Microsoft analyzes over a million PC failures...

Microsoft analyzes over a million PC failures, results shatter enthusiast myths

What they found is fascinating. The full study is well worth a read; we’re going to focus on the high points and central findings. There are two limitations to the data collected that we need to acknowledge. First, the data set we’re about to discuss is limited to hardware failures that actually led to a system crash. Failures that don’t lead to crashes are not cataloged. Second, the data presented here is limited to hardware crashes, with no information on the relative frequency of software to hardware crashes.

Click to expand...

Source

A Guy

But the analysis can only be as good as the crash data received. Nevertheless interesting.

We present the first large-scale analysis of hardware failure rates on consumer PCs by studying failures in the CPU, DRAM, and disk subsystems.

Click to expand...

Yeah well that may be intersting read for harware manufacturers

Selection, bias, 3.2.1, yeah, what I feared.

My machines are not there.

Figure 6 shows that white box machines have less reliable CPUs and dramatically less reliable DRAMs than brand name machines, although disks are just as reliable.

Click to expand...

Goes to show when homebuilders go for the cheap way out. I currently have 5 homebuilt desktops (two are modified HP U desktops) and I do not think they have any of the described issues. But then again, the stats are a bit overwhelming to the contrary.

How does this study help improving Windows? Ah, section 8. Interesting.

OEM has a locked down BIOS, no OC'ing.

Custom built rigs have unlocked BIOS, so we play with the settings, it's unavoidable.
To OC a system you advance the settings until it crashes, then adjust again or back them off.
So a custom built system will have more crashes but, will run better
This usually requires more voltage which produces more heat which degrades electronic devices, we all know that and OC anyway.
If you OC with EIST (speed step) enabled your system will run at lower voltage and lower heat most of the time, making it last longer
So it's all good, and we get to have fun.

lol...

To OC a system you advance the settings until it crashes, then adjust again or back them off.
So a custom built system will have more crashes but, will run better

Click to expand...

not me. Since i played around with KT133 a few years ago I never done any OC.

Since MS primarily deals with software, it's hard to fathom all the different combination of hardware components that create a near endless amount of variables to track down a definitive cause of these failures.

PC enthusiasts are generally those who pick their hardware very selectively (unless their budget can only afford an ultra-cheap setup) and for OC'ing, they tend to employ decent controls (better HSF, pastes, etc.) to manage the additional stress/heat. Since you don't have to be an IRQ/device conflict genius now unlike the Win95/98 days, putting a system together has gotten much easier. The casual user that goes to Costco or what not to buy the cheapest computer made by <insert random OEM here> may not know what quality of components reside in their purchase (I still have nightmares of people buying Compaq or e-Machine computers).

Skewing off the topic just a tad:
Even Apple users still have a very hard time accepting that iOS apps crash/fail significantly more often than Android apps. Unlike Microsoft who deals out only the OS, the hardware found in each iOS device is strictly controlled by Apple. Study: iOS apps crash more than Android apps do | TUAW - The Unofficial Apple Weblog

To me that research is useless. Not over a long enough span of time. Not knowing the products, equipment, and or programs also makes it useless. Hell, lack of a Java update can cause all kinds of problems. If you overclock and don't get a shut down your not trying hard enough.

Agreed. Personally I used to OC my Slot-1 machines back when the Celeron 300A could easily hit 450Mhz (and was significantly faster than the Slot-1 PII 450Mhz due to the Celeron's full speed L2 cache) with just a minor bus speed change in the bios or jumper (Abit BH6).

For today I see no need to OC seriously anymore, machines are often fast enough and the only motivation I'd have to even consider OC'ing is if I'm interested in being the next benchmark king. All the extra mods I'd implement just to keep things under control could easily be reinvested in simply purchasing a faster machine or key system component (graphics card for gaming, CPU, RAM, etc.)

No real surprises here.

Interesting results, but the methodology could be improved.