手机游戏性能大横评：厂商作弊太疯狂！

Alright, it's time for our annual smartphone comparison review!

Have you been waiting for this?

Many of you probably remember our big review from early last year, right?

We thoroughly dissected every flagship phone, from the standard to the ultra-luxury models.

This year, we've decided to go all out including not only flagship phones but also mid-range and near-flagship models.

The entire review will be divided into four parts: gaming performance, battery life, imaging, and system smoothness.

Our standards will be even stricter than last year , not only introducing more methods to uncover the various tricks used by phone manufacturers but also incorporating more dimensions for a more comprehensive evaluation of this generation of phones . To be honest... This workload is incredibly heavy,

. To be honest... This workload is incredibly heavy, so we hope everyone can support us by liking, commenting, and sharing. Supporting us isn't too much to ask, right?

So today , let's talk about a familiar topic, but one of the biggest areas where phone manufacturers cheat : gaming performance.

As we all know , phone manufacturers send us media units for review before launching a new phone , many of which are hands-on experience videos—essentially paid content.

These media units are quite subtle; many manufacturers like to manipulate them.

First, they meticulously select high-quality, low-power chips.

Then, they use targeted scheduling strategies based on the media testing process, specifically choosing… Since we can't buy retail units to verify the custom- tuned devices before they go on sale, anyone wanting to capitalize on the initial sales will have to swallow their pride.

Actually, we brought this up in last year's comprehensive review, criticizing various manufacturers.

Back then, to avoid suspicion, we only referred to them as Model A and Model B.

I thought everyone would be more restrained , but unexpectedly, last year's new devices not only didn't restrain themselves but went further down the path of cheating, becoming even more outrageous.

A major reason is that this generation of 8E5 and 9500 chips are incremental upgrades.

If they don't tamper with the media units, it's really hard to differentiate them from the previous generation 8E and 9400. This

year, consumers have completely stopped trusting the data from initial media units.

Some data is obviously fake, and everyone assumes it's all false.

I think if we don't completely expose the issue of specially tuned media units, these manufacturers will only become more unscrupulous this year , which will only affect our own credibility.

So, in this year's comprehensive review, we're not even trying to avoid any appearance of favoritism; we'll directly list which manufacturers and which models are being compared to see exactly how much the media units sent for testing differ from the retail units sold to consumers.

Let's reveal the answer.

First, there are a very small number of manufacturers that didn't use any special tuning at all, such as Apple iPhone , and even those that use iOS... The 26.1 and 26.2 systems are constantly being optimized.

In fact, the newly tested retail units generally perform better than the media units that were initially tested.

Of course, you could say that the 26.0 system was too early to be tested , but Apple is indeed relatively honest.

Almost all the major Android manufacturers that participated in the testing failed.

Since everyone's initial test resolutions were different , we won't make direct comparisons between different manufacturers here; we'll make a unified comparison later.

But what I want to say is that without exception, everyone was cheating with custom tuning; no one is innocent . For example, the Find X9 series

. For example, the Find X9 series media units equipped with the Dimensity 9500 could run "Clash Bottom" at 900+ P resolution with a full frame rate, while the retail units couldn't even reach that, dropping to just over 50 FPS.

Vivo wasn't much better; although the retail units could run close to full frame rate at 700+ P resolution , their power consumption was 1.3W higher than the media units.

The fact that a device with over 5W power consumption has a 1.3W higher power consumption than the media unit shows how outrageous the special tuning is.

For Qualcomm models, like the OnePlus 15 with its 8E5 processor, the media unit maintained a full frame rate for 30 minutes , while the retail unit only lasted 15 minutes.

This is considered a relatively small tuning, as it was the first model and lacked a reference point.

Later flagship models like the Ace6 and Ace6T also showed obvious media unit tuning.

In fact, we only did our review of the Ace6T after its release, using retail units for retesting . Comparing the power consumption shows how drastic the tuning was.

. Comparing the power consumption shows how drastic the tuning was.

(iQOO... ) The custom tuning of the OnePlus 15 is even worse than OnePlus. Not only

OnePlus. Not only does it only maintain a full frame rate for 15 minutes, but it also starts to throttle, and its power consumption is much higher than the media units.

Xiaomi's situation is more complicated.

The three phones in the Xiaomi 17 series had VRS (Variable Shading Rate) forcibly enabled on the initial media units and could not be turned off.

This sacrificed image quality to save performance, which we already reminded everyone in the Xiaomi 17 series launch video.

After the release, VRS could finally be turned off , but these retail units were significantly worse than the media units at the time, with much worse performance scheduling and higher power consumption. The

Redmi K90 Pro Max was one of the few phones with relatively minor cheating , but the standard K90 and Turbo 5 Max had very aggressive custom tuning, resulting in significantly lower performance for retail units.

Honor Magic 8... The retail version of the Pro not only has a lower frame rate

Magic 8... The retail version of the Pro not only has a lower frame rate but also higher power consumption (1W) and significantly more heat.

The much-anticipated Honor WIN isn't clean either.

Although the retail version still has strong frame rate stability , its power consumption jumps to over 8W, a staggering 1.7W.

This heat means the frame rate can't be maintained at its maximum throughout like the media version.

In the latter half, due to heat , the frame rate stabilizes and stops.

So you see, its 1% low frame rate is far inferior to the media version.

As for Nubia and Red Magic , they're repeating the same mistake.

While the retail version's performance is powerful , its power consumption is inevitably much higher.

The media version, with similar frame rates but significantly lower power consumption, is quite common. Ultimately

, when it comes to the device's temperature, you see a huge difference in the measured results.

Viewers see a first-release review and think , "Wow, the temperature is acceptable!"

But after buying it and playing with it , they find it overheats and explodes, completely different from the review. This kind of thing is all too common.

Besides, some people have also noticed that many later-released devices have more severe custom tuning than earlier-released ones.

This brings us to why manufacturers choose to custom-tune their media devices.

For example, if manufacturer A releases its flagship platform , and we test A's results, then company B's leaders, seeing our test results for A, will make demands on their engineers like this: "The performance of your tuned device must be better than A's," "We will not accept results worse than A's."

The leaders themselves may not intend to cheat; they just have too much confidence in their team and feel that their own capabilities shouldn't be inferior to others, so they proceed with this arrangement.

But even the strongest engineering team must follow the laws of physics.

Using the same chip, unless A's tuning is extremely poor , how can B possibly outperform A?

Ultimately, the engineers couldn't meet their boss's KPIs without cheating, so they had to resort to underhanded methods.

The boss of company C, seeing this, exclaimed , "Wow, how come company B is so strong?"

"If they can do it, why can't we?"

"We'll definitely beat B when we do it!"

Company C's engineers were helpless.

"Since they cheated, " they thought, "I have no choice but to cheat to catch up."

So, initially it might have just been a minor special adjustment.

However, latecomers are never willing to fall behind early adopters, so this cheating escalates and eventually becomes blatantly obvious.

Therefore I want to take this opportunity to raise awareness about media device cheating and urge all mobile phone manufacturers to focus on proper calibration and avoid wasting energy on these tricks.

It's really unnecessary to lose face if we find out.

But there's good news as well as bad news.

The good news is that this generation of flagship phones has basically eliminated drastic brightness drops in high-pressure games.

Even the latest iPhone didn't significantly reduce brightness in our tests , or at least not to a perceptible degree in our 300-nit testing environment.

It seems this trick is largely obsolete.

So, back to the main topic: which mobile phone has the strongest gaming performance this past year ?

And which has the best energy efficiency and heat management?

Let's start with a newly released game that hasn't yet been added to any of the whitelists : *Arknights: The Last Land*.

We did a special video about *The Last Land* a while ago, and we actually had in-depth discussions with our technical colleagues at Hypergryph, gaining a lot of useful information.

For example, the game's thread design is more favorable for dual-core CPUs, etc. Those who haven't seen that video can check it out.

This comparative review used ultra-high preset + ultra-high rendering ratio 1036P resolution for the test route, specifically the Wuling City cable car map, which puts the most CPU strain on the entire game.

So let's reveal the results.

Unsurprisingly, the best-performing device in *Arknights: The Last Land* in terms of overall energy efficiency is the iPhone 17 Pro Max.

Not only does it maintain a full frame rate throughout without any frame drops, but its power consumption is also kept to an ultra-low 5.1W, far lower than most Android models.

Combined with the 1% Low frame rate and our own experience, the iPhone 17 Pro Max's gaming experience can only be described as silky smooth, with almost no lag.

In comparison, the iPhone 17 Pro's scheduling is better than the Pro... The Max is much more conservative.

Although its heat dissipation is also good , its strategy is very conservative.

After running for about ten minutes, it uses temperature control to reduce the frequency, and the overall power consumption is only 4.9W.

In contrast, the iPhone 17 standard edition with the A19 performs even better, averaging 56 frames per second, which is very smooth.

On the Android side, the better performing models include the iQOO 15 and its siblings , the OnePlus 15, the Nubia Red Magic and its siblings , and the Honor WIN and Magic 8 series, all flagship phones with the 8E5 processor.

The iQOO 15 only experienced slight frame rate fluctuations after 20 minutes, maintaining an overall average frame rate of 58.

All cores were kept at an average of around 1.6GHz, resulting in a final power consumption of 6.4W , which is quite good for the iQOO 15. While the Ultra's pricing is outrageous , it is indeed one of the few Android models that truly achieves a consistently high frame rate.

The OnePlus 15's performance is somewhat similar to the iQOO 15 , even running the CPU around 1.6GHz , resulting in similar frame rates and power consumption . Red Magic and Nubia's approach is rather cunning;

. Red Magic and Nubia's approach is rather cunning; even with Awakening Mode enabled, they don't truly run at full speed , averaging around 57 frames per second with minor fluctuations , but at least power consumption is controlled and heat generation is reasonable.

The Honor Magic 8 series is also quite well-tuned , while the Honor WIN's power consumption is relatively high.

Both of these models maintain a frame rate of 56-57 frames per second.

However, what surprised me most was Honor's two 8E models, the GT Pro and WIN RT.

These two, even with the 8E version, can still achieve very high frame rates.

In fact, the WIN RT's energy efficiency is not much worse than its 8E5 sibling, the WIN.

Now you can see how small the difference between the 8E and 8E5 is.

Then there are some devices that can achieve high frame rates , but their power consumption is not well optimized.

The most prominent example is the Redmi K90 Pro. The Max

's average frame rate of 56.4 seemed decent, but upon checking the scheduler, it was shocking to find the CPU was running straight at 3GHz!

The power consumption soared to 10W in the first 10 minutes , and sure enough, it couldn't handle it and throttled down, averaging only 8.2W throughout . It seemed like it hadn't been tuned

. It seemed like it hadn't been tuned at all – just going all out!

The situation wasn't much better with the four Xiaomi 17 series phones . The Xiaomi 17 Pro Max also started very aggressively

. The Xiaomi 17 Pro Max also started very aggressively , but its cooling wasn't as good as the K90 Pro Max, resulting in a quick throttling.

The energy efficiency was poor, and the temperature skyrocketed to 50 degrees Celsius.

The two mid-range models also had similar frame rates and energy efficiency, running like crazy at the start but then throttling drastically.

Excluding the unreleased Samsung models , these are currently at the bottom of the 8E5 series.

In fact, many 8E phones perform just as well as the 8E5, besides the Honor mentioned earlier, including the iQOO Neo11, OnePlus Ace5 Pro, Xiaomi 15 Ultra, and Find X8. Ultra

models can all run at over 50 frames per second with low power consumption, while other 8E models generally run at around 40 frames per second.

You can check the performance level of the models you are interested in.

Then we'll talk about Dimensity models.

We've mentioned before that Endgame isn't very favorable for MediaTek's single-core flagship, and sure enough, the 9500 models are indeed not very good.

Although the X9 and X300 series don't have low frame rates , their power consumption is much higher than Qualcomm's, exceeding 7W. The

exceeding 7W. The energy efficiency is terrible, and the temperature is alarming.

Even more outrageous is the iQOO Z10 equipped with the Dimensity 9400+. Turbo+ and the Redmi K80 Ultra from iQOO are too powerful, while the Redmi, whose game wasn't on the whitelist and hadn't been adjusted, experienced a power explosion . We also tested several new 8Gen5 phones.

. We also tested several new 8Gen5 phones.

To be honest, the 8Gen5's power efficiency in *The End* is definitely inferior to the 8E.

These phones' power consumption soared to over 7W, and the frame rate was only around 40fps, which doesn't seem very good.

It should be said that the 8Gen5's pitifully small cache is most unsuitable for scenarios with high map streaming I/O pressure.

When moving at high speed, a large amount of assets flood the cache, which is really a disaster for the 8Gen5.

Finally we 'll give you a complete overview of these flagship and near-flagship phones in *The End*, from strongest to weakest . I wonder if it meets everyone's expectations?

. I wonder if it meets everyone's expectations?

The inclusion of *The End of the World* in this comparison, as a new game, caught some developers off guard.

These Android developers heavily rely on whitelist mechanisms, and we're testing popular games here.

If you're playing a game not on the whitelist, don't expect these developers to perform well.

Next, we'll test a mobile blockbuster that's definitely on the whitelist: *Mingchao*.

Version 3.1 of *Mingchao* is of good quality, and many players have returned.

Unfortunately, the game's workload is still quite high.

This time, we updated the test route, following the loop in La Helo, riding a motorcycle at full sprint speed.

There will be a lot of asset flow, which is relatively difficult to handle in open-world games.

On Android, we uniformly set it to the highest settings: 864P resolution and all graphics to maximum.

However, *Mingchao* on iOS limits the maximum resolution to 718P , so you need to be careful when comparing it side-by-side.

Even with a lower resolution than Android, *Mingchao*'s performance on iOS is still inferior to the iPhone 17 Pro Max on Android, unable to maintain a full frame rate throughout.

With an average frame rate of 57.2 and a power consumption of 6.2W, it's a bit high compared to the 8E5 model with 864P resolution . However, it's rare for an iPhone to reach 6W power consumption in a game.

resolution . However, it's rare for an iPhone to reach 6W power consumption in a game.

The heat dissipation of the iPhone 17 Pro Max is indeed much more aggressive than previous iPhones.

The other two more conservative iPhones, the iPhone 17 Pro and the standard iPhone 17, are not so comfortable.

They only run at 50 frames per second at 718P resolution, which is about the same as some good 8Gen3 models on Android.

"Mingchao" is developed based on Unreal Engine 4.

After all, EPIC and Apple are close friends, so the poor optimization of iOS is to be expected.

So, are there any devices that run "Mingchao" well?

It seems that among the familiar 8E5 flagships , Nubia and Red Magic both achieved a smooth full-frame rate while keeping power consumption under 5.6W , a very good balance.

Honor also performed well; both the Magic8 Pro and WIN achieved full-frame rate without any issues. The iQOO

iQOO15 Ultra also achieved a full-frame rate , and the iQOO15 also achieved a near-full-frame rate with low power consumption, making it quite good among cost-effective flagships.

The performance of the Mi series flagships in "Mingchao" wasn't as bad as in "The End of the World," as they are on the whitelist . The Xiaomi 17 Pro Max and Redmi K90 Pro Max both achieved near-full-frame rate while keeping power consumption at a reasonable level.

. The Xiaomi 17 Pro Max and Redmi K90 Pro Max both achieved near-full-frame rate while keeping power consumption at a reasonable level.

However, the Xiaomi 17 Pro Max 's heat dissipation is indeed not very good due to its rear screen limitation, while the K90 Pro Max's heat dissipation is quite good, with temperatures similar to Red Magic.

However, as we all know, the 8E5 doesn't offer much improvement over the 8E, which is clearly reflected in "Mingchao."

In fact, many 8E models also performed very well, such as the Realme GT8, OnePlus Ace6, Honor WIN RT, and iQOO. The Neo11

, along with the X8 Ultra and X200 Ultra, both flagship imaging phones , can run at close to 60 frames per second in "Mingchao". MediaTek models also perform well in "Mingchao".

The Dimensity 9500-powered X9 and X300 series , while not pushing the limits as much as some 8E5 models , can still achieve near-full frame rates with relatively low power consumption . Even phones based on the 9400,

. Even phones based on the 9400, such as the Redmi Turbo 5 Max and OnePlus Ace5 Ultra, offer decent performance . However, for every good phone, there's a bad one.

. However, for every good phone, there's a bad one.

For example, the OnePlus 15 performed very smoothly in media tests, but now retail units lock the frame rate at 45 frames per second after 18 minutes in "Mingchao", offering a worse experience than even its own 8E model, the Ace6.

OnePlus's Ace 5 Pro and 13T also employ similar frame-locking strategies, including the Realme GT8. The Pro version also suffers from this issue; after playing for 11 minutes, it starts locking at 45 frames per second , even worse than its own GT8 . The performance of these three 8E5 flagship phones – the Xiaomi 17 Pro, 17 Ultra, and standard 17 – is also subpar,

. The performance of these three 8E5 flagship phones – the Xiaomi 17 Pro, 17 Ultra, and standard 17 – is also subpar, with frequent frame drops and frequency throttling.

Even more absurd is the Redmi K90.

As an 8E model, it locks at 30 frames per second after only 7 minutes , then intermittently recovers to 60 frames per second before jumping back to 30.

Just look at this frame rate curve – what is this?!

It's a miracle this game runs smoothly.

Honestly, the K90 and its bigger brother, the K90 Pro Max, are from completely different teams. Ah, that's terrible.

We've also tested the other 8E models, so you can easily tell which is better and which is worse.

The 8Gen5 models, on the other hand, didn't perform as badly in *Mingchao* as they did in *The End of the World*.

All three 8Gen5 models maintained an average frame rate of over 58 FPS.

Except for the Ace6T, the power consumption of the other two was also relatively controllable.

Speaking specifically about *Mingchao*, the 8Gen5's performance wasn't actually worse than the 8E models. However,

compared to its performance against the 9500 in *The End of the World*, the Xiaomi Xuanjie's performance in *Mingchao* was relatively poor; it throttled and dropped frames after playing for a while, averaging less than 50 FPS . Next, we'll give you an overview of the performance of *Mingchao*

. Next, we'll give you an overview of the performance of *Mingchao* to see how it performs on the devices you're interested in.

Then, we'll move on to a regular on the show, *Honkai Impact 3rd: Star Railway* . *Golden Moment* remains a peak of mobile game pressure.

. *Golden Moment* remains a peak of mobile game pressure.

However, *Honkai Impact 3rd* limits the resolution to around 700p on iOS, some 8E models , and all 8Gen5 models.

These devices cannot participate in our 981p comparison, so we tested it in two rounds.

Let's first look at the high-load test at 981p.

Unsurprisingly, the top two devices in *Honkai Impact 3rd* are a Red Magic 11 Pro and... The Honor WIN Red Magic doesn't employ any fancy tricks; it simply precisely controls the CPU performance to a level that's just sufficient.

On average, the super core is 1.4GHz and the medium core is a little over 1.2GHz, which isn't high.

In the demanding game *Armored Iron*, the power consumption reaches 7.6W at this level, but the Red Magic still manages to keep it under control while maintaining a near-full frame rate.

This is clearly a performance-specific tuning, emphasizing that "gaming phones must prioritize gaming."

However, the Red Magic's performance isn't perfect.

Despite its top-tier performance tuning , we encountered touch issues during our *Armored Iron* testing . The touch controls on both sides of the Pro's screen completely stopped working;

. The touch controls on both sides of the Pro's screen completely stopped working; only exiting and re-entering the game could fix it.

This generation of Red Magic uses a Synaptics touch solution, and there might still be some optimization issues in the software; it seems it needs further refinement.

Honor's performance was also very good . Even without Phantom mode enabled

. Even without Phantom mode enabled , Windows could maintain a full 60fps for the first 25 minutes.

After 25 minutes, some minor fluctuations occurred , but the frequency drop was very slight, maintaining 57fps during this period . Over the entire 30 minutes, the average was 59.3fps with

. Over the entire 30 minutes, the average was 59.3fps with a power consumption of 7.8W.

The temperature was a bit tight, but it was still manageable. Enabling

Phantom mode didn't actually help with the subsequent frequency drop; its main function was to improve the 1% low frame rate at the initial 60fps, preventing occasional stuttering.

Even if other games boosted the frame rate to 120fps, this small frame drop issue couldn't be completely resolved.

The 1% low frame rate was indeed very noticeable in the initial experience.

It's almost always at 60fps, incredibly smooth , but the 981P version of "Clash Bots" consumes too much power, so it automatically quits Phantom mode when it gets hot.

Overall, the performance on Windows is still very good, but I want to say that Honor's Phantom frame stabilization isn't a completely free technology.

In our retail tests, we found that enabling Phantom mode introduces some latency , which is different from the frame interpolation latency we usually talk about.

This latency in Phantom mode mainly comes from triple buffering.

You can actually see that when dragging, Phantom mode sometimes becomes unresponsive , dragging 0-2 frames more than normal. This is

because the prerequisite for stable frame output in Phantom mode is that there must be frames in its reservoir.

So Honor increases this reservoir through triple buffering to ensure that frames are available when needed.

However, this increases latency and decouples rendering and output.

Although it does eliminate minor stuttering , the trade-off is that the gameplay feels a bit mushy.

In addition, Phantom frame stabilization in Magic... There seems to be a bug on the Magic 8 series.

It works fine when enabled , but cannot be properly disabled.

It turns off frame stabilization , but triple buffering isn't disabled, meaning even with Phantom Mode disabled, lag persists.

Windows doesn't have this issue.

Only the Magic 8 series has this problem , which I can only assume requires a bug fix.

One rather unusual device is the iQOO 15 Ultra ; based on our frame rate tests, it runs incredibly fast— even with 1% low frame rate, it reaches 59.

It's incredibly fast , but something felt off during actual gameplay.

While it was indeed smooth , it certainly wasn't at the level of a low 59 FPS. A

low 59 FPS would have a completely fixed frame rate of 16.67ms, resulting in a silky smooth feel.

However, the actual feel was more like V-Sync wasn't enabled.

While it was responsive , the frame generation time was unstable.

We decided to investigate further.

After carefully capturing the trace, we discovered the problem: its synthesizer (SurfaceFlinger) seemed completely disconnected from the game's rendering. The synthesizer

consistently refreshed the screen at a fixed 60Hz , without waiting for game frames . This meant that

. This meant that regardless of whether the game was ready for a frame, the synthesizer would refresh the screen every 16.67ms. Therefore, the testing software always read a perfect 60fps straight line.

However, for the game, without touch input, it didn't even receive V-Sync signals from the synthesizer to update the 60Hz refresh rate. Completely disconnected,

the game was actually running at its own pace.

Only when touch input occurred would the game receive VSync again. The signal

update interval gradually re-aligns with 60Hz , and iQOO also briefly boosts the GPU frequency to reduce the rendering time to within 16.67ms before waiting for the next V-Sync to continue rendering.

Throughout the process, the synthesizer remains stable at 60fps , but the actual frame time in the game is drifting, resulting in a feeling of one frame being faster and another slower.

We don't care whether this feeling of not using V-Sync is good or bad; perhaps it can reduce latency.

But you only need to know one thing : the frame rate measured this way is not realistic.

There will only be more and more tricks that testing tools cannot detect , so we will gradually introduce vision-based algorithms to test games in the future.

Anyway, there's always a way to catch you.

Returning to our comparative review, the Nubia Z80 Ultra with an 8E5 and the Honor WIN RT with an 8E both perform very well in running "Collapse Iron Man," but both of them sacrifice power consumption.

Under such an insane 8-9W power consumption, the temperature is completely off the charts, especially the Nubia Z80 Ultra.

Its awakening mode is even more aggressive than the Red Magic's awakening mode, with the temperature soaring above 50 degrees Celsius.

I would call it "familiar mode."

K90 Pro Max's performance is quite commendable; indeed, this generation of Redmi flagships performs quite well in games on the whitelist, ranking first among Mi series.

The Realme GT8 Pro and iQOO 15 are slightly behind but still quite good.

Further down the line, various 8E5, 8E, 9500, and 9400 models are all competing, with most hovering around 50 frames per second and power consumption exceeding 7W, which is considered average.

Are there any particularly bad models? Yes,

there are . For example, the OnePlus Ace6, iQOO Neo10 Pro+, OnePlus 13T, Realme GT8

. For example, the OnePlus Ace6, iQOO Neo10 Pro+, OnePlus 13T, Realme GT8 , and Redmi K90.

This isn't the first time this has happened; these 8E models, including the Xiaomi 15S, have also performed poorly . Even the Pro and other mid-range phones couldn't reach 50 frames per second in 981p of "Clash Bottom."

So let's lower the resolution to 763p and see how Apple and its near-flagship perform . Back in the Unity engine,

. Back in the Unity engine, the iPhone's power efficiency seems to have improved.

At around 700p, the iPhone 17 Pro Max's power consumption at full frame rate is just over 5W , still significantly lower than the 8E5 flagship.

The frame rate is basically full throughout, and the performance is quite good.

The frame rate fluctuations on the iPhone 17 and 17 Pro are slightly larger , but both are still relatively smooth.

At 700p resolution, most current mid-range phones with 8E5, 9500, 8E, and even three 8Gen5 processors can achieve this. It ran smoothly at near full frame rate, but there were a few disappointing exceptions, such as the Xiaomi 17 standard edition and, once again, the Redmi K90, both with very weak scheduling strategies.

For this test of *Armored Iron Man*, we decided to add a grueling 30-degree 5G test to the already runnable 700+ P/S setting.

Based on our self-built 5G core network—Geekbay Telecom— we successfully conducted our first 5G game power consumption test.

Under this 30-degree 5G test, we can also see if there are any differences in power consumption compared to a normal temperature Wi-Fi environment . Sure enough, in a high-temperature 5G environment,

. Sure enough, in a high-temperature 5G environment, many phones failed ; only the iPhone 17 Pro remained .

As expected, the Max maintained near-full frame rate even at 30 degrees Celsius with 5G, showing little performance loss compared to Wi-Fi at 25 degrees Celsius.

The improvements in heat dissipation in this generation of iPhones are evident ; the OnePlus 15, Ace6, and even the Ace6T with the 8 Gen5 didn't experience much frame drop.

All the tested iQOO, Honor WIN and RT, Red Magic and Nubia , as well as the Redmi K90 Pro Max, Turbo 5 Max, K80 Ultra, Find X8 Ultra, and Realme Neo 8 also performed remarkably well in the rigorous tests . Several Dimensity 9500 devices,

. Several Dimensity 9500 devices, the X9 series and X300 series , experienced slight frame drops at 30 degrees Celsius with 5G , but still managed to maintain decent frame rates.

Of course, some devices were significantly affected by outdoor conditions , such as the iPhone 17 Pro with its poor temperature control , the Xiaomi Mi 17 series, the Xiaomi Mi 15S Pro and 15 Ultra with their external modem and integrated modem, and the Realme GT8. The Pro

models were significantly affected by the 30-degree Celsius 5G environment, experiencing severe frame drops.

Surprisingly, the power consumption in 5G scenarios wasn't much higher than Wi-Fi.

While some models did have slightly higher 5G power consumption , others had similar power consumption to Wi-Fi, even with the added 30-degree Celsius heat.

Based on our experience, 5G becomes extremely power-intensive in scenarios with low signal strength , frequent cell switching, or games that require frequent network searches. In reality, unless it's cloud gaming , it doesn't consume that much network bandwidth.

Our lab tests...

Without very low signal strength , 5G and Wi-Fi might not have a significant difference in power consumption.

Finally, we also tested the relatively low-stress game "Honor of Kings."

Since iPhones only support 120fps HD , we won't compare it with Android phones running at 120fps.

Let's see how well each Android flagship performs in terms of energy efficiency.

Surprisingly, the crown for the best "Honor of Kings" energy efficiency was won by models from OPPO and Vivo using the Dimensity 9500.

The X9 series and X300 series both achieved a minimum overall power consumption of around 3W, including the OnePlus Ace5 Ultra and iQOO. The Z10 Turbo+ and these two 9400 models also have impressive power consumption.

In comparison, the power efficiency of various 8E5 flagship phones in "Honor of Kings" is not superior to that of the 8E , which is quite reasonable, since Qualcomm's low-frequency power efficiency has not improved at all this generation.

It should be noted that because "Honor of Kings" has a low load , the screen accounts for a relatively high proportion of the overall power consumption.

Please do not directly interpret the power consumption of the entire system as the chip power consumption.

The smoothness of "Honor of Kings" is mainly judged by the Low frame rate performance.

If you play "Honor of Kings" a lot, you can also refer to our Low frame rate test results.

In addition to performance testing, touch is also a very important part of the game.

How high is the touch latency of this generation of flagship phones when playing games?

How responsive is it?

We also tested the click latency this time.

First, as we did last year, we enabled the performance mode/game mode of each phone and tested the latency from touching to firing to the gun muzzle flash on the screen in "Peacekeeper Elite".

As we mentioned last year, due to the existence of vertical synchronization , the latency is not a precise value.

Sometimes it responds very quickly, and sometimes it may take a while to react.

Therefore, we have to do many rounds of testing and take the average value. This time, we will also test an additional standard deviation to see if the latency fluctuation range is stable.

Let's take a look at "Peacekeeper Elite".

Guess who performed the best?

It's the Honor WIN and iQOO 15 series.

These three devices all achieved a latency of only about 35ms, which is even close to the latency of many gaming PCs.

This is partly due to the Honor WIN's 480Hz touch sampling rate and the iQOO 15's 300Hz and 360Hz, which are also very high.

On the other hand, these two companies have indeed put a lot of effort into touch optimization.

So, is there anyone whose performance is relatively poor?

iPhone!

Our tests showed that all three iPhones tested this year had a latency of around 60ms in "Peacekeeper Elite," placing them at the bottom of the list among all the tested devices.

Other Android phones seemed to have similar latency performance in "Peacekeeper Elite," generally staying in the 40-50ms range.

However, looking at the standard deviation, a few showed relatively unstable latency performance, such as the Vivo X300 series , Xiaomi 17 Ultra, and Nubia Z80. Ultra

needs significant optimization.

I initially thought this was the general touch performance, but iPhone latency is indeed far from as good as rumored.

Then we introduced a new test : using the default battery mode for apps not on the whitelist.

We wrote a blank app using the same framework and ran it on Android and iOS to test latency again . This time, the results

. This time, the results were completely different from those in *Peacekeeper Elite*.

The latency on the three iPhones was reduced by nearly 20ms compared to *Peacekeeper Elite*, going from being far behind to being among the top.

Why is the latency so much lower ? Mainly because of the reduced latency from *Peacekeeper Elite*'s graphics engine rendering.

? Mainly because of the reduced latency from *Peacekeeper Elite*'s graphics engine rendering.

In fact, it's not that iPhone latency is high, but rather the iOS version of *Peacekeeper Elite* , or many iOS games, have high latency. Compared to the Android version

high latency. Compared to the Android version , the iOS version of these games has a significant difference in rendering latency : they use triple buffering.

This causes additional latency.

Apple's Metal API actually enables triple buffering by default.

Although it can be changed to double buffering , most developers don't prioritize latency and simply use the default recommendation.

So why does Apple still recommend triple buffering despite the increased latency?

This brings us to the benefits of triple buffering.

As we mentioned earlier, it increases the frame pool by queuing an extra frame to prevent frame instability.

Apple's iOS has always been a visually consistent, animation-driven system ; in their philosophy, strict frame rhythm is more important than low latency.

Anyway, after we switched to testing a blank app, the iPhone's latency wasn't as bad as in "Peacekeeper Elite." So

how did other phones perform?

Theoretically, testing our blank app, which has no graphics rendering, should have lower latency than "Peacekeeper Elite," right?

But in reality, the latency of most Android phones didn't decrease at all , and some even increased.

This is because Android's whitelist mechanism kicks in.

Only when whitelisted apps are detected will Android enable more sensitive touch sensing . You can see that

. You can see that because they're not on the whitelist, almost all Android phones have reduced touch sampling rates.

Many Android phones' touch settings are also affected by battery settings; only in performance mode is there a better touch experience.

For example, Xiaomi phones are like this; after changing the battery settings back to the default balanced setting, the touch latency of most Xiaomi phones increased significantly.

In our blank app test, which removed graphics rendering, you can also see an advantage of the iPhone : The touch latency range is extremely stable.

Our measured standard deviation is only around 2. This means

that after so many rounds of testing, the touch latency has hardly fluctuated , remaining within a very small range.

Among the tested models, this is one of the most stable.

This is why some rhythm game players prefer iPhones or iPads.

Because of the offset, rhythm games don't actually pursue absolutely low touch latency; as long as the latency isn't too high, it's fine.

However, iOS's touch latency fluctuation is relatively low, and the frame rate is very stable, which is more important to rhythm game players.

On the other hand, iOS's audio link is also better .

We had a technical exchange with the developers of *Muse Dash*, and they developed an analysis tool that can measure the audio link latency of both platforms. You can see that the iPhone not only has very low audio latency— the iPhone 17 Pro Max has an average latency of only 28ms —but more importantly, its standard deviation is much smaller than that of Android phones, only a few tenths of a percent.

This means that the iPhone's audio is extremely stable, and the latency won't fluctuate wildly.

Android's audio performance is quite inconsistent , although there are some Android phones with good audio.

Do you know which Android phone has the best audio?

The Redmi K90 Pro Max!

While not as stable as the iPhone , its latency is exceptionally low at 18ms, standing out from the crowd.

And as you know, the sound quality is also excellent; its audio system has clearly been well-tuned.

In fact, Xiaomi phones generally perform quite well ; the Xiaomi 17 series, along with the two iQOO 15s and several flagship Honor phones, all have excellent audio latency . However, some Android phones are truly terrible.

audio latency . However, some Android phones are truly terrible.

For example, the OPPO Find X9 has an audio latency of nearly 90ms – is that even suitable for playing rhythm games?

And OnePlus, except for the 15, none of its "low-end" models perform well.

Plus, many rhythm games might not even be on the Android whitelist.

Before buying an Android phone, players have no idea whether the audio latency is stable and normal.

Now you know why rhythm game players buy iPhones and iPads, right?

Actually, we originally wanted to introduce a new latency test : a device with good swipe latency and click latency performance may not necessarily have good swipe latency.

However, measuring swipe latency is particularly troublesome.

If we don't want to count frame by frame, we need to develop a dedicated vision-based algorithm to measure it.

We're already working on this, and we estimate we can use it in our next comprehensive review .

So , that concludes our comprehensive review of mobile game performance.

How was it?

You should have enjoyed it!

For mobile gamers, which of the new phones released in the past year is the most suitable for gaming?

Let me save you some money.

First the iPhone 17 Pro Max is definitely highly recommended. It

's the iPhone with the best gaming performance and heat dissipation in history , if your budget allows.

For Android, among the Snapdragon 8E5 flagships, considering both gaming performance and cost-effectiveness, I feel the iQOO 15 and Honor WIN are the most worthwhile.

Of course, as you saw in today's review, the Snapdragon 8E5 doesn't really offer much improvement over its predecessor.

The Dimensity 9500 next door also had its chance, but it didn't perform well.

So, for those with a limited budget...

I think buying the 8E model is the most cost-effective, but you have to be careful to avoid some duds, like the K90 standard edition.

Of course , this video only talked about gaming.

We will be testing all these machines in the future on battery life, imaging, and smoothness, so stay tuned.

Okay, if you think this video is informative and helpful, please like, comment, save, share , and follow our GeekBay channel.

I'm Yunfei.

Happy New Year , see you next time, bye-bye!