Slashdot Mirror
ASUS Secretly Overclocking Motherboards?
Hubert writes "It seems that the motherboard manufacturing industry is getting a little bit too competitive now that ASUS and many other manufacturers are secretly tweaking and overclocking the motherboard in default BIOS settings." A front side bus that's a mere 2 MHz faster may not seem like much of a tweak, but it's just enough to gain an edge over the competition.
Saves me the trouble
Pair this story with the little quote of the day "It's better to burn out than to fade away." Coincidence?
My humor is probably your flamebait
On the surface, it seems "cool" or a "so what" situation. However, when you're relying on your PC to do real work rather than just trying to eke out a few more fps in a game, random crashes matter. And that's what these kinds of tweaks will cause. And it will be particularly annoying for people who don't know about the "secret tweaks" since they'll immediately suspect things like the memory or the processor before thinking that the motherboard settings are being quietly manipulated without their knowledge. So while this might be neat for my game box, I'd want to know about this "feature" so as not to include such a board on a production workstation or server. Cheers,
Is it really overclocking if the manufacture does it? Isn't it just deciding the default settings? Components aren't made with a built in correct speed - there is a certain speed that going above means you've overclocked it. They decide the level of stability they want and set the components accordingly. All this means is that they've decided that stability is slightly less important in comparision to speed than they had decided previously. It's not overclocking.
Cool, now the warranty's been voided out of the box. :)
Sent from my computer.
Now GET OFF MY LAWN!
"A front side bus that's a mere 2Mhz faster may not seem like much of a tweak, but it's just enough to gain an edge over the competition."
People need to learn to read graphs. "Best" is too often judged on speed, to the exclusion of other important factors. And too often, performance graphs in magazines and articles are drawn to exaggerate the differences between the worst-performer and the best performer, when the actual performance difference may be 1% or 2%. In terms of PC performance, neglibible.
But a 2% performance improvement may make the difference between a component or system being labelled as "disappointing" and "out in front" by a lot of dumbed-down magazines and online articles.
If only people were better able to keep a sense or proportion, and view performance tests with a little more intelligence, manufacturers wouldn't be so tempted to pull silly stunts like this one.
Er, no. It scales. It's still only 1% of the reference clock speed,assuming we have a 3GHz or above CPU, and any CPU manufacturer that tried to release CPUs that were exactly marginal on stability at the designated clock frequency would soon be out of business.
My own usual experience, back in hardware days, was that a lot of old boards were badly designed and had out of spec built in delays, but that the tolerances built in to the main components allowed them to keep going regardless. This was as true in the days when EPROM had a claimed access time of 450nS but the board only gave it 400 from address and chip select going stable, to this case where the deviation is quite small.
To be really tedious, I'm going to point out that the defined frequencies are not what really matters. What matters is the access time, the time between the input parameters going stable (i.e. address, chip selects etc. staying below the zero threshold or above the 1 threshold) and the actual point at which data is either read from or latched into a register. This is governed by four main factors - chip to chip variation, clock frequency, supply voltage at the chip, and die temperature, and that is as true for latches and registers as well as for memory and processors.
Therefore, if manufacturer A is confident that all the system delays on his motherboard are consistently within the maximum safe values by a determined amount, he may perfectly well be able to drive the clock speed a little higher than manufacturer B, whose process variations are greater or who has a less well designed board. The actual time available to the bought in components to write or read data may be greater than on manufacturer B's board, despite the higher clock speed.
Personally I do not go in for overclocking- I work for a company that now standardises on AMD64 boxes and, for our work, performance is no longer a real issue - but there is nothing in principle wrong with it. It's just like auto making, where some manufacturers release models using the same engine but slightly different torque curves and outputs, for whatever reason. They don't change the water pump and the gas pump just because one model is rated at 98BHP in one market and, because perhaps of slight variations in fuel quality, 100BHP in another market.
Panurge has posted for the last time. Thanks for the positive moderations.
Some points critiquing this review:
* He's measuring using software. The error margin of software methods to measure this kind of thing dwarf 2% and head into the 6% range; typically more so with voltage measurement, which motherboards tend to measure about 0.1V lower than they really are; but for this kind of thing, we should demand testing with a correctly calibrated and maintained hardware frequency counter. I don't think CPU-Z qualifies to measure a change that small reliably.
* This is normal, and within expected tolerances; it's only running a bit high because the natural wandering of his motherboard's PLLs is a little high, and it's only 33MHz above because it's being rounded up when locked.
* Each individual motherboard will have a slightly different clock. Some vary WAY more than this. And only one motherboard of each model tested. That isn't statistically significant, particularly as this is an issue which will vary from individual motherboard to motherboard, as it relies on the tolerances of the clocks. He needs to take a lot more samples; over 100 really; and graph a bell curve from that.
* Also - only one control?
Furthermore, I think this reviewer simply doesn't understand the default settings of the motherboard. He's letting it select sensible defaults, then complaining they're not as sensible as he'd like. He's complaining that his particular motherboard is a little bit out on manual settings, but really if he's that concerned about such a small change, why isn't he testing using hardware?
I think the memory timings can be put down to ASUS's "AI". This is a motherboard feature... and it can be disabled. ASUS's concept "Normal" or "Slower" is a very small push, but if he wants to run truly at stock like a paranoid, use "Disabled", Manual timings on memory, and lock the PCI speed to 33MHz. That goes particularly for the PEG Link Mode. This is normal and expected behaviour for an ASUS (and everyone else).
However, the fractional overclock is actually well within what would be considered normal tolerance. 6% at worst, and that's only on the PCI bus if you didn't lock the PCI bus clock (but in fact it _does_ lock the PCI bus clock, he just didn't measure that bus).
If this caused any problems with system components, the components would not be binned at this level, as for example CPUs are required to pass all self tests at over 10% over a given bin speed to actually make the bin (to reduce returns and DOA); less than that and they will go into the lower bin, because there's a question mark about their ability to perform consistently at stock.
So yeah, his motherboard might be (according to software) running a trifle high; but only 1.1%-1.2% high. Woo, his motherboard's within normal tolerances. Whole lot o' nothing, from a guy who just wants blog traffic.
> Is this bad, unethical, or in any way illegal? What's the big deal? Why the slashdot story?
What are you, a CEO of some big company?
Why yes it's bad, unethical and likely illegal.
It is bad for various reasons, one of the biggest being that you have a market leader effectively performing unqualified tweaks on the timing of various system board components. I'm fairly certain that Asus isn't doing any chip qualification tests on the components they are overclocking.
It's unethical because they are doing that to receive an unfair advantage in the highly competitive (and extremely bogus) MB performance rankings. MBs differ in performance by extremely small amounts, so a 2MHz difference is plenty to differentiate one board from another (and again, I'm not saying that this has any noticable impact on the performance of your system, other than a 1% increase in some dumb benchmark).
It's likely illegal because when Asus says it has a 400MHz system bus they are not telling the truth. That would be false advertising (I mean heck, the number is written right on the MB boxes).
But the REAL point here that is MOST disturbing is that the poster doesn't think any of this is even worth posting. THAT'S what I find most appalling. Since when is lying to gain a competitive advantage OK? It is NEVER OK.