the sake of simplicity, assume four identical drives. If they're striped in pairs, any single access will hit one drive on channel 0 and one on channel 1, so each channel will only be talking to a single drive at any time. Is there a capture scenario where you'd really be talking to 2 drives on a single channel at once?

I could see the four channels being an advantage if you want to create two separate arrays, and access them simultaneously, as it could then overlap accesses to the two arrays, but that doesn't apply to capture scenarios (does it)?

Do both card use a single IRQ?

John

[IM_Riktar]

Why is that? For the sake of simplicity, assume four identical drives. If they're striped in pairs, any single access will hit one drive on channel 0 and one on channel 1

The scenario I am describing is taking all your drives and making them into one big array. Obviously the TX4 would win since it can handle each drive individually, while the TX2 would have to share the slave drive with the master that is on the same channel. As to the twin array scenario you are describing,,,,, not sure.

I do believe the TX4 does take up 2 IRQ's.

Hey Doc! Jump in and answer this better than what I am doing!!!

[John Wray]

The scenario I am describing is taking all your drives and making them into one big array. Obviously the TX4 would win since it can handle each drive individually, while the TX2 would have to share the slave drive with the master that is on the same channel.

Oh! Does that mean the TX4 can do a 3-way or 4-way stripe? I'd assumed it just striped across pairs of disks. For 3- or 4-way striping, the extra IDE channels make a lot of sense.

John

[Dr Mordrid]

The TX4 differs from other Fasttraks in that instead of having two headers with a master & slave drive on each it has FOUR master headers with NO slave drives.

Normally writes & reads with slave drives slow down the array because the masters have to stop for the slaves to be accessed. This doesn't happen when all the drives are masters.

The differences in speed;

"Normal" Fasttrak (FT100, TX2 etc)
1 drive array: 1x single drive speed (of course)
2 drive array: 2x single drive speed
3 drive array: 2.4-2.5x single drive speed (slave effect showing)
4 drive array: same as 3 drive array. Just a capacity increase.

Fastttrak TX4
1 drive array: 1x single drive speed (of course)
2 drive array: 2x single drive speed
3 drive array: 2.8-2.9x single drive speed
4 drive array: 3.6-3.8x single drive speed

Much of this depends on the particular drive you're using and a bit on how fast the system is overall, but you get the idea.

NOW....for one caveat!!

Close watchers of this forum will notice I have updated this system once again to use the Gigabyte GA-7VRX KT333 based board.

I have had to degrade to a Fasttrak 100 at least temporearily as I cannot get the TX4 working properly with this board/chipset under Win2K. I have not tried Win9x or XP yet. I suspect XP may have a similar problem.

The ECS K7S5A SiS-735 system had no problems at all with the TX4.

Basically Win2K reports not enough memory resources to mount the TX4's PCI Bridge chip, which is of course patently rediculous on this system. Since the Fasttrak100 has no bridge chip it's not having a problem.

More to follow once I get a better handle on what's causing this.

Dr. Mordrid

[dchip]

I want to get one of those "classified" cards.

[Dr Mordrid]

Start saving your pennies. It'll take about 550-600 pounds of 'em

Dr. Mordrid

[Brian Ellis]

I've just found out one thing with my TX-2. I originally partitioned the 2-disk array into 2 slices, one c. 70 Gb the other c. 50 Gb. The performance was disapponting. I then FDisked it into a single partition of 120 Gb and reset the Fasttrax setup: the performance improved somewhat over both the original partitions.

This does not happen with single IDE disks, although you always get the best performance with the lowest partition, no matter its size.

I'm still disappointed with the performance, though: certainly NOT the 2 x that Doc claims, more like 1.4 or 1.5 x.

The disks are identical Maxtor 60 Gb 7200 rpm jobs.

BTW, I had to reset the PCI usage of the RAID card down to about 60% of the default 100%. This made no difference to any of the benchmarked figures but it reduced drops considerably. Promise Fasttrak is a bus hog in its default condition.

Does anyone know how virtual cacheing works with Fasttrak?

[jomiwic]

I'm still disappointed with the performance, though: certainly NOT the 2 x that Doc claims, more like 1.4 or 1.5 x.

after 2 drives you start to max out the data throughput on the PCI bus(theoretically, 3 drives would give you alot better performance than, in reality, it does)

the tx4 would be better suited for the 64bit PCI bus than the current 32bit

[Dr Mordrid]

Did you also install the Fastcheck utility?

IF not then do so and turn off the S.M.A.R.T. feature in its Options tab. Make sure Fastcheck runs every time you start up so the setting will stick.

This is very important. Almost as important as turning off write verification on all ATA100 and ATA133 Maxtor drives.

Dr. Mordrid

[John Wray]

Almost as important as turning off write verification on all ATA100 and ATA133 Maxtor drives.

The newer Maxtor drives still turn off write verification by themselves after 10 power-ups, don't they? If so, it's presumably easiest to let the drive handle it itself, as the first ten power-ups might find a defective OOTB drive.

John