Slashdot Mirror
Next-gen PCMCIA: Expresscard
An anonymous reader writes "According to this article at WindowsForDevices, the PCMCIA trade association rolled out version 1.0 of its next-generation standard for modular mobile and desktop computer expansion at this week's Intel Developer Forum in San Jose, CA. Dubbed "ExpressCard", the new standard is "thinner, lighter, faster" than the group's previous PC Card standards, according to PCMCIA chairman Brad Saunders. ExpressCard achieves its space reduction by replacing the legacy parallel buses of the first and second generation PCMCIA card standards with state-of-the-art, high-speed serial connections, following a trend common in current computer system design."
To make cards any thinner that Type 1 PC-Card's you need to use chip's with special packaging as standard surface mount components on a dual sided PCB just barely fits into a PC Card enclose. For instance most CF or smaller WiFi cards are a radio on chip solution which is generally more expensive than a design based around discrete components (at least initially, if you can get the paramaters right and the process down then ultimatly the single chip solution is probably cheaper). For ram you need to use the highest density chips available which tend to be expensive instead of a small aray of cheaper, less dense chips.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
1x? 16x? Sounds like you're talking of PCI-Express, not PCI-X (which is simply 64bit PCI over a 133MHz bus)
I sense a name change in the near future.
"thinner, lighter, faster"
Sounds good, but I just took a look at the handful of CardBus cards that I already own, and they're already plenty thin, plenty light, and plenty fast. I've never said to myself "Man, these CardBus cards are really weighing my bag down." They're already small enough that I often can't find them, and I really don't have a problem waiting the few seconds that it takes to transfer 128MB of digital photos to my PowerBook via a CardBus adapter.
Frankly, I'd strongly prefer that industry stick to the current standard, and instead focus on coming up with nifty new CardBus products that add new capabilities to the computers I already have. Let's have a very affordable data acquisition card, for example. (I know data acquisition cards are already available, but AFAIK they tend to be pretty pricey.) Or a card that measures air quality wherever I am, or analyzes chemical samples.
ExpressCard stinks to me of planned obsolescence.
Yes it is simpler. Simpler circuitry, although at much higher speed. Less components, thus lower costs. Less wires, thus lower cross-interference. Smaller size, easier to fit into smaller devices. None of these taken separately would be enough, but when you put them together they're hard to beat.
If con is the opposite of pro, is Congress the opposite of progress?
PCMCIA actually stands for 'Personal Computer Memory Card International Association'. The parent was just a joke.
> Bitch and moan all you like, but can you hot-plug a usb mouse in a laptop running XFree and be able to use it immediately without quitting/restarting XFree and/or editing XF86Config? Thought so.
I'm using Red Hat 9.0, and the answer is a definite "yes, I can". I can plug in a USB mouse and use either the mouse or the builtin trackpad of my laptop interchangably, without touching XF86Config. It works exactly like it does in Windows XP.
Human/Ranger/Zangband
It isn't simpler. It's actually way more complex: You have to have PLLs or DLLs on each received lane for clock recovery, and a 2.5 Gbps serdes (for Express), which is not an easy beast to design. Next you need bit- and symbol-synchronization. Then you have to have elastic buffering to drop out or insert skip characters, to compensate for clock speed differences between the transmitter and receiver. And for multi-lane PCI Express, you need lane alignment logic, to compensate for the up to several symbol times of skew between the lanes, before you can unstripe the data. There are other difficulties as well but I hope you get the idea.
So why do serial? The answer is that parallel interfaces using synchronous clocking are running out of steam. The main reason for that is skew, which eats up an ever larger percentage of the timing budget as the clock rate gets higher and higher. Serial also has advantages due to the DC-balanced encoding scheme (8b/10b) that is used, which not only improves its electrical signaling characteristics (lower intersymbol interference) but also embeds the clock within each data stream so that skew isn't a factor (it's taken care of in the receiver's clock recovery).
4-lane PCI Express will roughly match 64-bit PCI-X bandwidths. Scale it to 32 lanes (64 differential signals) and you still have fewer pins (PCI-X is around 90 or so pins) but 8 times the bandwidth. Considering that there are already serial interfaces running at 10 Gbps -- 4 times Express's 2.5 Gbps per lane -- no doubt the clock frequency can be doubled for several generations to come. PCI-X will not be able to keep up.
I call BS, at least for network adapters. I've replaced my Xircom dongle at Fry's.