It's story time with Uncle Dusty, kids, so scoot in and I'll spin you a tale of terror sure to make spiders crawl up your spine. It's a tale of avarice and betrayal. How one man's hunger for power ultimately led to his undoing. It's a tale... of upgrading.
As I mentioned in my last post, I found myself with the opportunity to make the jump from a Core 2 Quad Q6600 (overclocked to 3GHz) to an AMD Phenom II X4 955 Black Edition, yet this upgrade proved unwise. Some have questioned or criticized my desire for motherboard-based RAID, but I assure you I can tell the difference on this machine. I've seen video encoding and rendering processes get throttled by even the RAID 0's bandwidth. Yet a dedicated RAID card costs at least $300, and I have yet to see one that's universally recommended. Intel's ICH south bridges are a known quantity for me, and many benchmarks where they compete against affordable RAID cards are fairly tight races, as far as I've seen.
Being in need of decent motherboard-based RAID, I wound up crawling back to my Gigabyte GA-X38-DS4 motherboard and Core 2 Quad Q6600 processor, figuring I could just return the Phenom II, sell the motherboard I was given, and upgrade to a Q9550 or Q9650. I know some of you must now be asking, "Why would you bother with another Core 2? Why not just upgrade to a Core i7?" Your question is a valid one, and the answer is two-fold. First, upgrading to Core i7 would be prohibitively expensive. I already have 8GB of Corsair XMS2 DDR2-800 that plugs along happily for me, and I'm certainly not going to go with less RAM to upgrade. The expense of upgrading to a Core i7 pushes me over $600, at least, which is just too rich for my blood. Second, I have a completely irrational and ludicrous bias against Core i7. I don't like the branding, let alone how muddled it's become with Lynnfield, and overclocking seems daunting. The platform is also nowhere near as mature as the Core 2's, where everything is also a known quantity. And finally, since I know you're going to ask, "Why not wait for Lynnfield," my answers for i7 are for the most part applicable there, too.
So naturally, before even checking to make sure my old setup worked properly—why wouldn't it?—I ordered myself a Core 2 Quad Q9650. Again, this seems like another silly decision in a string of them. Why didn't I just buy a Q9550? Was the extra .5 on the multiplier really worth another hundred dollars? Well, first, I was able to get a healthy discount on the Q9650. Second, I wanted to start out from where I left off with the overclocked Q6600. And third, the extra .5 on the multiplier was worth the difference to me. With a 1600MHz front-side bus, that takes me all the way to 3.6GHz versus the 3.4GHz I'd get from the Q9550. Lower overclocks don't ask as much from my older X38-based board and DDR2-800 memory. And finally, it's a psychological thing. I would have felt (and do feel) happier with the Q9650 than the Q9550. Part of the purchase, at least for me, is the experience of having, and I do attach strange sentimentality and feelings towards hardware (hence my aversion to the Core i7). Insane? Sure!
With my old X38 board back in my case and the Q6600 humming along, I noticed an interesting quirk on startup: the system had reset itself to stock speeds. The computer was perfectly stable otherwise, but my overclock just seemed to have reset itself, and the system was taking an extra ten seconds to POST. So I went back and, naturally, set my overclock again at the settings that I knew were rock stable. Computer exits BIOS, saves, shuts down. Restarts, the lights come on... and no one's home. No POST screen, nothing. Ten seconds later, it shuts down again, then starts up again, POSTs: no overclock. I went through this process a couple of times with no success in changing settings. If you're like me, you're thinking something's wrong with the motherboard, like maybe some BIOS corruption or something. So I cleared CMOS, flashed the BIOS, changed the battery, sacrificed a goat... and the problem persisted. Worry set in. Did I damage my board? Was I careless? I tested it with my friend's Core 2 Duo E6600 and got the same issue.
At this point, I faced a conundrum. I couldn't "go back;" I was officially on the road to upgrading. Should I buy a new Intel-based board or purchase a RocketRAID card for the AMD platform? The Intel option would cost me roughly $40 more, but performance would be higher and I'd still have ICH-based RAID. The AMD processor and board would be easier to overclock, but the RocketRAID card would need to use a PCI Express x16 slot. Since the board is 790GX-based, that means my Radeon HD 4870's bandwidth would be cut in half. All signs point to PCI Express 2.0 x8 still being fast enough for the Radeon, but there's that voice in the back of my head going, "But what if they're wrong? What if those two frames per second you lose are off the minimum and not the average?" I ended up going the Intel route.
The only brick-and-mortar store selling motherboards in my area is Fry's Electronics, whose selection of Core 2-ready boards was remarkably anemic—many weak MicroATX boards and few offerings with at least a P45 chipset and a RAID-capable south bridge. In the end, I was left with exactly one option: the Gigabyte GA-EP45-UD3P. I came home, checked reviews, found I had made a good decision, and mounted it in the case while I waited for the Q9650 to arrive. I proceeded to live off of my laptop for the next few days, and I have to tell you... the GeForce 9600M GS just doesn't cut it for gaming if you've been using a Radeon HD 4870 for a year. Still, at least keeping the laptop attached to my 27" Dell monitor (another odd purchase choice, but it was cheap on campus) made the computing experience plenty enjoyable.
Only after starting to mount the GA-EP45-UD3P did I notice a major problem: Fry's had actually sold me an open-box product. Oh, it was presented as new and not marked as open-box, but it was missing the backplate and parts that should've been packaged in. I was so exasperated that I mounted the board anyhow and just used my old backplate, which matched nearly perfectly. Neuroses eventually set in, though, and I proceeded to order the same board new from Amazon. The new mobo arrived the same day as the Q9650. I removed the used EP45-UD3P, replaced it with the Amazon-bought model, and set up the computer again. Once I had a stable Windows 7 RC installation going, I proceeded to try overclocking the Q9650. I'm sure you can imagine what happened next. I'll give you a hint: the exact same problem I had with my old board. "Is this a Gigabyte issue? Did I get two bad boards? What's going on?"
At my wit's end, I drove an hour to the next closest Fry's to see if it had another option—maybe an Asus board in stock—but no dice. They did have an open-box EP45-UD3P, and by then I was just thinking, "You know, I've got two bad boards already. At worst, what's this one gonna do? Not work?"
Naturally, the new board had the exact same problem as the other two. Something was definitely up. Could I have damaged my power supply's auxiliary 12V connector with my constant fiddling? I'd taken to using needle-nose pliers to get in there and squeeze the clip. Had I damaged the power supply, and was the BIOS reading this as a reason not to let me raise or lower my front-side bus by so much as a megahertz? With no other options and in for a penny, in for a pound, I tried hooking up my old X38-based board and my friend's E6600 to the power supply in my dad's desktop. I plugged a spare Radeon HD 3650 into the graphics slot, and powered it on. Sure enough, it let me change the FSB. The board worked fine! Apparently, my PSU was the cause of all my woes.
The next day, I swapped my X38-based board back into my case and put in the Q9650, convinced a new PSU would do the trick. However, I wanted to double-check with another unit to be absolutely certain I had to go out and purchase a replacement. As it happened, a new case I had ordered to build a secondary machine out of spare parts (and sell it at a later date) arrived. I tried connecting that enclosure's bundled PSU to my motherboard, and the problem suddenly reappeared—I still couldn't change the front-side bus. Even if I disconnected everything but the video card, keyboard, and mouse, the problem persisted. At that point, I theorized that it was because the PCIe connectors to the video card might not be working right (insane, I know), or that the video card itself may somehow have been causing the problem. I asked a friend to come over with his Radeon HD 4870, swapped that card in...
...and I could overclock again.
This is one of the weirdest quirks I've ever encountered, but there you have it. My otherwise perfectly functional and stable Radeon HD 4870 was somehow screwing up the POST process on my motherboard. With a heavy heart, I sent my 4870 back through the warranty service, knowing I would probably get it back after technicians failed to see anything wrong with it. Now, with the Radeon HD 4670 from my media center in my old my old X38 mobo's PCIe x16, and a mountain of other hardware to return, my computer finally works fine again, and the Q9650 hits 3.6GHz stable at stock voltage.