About two weeks ago I started building a GigaStudio. I thought this would be easy since I have built many for myself. Since then I have used 3 different motherboards many different types of RAM and all other types of tweaks I have read about.
My first motherboard configuration:
ASUS A8N-VM CSM (NVidia 4/939/Video/Lan/SATA/Firewire)
AMD XP 3700+ (San Diego)
2 GB RAM – (2x1gb Chips) Kingston KVR400x64C3AK2 (QVL Certified by ASUS)
2 x Maxtor 320GB SATAII – RAID-0
120 Seagate ATA – Boot Drive
Enermax 485 Noisetaker
Frontier Wavecenter latest drivers.
Windows XP, Service Pack 1a with all the tweaks.
Turned off Parallel port and Serial port in bios.
Updated the bios.
I put all these things together, installed GigStudio and got the BSOD first time.
I then continued to turn off all on board chipset features except for VGA, still BSOD.
I moved the Audio card to another PCI slot and changed IRQ’s, BSOD.
I purchased a e-GeForce 6200 PCI-e video card.
Turned off all onboard features, still BSOD.
I took out everything except for the Audio card, VGA and 1 hard drive, still BSOD
I tweaked the ram, still BSOD.
The particular BSOD was "Machine_Check_Exception". And it always occured for every setting and change.
So then I thought, "well maybe it’s the RAM."
So I purchased over 1,500 dollars of different types of RAM and tried each one at different settings and CAS latency’s repeating all of the above.
List of RAM:
Corsair – TwinXP 1024 – 3200C2 (2x1gb)
Corsair – CMX512 3200C2PRO (4x512MB)
Corsair – TwinX 2048 – 3200c2 (2x1gb)
Kingston – Same as model as original but tried a different set
Geil – PC3200 Dual C2, Model GLIGB3200 PDC (2x1GB)
PQI Turbo – 3200 – 2048DB (2x512gb)
PQI Turbo – 3200 1024DP – (4x512MB) From another GigaStudio that works
Patriot – PEP512 PC 3200 (4x512mb)
I repeated all the above steps and still got BSOD. (Machine_Check_Exception)
I decided to try another mother board.
ASUS A8N32-SLI DLX
And I added a
SIIG SATA-II PCI-e RAID Card
With my e-GeForce card from the previous system. I checked it on other systems and it worked fine.
Still BSOD. (Machine_Check_Exception)
I moved the audio card around. Did all the previous steps. Updated the bios. Still BSOD.
I tried all the RAM I had, still BSOD. I tried other audio cards, RME DIGI96/8 PST,
AudioPhile 2496, Delta 1010, Wavecenter. Repeated all above steps, still BSOD. (Machine_Check_Exception)
I tried different networking cards, D-link, Trendnet, Netgear. Still BSOD. (Machine_Check_Exception)
I then moved to another motherboard after everything failed.
I moved to a Gigabyte, GA-K8N51GMF-9
Essentially another version of the first motherboard I was using except made by Gigabyte and not ASUS.
I figured if I was having the same problems with both previous motherboards maybe it was a brand issue. I rebuilt everything, meditated, went for a run and prepared myself for another long couple of days.
I built this next computer just as the first ASUS one, A8N-VM CSM. The only thing different was the memory I slapped in the Geil:
Geil – PC3200 Dual C2, Model GLIGB3200 PDC (2x1GB)
I installed and updated everything that needed updating. I only turned off the parallel port and serial port. Placed the audio card in slot 1 and made its IRQ 9. Ran GigaStudio, BSOD. (Machine_Check_Exception)
I decided to start messing with the CAS latencies on the Geil. I went in to the bios and found there were no settings to adjust memory. Instead, there are only the multiplier settings for the CPU. So, I said what the heck, I’ll adjust the CPU. So I adjusted the multiplier from 11x to 10x making the fat 3700+ run at 2.01 instead at it stock speed of 2.2.
And now GigaStudio runs. Solid.
So, I was thinking when you guys get the BSOD (Machine_Check_Exception) and you think it may be RAM also consider adjusting the CPU.
Last edited by mutato5; 01-27-2006 at 03:46 PM.
Reason: Update, I left something out.
Congrats on your success. What a path to get there!
Your story helps confirm that GS is the ultimate stress test. If your hardware isn't up to the task, GS will find it's weak spot. But it's not the software's fault that PC hardware quality is so variable.
Good tip on slowing the clock. That takes stress off the CPU, RAM, busses, everything.
Good you got it solved but this need for underclocking is pretty bogus. If your CPU / RAM / whatever isn't computationally perfect, IMO it's junk. All semiconductors have clocking limits past which they start to flake out, and the buffer should be more than 10%, which is why some people feel that overclocking isn't a bad idea. But when you're this close to your limit, you never know when it's going to poop out, and perhaps worse that a full crash is subtle corruption (like of files you moved on a hot day a year ago) that you might not catch for a long time and might never be able to associate back to the real cause.
I'd be interested to hear if your system can run a 'Prime95' stress test for 24 hours. I bet it fails. If it does, it's probably the CPU (since you seem to have swapped everything else) and I would get an RMA. Gigastudio has proven to be a great stress test, but it's not as conclusive as prime95. A flaky system may appear to work almost all the time when it's not stressed, but a perfect system should be problem free for years.
Also you might well have heat issues. Asus comes with the 'AsusProbe' utility that reports CPU and motherboard temperatures, and other makers probably have something similar. Make sure you have a good heat sink properly mounted, and a proper application of thermal compound. Make sure your case is exhausting the computers heat effectively.
When I moved last year I had a couple of formerly reliable computers go flaky; the move had jostled the CPU heatsinks and apparently broken the thermal compound seal. There was no visible change but the CPU's ran hot and one rebooted regularly and the other froze. A new application of thermal compound reduced temperatures by 15C (yikes!) and they returned to their old selves.
Finally I see more poor case designs than good ones. It's not enough to have 4 fans moving hot air around, but a couple of slow moving 120mm fans that draw cool air across all the hot things and exhaust it from the case can be totally effective. Modern computers need efficient thermal transfer much more than older designs.
Currently I'm still running Prim95 and things are going well. The CPU never gets above 45c and I have only three fans in the case rotating about 1500. It's a super quiet box with a big copper heat sink, Hyper48 by CoolerMaster. I agree with what you both are saying, my problem is after building about 30 Giga's I run into more problems then perfection and I find tweaking a good option. I regularly test with MemTest and Prime 95 and I absolutely agree that items should always be returned when problems arise. If things start going freaky I’m going to RMA this processor back to AMD and get another one. And I always love a good case design so I pay particular attention to thermal conditions and internal convection. When I’m happy with the design I put the giga in my server closet and let it bake for a couple of days testing. I’ll let you guys know if it comes out with problems.
WOW !!....That's a lot of work !!....I'm really glad you got it solved.....There are so many variable's to deal with ,and to think about thermal adhesive making a 15 degree difference just boggles the mind !!....Congrat's , I know I would have been pulling my hair out !!....Sincerely, Jim