Intel's Z68 Chipset to Bring SSD Caching to Sandy Bridge

Obviously nobody remembers Turbo Memory. This is just embedding it into the board.

http://en.wikipedia.org/wiki/Intel_Turbo_Memory

It did very little to improve performance and actually hurt performance in several areas.

http://www.anandtech.com/show/2252

I don't think it's really valid to draw conclusions on this from that. SSD performance - and in particular size - have come a long, long way since then.

This is basically an attempt to implement ZFS's L2ARC and (hopefully) ZIL in the chipset. IMHO this is the best usage of SSDs at current (and near future) technology levels and I've been waiting for other OSes/filesystems to implement it for years (Microsoft kind of did with Vista's/7's ReadyBoost, but you can't use regular SSDs and it doesn't cache writes).

Per those that say just get a SSD, this would still be beneficial. You could have small (4GB?) super fast SSD's (think almost RAM speeds) using the new caching tech, and have a regular-speed SSD (80GB) for your "regular" data. That would provide the best bang for the buck.

Better to just spend the money on 4GB more RAM.

This sort of thing really needs 10s or 100s of GB of SSD to be effective.

The real problem (from a marketing perspective) is that it's extremely difficult to benchmark, since it requires some time (anywhere from hours to days of representative usage) for the SSD cache to properly "warm up" before the performance benefits are seen.

Why do we need a new chipset to support SSD Caching? It seems like this could be done with software on the OS level... Windows can already do this, can't it?

The OS level is the best place to do it (because it knows about the filesystem, not just blocks), but OS vendors are dragging their feet. Vista and 7 have this somewhat implemented with ReadyBoost, but the real standouts are the OSes using ZFS.

One TB port may be fine for a laptop but but a couple of years down the road and desktops will find the need for more than one port. Laptops may never have that need but that is no surprise.

I disagree. Laptops are _exactly_ where TB can be most useful due to their physically constrained expansion options. I'm hoping future Apple laptops - particularly the Air - will soon come with at least two TB, if not three or more, TB ports.

(In the PC world it's somewhat less important since docking stations with things like multiple DVI outs, eSATA and even PCIe slots are available, but the same argument applies.)

Exactly the question I had. I assume a hardware solution may be optimal but even a software solution should be able to provide a big improvement over current solutions (similar to hardware raid versus software raid).

The "optimal" solution for both these things (caching and RAID) is in software, not hardware, because the OS has greater insight into the structure, type and frequency of data being accessed, and can hence make more intelligent decisions about how to improve performance.