How About iRAM Reloaded? . . .

Gigabyte has released a new motherboard. The Inquirer mentions it here, and Gigabyte manages to say something about it, too.

It uses Intel’s next-generation 975X chipset, but that isn’t what makes it interesting. It features improved circuit cooling, which may prove interesting, but we’ll have to see about that.

No, what makes this interesting are two little slots this card has. PCI Express x4 slots.

What’s so interesting about that? Well, a PCI Express x4 slot allows transfers of 1GB/sec.

This is over seven times faster than a ATA/133 connection, over six times faster than a SATA1 connection, and over three times faster than a SATA2 connection.

Is there anything that isn’t a video card out there which could really use that kind of speed? I can only think of one: the iRAM card.

And who makes the iRAM card? Gigabyte.

What a coincidence!

Uh, oh, maybe it is a coincidence. We’d better do something about that.

The main problem with the iRAM card is that it is throttled, stunted, blocked by the speed of the SATA bus. It is this bottleneck which makes much slower but cheaper flash memory-based solid state drives competitive or better against it.

Really, if your coronary arteries were as obstructed as the iRAM card is by the SATA interface, you’d be in the operating room getting a bypass right now.

It shouldn’t be all too difficult to make a PCI-E interface for the iRAM. Yes, it doesn’t completely free the bottleneck, but let’s not make the perfect the enemy of the good.

Maybe it is tougher than it looks, but it’s at least something Gigabyte ought to look into if it hasn’t already, and if it is really tough, tell us so.

It may also be possible that there isn’t much real-life difference between a 150MB and 1GB transfer rate; CPUs may not be able to seriously process that much data that fast. It would be good to know that, too (if nothing else, it would hint that flash-memory based features like Intel’s Robson will be more of a software-based fast-boot stunt than a general breakthrough).

Nonetheless, even if you regard iRAM as a niche product, provided development costs aren’t too great, I think most people inclined to buy one would be much happier to buy a PCI-E version of iRAM than a SATA version, and if that means having to buy a Gigabyte motherboard with this nice x4 slot, I’m sure that wouldn’t break either the buyer or seller’s heart.

BTW, this isn’t a sneaky way of saying, “Buy Intel.” Everyone making PCI-E chipsets is providing for more PCI Express lanes in their second-generation PCI-E products, so doing the same with AMD-based mobos shouldn’t be a big deal.

Again, maybe Gigabyte has already thought about this, maybe they haven’t. If you’re at all interested in seeing this kind of product, you might want to go here and ask them about it. I did.

Ed

Be the first to comment

Leave a Reply