Havoline vs Castrol vs M1 vs GC

[glennc]

In response to one of the Havoline/anti-Havoline threads I was asked to post some actual data to back up my claim that Havoline had posted better UOAs here than Castrol. Since then I decided to do similar quickie rundowns for M1 and GC. Here is my original post followed up with the M1 and GC comparison info:

I figured I'd do a little quick experimentation to see if my impressions from browsing UOAs would jive with a slightly more systematic approach. To keep it simple I just looked at the first ten UOAs for each type of dino oil in the UOA section, and looked at miles on the oil, then three wear metals: Al, Fe, Pb. For Havoline I actually looked at any dino Hav/Chev/Delo UOA (no HM, blend, etc). For Castrol I likewise took only regular dino data. I took the first ten "normal" UOAs, meaning I ignored anything where there was a mechanical problem, an AR-X treatment, under 3k- or over 10k-mile OCIs.

Hav/Chev/Delo:

Miles Al Fe Pb
5 2 5 2
4 1 10 1
3 2 4 7
6 2 8 3
4 2 4 1
7 7 4 0
3 6 7 2
4 4 8 4
3 2 5 6
4 2 5 3

Castrol
Miles Al Fe Pb
3 6 7 0
7 3 8 5
4 1 12 1
3 3 13 0
3 0 0 1
8 8 25 1
4 11 12 4
3 5 10 0
3 3 5 0
5 4 8 3

Turns out the total miles are the same in each case, 43k, for an average OCI in each brand of 4300. Here is the average wear metal PER 1000 MILES for each oil type:

Hav/Chev/Delo
Al 0.70
Fe 1.40
Pb 0.67
Total 2.77

Castrol
Al 1.02
Fe 2.33
Pb 0.35
Total 3.7

And now for M1 (again, samples with 3k - 10k miles, no mechanical problems or ARX phases, also no EP formulations included):

Miles Al Fe Pb
5 2 10 1
7 4 9 2
7 2 5 0
6 2 19 0
5 2 15 1
4 2 13 0
6 2 7 0
6 3 14 4
7 2 12 13
8 3 23 0

Averages per 1k:
Al 0.39
Fe 2.08
Pb 0.34
Total 2.82

And GC:

Miles Al Fe Pb
5 2 7 2
7 4 10 21
4 2 7 1
4 1 7 1
9 4 12 0
7 2 7 3
4 2 16 5
6 4 11 3
6 3 8 1
8 4 18 24

Averages per 1k mi:
Al 0.47
Fe 1.72
Pb 1.02
Total 3.20

Again, this is different vehicles owned and driven by different people in different circumstances, but nevertheless, going by UOAs posted here it seems Havoline/Chevron/Delo at least hold thier own against some of the best oils out there. This is the source of the Havoline fanfare on the site. Certainly doesn't seem like empty drumbeating to me!

- Glenn

 

[like a rock]

Havoline is a dino oil with mobil 1 and GC being synthetic. You can not compare these oils in total performance in regards to cold flow starts, high temperature stability, tbn retention and other aspects. While havoline is excellent no doubt it is not the best oil by a long shot. Uoa's are trending data. If you ran havoline 10w-30 against M1 or gc of the same viscosity over 100,000 miles in a Patmans corvette and tore down the engine for inspection what do you think the results would be? Heavy varsinh and deposits with stuck rings and oil galleries would just be a start. Haovoline is good compared to other dino oils, but no fair compairison to synthetics in severe aplications.

 

[mechtech]

glen - Thanks for the homework and useful data.

 

[Al]

Statistically..comparing the 4 oils may not be that accurate bc of the low sampling number. Also adding the numbers of Al, Pb, and Fe isn't valid bc they each have different orders of magnitude (I'm groping for the right way to say it..its like adding apples, oranges and lemons)

BUT

I appreciate the effort you took and it was very interesting. Future food for thought. Thanks for your effort.

Average wear metals for all tests:

Pb= .6
Fe= 1.9
Al= .6

 

[Blue99]

Glenn, thanks for putting the effort into this analysis. It's indeed a different method of examining UOA data.

My intial comment is the same as Al's - the average of the total wear numbers for each oil is a very muddy number.

The synthetics will show higher averages simply because the OCI's are in the higher 5-8K range and wear metal amounts tend to grow with mileage.

And the fact that it is difficult to find 10 Bitog UOA's that are normal(Yes, I know, many have some type of problem).

An example is the GC group:

And GC:

Miles Al Fe Pb
5 2 7 2
7 4 10 21*
4 2 7 1
4 1 7 1
9 4 12 0
7 2 7 3
4 2 16 5
6 4 11 3
6 3 8 1
8 4 18 24*

Offhand, I'd say both these UOA's have a high lead issue not connected to the performance of the oil, and should not be included.

Also, copper is an important wear metal and probably should be included. It's always an issue with the GM 5.3L's and I'm happy to score in the low 20's with copper.

Again, thanks for posting this data and taking a quantitative approach to the debate.

 

[Triple_Se7en]

Where are these anti-Havoline threads? The only discussions where Havoline dino's quality gets questioned is when someone compares it against some other quality synthetics.... much like this thread.

 

[merlin64]

"Again, this is different vehicles owned and driven by different people in different circumstances"

Because of this we really can't use this data to draw any conclusions. We would need to use the same engine type (preferably, the same vehicle) driven in a consistent manner and with variences in climate held to a minimum. Then, we would need use the same lab and audit their methods to ensure they are consistent. We also need to control for total milage on the vehicle and variences in OCI.

I appreciate your work but regretably without a more scientific approach all of this is mere speculation, and thinly-based at that.

 

[Alan]

like a rock,

I don't think a dino like Hav 10W30 would sludge up in a corvette over 100K necessarily when used
sensibly.Their was some talk here a while ago that a LS1 Corvette was very clean just running Wolf's Head 5W30 of all oils.

I also remember the GM Engineer posted here once that the only reason why Chevy uses M1 in the Vette is because of the chance that some may track the car making the oil temps exceed 305f.

 

[glennc]

Thanks for the feedback everybody.

I chose Blue99's message to respond to since it included most of the objections raised by others.

Blue99: "My intial comment is the same as Al's - the average of the total wear numbers for each oil is a very muddy number."

Yes, I agree, and you put it well. It is a "muddy" number. As another poster stated it contains components that are different in magnitude, and simply adding the numbers effectively weights Fe higher than Al and Pb, and that gives an advantage to the oil that has the best Fe numbers - Havoline in this case. A better approach would be to weight the numbers according to their magnitude, or to look at each wear metal separately. You can do that yourself, and I may take another crack at this in the future. Even so, the number does have at least some rough-hewn usefulness in my opinion.

Blue99: "The synthetics will show higher averages simply because the OCI's are in the higher 5-8K range and wear metal amounts tend to grow with mileage."

I don't think this is the case unless the oils are used beyond their useful lifespan. There are lots of threads here speculating on what the "wear curve" is for different oils in different engines, and that is the reason I cut off samples of greater than 10k duration. I accept that synthetics last longer than dinos, and maybe it would have been more representative to cut off dinos at 10k and synthetics at 15k for example, but I wanted a direct apples-to-apples comparison without correcting for an already-assumed superiority of synths. It could be argued that this actually gives synthetics an advantage in this type of comparison.

Other data points to support this: There is no evidence among the samples that those with fewer miles have the best wear numbers on a per-mile basis; in fact the opposite may well be true, in both the dino and synth groups. Also it has been speculated here that wear might actually spike with a new oil fill. That's kind of tough to accept, but at the least it hints that the reverse isn't often seen.

Blue99: "And the fact that it is difficult to find 10 Bitog UOA's that are normal(Yes, I know, many have some type of problem)."

Well, yes, and that is where averaging many samples is important. I'll definitely agree that ten is not a large number, the question is whether it is large enough to begin to draw useful conclusions. Other people have pointed out, as I did, that this is averaged across various labs, drivers, cars and so on.

The truth is none of these will matter if the sample size is large enough. Drivers, cars, labs will eventually average out, with the exception of "selection bias." That is, those who drive American cars may tend to drive more conservatively, and may tend towards one type of oil - or, those who maintain their cars best might use synthetics - and so on.

I respectfully do not believe any of that invalidates the results, especially as the sample size goes up. I do agree that with a sample size of ten, conclusions are somewhat tenuous, and the level of certainy is more like "this shows there's no obvious advantage to oil yyy", rather than, "oil yyy clearly performs x better than oil zzz." In other words the data, I do believe, is useful in this form to show broad trends with a reasonable level of confidence.

Blue99: "An example is the GC group:

And GC:

Miles Al Fe Pb
5 2 7 2
7 4 10 21*
4 2 7 1
4 1 7 1
9 4 12 0
7 2 7 3
4 2 16 5
6 4 11 3
6 3 8 1
8 4 18 24*

Offhand, I'd say both these UOA's have a high lead issue not connected to the performance of the oil, and should not be included."

I agree, with reservations. I thought the same thing and considered eliminating those results. If I did, though, I would have to apply some method to eliminate results objectively, or else I would just be cherry-picking the data. I just didn't get that sophisticated, but if you wanted to cut those numbers out, it would probably be a more representative result, in this particular case.

Blue99 "Also, copper is an important wear metal and probably should be included. It's always an issue with the GM 5.3L's and I'm happy to score in the low 20's with copper."

Yes. Many other things I could measure. Again, I wanted to put maybe an hour into this. I agree.

Blue99: "Again, thanks for posting this data and taking a quantitative approach to the debate."

You're welcome, and the same to others who expressed similar comments. I do think this is useful and I've thought about trying to do it on a bigger scale. More oils, more samples, and incorporating new UOAs as they come in. I'll give it some thought. If we could build up a database of hundreds of UOAs on a dozen or more oils, it would undoubtedly have some validity and usefulness.

There are some other comments I'll answer:

To the statement by a couple that they didn't notice any Havoline vs anti-Havoline threads: That wording was an exaggeration in the interest of brevity. There have been three or four threads where people have questioned the attention given the Havoline family, and stated or hinted that the oil wasn't as good as its reputation here. Since I came along fairly recently, and came to the same conclusion as the rest of the Havoline advocates - that it was posting outstanding results comparable to even very good synthetics - I wanted to do a little bit of quantitative experimentation to see if my own conclusion was merited, or whether I had reached it erroneously, through the power of suggestion or some other bit of unintended self-trickery.

To those who say you can't compare dino with synth: Of course you can, and that is exactly what I'm trying to do. That is one of the main reasons BITOG is so great. We try to answer questions like "is synth better," "are boutique oils worth the money," etc. We can all rant all we'd like to about what we THINK, but if we actually want to find out what is TRUE, then we have to try to do it in some objective way. This is an attempt at exactly that.

Of course, there are qualifications. Studies of analysis results can only show what analysis shows to begin with, namely wear metals and some types of oil degradation. It doesn't show engine cleanliness, or protection in extreme conditions (unless extreme conditions were encountered, which would be impossible to track accurately). My argument is that "the proof is in the pudding." If the wear numbers are good (and this includes many high mileage engines run all their lives on a given oil), then those factors, whatever their importance, are not affecting the bottom line: how well the oil is protecting the engine.

I think I'll leave it at that, for now. It's a nice day here in Colorado and I'm going to go use up some of the 15w40 dino oil currently running the first rinse phase in my old 3-series. However, I'm finding this interesting enough that I'm considering putting together a much broader tabulation of results. I suspect if I can get two or three dozen samples for a given oil, most of the statistical variations will be sufficiently averaged out to reach some really useful conclusions. For my own part, I am curious whether Havoline's higher wear rate for Pb remains consistently so, or was a statistical aberration in these particular samples. So, those interested, stand by.

And to everybody, fire away!

- Glenn

 

[merlin64]

"This approach produced an anecdotal evidence.
Thanks for posting though."

Not really, anecdotal evidence comes from personal experience. I don't think these UOA's came from the posters car, therefore the evidence cannot be considered anecdotal.

It's really more along the lines of empirical evidence, or more correctly, the empirical method--the collection of a large amount of data on which to base a theory or conclusion. The problem here is we don't have a large amount of data, nor was the "experiment" controlled in any way.

Glen, why not use your "old 3-series" to prove your point? Drive in a consistent manner at all times, then run each oil in your own car for a set time using the same lab and sampling method each time. Ideally, make sure climate changes are minimal (hard to do in CO). That way we at least have some consistencies in the data (same engine, same driving style, same sampling stlye, etc.) The evidence will of course be anecdotal, but actually a heck of a lot better evidence than simply slapping some UOA results up when the engine types are varied and myraid other factors are unknown.

Use those nice days in Colorado and put on about 50k miles in the name of science!

 

[vad]

Well, sorry for using the wrong term.
English is my second language.

 

[MarkC]

"Heavy varnish and deposits with stuck rings and oil galleries would just be a start."
I don't agree. At leat with my Jeep.Ran the first 80K with Castrol GTX 10w30(and Fram filters, too), then the past 30K with Chevron(and Napa or Purolator filters)and I have no varnish anywhere. Well, I mean my engine has none.

 

[SSDude]

This really doesn't prove anything. A true comparison would be to compare similar engine driven in a similar manner under similar weather conditions. That might prove something.

 

[like a rock]

Well MarkC I was refering to using dino oil in a high performance role not in a every day ride like yours. Glad your engine is so spotless.

 

[vad]

This approach produced an anecdotal evidence.
Thanks for posting though.

 

[vad]

quote:

---

Where are these anti-Havoline threads? The only discussions where Havoline dino's quality gets questioned is when someone compares it against some other quality synthetics.... much like this thread.

---

Exactly my thoughts!

 

[miked]

quote:

---

Well MarkC I was refering to using dino oil in a high performance role not in a every day ride like yours.

---

My friend runs nothing but 10w 30 Castrol GTX in his 9 second car. He tears the engine down every week and not a scratch on it. I use too question dino's also until he showed me his.

 

[glennc]

Here's a really interesting one. New engine, and you see some of the wear numbers trending down as the engine fully breaks in. But look at the amazing difference between the synth runs and the dino runs.

Maybe this has something to do with break-in?

Anyway, it knocked my socks off when I saw it.

Look at Fe, Al, Tin, Lead:

http://theoildrop.server101.com/cgi/ultimatebb.cgi?ubb=get_topic;f=3;t=002856

- Glenn