This is not unprecedented -- similarly, there were performance variations between the Nexus One, HTC Incredible, and HTC Evo at the time of their release. *shrug*
-
After 15+ years, we've made a big change: Android Forums is now Early Bird Club. Learn more here.
GL benchmarks of the 3D's Processor...
- Thread starter Spdy
- Start date
-
- Tags
- htc evo 3d
Yep. Small firmware revisions might fix some bug or UI lag and simultaneously cause a couple FPS hit in a benchmark. Maybe they put more aggressive clock throttling in the EVO so it gets better battery life, but causes a little performance hit. Plus, there's some margin of error in most benchmarks.
Not too worried -- it's still smokin' fast.
Sensation has less ram 768 MB vs. 1 GB on the 3vo.
Ah, I think I figured it out. Check out Anandtech's own benchmarks from three weeks ago:
http://www.anandtech.com/show/4416/...8x60-phones-htc-sensation-4g-and-htc-evo-3d/2
From there, the EVO 3D was beating the Sensation:
GLBenchmark2.0 - Egypt:
- sensation: 20.8 FPS
- evo 3d: 21.0 FPS
GLBenchmark2.0 - Pro:
- sensation: 52.8 FPS
- evo 3d: 53.4 FPS
But from yesterday's benchmarks, the Sensation is now winning:
http://www.anandtech.com/show/4459/...-up-on-glbenchmark-a-verizon-xt883-with-omap4
GLBenchmark2.0 - Egypt:
- sensation: 23.6 FPS
- evo 3d: 21.0 FPS
GLBenchmark2.0 - Pro:
- sensation: 54.6 FPS
- evo 3d: 53.4 FPS
So if you didn't see what happened there, the Sensation went UP in performance in the past couple weeks, but the EVO 3D stayed the same. I suspect there was a firmware update to the Sensation that improved performance. I further suspect that the same update will be coming to the EVO 3D.
http://www.anandtech.com/show/4416/...8x60-phones-htc-sensation-4g-and-htc-evo-3d/2
From there, the EVO 3D was beating the Sensation:
GLBenchmark2.0 - Egypt:
- sensation: 20.8 FPS
- evo 3d: 21.0 FPS
GLBenchmark2.0 - Pro:
- sensation: 52.8 FPS
- evo 3d: 53.4 FPS
But from yesterday's benchmarks, the Sensation is now winning:
http://www.anandtech.com/show/4459/...-up-on-glbenchmark-a-verizon-xt883-with-omap4
GLBenchmark2.0 - Egypt:
- sensation: 23.6 FPS
- evo 3d: 21.0 FPS
GLBenchmark2.0 - Pro:
- sensation: 54.6 FPS
- evo 3d: 53.4 FPS
So if you didn't see what happened there, the Sensation went UP in performance in the past couple weeks, but the EVO 3D stayed the same. I suspect there was a firmware update to the Sensation that improved performance. I further suspect that the same update will be coming to the EVO 3D.
that is interesting....
also... i remember reading the snapdragon dual cores can run at lower speeds as needed independently....
this is just a GUESS!!!!
maybe..the test apps are not tailored correctly for the snapdragon dual cores.
the snapdragons are not ramping up too full speeds during these simple tests.
or one of the cores are running at a very low speed.
also... i remember reading the snapdragon dual cores can run at lower speeds as needed independently....
this is just a GUESS!!!!
maybe..the test apps are not tailored correctly for the snapdragon dual cores.
the snapdragons are not ramping up too full speeds during these simple tests.
or one of the cores are running at a very low speed.
Nice catch. If I remember correctly, the original testing in the first article (on both the Sensation and Evo 3D) was on preproduction units at HTC Headquarters.
It might be that the updated numbers in the newest article are from a production Sensation device that they are testing for a future review article (since the Sensation was released in the US about a week earlier than the Evo 3D).
Like what? If anything, they were closer in similarity than the Sensation and Evo 3D (if I remember right).
Nexus is plain android.. no sense to eat up some RAM / performance
it is all the block all programmers use to test/develop on, which might help its run things better.
Incredible has less RAM but more Storage... first ones had AMOLED screens which may or may not affect FPS.
EVO had a bunch of battery hog extras... (guess) which might mean they tweaked the kernel for it to run less aggressive to save battery, which might affect performance.
.......
e3d has more RAM
sensation does not have 3D (but 3D can be turned off)
if i was to guess.. i would think e3d should be faster; if not the same.
Not really the tests' fault.
The tests claim to say what the processor does - not strictly true, they do not make low-level hardware calls and assess response. That sort of thing could entail biasing the tests. This ain't that.
What these tests really do is show system response when an application makes specific calls to underlying graphics support libraries. These graphics libraries offer a very rich set of operations for the application programmers, allowing them many different ways to accomplish the same sorts of things from our point of view. Some of these paths (through software to hardware to user perception) is going to work better on some processors than others, yet other paths through the graphics libraries will turn things around and the slot for the top dog changes.
Unless and until app developers release source code for free (on that stuff they make their living on!) and you can check their programming methods against what calls these benchmarks make (assuming you've got a few years of experience of analyzing code) all these benchmarks can do is give rough order of magnitude.
What the Sensation vs. 3vo results tell me is that the 3vo was likely running more processes or something at the time of the benchmarking - entirely likely given we have zero clue if (example pulled from sky to illustrate) the 4G was on or something on the 3vo while the Sensation doesn't even have that component to manage.
But a frame or frame and a half here or there difference - within order of magnitude, that's the same result - and is showing x3 to +x9 increase over the Evo - by order of magnitude, makes sense.
From time to time, you can see what looks like a lesser processor doing better in a benchmark - happening above, too.
All sorts of reasons. Most apps have multiple things happening at once - those little - what shall I call them? here's a phoney name - applets within the app: those are called threads, and will immediately distribute between the cores without need of re-coding the app. But - if an app (such as a benchmark) doesn't use many of those, or has one that's very intensive such that the distribution between cores doesn't make much sense - then the good single core can outperform on that one metric.
Goes back to what I always say: benchmarks are tools for pros (like app developers wanting to choose the better graphics calls) rather horse-race results.
When you see a single processor outperforming others, on every benchmark you can find, that is suggestive that that processor is faster overall.
Otherwise, you simply see evidence of the balance of design - faster here, not so much there, faster someplace else, less so over there... and so forth.
Order of magnitude, these benchmarks are saying that the superphones of a year ago aren't able to hold up to the superphones of this year in overall performance - without having to re-code anything on the application side, that's just happening. Exactly as I predicted would happen if the SMP capabilities of the underlying Linux OS were properly exploited.
The tests claim to say what the processor does - not strictly true, they do not make low-level hardware calls and assess response. That sort of thing could entail biasing the tests. This ain't that.
What these tests really do is show system response when an application makes specific calls to underlying graphics support libraries. These graphics libraries offer a very rich set of operations for the application programmers, allowing them many different ways to accomplish the same sorts of things from our point of view. Some of these paths (through software to hardware to user perception) is going to work better on some processors than others, yet other paths through the graphics libraries will turn things around and the slot for the top dog changes.
Unless and until app developers release source code for free (on that stuff they make their living on!) and you can check their programming methods against what calls these benchmarks make (assuming you've got a few years of experience of analyzing code) all these benchmarks can do is give rough order of magnitude.
What the Sensation vs. 3vo results tell me is that the 3vo was likely running more processes or something at the time of the benchmarking - entirely likely given we have zero clue if (example pulled from sky to illustrate) the 4G was on or something on the 3vo while the Sensation doesn't even have that component to manage.
But a frame or frame and a half here or there difference - within order of magnitude, that's the same result - and is showing x3 to +x9 increase over the Evo - by order of magnitude, makes sense.
From time to time, you can see what looks like a lesser processor doing better in a benchmark - happening above, too.
All sorts of reasons. Most apps have multiple things happening at once - those little - what shall I call them? here's a phoney name - applets within the app: those are called threads, and will immediately distribute between the cores without need of re-coding the app. But - if an app (such as a benchmark) doesn't use many of those, or has one that's very intensive such that the distribution between cores doesn't make much sense - then the good single core can outperform on that one metric.
Goes back to what I always say: benchmarks are tools for pros (like app developers wanting to choose the better graphics calls) rather horse-race results.
When you see a single processor outperforming others, on every benchmark you can find, that is suggestive that that processor is faster overall.
Otherwise, you simply see evidence of the balance of design - faster here, not so much there, faster someplace else, less so over there... and so forth.
Order of magnitude, these benchmarks are saying that the superphones of a year ago aren't able to hold up to the superphones of this year in overall performance - without having to re-code anything on the application side, that's just happening. Exactly as I predicted would happen if the SMP capabilities of the underlying Linux OS were properly exploited.
xXSecretAgentXx
Newbie
You guys who already have the Evo 3d's should check out the newest benchmark software from chainfire that takes advantage of both cores properly. Even he himself said that the final score that comes out is again not what real world speeds whould be but the individual scores are interesting. Im definately impressed with the latest qualcomm processor!
wow this thread has moved along quite nicely since i was last on... and in lieuof the impending release date im glad to see that the 3vo has been reported to be "blazing fast" at least in terms of personal usage, and i really wonder about which benchmarks Could "accurately" take advantage of both cores going at full speed..considering their asynchronous dual cores. Also just one more day..finally! 
...i really wonder about which benchmarks Could "accurately" take advantage of both cores going at full speed..considering their asynchronous dual cores.
If a benchmark can't take advantage of both cores at full speed, are you saying that other programs can't take advantage either (meaning programs that are capable of using dual cores but not at "full speed")? Seems unlikely.
^This.
The dual cores are being used - but load balancing in SMP designs is a science unto itself.
I mentioned earlier (this thread? another?) that symmetrical multiprocessing isn't the most advanced approach for this sort of thing, but it gets you to higher performance quickly and tends to level the playing field.
Will Android always be expecting better hardware to get the job done as opposed to the most efficient and streamlined coding? Sure. Nature of the beast.
Can and will the chipmakers abide that path? So far, so good - so, yeap.
Can the app devs alone provide the needed load balancing for top performance? Maybe and maybe not or maybe not a whole lot - they run within the Dalvik environment and have to work with various kernels and different cores within the constraints of their dev kits.
The dual cores are being used - but load balancing in SMP designs is a science unto itself.
I mentioned earlier (this thread? another?) that symmetrical multiprocessing isn't the most advanced approach for this sort of thing, but it gets you to higher performance quickly and tends to level the playing field.
Will Android always be expecting better hardware to get the job done as opposed to the most efficient and streamlined coding? Sure. Nature of the beast.
Can and will the chipmakers abide that path? So far, so good - so, yeap.
Can the app devs alone provide the needed load balancing for top performance? Maybe and maybe not or maybe not a whole lot - they run within the Dalvik environment and have to work with various kernels and different cores within the constraints of their dev kits.
theres also the give and take issue...
how much can pushing both cores to the max speed in an already fast system..
most apps (NON-super resource hungry Games) are not that big... so.. to push the cores gets you a little (nano seconds) faster. but having the cores at full eats battery.
so.. you have to have an OS / kernel / app that knows when to push them and when to slow them... and how much...
and where is that happy medium or line is the sand???
how much can pushing both cores to the max speed in an already fast system..
most apps (NON-super resource hungry Games) are not that big... so.. to push the cores gets you a little (nano seconds) faster. but having the cores at full eats battery.
so.. you have to have an OS / kernel / app that knows when to push them and when to slow them... and how much...
and where is that happy medium or line is the sand???
I dunno if anyone saw this but AnandTech has their review of the HTC Sensation up, and it's chock full of great information:
AnandTech - HTC Sensation 4G Review - A Sensational Smartphone
If you read the review -- the HTC Evo 3D review is coming soon:
And something pretty neat is that they've identified a RESOLUTION AGNOSTIC graphics testing benchmark -- so more direct device to device comparisons can be made without the resolution influencing results:
Good times, I can't wait for the HTC Evo 3D review to come out.
There's also this nice tidbit about the Sensation's cameras -- I hope they figure out what sensors are being used in the Evo 3D as well:
AnandTech - HTC Sensation 4G Review - A Sensational Smartphone
If you read the review -- the HTC Evo 3D review is coming soon:
And something pretty neat is that they've identified a RESOLUTION AGNOSTIC graphics testing benchmark -- so more direct device to device comparisons can be made without the resolution influencing results:
Good times, I can't wait for the HTC Evo 3D review to come out.
There's also this nice tidbit about the Sensation's cameras -- I hope they figure out what sensors are being used in the Evo 3D as well:
At this point in the thread, I want to give the same caution I gave in the SGS2 and Sensation forums where claims of optimized or deficient benchmarks arise:
The benchmarks attempt to measure and show how hardware responds to a specific set of app calls to an OpenGL software library, usually made in some stressful way (if the benchmark is worth anything).
It's tempting to explain away unfavorable results, but in truth, if some app you need or want is coded in any way similarly to the benchmark in question, then that app is likely to run less well on your phone.
In the end, looking at all benchmarks is a good idea - but the best use of the graphics benchmarks are for app developers to choose which OpenGL calls to make to serve their audience - because there's more than one way to do about anything in graphics programming.
The way to not use the benchmarks is like results of a horse race.
There is no mystery whatsoever as to what the hardware can do. Sign a non-disclosure agreement with SoC maker as a recognized member of the hardware industry with a need to know and you can get the raw chip benchmarks straight from the horse's mouth. I absolutely promise that Qualcomm and Samsung and TI know precisely the performance of their graphics cores measured on bare metal.
At one time not long ago, they usefully published that in the open on the web. My favorite was the blog-published benchmark showing that Hummingbird could do more millions of triangles per second than Samsung measured and spec'd - by a wide margin. IOW - what the blogs reported was flatly unpossible for one particular measurement by one particular benchmark.
So - yep - it's a fine line. Look for benchmarks that exaggerate and through them out - but consider unfavorable benchmarks carefully because you might get an unfavorable app some day.
This whole rant goes back to my common claim - benchmarks have to correlate to the real world - and that ain't easy when you think about it.
Anyways - I promise if I had the answers, I'd tell you.
Genius sir. I wish there was a way to get all your informative posts on here (such as this one) on paperback or kindle, haha! As it stands 12k posts is a lot to dig through (akin to an patent attorney trying to find 'the one'). You should really create an expansive, indexed "Android" post of posts.
I appreciate the kind words and am truly humbled.
However - I don't know how I could do that without it being seen as a vanity thing.
Plus - a good deal of the time, many of my important discoveries simply come from our wonderful newcomers who don't know something is impossible and therefore they march off, find a way and post it.
I just spread the word.
That - and I post a lot because I never know what is helpful and what just seemed like a good idea at the time.
Maybe a community wiki will be an idea whose time will come - we've chatted about it on staff before.
Hmmmmm. Will update on that last one if I can see who on staff would have time to support it...
I've ordered my HTC EVO 3D, which should be arriving in a couple of days, and I'm really excited about it! One thing I noticed when I was doing some research on this device was that the benchmarks were considerably lower than the competition and side-by-side comparisons showed that the EVO 3D loaded apps and other things slightly slower than other devices. The EVO 3Ds specs are really competitive though, so why is it that this phone seems slightly slower than other devices?
mteezey
Android Enthusiast
I'm sure no matter what those scores say your gonna be happy with the phone and will have no problem with apps opening. I don't know if you plan on saving it stock or not. I know if you root and get a custom kernel it's going to raise those scores considerably above the other devices. The only downside to this phone that I have is random reboots. I've been doing some research and thought it was due to my rom/kernel but there is people with stock phones having this issue.
@Voltex,
I have merged your thread here. This thread speaks about benchmarks from other handsets and is definitely worth a good read. Early Mon's posts alone are quite informative.
The EVO 3D is a great handset, despite what you have seen or "read". IF you decide to root, you can open up more possibilities and unleash more performance. There is an extensive rooting section in this forum which will give you even further proof the horsepower in this handset is just as peppy as you want it.
You no doubt made the correct decision as the first impressions thread has shown and also from the discussions in this thread.
I have merged your thread here. This thread speaks about benchmarks from other handsets and is definitely worth a good read. Early Mon's posts alone are quite informative.
The EVO 3D is a great handset, despite what you have seen or "read". IF you decide to root, you can open up more possibilities and unleash more performance. There is an extensive rooting section in this forum which will give you even further proof the horsepower in this handset is just as peppy as you want it.
You no doubt made the correct decision as the first impressions thread has shown and also from the discussions in this thread.
Eaglesfan9106
Android Enthusiast
I updated and got a score of 2703 stock just curious as to what you guys get
Try AnTuTu - the free version gives a score breakdown the way that the paid Quadrant does.