No test is perfect. ...
I do not have an opinion on the GM studies accuracy. I think of it as just another indicator that better filtration is a good idea.
I would agree that no test is perfect. But I would counter with the fact that there can be much better studies done when real life is taken into account. It's just that most folks who do have the interest in doing so, don't have the time/money because the burden of proof comes from large quantities of data. Companies, OTOH, whom do have much deeper pockets, don't have the inclination to be patient; they want a fast return on the investment and not wait for long periods of time.
When a ALT/HALT test process is used, it can be a good indication of how things perform, one relative to another. But that relativity stops when you try to quantify it to "normal" daily use of something.
Example:
We have salt-spray chambers at work. They are used to induce corrosion effects when we test different materials which would be in our products in coastal conditions (near the oceans and gulf). What we can do is test product A against product B, and then conclude that A is either better, same, or worse than B. But we cannot say with impunity and assurance that A is going to last "XXXX" hours longer in real life. Product A may have lasted "yyy" hours longer in salt-spray than B did (to a predetermined state of some characteristic), but there's no way for us to say "yyy" hours in the lab equals "xxx" hours in the field, because we've never spent the time/money to find the correlation between those two.
For the GM filter study to be accurate and useful to us common weekend garage-wrencher types, we'd need GM to show accurate correlation that "xxx" hours in their test = "yyyy" miles on my odometer. They'd have to take their data from the lab test, and then do field testing to establish a ratio of lab hours to typical miles driven. They never did that, and in fact they knew darn good and well that the wear rates they induced were so absurdly high that there's no way they'd be able to recreate the same effect in the field. Hence their disclaimer statement.
To be a well conducted test, you want to remove or control as many variables as possible, so that you can manipulate the desired input in an "on/off" method, such that results can be attributed to the controlled input and not other uncontrolled things. By choosing to never change oil during the filter study, GM eliminated OCIs as a variable input. They held the OCI constant (never changed oil), and the controlled variables were filter pore size, filter pressure differential to indicate change needed, and particulate loading rate.
In fact, the particulate loading rate is even absurd. Not just in it's quantity, but in it's introduction. They didn't meter the dust in, they just dumped it in once an hour. That would be akin to you and I deciding to remove our air filter, then go do a bunch of wheel-ripping donuts in a dusty empty bean field, right after we change oil. They rate of particulate induction wasn't a small constant stream; it was a massive overloading at the top of each hour!
Conversely, there are "tests" which do exist which give an excellent indication of how well or poorly products last in our vehicles. We have anecdotal evidence (reasonably documented stories which establish results, but don't necessarily have a lot of data), and we have data from actual use (UOAs, PCs, etc).
UOAs are NOT perfect, either. But they are a representative look at the way wear presents itself over time. UOAs most certainly do have limitations, but just like any other tool, if you understand those limitations, you have a good chance of knowing what the data can and cannot tell you.
Want to enjoy a little sliver of irony? The GM filter study used spectral analysis to determine relative rates of wear, along with other methods, in their HALT results. But then they didn't want to invest the time/money of using UOAs to determine how those lab tests correlated to the real world! Kind of ironic, is it not? They can accurately claim that the 7um filter extends a project lifecycle 7x greater than the 40um filter, but ONLY under the conditions of the test. Once those conditions are no longer true, the test results are no longer true!!!
If GM had changed the oil during their tests every 4.2 minutes, AND introduced the particulate loading in a metered manner such that the loading effects were managed by the simulated mileage rates, THEN AND ONLY THEN would the GM filter study have merit in our garages. But ... GM knew that such efforts would bring forth such incredibly tiny, fractional differences that they would have no ability whatsoever to discern filtration effects. Why? Because the oil filter is not the only entity controlling wear during normal maintenance of an engine. OCIs and TCBs are also major contributors to the wear rate control.
Most simply put, the GM filter study is completely and utterly useless to establish any kind of meaningful representation of what happens in our engines in the daily lives of any typical Joe Average vehicle owner.
- What we know anecdotally is that there are LOTS of vehicles which run 200k-300k with just "normal" oils and filters.
- We also know that UOAs can show us that wear rates due to full-flow filter selection are moot; the amount normal wear variation trumps the "noise" of filter selection in statistical analysis.
YES, ABSOLUTELY, filtration is important. It's VERY important. But only to a point. After that point, the effect of filtration is overtaken by the OCI duration and the TCB. In fact, those three things act in concert together (filter, OCI, TCB) to reduce wear, and it's very difficult to test one without giving credit where due to the others.
The longer an OCI is run, the greater the need for better filters. The shorter the OCI is run, the less important filtration is. That, in a nutshell, is the (unmentioned) lesson to glean from the GM filter study. Anyone who practices "normal" oil changes, is NEVER, EVER going to see a tangible and meaningful shift in their wear rates simply because they chose one filter over another, presuming the filters they pick from are approved for the application. The effect of filter selection is so incredibly small in a normal OCI that there is no way at all it can be discerned in real life. And GM even admitted that fact in their conclusion statements; they truthfully admitted that the results from their lab study will never be seen in real life because what they did is not practiced in your garage.
HALTs are meaningful for relative performance delineation only. The results are NOT transferable to real life,
unless there is a subsequent correlation study done to show xxx lab hours = yyy life hours (miles, or whatever UoM is chosen). In this particular case, GM chose not to do that correlation study, because they knew they'd never be able to generate enough wear to show the wear rates were statistically significantly different in normal operation above the noise of normal wear patterns. Why? Because oil changes flush out the particulate loading soon enough that filter choices don't matter. It's incredibly important to have a filter in place. It's not nearly as important to worry about filter A having 80% efficiency at Xum over filter B having 95% efficiency at that same size. Or 95% versus 99%; even less so.
Once a filter's efficiency is "good enough", making it better has no real effect. I cannot tell you what "good enough" is; there's no study done that I know of that credibly establishes that level. But, there are tens of millions of cars driving around every day that anecdotally tell us that "good enough" already exists on the shelf of every parts store in North America.
My long winded point to all this? Ignore the GM filter study, unless you never, ever change oil at all and drive around in a dusty bean field all day, every day, 365 days a year, with no air filter in place.