Oct 3 2008
… or why I vow never to build a PC ever again 🙁
Having last built a new desktop computer (I do have an ever expanding collection of MiniITX servers, but these don’t count 😉 ) in about 2003, I decided that it was high time to upgrade my 256Mb Athlon 1800+ desktop to something more recent: I was finding that I wasn’t updating my Gentoo Linux installation on the basis that I’d probably be replacing it in the near future, and I really wanted to play a few more interesting recent games such as Bioshock.
I eventually took the plunge and bought an AMD Phenom 9850BE and a SB750-based Foxconn A79A-S motherboard from www.scan.co.uk and 4Gb Corsair TwinX Dominator 1066MHz DDR2 memory and a Sapphire ATI Radeon 4850 from www.overclockers.co.uk.
Here, the trouble started…
I assembled the system (which basically meant gutting my Yeong Yang case and replacing everything inside… except for the floppy drive. Seriously…) and powered it up… and immediately hit problems: Attempting to boot any flavour of Linux – and I did try about three – resulted in a hang before the kernel had finished initialising. Attempting to install Windows XP SP2 got a little further – after loading the various device drivers, it hung at the “Setup is starting Windows...” message.
I disconnected all additional peripherals, and the behaviour was still the same. I tried completely different memory (2x512Mb 667MHz DDR2 Micron/Crucial modules) and the problem remained. I took the graphics card to work and tried it in a PCI-Express machine in the office, and it performed without problems. This seemed to narrow the problem down to the processor or the motherboard. I packaged up the processor/motherboard with the heatsink and sent it back to scan as faulty.
scan got back to me only a few days later, and said that they’d successfully installed Windows and run Prime95 overnight, and that it was working correctly. I have no problem with that – perhaps shipping shook something lose – but what irked me was that scan’s site clearly states that the cost of sending back a non-faulty item is £15. This, it transpires, is incorrect: the cost of shipping the item is £15, but they also charge for their technician’s time, and so the actual cost is £35+VAT. This isn’t so high, and given the hourly rates that some companies charge is actually fairly reasonable… so why not be upfront and tell customers this? By hiding the true cost, I feel as though I’ve been misled. I paid them by credit-card and then left on holiday, only to have random problems using the card for payments, and then finding upon my return that the card had been blocked (so how did any transactions succeed?) because the second transaction with scan was flagged as suspicious. I couldn’t possibly comment…
In any case, they eventually shipped back the motherboard and processor – but the heatsink was nowhere to be found. I contacted scan again, and the operator said that he’d send through “a heatsink” (although not “my heatsink”, I noticed) on next-day delivery. This was sent to my home address rather than my delivery address, which meant that I wasn’t in when the courier called. Without the option to have them re-deliver to a different address, not wanting to stay in all day for the sake of a heatsink, and keen to see whether the machine now worked I decided to collect the package from the courier’s local depot.
Google maps puts the postcode for the depot as the middle of a field in broadly the middle of nowhere – but, I noticed, near an industrial estate. I thought I’d try there. After at least an hour of driving through the winding roads of deepest, darkest Cambridgeshire, I eventually found the depot – many miles from where the postcode indicated and entirely un-signposted until you’re right on top of it. I finally got the package, ripped it open, and found that scan had sent me completely the wrong thing:
The retail heatsink supplied with the Phenoms is actually a thing of (relative) beauty: copped-based and with a number of heat-pipes, it’s actually fairly efficient and quite quiet. What I’d received, on the other hand, was a large and loud Aluminium affair. Phoning scan the next day, they promised to find my original heatsink and again get it to me by the next day. This arrived this morning.
However, from trying the machine last night, there are still serious problems:
- Once the correct drivers are installed, Windows will hard-lock on boot if the on-board Realtek HD Audio subsystem is enabled;
- Brief presses of the reset button cause the display to turn off and the machine to hand – the reset switch must be held for 5 seconds or so to actually reset the machine;
- Randomly, the machine will hang during POST, seemingly during the detection of IDE or USB devices;
- If IDE mode is set to “Legacy IDE”, the machine freezes when detecting IDE devices before it is possible to enter setup, requiring a CMOS reset;
- The 1066MHz memory I have is detected and runs at 800MHz. Even with the DIMM voltage boosted from the default 1.9v to 2.1v (which is apparently required, but not documented outside of their support forums for these DIMMs), the machine freezes pre-POST if either 1066MHz is set or the EPP SPD data from the memory itself is used.
Finally, although this isn’t the hardware’s fault, I’m unable to boot Linux currently as none of the CDs I have (2.6.19 and 2.6.24 respectively) support the IDE/SATA chipset present, and so can’t see the hard disc. I’ll try Ubuntu tonight and, failing this, it may be possible that one of the secondary controllers might work correctly (as the two SATA DVD drives are recognised).
If another Windows reinstall (which will be three and counting) fixes the on-board audio problem, then I should finally have a machine which is at least usable…
But my conclusion from the whole affair is that if you’re looking for a modern, powerful desktop PC, then the Mac Pro is the way to go… I’d have been scaring myself silly playing Bioshock weeks ago by this point!
So why did the returned board and processor work when the ones I sent away didn’t? It can’t have been any BIOS tweaks, as the process of testing the memory required numerous CMOS resets. The hardware is unchanged and connected in the same way as previously. Infact, the only thing I can think of is that I may originally have accidentally plugged in a Firewire header to the motherboard offset by one pin. Could this have killed any OS which I tried to boot?