Reviewing a proudct at stock is - obviously fine, great even. What im calling flawed is reaching any conclusions about efficiency. I mean I brought up the fans and coolers example repeatedly. Who cares how a cooler or a fan perform at the stock fan curve or whatever? What matters is performance at normalized noise levels. Imagine if cooler reviews were only done with stock fan curves. What would be the usefulness of that for people that care about efficiency? None whatsoever. The coolers with the worst slowest fans would be topping all the efficiency charts, even though they in fact are terrible.
But ...
fans don't have stock fan curves.
Motherboards have stock fan curves. The fan has a range of speeds which is an inherent trait of the fan, but it runs at whatever speed it is told to run by the motherboard, and if told nothing, it runs at its max speed. CPUs
have built-in performance characteristics, even if these work in conjunction with the motherboard (the increased complexity when compared to a fan makes that an obvious necessity). Your comparison here is fundamentally invalid, as the characteristic you're discussing literally doesn't exist in one of the examples.
As for actual stock fan curves, do you have any idea how many people run them? Essentially everybody. True, a lot of people install the motherboard's software and maybe pick a fan profile from the few presets that comes with. But actually tuning a custom fan curve? Again: that's a niche, enthusiast activity, and comparable in its rarity to overclocking or undervolting (likely more common, but still comparable)
Exactly. Your position when taken to the extreme means that the card in my example would be entirely inefficient and you wouldn't buy it. That's what my argument showed
But ... your reductio ad absurdum thing didn't resemble my arguments whatsoever. Not even remotely close. It looked more like an extreme version of
your argument? I mean, if you think that looks like my arguments, please, go back and re-read my posts, as that makes it plenty clear that you're just not picking up what I'm saying at all.
It's not an oversimplification, you are just making a bigger deal out of it than it is. Any card is going to be less efficient at higher wattages than at lower wattages.
To some degree, yes, but that isn't a contradiction of what I'm saying - I'm saying that this (roughly) applies to any single card, but that doesn't allow for like-for-like comparisons
across cards at the same power levels, as they will inherently be at different points in their V/F curves. And, again, the thing I'm "making a big deal out of" here is the fact that you're making bombastic, all-encompassing pseudo-factual statements that
simply aren't true - unless you add
major caveats to them.
Remember, you didn't say "Any card is going to be less efficient at higher wattages than at lower wattages." You were talking about normalization of testing at either a given level of performance or power. Which means you were talking about comparisons
between different cards, not charting the efficiency curve of a single card. You see the difference there, right? And how, in order to normalize for something, there must be a comparison involved?
The thing you seem to be missing here: what you seem to want to be asking for/advocating for, but what you are consistently failing to do so through making bombastic, simplified-to-the-point-of-being-wrong claims is testing that demonstrates the efficiency curve of a GPU, which in turn would allow for comparing this to the efficiency curve of another GPU. And, as I've said above, if this existed, it would be fantastic! But we won't get any closer to this by, say, refusing to measure the efficiency of a 4090 at 800W (if that is its stock power limit) and instead limiting it to 400W. That gives us a single, interesting but unrepresentative snapshot that ... well, it says
something, but certainly not much.
Oh come on, again you are portraying a person that doesn't exist. There is a person out there who cares about efficiency, reads reviews about GPUs on tech forums, but doesn't know what MSI afterburner is or where to find it? I mean come on..
Really? I don't know, it seems like your experience of PC users is
purely based on forum users and enthusiasts, which is thankfully not the sole basis of my experience. Having worked in computer retail for quite a few years, I'm well versed in just how common these types of people are. Remember: whether or not someone "cares about efficiency" is rather random, and might be down to something as simple as this scenario:
someone who knows next to nothing about hardware but wants to upgrade or buy a new PC googles "best gaming GPU 2022". This leads them to one of the dozens of review roundup articles/videos on the web (articles/videos that are made frequently because they get an order of magnitude more eyeballs on them than individual reviews, which speaks to just how common this scenario is - these types of roundup/overview pieces are
incredibly popular, as they are informative for those not very into the thing being discussed). Now, in this article or video, at some point efficiency is mentioned. At this point, said person either thinks something along the lines of "hm, it's more efficient, that sounds good" - or they don't. If they do, then they belong in the group I sketched out above. Does that sound like "a person that doesn't exist" to you? If so: go ask random people on the street whether they care how efficient their car is. Or their TV, or whatever. Regardless of the group, you will get a significant number of people saying yes, despite this not being something important to them or that they know much about, but rather because they like the idea of things working efficiently and not being wasteful (a value that is deeply ingrained in many, many cultures in different ways), and by simply bringing up the subject, you have prompted a "yes" even though they might never have thought of this before at all.
Now, how many of the people randomly googling "best gaming GPU 2022" do you think know what MSI Afterburner is, where to find it, or how to use it? How many of them do you think even understands that a GPU can be tuned in any way? They absolutely have the potential to learn and understand these things, but these knowledges are not inherent to wanting to play PC games. Most PC gamers buy laptops or prebuilts; most PC gamers know nothing or next to nothing about configuring their hardware; heck, most PC gamers
barely know how to update their GPU drivers. That's just reality.
Setting a power limit doesn't lead to instability. It's literally a single button press on MSI AB, or whatever software of your choosing.
It can absolutely do so - it normally doesn't but that is entirely down to the specifics of your build and its configuration. Now, I haven't used an Nvidia GPU to any real extent in years and years, but does AB allow for infinite power limit adjustments? At least for Radeons, there are strict limits to how much this can be changed. My 6900 XT doesn't allow more than a 10% reduction of the power limit - though it can easily be brought lower through downclocking and/or undervolting. Either way: this is not trivial, it is not a one-button process, and is not something that doesn't require knowledge.