Win 7 x64 hangs at load screen after installing ANY drivers for HD 6990

Hey guys, I'm working on my first ever liquid cooled rig at the moment (worklog link in sig, but theres no info there about the problem i'm posting here about), and I've more or less finished the hardware side of things, did my 18 hours of leak testing, but then ran into problems when I got to the software side.

Just to start with, these are system specs, so you don't have to read the worklog.

Windows 7 Ultimate x64
i7 990x
2x Dominator GT triple channel 12GB packs (6x 4GB DIMMs)
Rampage III Black Edition
2x OCZ Vertex III 240GB (RAID 0)
Corsair Force 3 Series 240GB
2x Gigabyte Radeon HD 6990 (CrossfireX)
1500W Thermaltake ToughPower W0171 Modular PSU
Coolermaster HAF-X
Pioneer BDR-206BK 12X Blu-Ray Writer

And the cooling setup

Swiftech MCP-35X pump, reservoir, heatsink
Swiftech ApogeeXT rev 2 CPU block
Swiftech Epsilon HD6990 fullcover graphics card blocks (x2)
Hardware Labs Black Ice GTX360 (120.3 rad)
Hardware Labs Black Ice GTX480 (120.4 rad)
Scythe Gentle Typhoon 5400rpm fans (for rads) x7
Lamptron FC8 fan controller
1/2 ID 3/4 OD tubing and fittings
Distilled water with Feser corrosion blocker and silver coil for coolant

Anyways, the first thing I did once I had concluded my leak testing (with no problems or leaks detected) was switch the motherboard to BIOS2 (i'm leaving the first one default) and flash it to 0501 (which I found here) so that I could RAID my Vertex III's on the Marvel 6GB/s controller, and set the intel sata controller to AHCI. I managed to create the array and have windows detect and install onto it fine, with no problems.

This is where my issue began. I starting doing my usual settings blitz in windows to get things how I like them (nothing major just turning of UAC, changing start menu settings ect), then moved onto drivers.

I won't go over step by step exactly what i've done cause it would take to long but basically I was able to install any drivers with no problems except for the display drivers for the 6990's.

Chipset, eSATA, SATA, Thunder bolt LAN / Audio, USB 3 drivers all went on without any issues, but whenever I install ANY drivers for the graphics cards, (included gigabyte support disk drivers, latest gigabyte drivers online, 11.6, 11.7 from ATI / AMD site, 11.8 preview from ATI / AMD site, and even windows update 6990 display drivers), it asks me to reboot, gets to the loading screen (where the 4 coloured lights swirls around once and form the windows logo), then I hear a faint click coming from somewhere inside the case, which is normally when the login screen (or desktop if I haven't set a password) appears, but it just sits there indefinately (have left if for 5 hours and took a power nap, still there when I woke up).

Startup repair does not fix the problem, nor does sfc /scannow, or changing any other software in windows. I am able to get into safe mode, and by going into command prompt safe mode I was able to do 'start devmgmt.msc' to get the device manager and uninstall the 4 display adaptors (presumably 1 for each GPU?), which lets me log back into Windows normally no problem, but without drivers for the video cards.

I have done fresh installs 3 times and tried a multitude of things, taking 3 of the RAM DIMM's out, disabling each card thru the mobo's PCI-E switch function (both cards boot and will display a picture as single primary card, so its not a PCI-E lane), going through the BIOS settings, getting ALL the windows updates including SP1 and then trying again. Nothing has fixed or changed the problem in any way.

I'm really stumped, surely it can't be a faulty card or the offending item wouldn't display picture at all when set to single display on that card? I also don't think it could be the PSU as its well above the required wattage and the PCI-E 8 pin connectors aren't all on the same rail. The board has the supplementary 4 pin molex connector for the PCI-E lanes plugged in at the bottom, the crossfire strip is connected properly, and i've tried both BIOS on both cards.

If I've left something out i'd be happy to provide any other info you guys might want, Just let me know and i'll post it. Please guys, I could really use a hand here :(


EDIT - I have put this issue behind me, and while I am still unhappy about the outcome, I would consider the problem solved.

A quick summary:

PROBLEM
- two brand new Gigabyte HD 6990 cards display picture in new computer, but cause Windows 7 to hang at the loading screen if any display drivers are installed
- cards had swiftech epsilon hd6990 waterblocks installed by shop that I bought computer from, and were apparently working fine, but they never installed operating system so I'm not sure how they tested them

ATTEMPTED FIX
- Startup repair
- sfc /scannow
- uninstalling every other possible piece of software on the system
- fresh install of Windows multiple times, on three different ssds in RAID 0 and single configs, using both marvell 9182 SATAIII and intel ich10r SATAII controllers
- same thing as above, but with Windows disk from my Alienware laptop (same windows version)
- putting the cards in different slots, using each card by itself, with the other removed
- checking the power leads and spreading the load over multiple rails on PSU
- taking half of the system RAM out
- windows updates
- various BIOS settings
- all posted suggestions in this thread up to post #17, and a few other little obvious 'hurr durr' fixes like different cables, different slots, checking card is plugged in properly

FINDING WHERE THE PROBLEM WAS
- bought a cheap HD 5450 graphics card, ran with all the same settings, same drivers, worked perfectly
- took cards to local store to have stock coolers replaced, and cards tested in a different computer, still did the same thing on their rig

FINAL ACTION
- went with GTX 590's instead, still unsure whether the cards were defective to start with, or if installing the waterblocks damaged them, and expensive lesson learned to test things thoroughly before modding them

ATIs drivers sometimes have issues with two DIMMs per channel being installed. Remove one DIMM from each channel and try again. I've had a BSOD referencing a SYSTEM_SERVICE_EXCEPTION and atikmdag.sys myself, but a hang is also possible. If the memory reduction doesn't work, try moving the video card to a different slot.

Quote:

---

Originally Posted by lutjens

ATIs drivers sometimes have issues with two DIMMs per channel being installed. Remove one DIMM from each channel and try again. I've had a BSOD referencing a SYSTEM_SERVICE_EXCEPTION and atikmdag.sys myself, but a hang is also possible. If the memory reduction doesn't work, try moving the video card to a different slot.

---

I'll give the RAM another try tomorrow either before or after work. Just to verify though, my board has two channels, each with three DIMMs, not three channels with two DIMMs each, are you saying to remove all the RAM from one of the channels, eg the 2nd, 4th and 6th slots? That is what I tried earlier, but I think it might be worth another shot.

As for changing the slots, it won't be so easy, I have the cards in slot 1 and 3 which are the only x16 slots on the R3BE board, 2 and 4 are 8x for tri and quad card setups. Plus if the slot was dead, the card would not display at all, and it does, for both cards. I was able to test slot 3 as the main card by switching off the PCI-E 1 slot with the handy PCI switch on the mobo, and disconnecting the power / crossfire leads from the top card. It worked fine, as did the top one when I turned off slot 3. I can't easily move them because the blocks are full of water at the moment and are connected by phobya 3 slot variable SLI/CF connector pipes.

The problem only occurs after installing a display driver for either one, or both, of the graphics cards, and restarting windows. I can still boot into safe mode, presumably because windows doesnt load display drivers in safe mode, and I can get back into normal mode by uninstalling the display adaptors in device manager using safe mode with command prompt. I really hope this is just a software problem I haven't figured out yet, I've never a faulty graphics card or motherboard (except one that was physically broken before it went into any computer by shipping process, so I don't count that one).

Any other suggestions for me to try tomorrow?

Cheers!

If you remove one card, install drivers for one card. Shut down and install the second card, and activate crossfire. Any luck?

Stress testing both cards individually is also a good place to start.

Quote:

---

Originally Posted by Kyprioth

I'll give the RAM another try tomorrow either before or after work. Just to verify though, my board has two channels, each with three DIMMs, not three channels with two DIMMs each, are you saying to remove all the RAM from one of the channels, eg the 2nd, 4th and 6th slots? That is what I tried earlier, but I think it might be worth another shot.

---

X58 is triple channel, two DIMMs per channel. I looked at your manual, try installing DIMMS in the DIMM slots A1, B1 and C1. If you start counting the DIMM sockets from your CPU, fill sockets 2, 4 and 6 (DIMM socket called #1 closest to CPU should be vacant).

Did you try connecting your monitor to the other GPU?

Quote:

---

Originally Posted by wez

If you remove one card, install drivers for one card. Shut down and install the second card, and activate crossfire. Any luck?

Stress testing both cards individually is also a good place to start.

---

Yes, I tried this. I didn't physically remove the other card, just disconnected the power / crossfire bridge, and killed that PCI port with the mobo switch, which is essentially the same as it not being there at all. Once drivers are installed, for either card, no matter which one, or both, I can't get into windows (except in safe mode). I guess I could try installing the drivers in safe mode for the second card? And how would I stress test a card without drivers installed, or without going into windows, can you give me any info or help on that please?

Quote:

---

Originally Posted by lutjens

X58 is triple channel, two DIMMs per channel. I looked at your manual, try installing DIMMS in the DIMM slots A1, B1 and C1. If you start counting the DIMM sockets from your CPU, fill sockets 2, 4 and 6 (DIMM socket called #1 closest to CPU should be vacant).

---

Ok, thanks a lot, I will try that. The reason I said it might be worth another shot before is that the mobo didn't want to go at all when I took them out of the other three slots, now I know that it was probably because I took them from the first channel. I didn't fully exhaust that path before posting here.

Quote:

---

Originally Posted by CrazyNutz

Did you try connecting your monitor to the other GPU?

---

Yes, in various Mini DisplayPort plugs on both cards (MiniDP to DP cable from Monster). I get a display, and can log into windows fine as normal, until I attempt to install display drivers and restart the computer. I haven't tried installing them with both cards in, then plugging the cable into the second card though, so i'll give that a whirl too.

Thanks a lot guys, I don't have time right now, gotta go to work, but i'll see what happens in about 14 hours or so when I'm home again.

In the meanwhile, feel free to add anything else to the list for me to try later, cheers!

wow your system is hi-end

Quote:

---

Originally Posted by PatRaceTin

wow your system is hi-end

---

Thanks :D. Too bad it's only good for an extremely expensive browser and calculator without the display drivers :mad:

Alright, no dice on removing the second channel of RAM. The problem remains exactly the same. I also tried removing and re-seating the cards in their slots (both at once so that I didn't have to drain water from the loop), tried changing the 8 pin power cables around in various configurations to see if maybe a rail was being overloaded, and removed the Thunderbolt card to see if that had any effect. Interestingly after taking out the thunderbolt, the two faint clicks coming from the computer just before the desktop appeared stopped, but the system still hangs with display drivers installed. The keyboard and mouse light up, but it just sits there with the windows logo indefinitely. I tried plugging the monitor into the bottom card while both cards are active, and get no picture, as I expected, but when I disable the top card, the bottom one shows BIOS and will log in to windows (without drivers) fine.

I don't have any other graphics cards, computers or power supplies so I have no way of eliminating any of them as the cause of the problem, but a bit of Google searching led me to believe that my windows 7 disk might have some kind of fault on it. It seems unlikely, but I may later try installing again from my Alienware m17x laptops disk, and just use the product key from the new copy, and see if that makes a difference (both disks are Win 7 ultimate x64). I might also try plugging the drives into the SATA II plugs controlled by the onboard intel chipset rather than the Marvel SATA III ones and see if that makes a difference.

I have never encountered a problem like this before, where a brand new system with presumably no other software in conflict and still the drivers have issues. I can't bring myself to believe the cards themselves are faulty, because surely they wouldn't be detected by the BIOS in the GPU.DIMM post (they show up as ATI GPU @ 16x), and surely they wouldn't show a display at all, even without drivers installed? I guess just for the sake of being certain, I'll also check that the pins in the PCI-E leads are in the correct spots again, although I triple check every cable after sleeving them.

Any other suggestions for me?

Alright, so i'm fairly sure that there is actually something wrong with the 6990s themselves, even though they are both brand new.

I installed Windows from my alienware disk and experienced the same problem. I installed windows on my corsair SSD on the intel SATA controller (ICH10R), same problem. I went down the road to Austin Computers in Rockingham and picked up a Sapphire Radeon HD 5450 for $55 and it worked fine, with drivers installed (both the ones on the disk and the latest 11.8 drivers from ATI AMD website).

So I have drained my loop, taken the 6990's out, and intend to replace the stock air coolers, then see if I can get somebody to plug them in to a different computer and test them, just in case the problem has to do with my power supply or some other hardware conflict that would affect a 6990 but not the 5450. Then I guess I try either two new 6990's or a couple 590s. I'm not keen on having more than two physical cards on the computer, eg 3x 580s.

I'll post back if and when I am able to determine if the cards work in another system.

Did you try them with the stock coolers mounted in YOUR rig??? This could very well be a "block" issues as well.

I has some XSPC blocks that bricked 2 pairs of 4850x2's, and they were doing the same thing as you are experiancing.....

Quote:

---

Originally Posted by screwtech02

Did you try them with the stock coolers mounted in YOUR rig??? This could very well be a "block" issues as well.

I has some XSPC blocks that bricked 2 pairs of 4850x2's, and they were doing the same thing as you are experiancing.....

---

The cards were both tested with air coolers by the guys who put the blocks on them for me (before putting the blocks on, obviously), but I told them not to install the operating system because I still had a lot of hardware changes to make (literally took the whole computer apart again) and wanted to do that part of the setup myself. So they would never have run into the issue with drivers causing a hang.

I do intend to plug them back in my rig with air coolers before taking them anywhere else to see if the problem is magically fixed. What did you end up doing that solved the hang? Did you get new blocks, different cards, or do something else?

I have been looking at the GTX 590 as a candidate for replacement, but even with a waterblock, the PCI bracket and display connectors still take up two expansion slots, so there would be no room for the Thunderbolt unless I put it in the bottom full size PCI-E slot. I'm pretty sure doing that would split the bandwidth and therefore change the third slot (second GFX card) to 8x instead of 16x. The measurements do look big enough to allow me to put the Thunderbolt between the two cards, with the SLI/CF connector pipes clearing the edge of the card and 'caging' it in, but the cable routing would be difficult and ugly, and more importantly I'm not willing to risk trusting the numbers without actually getting my hands on one to try.

There are virtually no stores anywhere in Aus that seem to have 6990s in stock, the two I have I wouldn't have even bought had there been an Asus or XFX pair in stock. I'm not sure what cards to go with if I can't get the two I have working.

Actually, as I write this I've had an idea, I always assumed that because the second and fourth PCI-E slots were labelled as only 8x, that putting a card in either one with the slot above empty would still only give that slot 8x instead of 16x, but with this tiny 5450 it wont be hard for me to quickly test that. If I can get a card in slot 4 working at 16x then the double slot issue won't be a problem with the Thunderbolt.

Cheers

EDIT - The second and fourth PCI-E ports on the Rampage III Black Edition only work at 8x, even with the slot above vacant or switched off, and plugging a PCI-E 1x card into either of them does in fact cause the slot above to reduce from 16x to 8x. Bummer.

one sugestion if this problem persists is to add boot logging to screen instead of showing he windows splash screen.
that way you can see exactly what driver it hangs on...
it may or may not help. but just an idea.

Quote:

---

that way you can see exactly what driver it hangs on...

---

I have restarted the computer after installing drivers one by one and its always after installing display drivers, with or without the Catalyst Control Center, that the hang occurs. I have removed the cards from the loop now, so I can't test them on air until I get some thermal paste to put the stock coolers back on. However, I am positive that the display drivers themselves are not the problem, but rather that when Windows loads them, the cards don't like something that's happening and stall everything.

For example (warning - uneducated tech babble) the drivers could be telling the cards to go to a higher voltage than is used as a generic VGA adaptor, or something, which is where the presumably faulty cards freak out.

i would recomend hdd erase do a ssd format. reinstall OS. 11.6 farks up alot of people systems. which up top you said you have installed. not sure if you have done this on your latest install as it is not clear.

once you do the format and fresh install. then use 11.7 or 11.8 get them from the maker of your cards not amd.com.

if the card is a ref design 6990 it shouldnt make a difference on the place you get the drivers from but in my expierence it does.

you are using 2 crossfire bridges right?

Quote:

---

Originally Posted by SkullCracka

i would recomend hdd erase do a ssd format. reinstall OS. 11.6 farks up alot of people systems. which up top you said you have installed. not sure if you have done this on your latest install as it is not clear.

once you do the format and fresh install. then use 11.7 or 11.8 get them from the maker of your cards not amd.com.

if the card is a ref design 6990 it shouldnt make a difference on the place you get the drivers from but in my expierence it does.

you are using 2 crossfire bridges right?

---

I only tried the 11.6 drivers on my two first installs of windows. All of the fresh formats since then I've used only the newer drivers, from both Gigabyte and AMD. Its not the drivers themselves causing the problem, I'm positive of that. The exact same 11.8 drivers work fine for another card.

It is a reference 6990, I don't know of any brand that has made a non-reference PCB for that card yet.

They only have a connector for one CrossfireX bridge on each card. I assume this is because additional connectors are meant to be used for 3 and 4 card setups, and the 6990 is a dual GPU card, so you wouldn't use more than two of them.

Before you ask, yes I am absolutely 100% certain there is only one connector for the bridge, and by that I mean the physical part of the card only has one connector, not that it only came with 1 bridge ribbon.

I am in the process of having the cards' stock coolers replaced and tested in a different computer. If they start working then I'll try to resell them, I've decided to try my luck with a couple GTX 590's. Obviously this time I will be testing only one of them, with the air cooler, before I open the second box, and then testing them both in SLI on air before installing the waterblocks.

Quote:

---

I am in the process of having the cards' stock coolers replaced and tested in a different computer. If they start working then I'll try to resell them, I've decided to try my luck with a couple GTX 590's. Obviously this time I will be testing only one of them, with the air cooler, before I open the second box, and then testing them both in SLI on air before installing the waterblocks.

---

I got a call back from the shop, they say that the 6990's are doing the same thing on air coolers in their test rig. I'm going in later today or on Monday to pick them up and ask them some more questions about the tests they did. In the mean while, I got my Asus GTX 590's and some goodies from FrozenCPU, including two Koolance VID-NX590 waterblocks, and am in the process of testing them now.

So far, I have plugged the first 590 into the top PCI-E slot with it's stock cooler, and it works perfectly with drivers. That tells me straight away it shouldn't be an issue with power that was buggering around the 6990s. I intend to put the other in, set up SLI, and stress test them for a few hours, then install the waterblocks and redo my loop.

So for anybody who experiences the same issue, I am sorry to report that I never got the 6990s working, and the issue as far as I can tell is definitely with the cards themselves. If the shop gives me any interesting or useful info I will post back for the benefit of anybody else who runs into a similar problem, but otherwise, I'm declaring this issue a thing of the past.

I probably have the same issue with you...
http://www.overclock.net/t/1199305/p...#post_16199737

Still cheking, but i am not very optimistic any more!

I went to the GTX 580 over this issue (and the better WoW performance), but found that disabling NUMA can rectify the problem. Don't know for sure if NUMA options exist on 1P boards, though...

Reported the issue to AMD last fall....never heard of any fix being done in newer drivers or if they even care. Nvidia doesn't have the issue at all, either with NUMA enabled or disabled.

I have similar win7 boot bsod (0x3B on atikmdag.sys) experience on (i) sager 8180 notebook, a gaming notebook with dual 6990m; (ii) 6950CF on p67; (iii) 7970CF on p67.

What I notice is Win update will install "ATI display driver etc etc" if u still using standard vga driver. You can either (i) hide this update item (so not install MS version of ati display driver) or (ii) install catalyst b4 u do the win update. Avoiding this solve the bsod boot issue. I guess using driver sweeper/ccleaner to remove the MS version driver will also work.