New Taichi x370 Build Won't Post Due to Dr. Debug

I'm in urgent need of help in pinpointing the culprit of a failed build I put together the other day, of a Ryzen 2700x, ASRock X370 Taichi AM4 AMD Promontory X370 and Corsair Vengeance LPX 32GB (2 x 16GB) DDR4 3200 MK32GX4M2L3200C16 memory I ordered almost simultaneously around Black Friday. If something needs to be RMA'd, I need to figure out exactly which part is at fault, because there won't be time in the remainder of the 30-day RMA period to send back more than 1 part back and retest the replacements in succession. After the detailed detective work I?™ve already done though, I strongly suspect that the problem is graphics card-related.
     The crux of the problem is that I haven't been able to post successfully once, due to Dr. Debug error code B2. This only occurs whenever I plug either one of my graphics cards into any of the 3 available PCI 2.0 and 3.0 slots. If I remove them, the Dr. Debug codes rapidly cycle through a sequence ending in A2, D6, D7, OD, A9 and OD; if I plug in a USB keyboard then D7 disappears and if I hook up the SATA drives, the A9 code disappears and the 0D only flashes briefly once, after which Dr. Debug is blank. Am I correct in assuming that this means that the system successfully posted, but I just can't see it due to the D6 error code, which means that no graphics card is attached? If so, then that sharply reduces the likelihood that the memory or processor are at fault.

Dr. Debug codes
------------------------
     The debug codes blow by so fast that I had to take a video of the sequences and slow them down to 1/8th speed - but even then I had trouble catching them all. I assume that if a particular code didn't linger, the motherboard had no real problem whatever bootup stage was involved. The only ones that lingered a long time were 15 and 65, as well as B2, D6, D7 and OD of course, as discussed above. The two sequences don't seem to differ much till we reach 28 or so, near the end:

* The Dr. Debug Code sequence goes something like this when no video card, keyboard or hard drive is attached: 00, 14, 19, 15, C0, D2, 65, 68, 67, ?7, 03, 88, 21, 89, 54, 99, 33, 99, 55, 7A, 38, 36, 96, 4F, 88?, 60, 88?, 78, 89, E9, 28, 98, 9E, 8A, 98, 99, A2, D6, D7, 0D, A9, 0D. The D7 can be removed by attaching a keyboard, the final OD by attaching SATA drives (after which the Dr. Debug turns blank at the end). Does this mean we've reached the OS boot phase or BIOS, but just can't see it because of a lack of a video card? If so, the issue is less likely to be memory than a PCI issue.
* Here's the (inexact) sequence when a video card is attached: 00, 88???, 3E, 14, 19, 14, 15, C0, D2, 65, 68, 67, ?7, 03, 09, 88?, 80?, 54, 21, 54, 33, 99, 55, 7A, 38, 36, 4F, 88?, 79, 89, E9, 27, 28, 27, 88?, 94, 07, 97, 86?, then stuck forever on B2.
* I also tried starting the computer without any RAM installed, which ended in a code #46 error.
* Interesting: If I plug my 2-SATA port Anandtech card into either of the PCI-express slots, a B2 is inserted into the first sequence between A2 and D6, but it doesn?™t linger long. If I also plug a SATA drive into one of the ports with all the others unplugged, the sequence doesn?™t stop with 0D as it always does when no drives are plugged in. The Dr. Debug instead turns blank after the sequence is complete, suggesting to me that the SATA drive is recognized and by extension the card as well. If so, then perhaps the problem is limited solely to either a) the other 3 PCI ports themselves or b) the graphics cards plugged into them, rather than the whole PCI subsystem. It?™s too bad I don?™t have a PCI-Express video card handy to narrow the problem down further.

The meanings of these codes are poorly documented, but I was able to glean the following from this list:

97 - Console Output devices connect
98 - Console input devices connect
99 - Super IO Initialization
A2 - IDE Detect
A9 - might represent the "Start of Setup" - i.e. trying to get into the OS, which it can't do since there are no drives attached at this point. A9 disappears altogether if the drives are attached
94 - PCI Bus Enumeration 32
79 - CSM initialization
33 - CPU post-memory initialization. Cache initialization
36 - CPU post-memory initialization. System Management Mode (SMM) initialization
38 - Post-Memory System Agent initialization (System Agent module specific)
3E - Post-Memory PCH initialization (PCH module specific)
21, 27 and 28 are "OEM pre-memory initialization codes"
B2 - "Legacy option rom initialization" - may be related to the older video card issue

     From my prior experience with other motherboards, B2 usually signifies a problem with an installed graphics card. I have found several other references on the Internet to B2 referring to a graphics card issues, like this one at Tom's Hardware. The fact that none of the sequences ended with a known processor or memory error seems to rule those out. It's unlikely to be a processor/BIOS incompatibility, because the shipped BIOS was 4.40, which can run Ryzen 2700x. I've verified that the fans start up on the PCI 3.0 Nvidia GeForce GT 740 and PCI 2.0 x16 Gigabyte Radeon HD 4350 DirectX 10.1 GV-R435OC-512I 512MB graphics cards I'm using for testing and that they're seated correctly. Other troubleshooting measures I've tried include:

* Cleared the CMOs several times.
* Removed the CMOS battery.
* Removed the motherboard from the case and test it in the foam frame, to make sure contact with some element of the case wasn?™t causing a short.
* Reseated the RAM in almost every possible combination of positions, including both on the 3rd and 4th slots, both in A2 and B2 and one stick in each slot. The seating is tricky because the clips are one-sided (something I've never seen before + which is also contrary to the depiction in the manual) but I inspected them closely to make sure.
* My power supply is a 1050 watt EVGA so insufficient power is not an issue.
* Verified that both the 24 pin and CPU power connectors were correctly attached. This eliminates the problem described at this page https://pcpartpicker.com/forums/topic/5075-asrock-z77-extreme-4-error-code where a user experienced a similar issue due to first inadvertently leaving the CPU power connector and second from also not pushing down the main power connector all the way.
* I did not and will not remove and reinstall the processor, because I checked it thoroughly before attaching the heatsink, which was a chore as usual thanks to the antiquated clip system (that's always a problem - I'm more at risk of breaking my processor or motherboard by reseating it). Furthermore, none of the Dr. Debug codes aren?™t remotely processor-related.

     What this reminds me of is the post Warning to Users with older Video Cards in this forum, which specifically ties BIOS 4.40 and above to graphics card BIOS issues and ROM initialization, i.e. precisely what the B2 error refers to. The Gigabyte Radeon card is from 2011 and thus may have one of the older BIOS' described in that post, but the Nvidia card has a UEFI BIOS according to TechPowerUp's GPU-Z utility. I found another post at Tom's Hardware where a user had a B2 error that was resolved by updating the BIOS to both his ASUS GTX 770 BIOS and the BIOS of his graphics card. Another offhand remark at Tom's Hardware says that "nvidia gpu + asrock = every 2nd bios version gpu fails its CSM setting related," which seems to match my configuration well. I guess I need to answer 3 specific questions:

#1 Given this detailed description, am I right in concluding that it's a graphics card issue, probably related to the BIOS rather than the PCI subystem on the motherboard? If so, then I don't need to worry about returning the RAM or processor
#2 Is likely that I'm posting when I remove the graphics cards and plug in the USB keyboard and SATA drives, in the sequence described above?
#3 If so, how do I fix the issue? I don't have any onboard graphics. If I can't post without disconnecting the video cards, then I can't flash the BIOS either. I can't afford to shell out $150 for a new graphics cards just to fix this. I don?™t have any PCI-Express video cards handy to narrow the problem down further. One potential workaround would be to attached a cheap external video card to the USB port - but would the motherboard recognize it and allow me to view the BIOS settings, prior to booting into Windows?

     What else can I do? I need to resolve this as quickly as possible, so any help would be appreciated. I'm in a bind given the RMA time deadline pressure; if I don't handle this deftly, I could end up with a few hundred dollars of equipment sit there for months before I can buy new hardware to test it with, or be out a few hundred in equipment that's defective and can't be returned. I need to find the right workaround on the first try.

Thanks,

Steve
[URL=http://www.tomshardware.com/answers/id-3595392/error-code.html ]this one at Tom's Hardware[/URL][URL=https://forum.level1techs.com/t/list-of-dr-debug-bios-codes/114364 ]this list[/URL][URL=http://forum.asrock.com/forum_posts.asp?TID=8184&PN=2&title=warning-to-users-with-older-video-cards]Warning to Users with Older Video Cards[/URL][URL=http://www.tomshardware.com/forum/393632-33-asrock-debug-error-legacy-option-initizalation]another post[/URL][URL=http://www.tomshardware.com/answers/id-3780278/motherboard-post-gpu-x16-pcie-slot-pcie-slot4-debug-error.html]offhand remark[/URL]

Tip #1 - avoid walls of text. Almost no one reads them.
Tip #2 - post your HW configuration and info.

B2 code sounds like a CSM issue with an older GPU and its initialization. Your GPU's are way too old to support UEFI & SecureBoot.

If you can enter BIOS, maybe there is chance. Try updating BIOS to L4.81 ( beta ) from JZ. Clear CMOS, leave just the 740 and try booting again.

I personally don't think you have a bad CPU, mobo or RAM. It's the old GPU's, me think.

I was able to get the board to recognize my main graphics card, an EVGA NVIDIA GT 740, but only by buying a newer card in order to get into BIOS. From there, I was able to change the PCI Oprom setting from the default of UEFI *or* Legacy to Legacy Only. The name of the setting is misleading, because the new ASRock graphics card I had to buy to get into BIOS is UEFI-based, yet still works in Legacy Only mode. In fact, the older EVGA card is UEFI as well, but is only recognized when either a) both cards are installed or b) the PCI Oprom is set to Legacy Only. This post on the NVIDIA GeForce forums https://forums.geforce.com/default/topic/788289/geforce-900-series/computer-won-t-boot-past-bios-after-installing-new-gtx-970/3/ describes a similar problem affecting other ASRock motherboards. This workaround for ASRock's faulty BIOS code cost me roughly $100; I can't simply return the second card now, because if the CMOS is ever reset for any reason, the PCI OProm setting will probably revert. I'll have to mention this unexpected expense in my product reviews.

I will try to remember to post a support ticket as you suggested. It never occurred to me that the EVGA card wouldn't be considered "modern" - it was released just 4 years ago and performs on a par with the brand new ASRock card I just had to buy. The cost was almost the same despite the passage of time. Then again, I'm not a gamer; this motherboard is being used strictly for workstation applications. I need all the CPU power I can get, but the GPU is immaterial to me as long as it can run my monitors.

The GPU approach to number-crunching undoubtedly has its benefits, particularly with certain types of matrix math, but it also has many drawbacks that have been forgotten in the current wave of enthusiasm. It's a useful tool for certain use cases that has been blown up into a fad, unfortunately, in part because of bitcoin mining and the like. The same can also be said about the phone market, where consumers have unfortunately been conditioned to buy the latest gadgets without receiving much benefit in return. Both are cases of marketing triumphing over actual technological benefit. It might also be true in the case of mainstream PCIE 3.0x graphics cards no longer being compatible after just 4 years. In any event, although I've usually always relied on ASRock's products, this is a major fail on their part, one I'll have to pass on to other potential purchasers.