I feel the need for speed! Make that a Quadruple Espresso!

Larry Desjardin- July 27, 2012

PCI Express holds major promise as a test and measurement interface.  Those of you who read my previous blog on the subject, PCI Express – Espresso for test systems, know that this interface can dramatically increase the speed of functional test.  It is the key fabric behind PXIe, AXIe, and the VXI 4.0 modular instrument standards.

Fellow blogger Ransom Stephens reports that a fourth generation version of PCI Express is on its way.   Since each generation essentially doubles the speed of the previous generation, that is an 8-fold increase over vanilla PCI Express. Or should I say espresso-flavored PCI Express, given the phenomenal speed benefits it brings by itself?  

This is good news for the test and measurement industry for several reasons. First of all, it shows the long-term commitment of the computer industry to the PCI architecture and to PCI Express in particular. This gives longevity to this very important interface, and keeps PXI, AXIe, and VXI 4.0 aligned with computer standards.

But it also enables new applications for test and measurement where data bandwidth is critical.  In my previous comments on the subject I observed that the major speed advantage of PCI Express in functional test systems wasn’t the data bandwidth, it was the low system latency associated with high speed drivers and register-based memory-mapped access.  The correct tool for measuring speed in most functional test systems is a stopwatch, not a speedometer. How long it takes to do a single measurement is the gating factor. This is further explained in the article Integrating Traditional and Modular Instruments.  Apples to apples comparisons have shown as much as a 50 times improvement in single measurement speed using a PXI product instead of its box equivalent.  And that was standard PXI, not PXI Express.

In the case above, moving to PXI Express would have brought only marginal benefits to overall throughput, since it was the memory-mapped architecture that brought the speed benefits, not the increased bandwidth. However, as soon as we move from functional test to very high-speed data acquisition, the bandwidth becomes critical.  Think of multichannel high energy physics experiments or IQ data stream capture of a wide-bandwidth signal.  For these, total bandwidth is the critical factor, and we need to use a speedometer instead of a stopwatch.

Here’s a rule of thumb. PCI Express Gen 1 allows 1 GB/sec data transfer for a 4-lane connection found routinely on PXI or AXIe. Note that these are bytes, not bits. Gen 2 doubles that to 2GB/s. Gen 4 would bring this to 8GB/s.  There’s no other test and measurement interface that approaches these values.  That’s a quadruple espresso!  Current state of the art is delivering Gen 2 speeds in PXI and AXIe, but it is good to know there is more headroom if needed.

There’s another advantage to PCI Express. Each link delivers the speeds above. With a properly designed switch (on the backplane in the case of PXI, and in the system module in the case of AXIe), several of these streams can occur simultaneously. You won’t see this with LAN or USB based systems. Multiply this up and you get bandwidths in the 10s of GBs!  This is why you are seeing more and more streaming solutions being based on PCI Express enabled architectures.

Feeling the need for speed? Whether you are creating a functional test system or high-speed data acquisition, you can’t go wrong with PCI Express.