Does PXI has the processing power of a desktop?

[master]

Hi

I have an application that acquired 36 channels at 1 KHz. No sample must be lost for 24/7.
I need to convert these signals to their units as I am using SCXI-1102C current reading.
After this is a compare whether these 36000 data is out of the tolerance with each channel have their spec.
If any is out of the spec., log a 1 KHz data for next 5 sec and log a 50 Hz (derived from 1 KHz) for next 15 minutes.
If everything is fine, just log 50 Hz every second.

My hardware configuration is:
PXI-8186 ( P4 2.2GHz, 256 MB RAM, 30 GB HDD) Labview RT 7.1, DAQ 7.3.1, Labview 7.1, RT 7.1
PXI-1031 chassis
SCXI-1000 chassis
SCXI-1102C
SCXI-1125

I am using consumer/product structure, but product loop count lags consumer and gets worst when when the data get out of the spec. This worries me as there will eventually be a queue overflow as product lag too long.

When I use the same code in my desktop (P4 2.4GHz (533 front-end bus), 256 MB RAM, 40 GB HDD, 32 MB Video card, windows XP, DAQ 7.3.1, Labview 7.1, there is just 1 lag count which is correct as consumer is done first (even maintain on out of spec.)

This let me think that is PXI systemm, due to the same size, has not much processing power for 36000 data every second? How is the comparsion of desktop to PXI processing power?

[Shea C]

Master,

Our PXI controllers are built using laptop components and typically they may compromise performance for their compactness and ruggedness. Comparing your desktop processor to your PXI processor, despite the fact that they are different speeds, is not a true comparison due to the fact that the P4 that the PXI controller is built with uses Pentium's Mobile technology which your desktop most likely does not. Memory speed is also a huge factor in the rate that data can be streamed to the processor for computations as well as the front-side bus speed. The controller that you are referring to may use slightly slower parts for these components. All these factors would lead to the reduced performance that you may be seeing. Benchmarks do exist but not for your specific application. I hope this clears up the differences in the two CPU's that you are considering. You may want to consider the NI PXI-8196 for the latest PXI perfomance available.

Regards,

Shea C
NI

[Adam_U]

As a rule of thumb, the PXI controllers will lag desktop performance slightly because of the size constraints as Shea mentioned. National Instruments strives to shrink this margin to as small as possible by leading development before chipsets and other hardware are on the market. This is evident in our newest offering (the PXI 8196 controller) which has 50% better performance than the 8187 does! In some benchmarks, it even outdoes 1U P4 based servers in some tasks.

You may want to break down your application into some components that you can benchmark on your system. If you're looking to get an overall feel for your performance, is it possible to create a "worst case" scenario where you falsely send out of range data so that it must be processed? Additionally, you need to ensure that both the producer and consumer loop are independent of one another. Perhaps you have a shared VI that is not marked as reentrant and it is stalling one of the loops (hence killing your performance). The "Profile VIs" feature in LabVIEW could help determine this.

If you break your application into specifics that focus on each bottleneck it can help narrow the problem. Something like "I need to write 15MB/s worst case" and "I need to perform an FFT on 4000 points/sec" can help you find the restricting point and work around it.

-Adam