Yes to lastest BIOS, 0040.
Drivers don't matter because the PCIe device is not even detected (because of BIOS).
The only two video capture devices listed among tested peripherals are USB capture sticks. My company's product captures high-definition video at 1920x1080/60fps, which those devices are not capable of handling. As I have explained, we use an HDMI-compatible PCIe capture card, which interfaces with the PCIe bus via an SAA7160, which meets PCIe 1.0a standards. Intel's own spec for PCIe 2.0 states that it is backwards compatible with 1.0a, and yet so many Intel motherboards are unable to detect such cards.
I will give you the solution to pass on to your BIOS people - Please provide an option, on a slot-by-slot basis, to change PCIe configuration register 0x70, the Link Control 2 Register in PCH, from 0x02 to 0x01, which limits the maximum PCIe speed to GEN1. That's it, very simple. I have tested this solution, now it just needs to be incorporated into the Intel Visual BIOS, and I and thousands of HTPC builders will be satisfied. You are welcome. ;-)
Really, it's that simple!
Any response yet from engineering about this issue? I have found a workaround that requires up to 2 warm boots and some low-level PCH rejiggering. This is clearly a BIOS issue.
A similar report was made over a year ago, with details as to a root cause:
In that case, the PCIE RSET timing was incorrect in BIOS version 0025. Intell representative Lois H. (lhill) sent the report to NUC engineering, which confirmed the defect, but it still seems unresolved in version 0040. This seems to be SOP for Intel.
According to engineering team the NUC is designed to be backward compatible to PCIe 1.0 where no bios tweaking is required to achieve this. Based on their analysis on the some of the card, the issue is related to "buffer strength". These cards are demanding for higher buffer strength, more than what is specified in the chipset specification, to operate. There will be some other reliability issue if the buffer strength is set to a higher than allowed value, thus we will not do anything to make these cards compatible. Hope this explains.
With respect, I think there is a problem with your translation of engineering-speak. Is it possible that "buffer strength" refers to CMOS voltage or current levels? I find it hard to believe that the SAA7160/NXP7160, one of the most widely-used PCIe interface chips for video capture cards, has a problem meeting the 1.0 specs. A timing issue, maybe, but not an incompatible electrical specification.
As explained above, I have found a simple solution to the compatibility issue by changing PCH parameters outside of BIOS, with no adverse effect on reliability in several hundred hours of testing with a NUC D54250WYKH. It does not require changing any electrical characteristics.
Would it be possible for me to discuss this issue, by phone or email, with the engineer who came up with the "buffer strength" theory? I would happy to discuss my solution with him or her, to demonstrate that a simple BIOS change would work without ill consequence, and with no cost to Intel beyond issuing a new BIOS release. I will drop the issue if someone can demonstrate that it would cause stability issues.
Mike, here is proof that your engineer is mistaken about the SAA7160 "buffer strength" issue. That device is fully certified by the PCI-SIG compliance program, having successfully completed the rigorous testing procedures of the Compliance Workshop. Please see PCI-SIG - PCI Express and search for SAA7160.
Also, again, I refer you to this explanation of the root problem from one of the capture card manufacturers: TBSDTV Community Forum • View topic - TBS-6984 not detected on Dell R310
If you haven't seen my private email with Joe_Intel about this issue, here are the supporting links I sent him:
Sent July 6, 2015
I have tested on NUC5i5RYK (Broadwell), D54250WYK and D34010WYK (both Haswell), and all have this problem.
I'm sure the BIOS designers did not consider that older devices might be used with mini-PCIe, but it is possible!
The 7 Series chipset had this problem, you can see the complaints at the links below. I think I am the first to report incompatibility in NUCs.
More can be found in a Google search of : TechoTrend "communities.intel.com", or simply: TechnoTrend pcie
The official PCI website states that all PCIe 2.0 is backward compatible with PCIe 1.1 and 1.0. The same applies to PCIe 3.0.