PCI-E Test card announcement

[David (PassMark)]

For the last year we have been developing a new PCI-E Test card. This diagnostic card is the replacement and successor for the POST card (Power On Self Test) card that we used to stock some years ago.

The old POST card were only designed to work on the ISA and PCI buses, but both these buses are now obsolete. These cards were extremely basic, their function was to read a value from a memory location (known as PORT 80) and display it for the user to see. More details about POST are on Wikipedia. Over the year the beep codes and integrated display functions built into the motherboards made these POST cards less necessary. Further the BIOS vendors stopped publishing the codes needed to make use of the POST cards. Combine this with the advent of UEFI and the PCI-E bus, and we are calling these POST cards officially dead. R.I.P.

Nevertheless there is still a requirement to test motherboard slots for electrical issues and performance. So somewhat in stealth we embarked on a development process a year ago to develop a more sophisticated PCI-E test & diagnostic card.

This is the picture of the 1st prototype of this new PCI-E card that was the result of the development process. Only two units of this prototype have been made.




Tomorrow I hope to go into detail about some of the card's functions and challenges in designing the card.

Update: PCI-e Test card now available for purchase.

[David (PassMark)]

PCI-E diagnostics card functionality

Feature list of the PCI-E diagnostics card.

• Double edged PCI-E card with x1 connector on one side and x4 connector on the other side. Card can be flipped over to test different sized slots. Card auto-senses the connector in use.
• Support for the PCI Express gen 2.0 standard (500 MB/s per lane, in theory). In real life these speeds aren't obtainable. But we should get close. Especially for x1 where there will be less of a bottleneck.
• Capability to read and check voltage levels of 12V line, 3.3V line and 5V line. There are other cards on the market that just tell you if there is voltage. This card will measure the voltage in the same way as a multi-meter or oscilloscope. Allowing you to check if the voltages are in spec.
• Dual channel high speed sampling of voltage levels, up to 2Mhz at 12bit resolution. 2 million voltage samples per second for voltage spike and brown out detection.
• Capability to monitor temperatures inside the case. (-40C to 125C, +-2C)
• Red, green yellow and orange LEDs to display
  • Data read activity
  • Data write activity
  • Various error indications
  • 12V PCIe power (with a LED per pin on the PCIe connector)
  • 3.3V PCIe power (with a LED per pin on the PCIe connector)
  • 5V Molex power
  • 12V Molex power
  • PCIe Sleep state
  • PCIe Gen 1.0 or 2.0 / 3.0 support available
  • Test mode (Loopback, Benchmark or Oscilloscope)
  
  
• Dual channel AXI DMA transfer support (Advanced eXtensible Interface - Direct Memory access). Which allows high speed transfers between the card and the system RAM without the involvement of the CPU.
• 96 KB of high speed LSRAM integrated on the card.
• Benchmark mode to test throughput Read, Write, Read then Write and Read & Write
• Loopback mode to test reliability and bit error rates on the PCI-E bus
• Support for BurnInTest for Windows and Linux
• Standalone Windows test tool will be included with card
• Powered from the PCI-E, with no external power required.

Note: Some of the above features are still to be implemented & will be confirmed upon the final hardware / software release.

[csdarknightcs]

How much and where will it be sold? I am looking to buy a couple of post cards for my students to use in my IT classes.

[David (PassMark)]

We'll be selling it off our web site with our other test hardware. Shipping will be done out of the USA and Australia depending on the destination of the order. We have a number of reseller's, but often they don't hold stock of the full product line, so can't give an exact answer on who will be holding stock at the moment.

Exact price is yet to be determined. But we are estimating around $250 USD initially. Based on the assumption that volumes are going to be relatively small. Main cost is the custom chip (a FPGA) on the card and the engineering effort in designing and testing it. If the card is popular, then like all electronics the price comes down significantly as the volume goes up. For example, in the unlikely event we get manufacturing volumes up to 50,000 units the price would be more around $100.

Release date should be 2nd Qtr 2015.

[jsimoes]

Great News!

Just can't wait! Don't forget to ship to Brazil too!

Best Regards.

We'll be selling it off our web site with our other test hardware. Shipping will be done out of the USA and Australia depending on the destination of the order. We have a number of reseller's, but often they don't hold stock of the full product line, so can't give an exact answer on who will be holding stock at the moment.

Exact price is yet to be determined. But we are estimating around $250 USD initially. Based on the assumption that volumes are going to be relatively small. Main cost is the custom chip (a FPGA) on the card and the engineering effort in designing and testing it. If the card is popular, then like all electronics the price comes down significantly as the volume goes up. For example, in the unlikely event we get manufacturing volumes up to 50,000 units the price would be more around $100.

Release date should be 2nd Qtr 2015.

[David (PassMark)]

Work continues on the PCI-E test card. Here is a photo of the second prototype.




Significant differences from the initial prototype are,

• The PCB (Printed Circuit board) has been made slightly longer in order to make space for a collection of LEDs. These LEDs indicate the presence of voltages on the various pins, plus some red LEDs for error conditions.
• A notch was cut into the end of the PCI-E. So you can hook your finger under the card to help remove it from the motherboard.
• Various wiring hacks from prototype #1 were build into the tracks on the PCB.
• Some printing (in white text) was added to the board.
• Changed molex connector (not visible on the rear of the board) for a better quality connector.
• Added a by x1 & x4 jumper selector to manually force a particular edge connector to be used. (But we have since found a way to automatically detect and switch between the slots).
• Removed the power switch, as it wasn't needed.
• Spec'ed 100 micro-inch gold plating (over 200 micro-inch of nickel) on the two PCI-E edge connectors. This is to help increase the number of insertions & removals you can make before wearing out the connector. For comparison the typical thickness is 30 gold over 100 nickel.

[David (PassMark)]

This week we have been doing EMC testing.
EMC = ElectroMagnetic Compliance.

This is where they stick your device into a Anechoic chamber and measure the radiation being emitted from the device (through the air and via the power line) to ensure it complies with the standards. All electronic devices need to do testing of this nature.

Sometimes devices are accidental emitters of radio waves, which can then interfere with other devices, like your TV and Radio.

At the end of the process you get back a collection of graphs like this one. Hopefully confirming that the output from the device (green line) is less that the maximum allowed (red line) over a range of radio frequencies.



Fortunately the card passed the tests. It would have been terrible to need to redesign it at this late stage.

In the next couple of weeks we'll be looking at the factory testing process and then packaging.

[David (PassMark)]

Production run for the final hardware starts next week.
Development continues on the software that will accompany the card plus bug fixing in the device driver.

Here is a close up of the PCI-E voltage level checking LEDs on the final card. In case you are not aware the PCI-E connector has multiple power pins. For a x1 connector there are five separate 12volt pins (called B1, B2, B3, A2 and A3) and three 3.3V pins (called B8, A9 and A10). You can see how these correspond to the LEDs on the board. Further there is the optional 3.3V aux voltage line, that only needs to be present if the wake up process is used.

[jowellco]

Is this product already available? Love to buy one...pls provide a quote to me at jowell.saezco@gmail.com

[David (PassMark)]

jowellco,

I'll E-mail you details later today.

[David (PassMark)]

Another update on development progress.

Development has continued over the last couple of months. We had some technical issues that pushed the release date back a bit. While there were dozens of minor changes and improvements, the most significant issues encountered during testing were,

• Getting multiple cards in multiple slots to work at the same time.
• Getting the voltage readings to be stable. As it turns out the ADC (analog-to-digital converter) doesn't work in a stable fashion at the specified speed (2Mhz). Some examples work and some don't. Meaning we have had to underclock them to get them all stable.
• Dealing with various timing issues to get the card to reset between test sessions. Previously the was a bug causing DMA transfers into the wrong memory address.
• Digging into speed optimisations to see if any more speed could to be extracted from the hardware and understand the speed bottlenecks. At small block sizes we saw speeds that were above what was expected (due to an internal transmit buffer it turned out). In another case we saw speeds lower than that was expected (at PCI-E 2.0 at x4), this was due to internal bus speed limits. Knowing these expected speeds allows us to report on if a particular motherboard can obtain these speeds.

But all that seems to be resolved now (within the fundamental limits of the hardware) and we are finishing off the software.

Three software tools will be created / updated to support the PCI-E card.
- PCI-E-Test, a new standalone application that will be free with the card.
- BurnInTest for Windows, this package will be updated to V8.1 to support the card.
- BurnInTest for Linux, this package will be updated to V3.3 to support the card.

[David (PassMark)]

Here is a screen shot from the new PCIeTest application (the final release will look slightly different).

[niteman44]

Can't wait till this is for sell. It looks like a pretty cool card.

[David (PassMark)]

It should be available for sale mid July (~2 weeks from now).

Updated: The PCIe card is now available for purchase from our web site.

[niteman44]

Do you now have a better idea what the cost of the card will be since you now know the cost of your NRE and the parts and artwork for the CCA?