Video card no longer detected following physical crash

[AnitaPeterson]

Hi guys,

I didn't know where to post this thread, because it's got so many issues... please bear with me!

Some time ago, I received an Asrock 770DE+ with one of those AMD Phenom X3 CPUs, unlocked to a quad-core state. I used it as a base for a dedicated gaming system, in a Bitfenix case, with 8 GB of RAM, a SSD and a HDD, plus a Radeon 7770... and everything was well.

Last night, I tried to add a secondary 7770 for a Crossfire configuration. I took the computer from its usual place, propped it on a couch, and momentarily left the room to get the toolkit. And then, I heard a noise!

...the computer had fallen on the floor from the couch (about half a metre), landing on the side on the carpeted floor. It seemed to be very stable... alas, it was not.

And now, the system won't boot anymore. All I get is something like two rapid beeps in succession (in fact, more like one long beep with a very short stutter), followed immediately by eight short beeps, followed by a couple of seconds of silence, then another beep. From what I've seen online, this is an error code related to display memory...

I took out the motherboard from the case, unplugged everything except the CPU, one RAM stick and video card. Same thing. I tried two other (known to be good) video cards. Same thing.

After a couple of hours of testing, which also included a new PSU, I came to the sad conclusion that any video card inserted in the primary (green) PCI-E slot would produce the same result (eight beeps, no post). In fact, whether I put a video card in that slot or not, I'd get the same eight beeps.

Moving the video card to the orange slot (secondary PCI-E) makes the machine boot normally. But... there's no more room for extra cards (e-SATA and Firewire), PLUS the secondary slot is PCI-E 8X, not 16X...

The motherboard appears intact, no visible damage is evident under a strong light and a strong magnifying glass... I tried to clean the contacts in the slot, to no avail. And everything was well bolted down originally, so I really doubt any component moved when it fell... but who knows for sure? Yes, I was an idiot to prop it on the couch instead of putting it on the floor directly... let's not delve on that, shall we?

I tried to call Asrock, but they DON'T have a phone number, just a sales number (how's that for good customer service)?
I wrote to them, but who knows when they'll answer?

So... How screwed up am I?

I've put together several scenarios in my mind - but I can't even begin to answer some of the more complicated questions.

The first one would be to send the board back for RMA. As it was a hand-me-down, I don't have proof of original purchase, and it's probably out of warranty anyway, so - more than likely - I'd have to pay for it.

The second would be to just buy a new mobo... All things considered, it might even be cheaper than to RMA the old one. BUT....If I were to completely replace the motherboard, there's the added complication of the unlocked CPU. Would the CPU continue to be seen as a quad-core instead of an X3, on a different board? From what I've read, that's not likely - but maybe I am wrong, so please enlighten me, if you have more experience with these matters.

Scenario three: put a GPU in the secondary slot, and forget about the rest of the add-on cards. It would probably take a performance hit, but I could repurpose the machine as a HTPC - or even pass it along to a non-gaming person...

What do you guys think? Do you have any experience with Asrock? Am I flogging a dead horse?

*edited thread title for clarification*

 

[DSF]

I actually don't think you're going to see any real performance hit from a 7770 running at x8 if you just want to tough it out with the current board. It would've been running that way if you tried to Crossfire them anyway. (Actually, Newegg's page lists the second slot as x4, for what it's worth.)

The first step is to see if you get a response from AsRock. Who knows, maybe you'll get lucky and they'll take the board back. Considering that you broke it through negligence (admittedly by accident) and you aren't the original owner, I wouldn't hold my breath.

 

[VirtualLarry]

I don't know how many of the newer AM3+ boards support "Core Unlocking". It was mostly with boards with the SB710 southbridge, not the newer 990X/FX and SB950.

That said, I also don't know of many (any?) SB710 boards that also support Crossfire.

You might check Geeks.com though, they stock older boards.

 

[chubbyfatazn]

I don't know how many of the newer AM3+ boards support "Core Unlocking". It was mostly with boards with the SB710 southbridge, not the newer 990X/FX and SB950.

That said, I also don't know of many (any?) SB710 boards that also support Crossfire.

You might check Geeks.com though, they stock older boards.

If she (he?) goes with another board, s/he'd have to buy new memory. The 770DE+ she has currently uses DDR2.

If you buy a "new" board, you'd have to make sure it's a DDR2 board. Which, if I'm not mistaken, means you'd be looking for an AM2+ board, like this one. I'm not aware of any AM3/AM3+ boards that support DDR2.

I'd consider the price of getting a cpu/board/memory combo at Newegg or Micro Center, if you have one nearby. If that's not feasible, using one 7770 in the x8 slot won't make you take a massive FPS penalty... it's a couple percent/2-6 fps hit. You could also swipe up a 7850/7870/7950, sell the 7770s, and call it a day.

Otherwise, I have an extra Gigabyte board (same as the one in my sig) that I may be able to sell if you're interested. Would be a drop-in replacement to what you currently have and it supports Crossfire.

 

[AnitaPeterson]

WTF.... This is the entire text of the response I got from Asrock:

"Since your motherboard makes 8 beeps (video card as not detected) and when you install the video card into the second slot works , the ASRock motherboard got the problem."

 

[mfenn]

As has been mentioned, the 770DE+ is a DDR2 board. It's almost certainly out of warranty based on age. You're also not the original owner, which means the warranty is void anyway (non-transferable warranties are the norm for mobos).

Whether or not the system is worth buying a new board for depends on whether it is a Phenom or a Phenom II. I certainly wouldn't put any money into an original Phenom setup.

 

[chubbyfatazn]

As an addendum to what I previously posted...

If you buy another board, you'll be able to unlock. I haven't seen an AM2+ board use an 800/900 series south bridge.

Before you do anything, though... have you tried actually playing your games with the single 7770 in the x8 slot? If that doesn't cut the mustard, then short of getting a new cpu/mobo/ram I think selling the 7770s and moving up to a 7950 is a viable option. They're pretty similar in performance.

Otherwise you can scour the fs/ft forums here for someone offloading a 2500k and board.

 

[VirtualLarry]

Has it been determined whether the OP owns a Phenom I or Phenom II ? I agree with mfenn, if it's a Phenom I, best not to waste any money on it.

Better off getting a B75 + G1610 rig with SSD.

 

[_Rick_]

Sounds like the soldering/connectors on the first slot let go, due to the forces exerted by the weight of the video card on the slot during the crash.

considering the error mode, it's probable that one of the front-most pins got it, as otherwise it should be able to fall back to x8/x4/x1 mode, if PCIe does such negotiations. You might be able to force lower level detection by temporarily masking one or more of the hotplug detect pins (http://www.interfacebus.com/Design_PCI_Express_16x_PinOut.html)

Where's the PCIe controller on those chipsets? Integrated into the CPU, or into the southbridge? I suppose with the Socket AM2, you don't get the usual LGA-issues of slipping pins, so that's out. I'd guess the mainboard traces are also fine, if the lower PCIe slots still function.

You speak of a SATA/IEEE1394 card, that you used - if that's a PCIe card, try it in the slot. That could also help identify which end of the slot has the broken connection.

 

[AnitaPeterson]

Sounds like the soldering/connectors on the first slot let go, due to the forces exerted by the weight of the video card on the slot during the crash.

considering the error mode, it's probable that one of the front-most pins got it, as otherwise it should be able to fall back to x8/x4/x1 mode, if PCIe does such negotiations. You might be able to force lower level detection by temporarily masking one or more of the hotplug detect pins (http://www.interfacebus.com/Design_PCI_Express_16x_PinOut.html)

Where's the PCIe controller on those chipsets? Integrated into the CPU, or into the southbridge? I suppose with the Socket AM2, you don't get the usual LGA-issues of slipping pins, so that's out. I'd guess the mainboard traces are also fine, if the lower PCIe slots still function.

You speak of a SATA/IEEE1394 card, that you used - if that's a PCIe card, try it in the slot. That could also help identify which end of the slot has the broken connection.

Thank you for the reply! I think you make some good points, here! Could you please elaborate on the masking procedure you're suggesting? The link you provide is great, but I'm not sure how to proceed. I figured out that I have nothing to lose anyway, so I might as well turn this experience into something useful...

I also found this page, which, in addition to the pinout identification, also provides a visual reference: http://thisko.tistory.com/entry/Pinout-PCI-Express-x16-Pinout

Unfortunately, the eSATA/FireWire cards I mentioned are plain-jane PCI cards, not PCIe....


*edited thread title for clarification*

 

[mfenn]

considering the error mode, it's probable that one of the front-most pins got it, as otherwise it should be able to fall back to x8/x4/x1 mode, if PCIe does such negotiations. You might be able to force lower level detection by temporarily masking one or more of the hotplug detect pins (http://www.interfacebus.com/Design_PCI_Express_16x_PinOut.html)

PCIe can auto-negotiate, and in fact should do so automatically in the face of poor signal on a channel (an open circuit certainly qualifies!).

Where's the PCIe controller on those chipsets? Integrated into the CPU, or into the southbridge? I suppose with the Socket AM2, you don't get the usual LGA-issues of slipping pins, so that's out. I'd guess the mainboard traces are also fine, if the lower PCIe slots still function.

AMD doesn't have PCIe on the processor die yet, so the lanes are coming from the north bridge.

 

[_Rick_]

Thank you for the reply! I think you make some good points, here! Could you please elaborate on the masking procedure you're suggesting? The link you provide is great, but I'm not sure how to proceed. I figured out that I have nothing to lose anyway, so I might as well turn this experience into something useful...

I also found this page, which, in addition to the pinout identification, also provides a visual reference: http://thisko.tistory.com/entry/Pinout-PCI-Express-x16-Pinout

Unfortunately, the eSATA/FireWire cards I mentioned are plain-jane PCI cards, not PCIe....


*edited thread title for clarification*

The easiest way to mask, is to just put some reasonably robust (so that it won't tear when inserting the card) tape on the rear-end of the card's contacts, masking the pins corresponding to the rear 8 lanes and corresponding detection etc.

Although mfenn did point out, that PCIe usually should auto negotiate to the largest working configuration, it's not clear whether the implementation on both ends (card and northbridge) is actually fully implementing that part of the spec.

Alternatively you can also check if some of the pins in the slot have gotten misaligned or bent back by the crash, but it's more likely the actual connection between the traces on the board and the pin-out of the slot.

If an x1 card does not work, and you're willing to mess around a bit, you can turn the motherboard over, and carefully heat, pin by pin, the solder with a soldering iron and reasonably small tip, until it just changes phase. If you're better equipped, you can also wick the old solder up and put some fresh on, because re-flown solder does not have durability as its hallmark

 

[AnitaPeterson]

Well, here's what I ended up doing, in case anyone's curious....

I examined the slot several times, but was not able to detect anything suspicious.

With the second PCI-E still working, I decided to change the strategy and get rid of the computer altogether. I downgraded the processor to an Athlon 64 X2 4400, slapped a low-power video card in the working PCI-E slot, and put everything in a HTPC case. The next friend or family member in need of a computer will get it