I’ve been upset with motherboard manufacturers before over overclocking options I felt were a little too exposed, but this is the first time I’ve ever seen one company’s product sabotage the operation of another. This is, however, precisely the scenario I was afraid could happen if enthusiasts began pushing boundaries.
Based on my test results prior to this research, I had concluded that Mushkin DDR500 wasn’t very stable on this platform. I suspected the RAM itself was *not* at fault, but regardless, I would have thrown a caution flag on some aspect of the combination. As it turns out, I’d have been badly off-target.
Why did Mushkin’s RAM fail?
Because at a combination of multiplier 10.5x and a (supposed) 250 MHz bus speed, the motherboard was actually running at 263 MHz. At DDR400 / PC3200, a motherboard manufacturer can typically afford to run 5% out of spec without worrying about damaging machine stability; the same is not true at high frequencies.
The true irony here is that an end-user gets into trouble by attempting to keep his or her processor IN spec. I ran into problems when I held the memory controller frequency constant by lowering the multiplier. Unfortunately, an end-user who bought DDR500 and wanted to run that speed (without overclocking the CPU) could end up in precisely this scenario.
As a result of MSI’s attempting to market a product feature they didn’t actually have, both Mushkin and AMD easily could’ve come off second-rate, as could the Neo2 itself, billed as a “non-enthusiast” board. The ease with which this could’ve happened is disturbing, given the fact that I essentially stumbled on Everest’s correct diagnosis. I’m disgusted with MSI over pulling this without thought to the consequence for other company’s or themselves, and I’m disturbed at the lengths to which they are obviously willing to stealth overclock an end-user’s system without their awareness. Keep in mind, I had the “Aggressive Timing” and “Dynamic Overclocking” features disabled.
And if I was Mushkin, I’d be furious.
UPDATE: As of noon today I realized that our 90nm thermal results were ALSO run with this board using the same wonky mutipliers. Those will have to be re-run and re-compared.
Fonte : Sudhian Media
