Re: The New Ipad (Ipad 3? what the hell is the name of this thing) Twice as many frames at twice the frame rate you say? What the hell are you talking about? Naturally something that renders at twice the rate renders twice as many frames, that's not a further accomplishment on top of the already doubled rate. Yes many of those measurements are almost 4x, so where are the real world tests that reflect that? You can't claim 4x general graphics performance vs the competition based off a couple of tests that don't reflect real world use situations of the gpu. http://eee.asus.com/eeepad/transformer-prime/features/ You'll notice they are referencing effects used in the particular game shown, which happens to not be optimised for iPad, as seen in the Laptopreview article. So yeah, their wording is misleading, this is marketing, but the performance part seems to relate more to the enhanced graphical experience rather than raw performance. That's not flat out saying our product is 4x better than the competition. Really? Twice the fillrate for 4 times the resolution, you can't do that math?
You've outdone yourself. In the first half of your response, you refute benchmarks of SoC's as being applicable because it's not "real world" performance, then in the second half of your response you refute "real world" performance (i.e., apps updated to retina display without loss of detail or frame rate) because it doesn't line up with your own theoretical math for the SoC. : I think most rational people accept benchmarks as a way to compare SoC performance for something like graphics, and I also think they understand that generalized statements like Apple's don't imply absolute 1:1 correlation to every benchmark imaginable, just as you seem to understand that Asus didn't LITERALLY mean that things like "realistic textures" were "previously unseen on a tablet", despite the fact that their marketing presentation reads that way.
Anandtech has supplied their own explanation as to why Apple used "4X" in their presentation, and it's a reasonable one: they think it was based on GFLOP numbers. ] Tegra 3 doesn't necessarily run at the same GPU clock speed in all devices, so the low end would, in fact, be 4X lower than the A5X when it comes to GFLOPS. The high end would be less than that...more like 2.7X lower. So, in the end, the A5X theoretical performance is still quite good even at half the GPU clock speed of the Tegra 3. http://www.anandtech.com/show/5663/analysis-of-the-new-apple-ipad
Except that fillrate is directly linked to resolution, so yes, in a case where you want to increase the resolution 4x it is pretty much your major limiting factor in real world applications. As for real world apps being updated, well if we don't just take their word for it that they're running at the full res and just as fast as they did on the iPad2, then you'll find either one of two things: 1. the app's ran at an acceptable speed on the iPad 1, so at the higher res the frame rates are pretty similar. 2. The app's don't run at native resolution (higher, but not native), and there is limited improvement that can be made in other areas (textures, polygon detail etc) to keep frame rates acceptable. See example: http://www.theverge.com/2012/3/15/2876161/infinity-blade-ii-update-ipad-retina-display Update: As pretty as that 2048 x 1536 screenshot might look, Chair tells us that it isn't quite running at the full native resolution of the iPad 3. "It has received a nice boost," a company rep tells us, but it sounds like that's in conjunction with some kind of upscaling as well. We'd welcome such a "boost" in all games from here on out. All other games I'm finding run just like they do on the iPad 1 by the looks of things, so it's not surprising that they can run great at 4 times the res. On another note, in the comments section people are doing some good comparisons of their own using racing "sims" and concluding exactly what I have earlier in the thread: To simplify, the iPad 2 outputs less polygons than even the PS2 version, but makes use of a whole host of new, hardware-based tricks that have been developed in the 13+ years since the PS2 came out. iPad2/3 are slightly behind last gen when rendering at their native (higher) resolutions, but make use of newer techniques to make games look much more stunning, even when the overall detail is so low. That's all well and good, but when you're comparing to the competition, you'd want to make sure your comparison is valid for real world results. Especially when you word it so generally. It's reasonable to state a theoretical peak compute number as general "graphics performance"? And I'm the one flip-flopping here? HA! Anand couldn't recreate the results in real world tests so has concluded that they must have meant a theoretical number. Great, doesn't this line up exactly with what I said earlier in the discussion as something other companies NEVER do? AMD have been well ahead of Nvidia for years in FLOPs, but you don't see them stating in their press event marketing slides anything other than exactly that, more FLOPs. They wouldn't dare try and pass that off as general graphics performance, because no one other than Apple could ever dream of trying to get away with such purposely misleading crap.
I tried one today and the screen is damn impressive. There's something to be said for being able to watch 1080p content on a tablet. Good job Apple and Samsung. I'm hoping this means we'll start to see insanely high resolution screens become common on phones and laptops before too long as well. Although the back did get a bit hot under load as everyone has said, but nothing too bad.
I really want to see these sort of resolutions in all devices also. Desktops/Laptops need native resolution scaling of assets though, and they also need to be designed with these resolutions in mind. Using a 27" 2560x1440 monitor in windows is pretty limiting right now, and IMO even worse on OSX when using the 27" iMac at work simply because of design choices. Mouse acceleration sucks in OSX, so simply getting around a screen that size is fairly bad and every window you open defaults to this tiny default size. Resizing windows once doesn't seem to let the OS know that you want all new windows to be this size like it does in Win7. OSX really is horrible in it's management of multiple finder windows - which is odd considering the massive focus on drag and drop. On tablets, sure, the iPad screen really is impressive. I hope Win8 supports 2560x1600 for tablets as I think it's really going to be a fantastic OS for tablets. It looks like there will be Android devices with such a res but I'm not too excited about Android tablets TBH.
Not really. You claimed that Apple was being "misleading" by saying the A5X SoC had 4X the graphics performance of the Tegra 3, and Anandtech essentially confirmed that they weren't due to the GFLOP numbers lining up EXACTLY to Apple's slide...1X for Tegra 3 at 333 Mhz, 2X for A5 at 250 Mhz, 4X for A5X at 250 Mhz. Apple was making a comparison of the SoC's themselves, and it's standard for SoC comparisons to involve theoretical performance. Apple didn't make any claims that "4X" translated 1:1 into real world performance. That's simply your own straw man argument. As for the graphics being upscaled, that was never disputed. Current generation consoles upscale graphics to 1080p. What WAS disputed, and what turns out to also be false, is the "iPad 1.5" comment, where you implied that users would have to accept a downgrade in detail level from iPad 2 apps in order to have them run at retina display quality.
Anand didn't have exact GFLOPs numbers as they didn't know the clock speeds. He simply figured that the chart must have been referring to GFLOPs and extrapolated the clock speeds from that info. Then, he states, For most real world gaming workloads I do believe that the A5 is faster than Tegra 3, but the advantage is unlikely to be 2x at non-retinadisplay resolutions. The same applies to the A5X vs. Tegra 3 comparison. I fully expect there to be a significant performance gap at the same resolution, but I doubt it is 4x in a game. Obviously not stating that the chart is misleading no, but I think it's pretty clear by this point that the chart does not represent real world performance, neither in one select real world test or many. Theoretical performance is fine to state when labeled as such. When stated as something else then it is clearly misleading. They didn't have to make any claims that the 4x figure translated into real world figures, when you don't outline any facts what so ever you can rely on people less in the know to draw the (wrong) conclusion themselves. In other words, they were purposely vague as to mislead people. It worked - do you see anyone saying "Wow, iPad3 is 4x faster than tegra 3 in theory!" besides experienced tech reporters? No, because they haven't done the research to clarify the claim and nor should they be expected to. There's no straw man argument about what they apparently did say, my argument is about what they purposely didn't say - something which no other company in a similar field would dare leave out. My comment was directly related to the scenario of running games at the higher resolution. If you want to take my comment out of context then good luck to you but obviously I won't indulge that. The comment was more in regard to your claim of tablets approaching current gen HD consoles. So the comparison was made between a theoretical app using all of the iPad 2's potential vs the same app at a higher resolution on the iPad 3. It was intended to show that progress would be stagnated due to the resolution bump, or on the other hand, they wouldn't run games at native resolution. Your claim of iPad 2 app's running just fine on iPad 3 at native res is unproven. So far all I have seen is app's that also run fine on the original iPad or do not run at native res. So the statement still stands, at native resolution, the iPad 3 would struggle to run apps making full use of the iPad 2, at least without sacrificing frame rate or something else.
In a fairly separate argument now, I'll revisit the following, This is the kind of crap I'm talking about. You take marketing claims out of context and then use real world tests to extrapolate false data about where future hardware will be. One second what a marketing guy says means everything and theory is BS, then it's all real world benchmarks, then non-real world benchmarks suddenly matter just as much, then theoretical numbers are fine and don't need to be stated as theoretical. It's utter nonsense, you keep flip-flopping all over the place and combining different sets of data, theory and claims to back up whatever end result you desire. In the above example you took a vague marketing claim about iPad being roughly 1/3rd the performance of the Xbox 360 CPU. So you compare a system to one part of another system. Then claim by some benchmarks that the iPad 2 is already delivering 3x the performance of the iPad so it must be close to the whole Xbox 360 system, and the iPad 3 is double that, so... it's faster already? But hold on now, the complete xbox 360 system has a compute potential of 355Gflops, while the iPad 3 complete system can only do 34Gflops. Since you've made it clear that you now accept FLOPs as an acceptable form of measuring general "Graphics Performance", do MS have to make a slide stating that Xbox 360 has over 10x the "Graphics Performance" of the new iPad before you will accept the Xbox 360 as a far superior system? Does Mark Rein have to say the Xbox 360 has over 355 times the performance of a single CPU core in the new iPad? This seems to be the only way you will accept anything as truth - too bad the backwards science you partake on after finding these "truths" is insanely flawed.
It was never presented in the context of "real world performance" for anything other than driving the pixel density of the display. If you watch the segment where the 1X-2X-4X slide is shown, Phil Schiller never mentions apps or software or games. He doesn't make any claims about that type of performance. It's entirely about the resolution of the display and the need for an SoC that is more powerful than the A5. That's the sum total of the context for that slide. See above...the actual context of the slide was not related to software, which is what you're talking about when you bring up "real world performance". That's why it's a straw man. Apple never claimed every app available from every developer would run at native resolution and neither did I. It's not a point that's being argued. Books? Magazines? Comics? Photos? 2D games? Non-gaming apps? There will be plenty of applications that run at native resolution just fine. I'm sure many 3D games will require upscaling, but again...that's par for the course already with current gen consoles running 3D games at a lower display density than the iPad, and you obviously consider those to be significantly more powerful for graphics.
Because driving the higher pixel density of the display has nothing to do with real world performance? What are they stating then, pixel fillrate? No one seems to have any idea, all they know is some spectacular generalized claim. If he's merely trying to innocently demonstrate a point about how they haven't been able to drive such a display until now, then why bring the competition into it? An innocent guise doesn't make such comparisons fair play, especially when we all know what the general net tech community brought away from the slide. It provoked a competitors product yet was extremely vague. I don't know how that isn't misleading. Right, because driving more pixels in actual software was never the goal... they just wanted it to be theoretically possible... I didn't think I had to elaborate beyond saying "app's that ran fine on the original iPad", but apparently I should have. I'm talking about 3D games that push the hardware potential towards the holy grail of being a console replacement. Current gen consoles only prove my point. If you want a modern gaming experience on such limited hardware you have to sacrifice resolution and more often than not frame rate. If you're happy with last generation level games, you may be able to bump up the resolution a fair way or increase the frame rate. All current iOS games are very much comparable to last gen console games with a few more modern techniques thrown because of being able to take advantage of newer API's and broader dev experience. Gameplay is the biggest differentiation between the two platforms though, without the storage space or general hardware ability there's just too much you cant do. Unfortunately too many devs aren't making proper use of this on current consoles, but I feel as if it is something that will really start to bloom next gen - and tablets will only be further behind.
Do you remember posting this earlier? You're quoting unidentified people from the comments section of another web site and literally treating it like it proves something, but then you're upset that I've quoted Epic (who developed games for both the 360 and iOS) regarding iPad vs. 360, and then drew some of my own conclusions based on how much better the SoC was for the iPad 2 vs. original iPad?
It was a graphical comparison of games, so yes it's subjective and I never meant for it to be anything other than that. I just pointed out that I agree with what they are saying because the visual evidence supported it. I guess I should have posted the comparison pics also. It was merely a side note, I wasn't trying to say these anonymous people know more than game developers, that would be stupid, merely that I agree with their graphical analysis on a purely subjective basis. This was more to back up why I think the games look like last gen games, and to point out that I'm not the only person crazy enough to have this opinion.
Again, Anandtech has supplied the reason: the GFLOP numbers are basically identical to the 1X-2X-4X comparison when compared to the low end GPU clock speed of the Tegra 3 (which is still a higher GPU clock speed than the A5X). And again, using theoretical performance to make comparisons between SoCs is not misleading. Apple put it in the context of the display density needing a more powerful SoC, and didn't reference it again during the rest of the presentation. It was a single slide during a keynote that lasted close to 90 minutes. It's not being used in the marketing material on their web site, and it's not appearing in their commercials. I think the real reason you're upset about it is because it bothers you that Apple has a more powerful chipset than most of their competitors. It turns the old stereotype about Apple products vs. other tech companies on it's head. Games like Dead Space HD and Real Racing 2 HD already looked significantly better than last gen console games on the iPad 2. For example, I don't think you're going to find many people saying that the GTA III port to iOS is one of the best looking games on the platform. I think it's fair to say that the iPad 2 was already falling somewhere between last gen consoles and current gen consoles. The new iPad doesn't represent as big of a jump in the SoC as iPad 2 vs. the original iPad, but it's still likely to increase the quality. I don't think most developers were even close to maxing the iPad 2 games after only a year on the market.
And again, when you don't state it as theoretical (or define it in any way for that matter) and compare to the competition, it is misleading. Purposely? Well, it spread over the internet and seemed to get a pretty big reaction for an innocent slide, one of many in such a long presentation. Apple love to have self-righteous little digs at the competition. Do you think they legally could use it for actual marketing? I don't. Do you think the people they directly market to care about tech comparisons? I don't. So no, why would they use it in actual official marketing. Hitting the enthusiast core with an almost viral slide on the other hand, that sets off the halo effect in the tech enthusiast world - that's a market usually unswayed by regular marketing. Thanks for the insight, Dr. Phil. Dead Space HD and Real Racing 2 HD have pretty low poly models for their genres, definitely in the realm of last gen, in-fact lower than last gen for their respected genres. Textures range from worse to equal to better, but never approaching current gen. Resolution, effects and the like, yep, they're better, that's where the newer part I've been talking about comes into it. Basically, in the core ways (polygon count and texture resolution) even the best iOS games are very much last gen. Again, newer API's and dev experience allow for nice effects on top of the last gen core graphics - but when gameplay elements that requires processing power or larger storage space comes into play, tablets are going to be well behind for a long time. For me, that is the real exciting thing about what new hardware can bring to gaming. Going back for a minute to the discussion of the Samaritan demo. Mark Rein also stated that it would require 2.5TFLOPs of compute potential to run this demo. Clearly such a statement is very misleading as it's just marketing crap talk to sell some nvidia video cards, but since you take this crap talk so seriously, how does that gel with his other claim of the demo running on an iPad in a few years? Is an iPad going to be almost 100x more powerful in a couple of years? It is this sort of contradiction that makes your supposed evidence and reasoning ridiculous. You can't see these stupid claims for what they are and get caught up in the hype, and make further stupid conclusions based off these original fallacies.
When you actually watch the presentation, rather than simply look at the slide without context, it doesn't give you the impression of anything other than a general comparison of the chipsets. Do they literally say it's theoretical? No. Do they literally say that it's not theoretical? No. Which means you have to look at it in the context of the presentation to infer the meaning, and the context has nothing to do with software. The context is "we needed a more powerful SoC to drive the display". The fact that they don't continue mentioning 4X throughout the presentation is the clincher. It's not meant to relate 1:1 to software. They don't claim the games are 4X faster or the OS runs 4X faster at any point of the keynote. I'm sure they could use it for their central marketing if they thought it was that important. And Apple does include a tech specs section in their marketing on the web site, so the idea that they think all their customers have no interest in that is not correct. According to Epic's mobile design tips, the characters in the original Infinity Blade generally have 8000 polygons. Are there last gen games that have more polygons for a character? Yes (RE4: 10,000 polys for Leon), but most didn't (Halo: 2,000 polys for Masterchief), and there are also current gen games that have characters with very similar poly counts (GTA IV: 8,000 - 10,000 for story characters). So as far as polygons go, even original iPad games like Infinity Blade had pretty good poly counts. It's obviously going to vary based on developer and game type. I also just visited an Anandtech forum that had people verifying that Real Racing II HD and Galaxy on Fire II HD are both running natively at 2048 x 1536 with their new iPad updates.
I really can't understand where your thinking is on this. They brought in a competitors product when they could have clearly used the iPad 1. It was a jab at the competition and when you do that, you'd really want to outline exactly what metric for comparison you are using. Yet they did no such thing. It was a cheap shot, and pretty much no other company would get away with this sort of crap. The context seems very clear to me, they're stating SoC graphical power required to run current level content at such resolutions. Which is exactly what you are saying, yet, I don't see how that changes anything. They willingly chose to drag their competitors product through the mud, in a vague, misleading way, because they did not specify whether it was theoretical or not. If it wasn't that important to them then why include a competitors product in the first place? An iPad 1 would have made more sense and been more fitting. It's pretty hilarious how much you are downplaying the claim given the reaction it got on the net. It's seems to me that such a situation was the only place they felt they could get away with making such a claim, so they went for it. Way to switch between genres. Comparing a beat-em-up to an open world sandbox for poly count, really? Ninja Gaiden/2/Black on the original Xbox is a very comparable game to Infinity Blade, and it had character models of 10-15,000 polygons. I've seen Dead space HD in action and I found the polygon count and textures really struck me as the major difference between it and current gen games. To me, it really looked like an Xbox game. But like I keep saying, tablets are ahead of last gen consoles in many ways, it's just the core building blocks mean so much, and they are very much last gen. Like you said earlier, developers obviously haven't made the most of iPad2 tech just yet. My example always has been of a theoretical game that pushed the hardware to its limits. Never the less, it's very impressive if these games actually are running at this res, but at the end of the day, you're still only getting what the iPad 2 was already delivering (after such a short time too) just at a higher res, so progress is still stagnated. When games that really push the hardware like Infinity Blade 2 can't run at native res, I think that says it all - the new iPad can't run games that made great use of the iPad 2's hardware at native res. Any game it is running at native res was basically an iPad 1.5 game to begin with (not using the most of the hardware).
I saw an ad on TV this past weekend where a Dodge SUV is towing an endless line of various vehicles (motorcycles, ATVs, jet ski's, a boat) and at the very end of it is a Ford SUV that's also being towed. All they say in the ad is that the Dodge SUV has "best in class towing" vs. Ford. That's a national TV campaign. They don't show any charts or graphs comparing the specifics of the engines/drive trains or the towing tests performed or what the "real world" towing difference between Dodge/Ford actually is. In reality, Apple's slide was not the shocking affront you make it out to be. It's very tame, and it's also backed up by the GFLOP numbers, which is not an unusual comparison for SoCs. Very few last gen games had polygon counts that high, and I've seen iOS developers talking about the A5 SoC being able to handle 15 on-screen characters with 15,000 polygons each...basically that polygons aren't a problem with the A5. The A5 was a H-U-G-E jump forward compared to the A4, and the A4 was the high end for the original Infinity Blade. The A5 was most likely in-between last gen/current gen consoles, and the A5X puts it even closer to the current gen. Real Racing 2 and Galaxy on Fire II were developed to run on the A4 SoC. All of the updates for the iPad 2 and new iPad are basically add-ons, rather than representative of a ground-up effort. There's no stagnation. In reality, neither of those games is probably even close to maximized for the A5, which was a 7X jump from the A4.
I've been saying all along that the only apps the new iPad will be running at the native res are ones that ran just fine (ie, designed) for the original iPad. You disputed that by saying "here's some iPad 2 apps running at native res on the new iPad", then turn around and claim "oh, but they were designed for the A4 used in the original iPad so new apps will look even better!". So basically we're back were we started. The new iPad can't keep up with its predecessor when running at the new resolution....hence progress graphics (besides IQ) has stagnated at best. Nice journey we went on there. Your logic is really strange. You conclude that the A4 was already delivering games solidly midway between current and last generation consoles, and then say the A5 is 7x faster than that. So how much faster do you actually think next gen consoles are than the last? General specs average to around 8x, while compute potential is around 50x higher. Are you attempting to once again mix your subjective observations with marketing claims and benchmarks to come up with a ball park figure? You're backing up and observation with a measurement that discredits the original observation. Every measurement states that consoles are only around 8-10x faster than their predecessors on average, so if the A4 was half way between, and then 7x faster than that, the A5 would already be faster, and the A5x probably end up faster than the WiiU! The only measurement that would fit here is compute potential, but we already have numbers contradicting that, the Xbox 360 is 10x faster than the A5x, and 20x faster than the A5. Do you see the backwards science now. You establish a position of comparison with a bias subjective observation or misleading marketing claim, then back up your case for future performance increases based on actual measurements. The problem is that the initial base info is not a valid form of proof, and real world results even contradict it.