TR Forums

Hey all, good to see this website still around.
I just finished a build basically based off of the sweet spot build in the May 2014 system guide.
After putting everything together, it initially would not post. System would power on, fans would run, and subsequently nothing would happen. After removing the stick from the B1 slot System would boot fine. After the initial boot, behavior with RAM in b1 changed to a constant power cycle: fans spool up, fans die down, repeat. Ran memtest with a1 & a2 occupied, memory passed, no faults. Removed the CPU, checked for bent pins, none were found.

Evidently both of the b slots are inoperable and will cause a failure to post in all observed cases, using a1 & b1 or a2 & b2 have the same effect.

At this point I am hoping this is simply the result of some strange incompatibility between motherboard and RAM, and I have a new set of RAM straight of the MoBo's QVL; so, hopefully this will solve the problem but I am entirely out of ideas after that besides just replacing everything.

Any further leads to pursue, or even just advice on which particular items I should replace?

Intel 4770K
Asus H97 plus
16Gb (8Gbx2) G.Skill -not on qvl

I assume you've already tried clearing the CMOS? Yes, motherboards are supposed to come with all settings defaulted, but I've seen exceptions to this.

Also try updating to the latest BIOS (with the RAM in A1/A2, of course).

I have started to see this behaviour with recent Haswell chips, both on 8-series and 9-series chipsets. Must have happened to me five or six times over the last 20-30 builds or so.

With the 9-series chipsets it's unlikely to be outdated BIOS versions so I'm wondering if Intel is screwing up something with the revised Haswell parts.

Certainly CMOS reset and minimal boot would be a good starting point. I'd also suggest minimal boot (onboard graphics, 1 ram stick in the first slot, install a PC speaker if you have one lying around)
Modern boards often have a "three tries and reset" approach to failed boots, so be persistant. What confuse me a lot at first was inconsistent behaviour between boots because the board was reverting back to failsafes every third try. To reset this "three tries and you're out" auto-config behaviour without resetting the CMOS, pull the power cord out, and hold down the power button for 10 whole seconds.

have you run memtest against the DIMMs individually (or both on the A channel) to ensure it's not something with the memory? If there's a problem with the DIMMs it's possible it's only coming out in dual channel operation.

Testing with other RAM would be good, too, as it would isolate whether it's the board/CPU or the RAM.

deruberhanyok wrote:

have you run memtest against the DIMMs individually (or both on the A channel) to ensure it's not something with the memory? If there's a problem with the DIMMs it's possible it's only coming out in dual channel operation.

He already indicated that he did that, back in the original post.

deruberhanyok wrote:

Testing with other RAM would be good, too, as it would isolate whether it's the board/CPU or the RAM.

If he's got other RAM on hand, yes that would still be worthwhile just to get another data point regarding compatibility with the B slots. But I think he's already verified that the RAM isn't outright defective.

yep, that was a "too early in the morning" post. I must have missed that part.

Alright, thanks for all the replies.

My new memory came in straight off the asus qvl and exhibits the same behavior.
I will try clearing the CMOS, but what state should I initially reboot the computer in?
I guess it goes CMOS rest -> Minimal Boot -> Min. with RAM in A1 and B1?

Since the new, and so far assumed fully funtional RAM has exhibited the same behavior; I feel I can rule out the RAm as the culprit, but how am I to diferentiate between a processor or motherboard issue at this stage?

Edit:
Problem persists after clearing CMOS, updating BIOS. At this point I think I'm looking at returning both the processor and the motherboard.

You're problaby right to return both, but just do every step at least three times as you work through your minimal boot, because the Haswells I've seen exhibiting this behaviour have had some kind of auto-recovery.

CMOS clear, then minimal boot for at least 5 minutes even with a black screen in case it's doing a BIOS recovery from backup BIOS. Power it down by holding down the case button, not by killing power at the PSU, and then try again at least twice before giving up on that option and moving onto something else.

It could still be time to RMA, but I've thought that on more than a few occasions since the 9-series and second wave of Haswell chips came along (notably the 4790s) and been unable to reliably identify the faulty part. ie, the 4790 worked in another board, and the board worked with another 4770, despite having the same BIOS versions on both. Persistance and repetition of boots seems to have saved the day with no actual RMA's needed in my case. Being an issue on both Asus and Gigabyte boards, I'm going to assume its something Intel changed and the motherboard vendors didn't get the memo

Chrispy_ wrote:

I have started to see this behaviour with recent Haswell chips, both on 8-series and 9-series chipsets. Must have happened to me five or six times over the last 20-30 builds or so.

Yikes 5/20 = 25% kinda scary, even at 5/30 = 16.67% still scary.

Sounds like quite a frustrating problem. New semi-working builds are the worst when you don't have spare parts for testing.

First impression based on the problem description was that there are bent CPU pins in the socket. Reread your initial post and saw that you'd already checked that. Not at all trying to sound condescending, but I've worked with dozens of customers over the years who were certain that there weren't any bent pins, simply because it can be subtle and difficult to determine if they are indeed bent if checking out sockets for damage isn't something you do that often. Again, no offense intended.

Assuming that the BIOS is up to date, the CMOS has been cleared, and new QVL RAM is still having problems, I'd try the following:
-Make sure that the BIOS is populating the correct speed, timings, and voltage for your RAM
-Bench test the board outside of the case if at all possible
-Intel is generally good about cross-shipping a replacement CPU in one or two days - might be worth trying. However, getting Intel to actually consider that the CPU merits replacement given the behavior you're seeing is going to be a bit of a challenge.
-If all of the above fails to remedy the problem, a flaky mobo seems like the only remaining possibility. Rest assured the RMA department of wherever that board gets sent to will very much be on the lookout for bent pins. If they determine there's a physically damaged socket, IIRC the going rate to replace it for you is ~$90 for most manufacturers.

Good luck!

anotherengineer wrote:

Chrispy_ wrote:

I have started to see this behaviour with recent Haswell chips, both on 8-series and 9-series chipsets. Must have happened to me five or six times over the last 20-30 builds or so.

Yikes 5/20 = 25% kinda scary, even at 5/30 = 16.67% still scary.

None of them have needed an RMA, but the Gigabytes in particular "recover" with an autoflash from the backup BIOS after some black magic resetting/CMOS wiping/power-cycling voodoo. That's the first time I see any output on the display when it's doing a BIOS recovery. After ruling out all the components I moved to a different workbench in case the power at my bench was flaky and screwing boards, but the issue persisted with other builds using all-different parts. It's as if there's a bad batch of BIOSes out there, or more likely, there's something in the new Haswell refresh that isn't playing well with some of the BIOS revisions in these newer 9-series boards.

I've (well, we, as a department) have built around 60 1150 PC's now, but the problem only started when the 4790s started getting used, though it might have been as 8-series boards started to go out of stock and I switched to ordering 9-series. It's a far cry from being able to assemble a machine and have it POST first time without even thinking about the build which is what happened for all the earlier 8-series builds.

Ok. Still unable to resolve the problem. I have tried clearing the CMOS and rebooting with both sets of ram in A1/B1 three times, left waiting just leads to a continual black screen. I've also tried loosening the timings on both sets of ram, under locking, and undervolting. Just running with A1/2 filled the system is extremely stable, and I have not produced a single BSOD so far. I've also been playing skyrim and planetary annihilation for hours, and seen no deviation from expected performance.

I was wondering if anyone has some recommendations for good, failure-inducing benchmarks or tests I could run to see if the CPU throws any faults? I'm planning on running memtest overnight on my original set of ram.

I'm also taking this down to the local shop tomorrow, see if they can off any insight. If everything seems alright and I can't find some other problem I'm considering keeping the PC, unless that seems

Prime95 and Linpack were usually my default stress tests in the past. I've since moved on over the past couple years to OCCT quite a bit more regularly. It's essentially just a front-end that lets you choose between Prime and Linpack, with setting options that are a bit less vague than those found in the individual testing programs.

If all the RAM is good in the A slots, it's either a defective memory controller on the processor (really really rare) or a faulty motherboard (pretty common).
The chances of it being an incompatibility are pretty low because if Asus' H97 Plus didn't work with 4770K the entire internet would be mouthing off about it.