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.
iPad A5x vs Tegra 3 - Android Forum's perspective
- Thread starter MacAttacka
- Start date
- Sort by reaction score
The aspect ratio of the iPad being portrait focused is pretty much moot now that they resolution is higher than 1920 across so you just have slightly larger black bars but don't miss out on any of the pixels in 1080p video. In fact, you could display actual 2k video without problems if a video player supported it on iOS.
Android forums...? Give me the link!
2) That's not true. ShadowGun has some extra graphics which the developers added, which haven't been added to the iOS version. It'll run much better on the iPad 2 than the Prime.
3) That's not true. Some games run at 2048x1536.
4) It's okay for video, but great for everything else. With 16:9 or 16:10 you get an awkward shaped device which you can only use in landscape
Hey psonice, love your nightcap update
Back to the topic, the A5X does on board Z-buffering? Where can I read more about this?
Thanks! There's more on the way
The A5X doesn't do z-buffering, but "tile based deferred rendering". Basically it splits the rendering up into small tiles (can't remember for sure, but they may be 16x16 pixels). Each tile has a list of polygons that overlap it, and because the tiles are small the list of polygons is generally small too. It then very quickly determines which polygon is at the front in each pixel, then renders each pixel as normal (it's "deferred" because it sets up the polygons first, then actual texturing and shading is done later). There's some cost associated with this of course.
The benefit is that each pixel is rendered just once. With nvidia's solution, where several polygons overlap on a pixel it has to either render all of them (on average this works out at something like rendering 3 times per pixel) with associated extra costs from fetching textures and running fragment programs or it has to do z-buffer testing (which is expensive) and still do some over-draw. It's pretty obvious then that for actual rendering in a real game, the A5X is going to be 3x faster on average with the same hardware.
There's another big benefit too: the rendering is done in tiny pieces, and as it renders each tile it knows in advance what polygons are needed, and therefore what textures are needed. Because it's all done on a very small scale, the textures can be fetched in advance and held in very fast cache during rendering. That makes (most) texturing VERY fast because it doesn't need to go to main memory. When you do "dynamic texture lookups" (where the texture location isn't known in advance) it can't pre-fetch though, and then texturing is very expensive.
The powervr page on wikipedia has a small overview of the tech: http://en.wikipedia.org/wiki/PowerVR#Technology and this is the GPU used: http://en.wikipedia.org/wiki/PowerVR#Series_5_.28SGX.29 and the GPU specs: http://en.wikipedia.org/wiki/PowerVR#Series_5XT
If you want to see how it stacks up against the tegra 3 gpu, check the first table on this page: http://www.anandtech.com/show/5072/nvidias-tegra-3-launched-architecture-revealed/2 It doesn't show the SGX543MP4 used in the ipad3, but you can simply double the MP2 numbers. In raw performance, the MP2 was 2.66x faster than the tegra 3. The MP4 doubles that to 5.33x faster. The MP4 has more overhead when it sets up the tiles, but then it uses the rest of its time much more efficiently. End result is what you see in the benchmarks - it's still several times faster than the tegra.
Also note the core count on that table: what nvidia calls a GPU "core" is the SIMD unit (that does all the real work on these things). The tegra 3 has 12, the MP2 has 8 - but the MP4 doubles that to 16. And they're more powerful - 4 MADD operations to every 1 on the tegra.
First off regarding Android forum posts, it's no different from what I read here. Each are proud of what they own and critical of others. Let's face it, from this thread alone I see nothing but biased reactions from some posters.
The A5X is just a necessity to support the new display, being that it has twice the graphics cores than the A5 doesn't mean it has double the graphics power and the benchmarks prove it. Is it a performance boost most users can tell? Probably not. Most are not familiar with the association between having a higher resolution screen and the need to scale its fill rate in order to achieve the same visual experience.
Whether it's an A5X or Tegra 3, neither can realize its full potential if the software isn't making best use of it. I could care less if it's dual or quad core if my user experience is anything but good. I've heard folks saying the Android experience can be jerky, but I've seen all generations of iPads stutter under some circumstances so no tablet product can claim to be "baby smooth" in that area.
What I have to bring into my comments is memory performance since it must work with the CPU of either platform. Personally I believe the memory performance of both the Tegra 3 platform and iPad 3 are lacking, I can see it when I start to see jerkiness under certain conditions. Simply adding more memory isn't always the best solution as CPU's scale towards higher performance, the speed must also match properly in conjunction to its graphics performance.
To skew off the topic a little (since the thread is about the A5X and T3), my biggest headache with the iPad line is the quality of apps. I honestly don't play Angry Birds and all that junk, I use the iPad for work and I'm always on the hunt for good work Apps. I don't like any of the calendar/scheduling apps, the default Mail app is too basic, and a lot of iOS apps crash a lot.
Compared to the Android apps which I can definitely say is significantly better, offers better work apps than Pagis/QuickOffice Pro, better Sharepoint/Powerpoint/Excel compatible viewers/editors that're much more stable and don't crash (back to the desktop) nearly as much. This is why I can't stand using any iOS device unless it has access to JB apps (not Installous) since Apple's too stupid to realize that they're limiting users from having access (via App Store) to some of the best stuff written for iOS devices.
The A5X is just a necessity to support the new display, being that it has twice the graphics cores than the A5 doesn't mean it has double the graphics power and the benchmarks prove it. Is it a performance boost most users can tell? Probably not. Most are not familiar with the association between having a higher resolution screen and the need to scale its fill rate in order to achieve the same visual experience.
Whether it's an A5X or Tegra 3, neither can realize its full potential if the software isn't making best use of it. I could care less if it's dual or quad core if my user experience is anything but good. I've heard folks saying the Android experience can be jerky, but I've seen all generations of iPads stutter under some circumstances so no tablet product can claim to be "baby smooth" in that area.
What I have to bring into my comments is memory performance since it must work with the CPU of either platform. Personally I believe the memory performance of both the Tegra 3 platform and iPad 3 are lacking, I can see it when I start to see jerkiness under certain conditions. Simply adding more memory isn't always the best solution as CPU's scale towards higher performance, the speed must also match properly in conjunction to its graphics performance.
To skew off the topic a little (since the thread is about the A5X and T3), my biggest headache with the iPad line is the quality of apps. I honestly don't play Angry Birds and all that junk, I use the iPad for work and I'm always on the hunt for good work Apps. I don't like any of the calendar/scheduling apps, the default Mail app is too basic, and a lot of iOS apps crash a lot.
Compared to the Android apps which I can definitely say is significantly better, offers better work apps than Pagis/QuickOffice Pro, better Sharepoint/Powerpoint/Excel compatible viewers/editors that're much more stable and don't crash (back to the desktop) nearly as much. This is why I can't stand using any iOS device unless it has access to JB apps (not Installous) since Apple's too stupid to realize that they're limiting users from having access (via App Store) to some of the best stuff written for iOS devices.
Line of quality apps? Are you kidding? I have owned 3 different android tablets, all sold/returned, and have moved on to the ipad. I am a PhD student in molecular biology...so productivity is a big thing for me. Android has nothing that iOS doesn't have. Plus pages/numbers/keynote for ipad are microsoft office compatible (both importing and exporting). There are way above anything android can offer.
What kind of apps are you in need of for M-Bio? I have a dbl masters in mechanical engineering and information management so productivity is very important to me as well.
For me Android has many tools that iOS Apps can't offer due to how Apple regulates what apps can and can't do. Something as simple as a good WiFi analyzer can't be found in the App Store and you can only find it using 3rd party sources via jailbroken iOS devices. These apps are readily available on the Android Marketplace and it's common to find many engineers use these apps in place of standalone testers/devices. One of the many tools we use is internal penetration testing which looks for holes in our LAN security and for many years relied on $3000-$8000 Fluke testers. However keeping them updated is a chore and the devices are expensive. Find me an iOS app that remotely comes close to "anti" (android network toolkit) which basically does 90% of what we need for a fraction of the price then we'll talk about which platform doesn't have what.
As for Pages/Numbers/Keynote, I never said they weren't compatible with MS Office, I said there are better apps than those. For example Pages is a very nice app until you realize it's so limiting that it'll make you run to MS Word. I didn't read user reviews before I bought it and now regret it since they point out the flaws very clearly. If all you need to do is make a book report, Pages is great. Want to do a real report with objects and macros, won't happen. It's really a shame because it's really a pleasure using those apps however I expect much more considering they want $10 per app versus QOP which offers a full suite for the price of $20.
QuickOffice Pro for Android is very stable, works reliably and has never crashed in my experience. The iOS port on the other hand, have crashed back to the desktop numerous times, and lacks a few features (not a big deal but still worth noting considering apples vs apples).
Last edited:
1) No.
2) Shaddow gun is Nvidia funded so its bound to be optimised for tegra. Its lower res but smoother on A5/A5x.
3) The big real racing/Infinity Blade games dont run at at 1536p but who cares? Theyre astounding as they are.
4) When you can display full 1080p with pixels to spare who cares about aspect ratio? On balance the screen out reses 1080p and has the exceelent 4:3 productivity ratio.
----------
This forum along with the other forum is very one sides. We all tend to agree with each other. Its nice to get a different perspective sometimes.
Exactly, all they can do is clutch staws. 4:3 is optimal for anything other than video and even so when you can do full 1080p with room to spare, they have nothing on the retina.
Cmon guys, aren't there enough of these threads on the forum to not clutter up this one with the same old arguments? The A5X beats the Tegra 3. You say:
I've already posted many benchmarks. Some are close to being double, and some are more than being double. All benchmarks are far and away above the Tegra 3. This thread is about the A5X vs the Tegra 3, not A5X vs A5. This benchmark shows more than double the performance of the Tegra 3
You are clearly a technical person, so I'm very confused by these comments. I don't think I need to post any side by side screenshots here. I'm sure you consider a dramatic increase in resolution, with the same frame rate and in some cases increased graphical effects (shadows...etc) as a performance boost most users can tell. PC Gamers spend thousands of dollars upgrading so they can play their games at a higher resolution, how can you say that's the "same visual experience"?
This is a thread about performance. Android tablets stutter 100% of the time I've tried using them, even when doing simple tasks like scrolling a page, or switching screens. If the iPad stutters 1% of the time, and most users don't notice, then yes, the iPad is "baby smooth" to them. I have yet to meet an Android user that described their experience as "butter smooth".
This isn't a fair comparison, however, because many factors besides the main chip can affect this (Android background processes and such). The best way to compare the two chips is via benchmarks, and it's not even close. Even the A5 from iPad 2 beats the Tegra 3.
I'm really not sure what you're arguing here.
Last edited:
We already know the a5x is better than the tegra 3 in the most important ways. What about the vita. The vita is a dedicated video game console with the same gpu as the new iPad and a smaller screen. Does it have better graphics than the new iPad?
First of all if you're going to use a technical argument to come at me, I'll have to point out some flaws in your argument, however you need to be objective so you can understand where I'm coming from.
I'm not at all trying to state Android is better than iPad and if you intention is to sell to me the perfection of the iPad, you might as well stop because you won't make any ground with me with that argument. Android does have better apps/tools than Apple App Store iOS apps, however that isn't what this thread is about (I just threw it in there because software titles do matter when it comes to these tablets).
Benchmarks aren't the end-all to all discussions and they don't tell everything we need to know. The GL test shows offscreen performance, not on screen. As I've stated before, how is my info wrong about how the A5X having 4 cores doesn't quite equate to 2x the performance of the A5? I was simply pointing out a very common myth lots of iPad 3 owners say about its graphics performance in comparison to the iPad 2. Never did I ever mention about how the T3 stacks up in graphics performance, that's pretty much a dead discussion. http://www.tomshardware.com/reviews/ipad-3-benchmark-review,3156-6.html
Let me clear up the confusion and it wasn't meant to be "technical". I doubt users can tell the difference in performance under most circumstances. Here's an example, if I ran a non-retina app on both an iPad 2 and iPad 3 and didn't allow you to physically hold/touch either except just looking at the visual quality, could most people tell the difference between the iPad 2 and iPad 3? The differences could be around 85-90fps for the iPad 2, 135-140fps for the iPad 3. At those framerates, could a typical person clearly point out which one is which? From my experience, no.
You're right about how it's hard to do a real side-by-side comparison (real world) between the T3 and A5X because of many other factors however my point is to bring iPad/Apple cultists down to earth showcasing obvious flaws in the argument about which is better. The iPad can be a toy if that's how you choose to use it, while others like me use it as a tool. Tools come in the form of Apps and unfortunately Apple regulates what these apps can and can't do. I didn't buy the new iPad so that it can play Angry Birds and movies, I use it as an extension of my laptop for work and it's of no use to me if I can't find the right apps for it. Fair enough?
Your info on double the cores equalling double the performance is not wrong, but how do benchmarks not tell everything we need to know? The GL test shows offscreen performance, as that's the easiest way to get both machines on the same resolution. Since the iPad has the ability to render to intermediate resolutions even while on-screen (Infinity blade runs at 1.4x iPad 2 resolution for example), it should a valid test.
Honest question, when comparing CPU and GPU speeds, what part of the story is a benchmark like this missing?
In your example, no, I do not think a difference could be discerned. If I spent a few thousand on a new PC gaming rig, and ran the same test (same game at the same resolution with the same graphical effects on) going from 90fps to 140fps, I wouldn't see a difference there either. That's not the use-case.
The first thing I would do is bump the resolution up, increase the graphical effects until I had a stable 60fps in-game (maybe 30fps depending on the game), equalize the old PC to the same fps, then compare. This is the equivalent of running a Retina app vs a Non-Retina app.
If the devs decided to run Infinity Blade at iPad 2 resolution on the new iPad, they could theoretically increase performance by 70%. Not by simply increasing FPS, but by increasing draw distance and other graphical effects that most PC games allow us to change with a slider.
So I agree, it doesn't provide "twice the graphics", but it's a huge bump.
1) YES! (I didn't put in a #1 last reply so I'm not sure what you're responding too, but yes nonetheless.)
2) I'm aware they had something to do with it but I wasn't entirely sure that they actually funded the game, or did they just give them financial support to add extra's for Tegra devices?
3) Real Racing 2 runs at 2048x1536 w/ 4x AA (which is twice the AA it runs with on iPad 2), and Infinity Blade 2 is around 1430x1050, Modern Combat 3 which is a stunner is at 2048x1536.
4) It'll be a bit small though, that's all. With a 16:9 device the entire screen is the movie, whereas in the iPad 3 it'll be only about 70% of the screen, unless you zoom in.
Actually the iPad 3 does give twice the performance of the iPad 2, it just doesn't do it at 720p. The GLBenchmark tests are likely bottlenecked by the CPU or other factors, if you were to run GLBenchmark at 1080p or 2048x1536 for both the iPad 2 and iPad 3, you'd see a two fold difference between them.
Last edited:
A benchmark is just that, a benchmark or a point of reference based on a specific set of circumstances. If it was the only information we needed there would be no need for real world tests with real apps we all tend to use. Hope that clarifies things. While I do use benchmarks as part of my evaluation the reality is I didn't buy an iPad to run benchmarks.
Perhaps we're talking about 2 different things so I'll clarify my standpoint. I don't just look at any CPU and I could care less what those numbers are because it alone doesn't dictate the total result or the user experience. Software plays a huge part in how well that CPU/GPU is being used, and there's many layers of software from the bootstrap, kernel, OS, drivers and application that all must be harmonious in order to maximize its potential. It's like a 4 legged stool, it doesn't matter if 1 or 3 legs are perfect, all 4 needs to be balanced properly or else it won't be steady.
A gamer might be the extent to where the change could be noticeable but it needs to be under certain situations. I won't go into the details since you appear to know how gaming can be affected by changes in FPS performance.
Each application (I'm under the assumption that this is about a game) performs differently as quality of apps vary. Some games are written quite well and you can find consistent results from providing your unique settings in conjunction to your machine's hardware setup. However some aren't that way, some may require more or less depending on how that game is designed and written, whether deliberately or not.
no argument there.
Let's be serious for a moment, neither you nor I have the credentials to speculate where the bottlenecking is occurring other than taking wild guesses.
GLbenchmark is just 1 app, it involves a very specific set of circumstances. Just because the test results say it has twice the performance, the reader above already clarifies that under a real app/game designed to make use of the Retina display, the performance improvement is more like 1.4x over the iPad 2. This is what I'm talking about when I said you can't rely solely on benchmarks to tell the whole story.
I'm sure anyone here that has basic experience with graphics performance on a computer knows that when you start cranking up the resolution, having a faster GPU isn't enough. It isn't enough to simply throw in more graphics memory to ensure fast fill rates. Anyone who's ever overclocked video cards know that increasing the GPU clock speed doesn't yield nearly as much results as increasing the speed of the graphics memory clock.
Has the iPad 3 received enough of a bump in memory speed to go along with the performance of the A5X? I don't have the info comparing the changes, if any other than a 512MB bump between the iPad 2 and 3.
So what am I trying to get at? I hope I clearly illustrated that merely looking at CPU vs CPU or watching the results of GLbenchmark isn't enough, you need to look at the whole picture, not just bits and pieces of it.
I am a developer so I am reasonably familiar with what I'm talking about. I've not delved into any OpenGL ES games yet, but I still know enough to be able to take an educated guess.
Firstly, the poster to whom you replied said that Infinity Blade 2 is running on the iPad 3 at 1.4x the resolution as on the iPad 2, not that the iPad 3 is performing only 1.4x better than the iPad 2 -- 1.4x 1024x768 is not too far off 1536x1152 which is exactly double the pixels of 1024x768. You'd expect it to be able to do about that without losing any performance.
Secondly, I've looked a lot of things, but I do know that going from 2 SGX543's to 4 SGX543's will double performance, assuming perfect scaling. (Scaling should be quite high though.) How that applies to games varies by game, for example Infinity Blade 2 runs at roughly 1430x1050, but Real Racing 2 is running at 2048x1536 with more AA than it does on the iPad 2, and Modern Combat 3 is running at 2048x1536 as well.
From my understanding, the PowerVR SGX543MP2 on the iPad 2 is being bottlenecked by shaders, not by textures which it's very effective at doing, so games with heavy shaders like Infinity Blade 2 are really going to be capped to 1536x1152 without dropping in frame rate, but texture heavy games will run just fine at the higher resolution.
A software developer is nothing like the actual engineers involved with the development of the A5X or any CPU, let's not kid ourselves here. Don't take this the wrong way but I'm a developer as well and even I don't consider myself qualified to be remotely compared to the actual engineers of those CPU's.
Fair enough on the resolution remark however it still doesn't showcase double the performance where it's noticeable to any user. Even if there was a way where you could show FPS or other technical information while running the game, could you really notice that difference/improvement without it? I'll state based on my own experience that even diehard iPad 2 users have not been able to tell the difference between my iPad 3 and their iPad 2's.
Scaling performance relies on more than just having twice the cores of the previous generation GPU. As I've stated before all the graphics processing capability is nothing without a subsequent performance match from whatever memory it needs to use along with it.
Just to pick a nit here: the iPad DOES have HDMI, as does the iPhone 4 and 4S. Granted, you have to buy a dock connector accessory to make it happen, but I have it, use it often, and it works really well. Videos from netflix, hulu, iTunes and any other app works nicely, and it's great for mirroring the display for presentations or showing off websites.
There is also a USB port accessory, and a memory card accessory. They are limited in what they are opposed to be able to do of course, but, just like you need to root an Android device to get it working right, a jailbreak easily turns these connectors into something fully functional.
Last edited:
In terms of general speed, there shouldn't be a difference, the CPUs are the same. This has been confirmed in CPU based benchmarks. The real change from the A5 to the A5X is in the GPU, which is why only a gamer would notice the difference if they were measuring solely based on performance.
Of course the high resolution screen should give it away pretty clearly
In terms of general speed, there shouldn't be a difference, the CPUs are the same. This has been confirmed in CPU based benchmarks. The real change from the A5 to the A5X is in the GPU, which is why only a gamer would notice the difference if they were measuring solely based on performance.
Of course the high resolution screen should give it away pretty clearly
I'd have to agree. If this thread was about which platform offers the best experience with tablet gaming then I wouldn't hesitate to say that the iPad is second to none.
I posted this in another thread.
The Tegra 3 isn't actually all that good. If you look at the new HTC reviews, the One S (with the dual-core S4) consistently benches higher than the Tegra-based One X.
The only people trying to dispute this are Transformer Prime owners and nVidia. That's it.
The Tegra 3 isn't actually all that good. If you look at the new HTC reviews, the One S (with the dual-core S4) consistently benches higher than the Tegra-based One X.
The only people trying to dispute this are Transformer Prime owners and nVidia. That's it.
I wasn't comparing myself to the engineers who design these things, but you don't need that kind of knowledge to know what kind of performance you'll get from them, and the break down of how it all works.
It absolutely does, if the iPad 3 only had an SGX543MP2 it'd be running Infinity Blade 2 (at ~1430x1050) at about 17 FPS, which is very noticeable -- the extra two SGX543's allow for it to maintain normal performance.
There is a way to do that though, just plug it into XCode and profile it. (I'm not sure how much overhead there is when doing that, if any, but it shouldn't be anything drastic.)
I'd suggest you read the part of Anandtech's review of the iPad 3 about the A5X and memory, but from what I remember, Apple's moved the memory interfaces closer to the GPU, and they've added two of them. So they've now got a 128 bit memory interface (4 x 32 bit) compared to 64 bit (2 x 32 bit) for the iPad 2.