>Ed,
>
>Looks like what I really want is an Adaptec 29160 which apparently superceeds the 2940U2W. The 29160 evidently allows for narrow and wide adapters
at the same time without having to degrade the wide speeds downward.
>
>I found the 2940U2W card locally for US$197.00 + tax which isn't bad so now the dilema is newer/older technology and higher/lower costs..
>
If it's a single wide drive and a bunch of older narrow drives, even the original U2W can run the 68 pin connector LVD at U2 speed and the 50 pin at Ultra speed; there was a change in the U2W at one point where they integrated independent channel termination, just like the 29160, allowing for both a single-ended 68 pin and an LVD 68 pin connection - they coexist, so that the LVD drives aren't forced to be compatible with the SE drives; this uses some capabilities first exploited by some companies that built stuff around Adaptec's 7895 chipset using the 6380 to electrically decouple the LVD 68 pin channel termination from the standard SE termination. The 29160 is needed to take full advantage of Ultra-160, but without multiple simultaneous LVD targets active, it's unlikely that you'll saturate the Ultra2 bus to see the difference, and honestly, if I were going to look at really high-performance storage, I'd want to look at either fibre channel or the 1394 FireWire stuff.
You can have both narrow and wide active on the older U2Ws, but it's all commonly terminated, so only 2 connectors can be active at the same time - one continuous bus with termination at each end, and any SE forced all devices to clock to be compatible with the SE channel. The 6380 provided for independent termination of each connector, allowing for both LVD and SE chains, and >2 connectors active. The 29160 has the even higher clock support, but single device you'll never see the difference. Software RAID is another story.
>That and the cost of the adapter - 68<--->80 pin from the folks at
www.scsipro.com.
>
>Some things never really do change... <g>
>
>I guess I need to see if the 29U2W can do this. If not I'll have to open my safe. <g>
>
FWIW, I still maintain that having multiple SCSI channels, even at a slower clock, will tend to yield better throughput because of reduced contention for the bus, and the fact that most drives are limited by the bandwidth at the platter not the SCSI bus bandwidth. YMMV, but I consistently get better performance off the older 7890 dual-channel Ultra with the 1130CA RAIDPort on the SuperMicro P6DGS than a single channel U2-compatible 7895 chipset with the RAIDPort III option like the Asus P2B-S used.