How is the screen resolution off? (Note that it'll report at 1024x768 regardless of whether it's got the retina display or not.)Well, 1GB RAM officially confirmed. Looks like loading up all those tabs in Safari won't be an issue anymore.
As far as GLBenchmark goes, some of the specs look off (screen resolution for instance).
And the benchmark seems insane. If it's true, they're saying the Apple A5X outperforms the A5 multiple times when it comes to rendering triangle. The A5X shows a whopping 92MTriangles/s vs A5's 19MTriangles/s. That's over 4x the performance with just a 2x increase in core count!
Something seems very off here.
In end effect, it wasn't an ideal solution to quadruple the pixel demand and try to deal with that by only doubling the pixel rate. That is very simple math, and something any reasonable person should be able to see.
Do you think the on screen results are being tested at 1024x768?It should show 2048 x 1536. GLBenchmark reports 960 x 640 for iPhone 4 and 4S.
Ah right... Anand used GLBenchmark 2.0.In any case, check the Triangle Test: Textured, Fragment Lit here:
http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro21&D=Apple+iPad+3&testgroup=lowlevel
Then compare it to iPad 2's results here:
http://www.anandtech.com/show/4216/...rmance-explored-powervr-sgx543mp2-benchmarked
Of note, though, is that the GLBenchmark results for iPad 2 on the site shows 42MTriangles:
http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro21&D=Apple+iPad+2&testgroup=lowlevel
So perhaps Anand used an older test.
Here's one for the Transformer Prime for kickers:
http://www.glbenchmark.com/phonedet...ad+Transformer+Prime+TF201&testgroup=lowlevel
It shows 23MTriangles even with the new test suite.
Ah cool, thanks.No, I don't think the texture tests are resolution-dependent. They might be memory-dependent, if anything, so the increase to 1GB RAM in iPad 3 might have helped.
That's not how it works at all. The iPhone 4, and 4S, and iPad 3, all report the base resolution internally when queried, and the scaling factor for those devices is two, and one for all non-retina devices.And the iPad 3's internal resolution is simply 2048 x 1536. Scaling factor is applied automagically by the OS when the app being used has a lower native resolution. That's how I see it. So technically, you can write a 1024 x 768 app and let the OS do the scaling, but your graphics won't be as sharp, or you can write a native 2048 x 1536 app and include 1024 x 768 assets so the OS knows to scale down.
What you say may be true, but I doubt Apple would release something with subpar performance.
Regardless of what they render the game at it'll look better though, won't it? At 1024x768 they'll have twice the amount of power to play with, and any resolution up to 2048x1536 will still look better than on the iPad 2 because of the pixel density right?
So what would you have done if you were Apple?
And the benchmark seems insane. If it's true, they're saying the Apple A5X outperforms the A5 multiple times when it comes to rendering triangle. The A5X shows a whopping 92MTriangles/s vs A5's 19MTriangles/s. That's over 4x the performance with just a 2x increase in core count!
Something seems very off here.
OLED, same GPU they went with. Push the fact that you have, *by far* the best contrast of any 10" tablet and the GPU power to push all sorts of new gaming goodness. Handle the resolution swap later. Now when it comes time for Apple to make the move to OLED it could be pushed back by years due to them setting the resolution bar so high for themselves. I think it was a poor short term and long term choice. Am I saying this will make the new iPad fail? Absolutely not in any way. I am saying they would have had a better product and better positioning if they had went a different route.
Though OLED offer significantly better blacks, IPS LCD screens can still offer decent competition simply by offering more accurate colors (if the smartphone market is enough to judge the tech).
In any case, the move from (high quality) IPS to OLED is an incremental upgrade at best. Apple chose the blow-your-pants-off visual fidelity jump.
They never said that though. According to Apple, they've got a GPU twice as powerful. Whether that's enough to handle the load, well we'll see soon enough. I define par as current performance of games on iPad 2, or 30 FPS, and subpar is 20-25 FPS or less.According to Apple, the iPad 3 will be half as fast as the iPad 2 when both are run at their native resolution. If that is sub par or not depends on how you define par.
Mmm, I'm not so sure about that. I want everything at the native resolution. Look at Infinity Blade: Dungeons and Sky Gamblers: Air Supremacy, both touted their graphics, and they do look great, but they also said that the retina display makes it look amazing, and that because of the retina display... such and such, etc., which implies it's running at a higher resolution, if not native. (The Sky Gamblers video I linked I found to be particularly impressive looking.)A great deal of this is going to come down to personal perspective, in the abstract I would say that any 3D game is likely to be best served running at 10x7 and using the extra power while anything 2D is going to be much better running at 20x15.
Keep in mind that the test he was comparing it to was an older version, the iPad 2 scores about 42,000 in that same test in GLBenchmark 2.1.2, and the iPad 3 scores about 94,000.PowerVR GPUs take all of the geometry data for a segment of the screen and map it out then cast 'rays' on it to see what is visible and what isn't so it knows exactly what it needs to draw. Normally you call this chunk of memory a 'bin', and it normally is the main limitation for TBR setups when compared to immediate mode renderers(TBR supporters will argue this, but it is the only *additional* step in the pipeline they have over IMR GPUs to in relative terms it *must* be the limiting factor). It is possible that they could have increased the amount of memory on die devoted to binning and ended up with a significant increase in geometric throughput due to having to recalculate vertice data multiple times for geometry that exceeded the edges of a bin test batch. I'm not saying this *is* the reason, just that seemingly huge jumps in geometric throughput can be seen on these types of GPUs because they have very different bottlenecks then what we are used to seeing(which is just raw computational intensity).
I thought SSAA (super scaling) was the only form of AA that actually rendered at multiples of the original resolution and then downscaled.Well, back to A5X, I believe that if A5 can do AAx2 at 1024x768, then technically, A5X should be able to do AAx4.
Considering FSAAx4 is just rendering at four times the resolution and scaling down, that may technically mean that A5X has just barely enough power to handle "certain" 3D content at 2048 x 1536 with about the same visual quality as A5.
But do correct me if that assumption is wrong.
Though OLED offer significantly better blacks, IPS LCD screens can still offer decent competition simply by offering more accurate colors (if the smartphone market is enough to judge the tech).
An OLED would have been a minor upgrade at best.
If there was a high resolution OLED screen available, Apple would have used it, but what you get is all the pentile devices that we have seen as of late.
They never said that though. According to Apple, they've got a GPU twice as powerful.
Is there even enough OLED production capacity to satisfy Apple's needs?
Considering FSAAx4 is just rendering at four times the resolution and scaling down, that may technically mean that A5X has just barely enough power to handle "certain" 3D content at 2048 x 1536 with about the same visual quality as A5.
I see, but the fill rate of the PowerVR SGX543MP2 is pretty high -- I'd doubt anything is coming near to taxing it right now.With four times the pixel requirements. For PowerVR GPUs more then anyone elses, that determines the load. They don't deal with objects that aren't displayed on the screen, their fillrate requirements are exceedingly linear.
I'm not sure if it's "free" but I don't think it's very intensive. Real Racing 2 for example runs without AA at 1080p at 30 FPS on an HDTV, if it had no impact or very little impact they could've left it on.PowerVR GPUs use MSAA and it is in essence 'free'. While they have the geometry binned they simply up the samples on the geometry by whatever factor they are using for AA and then sample additional sub pixels on those precise edges. 4x AA on either part should be 'free' to any developer worth their keyboard(just turn it on, it is free).
That is an issue of calibration which Apple would have under their control. The color gamut advantage of OLEDs combined with their contrast advantage puts them in an entirely different league then LCDs if the screen is calibrated to be accurate. LCDs have no chance of coming close, their technology is too primitive.
50 fold increase in contrast, or a four fold increase in pixel density. There are almost no measurable metrics of display quality that would make your comment anything buy laughable. TN displays are much closer to SIPS panels then SIPS panels are to OLED at all measurable quality standards. How many people would want a higher resolution TN panel over a lower resolution IPS display? Not too many that care about image quality- that comparison is *far* closer then the disparity between OLED and SIPS.
See a lot of 2048x1536 10" panels around right now do you?
The Samsung 7.7 is not pentile and has a higher PPI then the iPad 2 display- larger displays don't benefit(*) from a pentile arrangement nearly as much as their smaller counterparts.(*In terms of cost and brightness factors)
With four times the pixel requirements. For PowerVR GPUs more then anyone elses, that determines the load. They don't deal with objects that aren't displayed on the screen, their fillrate requirements are exceedingly linear.
Globally? Yes, a lot of it would need to be refocused. LG and Samsung are both ramping up production of 55" OLED panels- the production capabilities of OLED has increased a huge amount. Apple would have certainly had an impact on production in that range for panels of that type, but they already are doing precisely that with their 20x15 SIPS panel.
Photo professionals and the like use S-IPS, I see no reason to deny its usefulness in a mobile device especially if it enables us to enjoy the high resolution that is available to us now.
They use S-IPS because it's the least bad option available to then not because it is actually that great.
Saying that an S-IPS display is bad is further from the truth.
That's your opinion, it also doesn't make S-IPS bad.