Whatever happened indeed. To the PC that is. Here's the story you've all been waiting for :p
Just a refresher in case you've forgotten: I ordered a few bits and bobs, with the purpose of watercooling both my GPUs (2xHD5870) and my NorthBridge. The watercooling solution is an external all-in-one by Zalman, called the Reserator XT.
After getting everything I needed, I sat down to do something VERY scary: remove the stock fan from the HD5870s and change it to a waterblock. This, though easy, has to be done precisely because you don't want to get it wrong and risk components overheating.
It took ages, but eventually I managed to get everything set up and connected. I had to "degas" the Reserator and then run the whole system while the PC is off to make sure it's leak free. And it was.
The computer was running beautifully, a lot quieter (A LOT), and most importantly the GPU temps were fantastic:
NorthBridge: before watercooling 60 degrees; after watercooling 50.
GPUs (both): before watercooling idle at 35-40, active at 90-95 (with very loud fan!). After watercooling, the idle temp is around 20, but most importantly the active temp is only 45! (and by active, I mean a prolonged gaming session of mainly new stuff, like Shattered Horizon, Borderlands, etc etc).
SO - all was fine and dandy. BUT, before I go to the big BUT, here are some pics:
A scene ALL OF YOU are familiar with :) Waterblock installed on VGA NorthBridge waterblock installed Complete "loop": water enters NorthBridge loop, then exits to first VGA, then from there to the second one, then from there back to the Reserator. Hopefully :| Everything attached to the Reserator
Water flowing - no leaks. Yet.
BUT - DISASTER STRUCK!!!
The 3rd day in to the project, and the computer was humming away fine and everything was OK, I felt like a spot of Shattered Horizon. I ran the game, and after the intro videos it crashed. I didn't think much of it, but tried again and that crashed, then I rebooted and it crashed again.
My initial thought was that I'd broken the PC by recently removing both IE and WMP (something Windows 7 let's you do). Nevermind, a quick system restore, or, at worst, a complete C:\ drive restore (I backup weekly) would solve things.
I had to work a long day following day, so didn't have a chance to play around till the day after. I turned the PC on, and 10 minutes later was staring at a black screen. I tried again and again, and the damn thing just won't boot. The BIOS performed all the POST fine, and Windows was attempting to load, and then... nothing. NOT GOOD.
Stupidly, I still thought it was a software problem - so I decided to do a system restore. Nothing. So then I thought I'd restore my backed up C:\ drive from the network backup. I ran the Windows 7 "emergency" CD, and chose to restore from backup. Though I had specified only C:\ to be backed up, Windows had decided that my E:\ drive (a seperate drive that contained ONLY game installs. Nothing else) should be backed up too. So, I had to restore both drives.
I clicked OK and thought this is going to take at least 6 hours (restoring about 1tb of data over LAN). I came back about 2 hours later, to find a dreaded "THERE'S BEEN AN ERROR - YOU'RE FUCKED" screen. Not exactly those words, but you read it like that because you know full well your drives are now unusable.
And indeed they weren't.
SO - back to the good old let's reinstall windows plan. I did it, and luckily I use another seperate drive for the really important stuff (pics, mp3s, vids, downloads etc etc) and as such my software disaster only lost me a couple of saved games (torchlight among them!) and windows. BUT at least the PC was working.
I went about installing the ATI drivers first, rebooted, and BLAM! BLACK SCREEN OF DEATH.
OH SHIT.
It's around this time I thought that this is probably much, MUCH more than a software problem. I stumbled in to safemode again, and eventually found out that one of the HD5870s wasn't initialising. The problem was, when the ATI drivers came on, it automatically enabled CrossfireX (ATI's name for SLI) and as such the system was breaking because one card was dead. DEAD I TELLS YA.
I tried, and tried, and tried in vain to try and resurrect the broken card. You've all been there, and you know what I'm talking about: take the card out and put it in again, change it's slot, attach the DVI cable to all it's slots, change the power connectors, even BLOW on the damn thing. NOTHING. This process was made much worse by the fact that both VGAs were handcuffed together by the waterloop, as you see in the pictures above.
After a desperate day of struggling, I resigned to the fact that I had probably broken a card with my watercooling experiments. Why it had decided to break after 3 days is beyond me, and I still believe it's just a coincidental failure. I will RMA the card soon so we'll see.
With a semifunctioning PC (it was hilarious cos I needed the broken card in the PC to keep the waterloop flowing, but it had to be detached, so I had it just hanging there for a few days :|) I thought I'd order a new card. It was then when I thought I'd take the opportunity to do a minor upgrade - to a HD5970.
A little background here: The HD5970 is ATI's dual-gpu-on-one-card solution (akin the the 4870x2 before it, or nVidia's GTX295). It's made out of 2x5870 boards together, with internal CrossfireX magic. The only problem is, the cards are actually underclocked as compared to a single HD5870. The reason for this is to save on heat generation, and power consumption presumably. It also explains why one HD5970 performs worse than 2x5870 standalones.
To mix a 5870 and 5970 successfully, they all both have to run at similar clocks. While you can't downclock a 5870 (and that'll be stupid anyway), but you CAN overclock the 5970. This involves increasing it's voltages, and then overclocking it to at least the stand alone HD5870 speeds. This way, when the system enables CrossFireX, all 3 GPUs will run at the correct speed, hence improving performance AND stability).
Of course, to overclock the 5970, you really should have 2 things: a watercooling solution, and a PSU that can take the increased voltages. Luckily, a good PSU investment in the past (Toughpower 1500w with 3 12v rails) and my Reserator can cover these requirements.
I ordered the 5970 (XFX black, because it comes with it's own overvoltage tool) and also a waterblock for the new card.
Again, I was at that point of installing a waterblock on an expensive piece of VGA. I checked and doublechecked everything was fine, and then put the new 5970 instead of the broken 5870.
With my breath held and my heart pounding, I ran the PC for the first time and it all booted fine! All cards identified fine, and a quick test of any game showed that it was working OK. Phew!
I followed the excellent HD5970 overclocking guide on guru3d, and made sure the 5970 was running at the 5870 speeds. It's then that things started going wrong.
I was getting random hard-crashes (even out of game), and games were crashing as well. FUCK.
A bit of investigation and I found out that Everest Ultimate was a bit messy with 5970s, and caused crashes. So I uninstalled that, and hoped for the best. This time, out of games crashing stopped, but games still crashed.
It's at this stage, that looking back at it, I think I was at the bottom of a very deep pit of despair. You see, for about three hours, I had broken the "working" 5870 AND the new 5970. Let me explain:
I was trying to work out why the PC was crashing when any game loaded. I then stumbled upon the XFX support and download page and found that they had released a BIOS for the 5970, to make overclocking more stable. AHA.
I downloaded the BIOS and the flashing utility, and made sure I follow the instructions TO THE LETTER.
Rebooting, I went in to DOS (as the utility wants), and ran the flash.bat file. It backed up the BIOS images first, then flashed the cards with the new BIOS. All OK.
I removed the USB stick, and rebooted. BLACK SCREEN. BLACK FUCKING SCREEN.
This time, the black screen was at the VERY beginning. I was seeing NOTHING. No POST screen, nothing. My mobo (Asus Rampage II Extreme) has a small LCD that shows you POST details, and it was locking at "VGA BIOS". OH FUCK.
What could have gone wrong? Perhaps I'd bad flashed the 5970? I removed it and found out that I still couldn't boot. I removed the 5870 and inserted just the 5970 - same. Right.......
I got my old 4870x2 (Lesson: never sell anything old, ever. You never know when you'll need it) and inserted it. Voila - PC boots. I had a look inside the flash.bat file, and noticed that it was flashing card 0 and card 1 in the PC with the 5970 BIOS.
It was at this moment that a sudden feeling of dread descended on me. My card 0 is my standalone 5870, my card 1 is the master 5870 (within the 5970) and my card 2 would be the slave 5870 (within the 5970). What had happened is the flash file had flash the standalone 5870 with a 5970 BIOS, and the master 5970 with the SLAVE 5970 BIOS.
ie I had bricked, within 10 seconds more or less, 2 VGA cards. Bricked them.
At this point, I genuinely wanted to quit the whole PC thing, and live on consoles for my gaming fix.
After dinner, and a bit of rational thinking, I had an idea: I remember the flash utility saved the BIOS images before flashing. So, if I booted the PC via the 4870x2, and then ran the flash utility (making sure it flashed the correct card), then I should fix the VGAs.
Again, I read and read the ATIFLASH.EXE commands, and very gingerly I booted to DOS and flashed my 5870 with it's original BIOS. I then flashed the 5970 with the new BIOS. Of course, I didn't use the silly flash.bat file, but manually flashed everything myself.
AND IT WORKED.
Amazingly, everything worked. After about 3 hours, I fixed everything. I kicked myself for not paying more attention before running that dreaded flash.bat file, but at least we got there in the end.
The question is - will this new BIOS fix the in-game crashes? Of course it didn't - don't be silly. Nothing is easy in the world of PC gaming.
For the second time, I resorted to the let's-reinstall-windows plan. I reinstalled, and installed all drivers etc, and I installed MSI afterburner (as recommended by guru3d) to check on the cards. It's then that I realised that the silly XFX overvolt utility FORGETS IT'S SETTINGS after each reboot!
For the 5970 to overclock stable, it's needs more power. With each reboot, the card was resetting it's volts to default, but the clocks were overclocked and as such it crashed instantly, crashing the PC. MSI afterburner retains it's settings, so I use that now to monitor and overvolt/clock the card. This solved the whole crashing business.
THERE IS A HAPPY ENDING - PROMISE. That being, that now everything is fine. As in, I have a watercooled PC, the temps are perfect, I have Tri-CrossfireX enabled (3x5870, split over 1x5870 and 1x5970) and running at 5870 clock speeds. Games are running perfect, and all in all it was worth it. Just.
This latest PC has really ben an experiment for me. From overclocking a CPU, to watercooling VGAs, to overclocking VGAs, to bricking and then unbricking VGAs.
I sometimes think, nay BELIEVE, that it's a curse, not a hobby.
Good delicious!
