Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Samsung Again Caught Inflating Benchmarking Scores, Phil Schiller Calls 'Shenanigans'
- Thread starter MacRumors
- Start date
- Sort by reaction score
I'm not surprised, I've run benchmarking tools before... Driver profiling is basically standard for the 'big name' tools for example.
We've all seen vendors stick graphs in their marketing blurbs, take them with a pinch of salt - They're there to sell. Real world comparisons are very hard to achieve when benchmarking a system.
Either way, this report and these tweaked numbers has made me want to is to see the numbers against competition like apple. Seeing a lineup of different devices from a single vendor isn't very interesting.
my, what fine mental gymnastics! amazing ability.
it's simple: because theyre hardcoded for this computing task based on *app name*, not computing power required. were it a true "turbo" feature that kicked in for intensive tasks, it would kick in for the benchmarks automatically...thus no need to hardcode it by name.
it's a hack.
So if they developed code to run this around other apps (like the article vaguely mentions it does) it would be OK?
You're trying to tell me someone on XDA isn't going to be able to write up some code to inject into the system that would give the app they develop a boost if they wanted?
CPU governors have been available on Android since pretty much Day 1. I have been able to *artificially* boost the CPU power for a task based per app or overall. Yes, battery life suffered. Yes, performance increased. But at its core, which is the entire point here, the CPU was capable of doing it - just like this benchmark proved. The CPU did the work it was asked to do. The app was NOT compromised to get around tests, the CPU did the work.
This thread -> LOL.
I think someone at Samsung snuck a peek those 64bit (dayum!) scores and felt obligated to do something drastic.
I think someone at Samsung snuck a peek those 64bit (dayum!) scores and felt obligated to do something drastic.
This is pretty hilarious actually. Why not just always run the CPU this way
Battery life... It's the same concept as Intel's SpeedStep. Pretty sure when benchmarks are running, SpeedStep is pretty much disabled.
However, on Android, one can change the speed of the CPU on a rooted phone; so while stock phones may not be able to achieve these speeds, a rooted phone should be able to with some additional software.
links, or it doesnt exist.
either you have data for this, or youre just making things up to smokescreen samsung.
This is nothing different than any other company, including Apple. Of course Samsung it trying to show the best case scenario, OF COURSE. APPLE DOES THE SAME DANG THING!
Someone please show me just ONE macbook pro that gets 5 hour battery life under normal use, let alone the "Up to 7 Hours" claimed. I have worked on 1000+ Macbooks, and they DO NOT EVER GET THAT BATTERY LIFE. PERIOD!! So Apple apparently tweaked whatever they could to come up with the 7 hour number. Clearly not realistic.
Take off your fruit shades. Let Apple win with better products, rather than with their executive donkeys.
Someone please show me just ONE macbook pro that gets 5 hour battery life under normal use, let alone the "Up to 7 Hours" claimed. I have worked on 1000+ Macbooks, and they DO NOT EVER GET THAT BATTERY LIFE. PERIOD!! So Apple apparently tweaked whatever they could to come up with the 7 hour number. Clearly not realistic.
Take off your fruit shades. Let Apple win with better products, rather than with their executive donkeys.
That's because its not using the booster mode. Which was the point of the entire shenanigans. And my point. Your daily usage and how well the device runs games, etc. THAT is what users want benchmarked. The booster mode is unsustainable and only there to spike the score.
Geekbench has lost all credibility. The new version 3.0 scores are different than the version 2 scores. A Galaxy s4 scored a 3200 on multicore on v2 and now a 1200 on v3. Pft. Seems written for apple now. Yes, V2 didn't have a single and multicore reading, it was just multicore only.
what on earth are you talking about?
cheating is cheating. only cheaters think cheating is ok.
This is EXACTLY like taking a banned substance in a sport. On one hand, you have the Galaxy Note 3's real performance....the way the processor runs on ALL apps. EVERY SINGLE ONE...except for benchmark apps, where it then overclocks the processor (and, based on some of the code, may even alter internal framerate rendering, which could cause it to skip frames to render) in order to get an artifically higher score. It's exactly the same as blood doping: it's still the cyclist's body driving the bike...they just supercharged it for the race...what's wrong with that, right?
If the Note 3 had a turbo mode for demanding tasks, and this used that...there would be no outcry...but this is overclocking and modifying performance code for the SOLE PURPOSE of inflating their benchmark numbers. In real world use on ANY app, this power is not available. Do you not see this?
I'm sure the A7 could run at 1.5-1.6GHz without a lot of trouble for short periods of time too, but the heat/battery loss/potential stability make 1.3 the chosen speed....it would be just as wrong for the 5S to overclock itself only for benchmark apps too.
say wha? 64-bit mobile computing, 100% increase to CPU and GPU, kickass fingerprint scanner...desperate acts of fear? i dont think so.
I don't know which one provides a more dismal view of humanity... the unbelievable content of your post, or the fact that at least 2 people have up-voted it.
You're kidding, right? Of course you can't compare scores between benchmark versions - benchmarkers are constantly updating their programs to better test the system. Did Geekbench promise cross-version score equivalence? I don't think so. These are all Geekbench 3 vs Geekbench 3.
My computer scores in the 50,000 range on 3DMark 05...amazingly it is far lower on 3DMark 11...gee, maybe it's because they update it to be more intense for the more powerful machines.
Your view is plain incorrect.
The reason to run benchmarks is to measure what performance the device will produce in normal use. Samsung took steps to show 20% higher values without telling anyone. That's plain old cheating.
If Samsung had gone to Ars Technica and told them "please test how fast this phone can run in a special benchmarking mode", that would have been absolutely fine.
I happen to drive a car with a Diesel engine. The engine is quite deliberately limited. Over 2500 rpm, torque and horse power both drop. The engine itself could deliver the same torque at higher rpm, but that would very quickly damage the engine. The same car could run significantly faster if you were willing to accept that the engine won't last 200,000 miles but only 20,000 or only 200. If the manufacturer wrote "we removed all limits from the engine and achieved 150mph and managed to stop before the engine exploded" that's fine. If they wrote "the top speed is 150 mph" without mentioning engine damage, making their customers believe that they can also achieve that speed, that would be cheating.
Because it's a benchmark app where it is a well known fact that product reviewers will use it to compare to other devices. You're really distorting the purpose of what benchmarks are for if this is how you feel about it. This isn't the worst thing Samsung has ever done, it isn't murder, etc. But it is deliberately misleading.
It better have a higher overall score than an iPhone 5S...we're talking about a phablet here.
Think they were referring to market share, not the ability to sell 5-10% more I-devices year over year.
As a normal customer I don't have enough resource to constantly audit the lies like this one revealed by Ars, so the only safe way for us normal customers is to totally avoid big liars like Samsung.
As to the components inside the product, I have to assume that the component buyers like Apple (as big as the Samsung liars, but yet still much less corrupted) to seriously audit all important aspects of those components, for their own fame.
It doesn't have a higher overall score than the 5S...unless they resort to cheating. The stock Note 3 is ever so slightly slower than the 5S in Geekbench.