I should have known better. I’d read enough online posts to know that the in-place upgrade from XP to Vista can be problematic, particularly if your XP system registry has grown to huge proportions. My own personal experience with a Vista upgrade had been moderately successful, but even that system was eventually nuked and a clean Vista install performed.

So here was a PC, with lots of fairly useless (from my perspective), but fun (from her perspective) background tasks running under XP. In goes the Vista upgrade.

Did I mention that this system had 2GB of system RAM?

Within two weeks, her system had pretty much become unusable. I scanned it with several malware detectors, and other than complaints about a couple of adware items, there were no Trojans or viruses found. So the problem really came down to an already overloaded Windows XP system that had developed massive registry and hard drive bloat after the Vista in-place upgrade. So I decided to nuke and start over.

One of the cardinal rules of troubleshooting is: change only one variable at a time.

Unfortunately in my hubris, I ignored that rule. I put her system on the bench, swapped out 2GB of Kingston DDR3 RAM for a pair of Corsair 2GB DDR3 modules. I then swapped out the QX6850 for an E8500 3.16GHz dual core CPU.

When I powered the system up, I got no POST, but the fans spun up. Also, the PC beeper didn’t beep. This is, as anyone who has built systems will tell you, one of the most ambiguous and frustrating types of errors.

So I popped her old CPU back in and rebooted.

Same thing.

So I put the old memory back in, and the system POSTed properly. I checked her BIOS on the P5Q3 motherboard, and found out it was still running the 0704 BIOS. That BIOS predated Intel’s 45nm CPUs. On top of that, if you read the summaries of the various BIOS updates, you also see a number of them with text that reads “enchances compatibility with certain memory.”

Flashing the BIOS fixed both the CPU and memory issues. At that point, a clean install of Windows 7 proceeded without any problems.

Lessons (re)learned:

  • Before upgrading CPUs or memory, check the BIOS version
  • Never change more than one item of hardware at a time without testing it.
Index The Constantly Crashing P7P55D


View All Comments

  • mindless1 - Thursday, January 14, 2010 - link

    ... and the aforementioned "different partition than the OS partition" can conveniently be on a NAS or fileserver when dealing with several client systems, even in a cloud if you've the throughput to the *sky*. Reply
  • vascos - Thursday, January 14, 2010 - link

    thank´s very much for this article. as a systembuilder i am used to this like every day.hardware does all test(like-prime,linx,furmark+atitool,memtest++)but we had ram(works fine on other mb) running for 2-3weeks and then instantly crashed-swapped other ram in and fine for 3month´s and hoping longer.this is with a lot of p55 mobo´s.

  • Etern205 - Thursday, January 14, 2010 - link

    While I agree the HD4870 is a old card to some extent, but I don't see it as a old card where no one wants to buy it. The card has dropped significantly in price which lets users on a budget get a card that's worth the performance within their price range.

  • bob4432 - Wednesday, January 13, 2010 - link

    this is why i always use mature to 1gen old hardware and os w/ at least a decent service pack out if i need to upgrade the os at all. w/ the exceptions of gpus, nothing else is really equal for the return on investment when you consider in the time wasted on troubleshooting bs due to "out of the gate" brand new items - let others waste their time and $$$ and when it is time for me to buy, i will build a machine using older tech that will be faster than people buying the latest because the drivers/os/support has been had a few once overs and i will save a ton of time chasing "beta" hardware output. Reply
  • Andreos - Wednesday, January 13, 2010 - link

    I found this article very interesting, also the comments. Lets have more like this! Reply
  • cesthree - Wednesday, January 13, 2010 - link

    I had very similar issues with my EVGA P55 FTW E657 and DRAM.

    My OCZ3P1600LV4GK wouldn't work at the rated 7-7-7-24-2T @ 1.65V AT ALL. Tried a slew of BIOS's and settings.

    Purchased some Dominator-GT's CMG4GX3M2A2000C8, set the XMP, and was flying from the get go. 12+ hours Prime Blend AND 20+ Passes LinX stable easily.

    DRAM compatibility is strange with the P55. I have seen all manufacturers forums filled with inconsistencies. Mostly there has been trouble with 1600Mhz kits.

    I have also noticed where the kit would work on an AMD, and not on the i5 i7 part the kit was "designed" to work with.

    Hope memory manufacturers and mobo manufacturers can get together to get this solved, or prepare for an onslaught of RMA's.
  • cesthree - Wednesday, January 13, 2010 - link

    Looks like you were dealing with a Xenomorph. Reply
  • nubie - Wednesday, January 13, 2010 - link

    Yep, it can sneak up on you and then bite you when you forget your own rules.

    Kudos on Windows 7.

    Please don't say things like "not bleeding edge system. Core components include an Intel QX6850" and "older 512MB Radeon HD 4870". It makes me want to cry, I have an e5200 and 8600GTS (could be worse, the GTS is quite a bit faster than the 8600GT, it has 32.3 GB/s of RAM bandwidth for one.) I had to search and search to get this system put together, even using an OEM MSI board (missing 2nd PCI-e and parallel ports, etc) and RMA'ing it twice to get it to POST with a 45nm processor. Some people have no jobs you know.
  • xeopherith - Wednesday, January 13, 2010 - link

    I experienced the same memory trouble you had when building a system that was for work purposes. I ended up having to trade the memory in my gaming machine with the work machine just to resolve some memory problems.

    Honestly it felt like a timing issue but I wouldn't get results in memtest only when running prime95 for long periods of time. After reading on OCZ's site it specifically said that the memory wouldn't work with anything but intel chipsets! I was building an AMD Phenom 940 when that was new with a Asus M4A78 Pro motherboard if memory serves. It didn't matter how much I underlocked, overvolt, undervolt, modified timings, ect.

    I traded the memory for some Crucial that was specifically marked compatible on the Crucial configuration utility and holy crap it worked fine.

    I have since had friends with the same problem but they certainly don't want to believe that if they pick the right frequency memory that it could be incompatible. It just goes to show you should be reading the PDFs for the motherboard and tested memory modules and or the memory manufactures tested ok modules. Both Crucial and Corsair have good documentation. I'm sure there are others but I prefer those two or OCZ.
  • Taft12 - Wednesday, January 13, 2010 - link

    First off, good article!

    You are correct when you say "the state of the art in DDR3 modules seems to be evolving rapidly". I find RAM more than anything to be impossible to keep up with and fraught with incompatibilities. I now recommend only Kingston Value RAM for any system. They advertise that it is "by spec" and I see it on EVERY QVL I've ever looked at. The 1.5V is key, every vendor bumps this up. High-end RAM will gain you a couple % points in any benchmark at best, it is absolutely not worth buying anything else IMHO for anyone other than an "enthusiast" (hate that term) and even then it is money better spent on other components.

    Thanks for bearing with my rant!

Log in

Don't have an account? Sign up now