No, you keep saying it is physically impossible. It has been shown that the A6 GPU is as fast as the A5X and the iPhone 5 despite a much smaller battery lasts 10 hours.
So the delta between the iPhone 5 and a retina iPad Mini power usage wise is a larger panel and perhaps a speed bump. It should run as fast as the iPad 3 and maintain the same dimensions and weight.
How do I know this? Because the iPhone 5 is powerful, thin, light and has good battery life.
You keep using that word. I don't think it means what you think it means.
Fact: the iPhone 5 lasts 10 hours.
Fact: the iPhone 5 is thin and light.
Fact: the iPhone 5 CPU is faster than the one in the iPad 3
Fact: the iPhone 5 GPU is as fast as the one in the iPad 3 according to Anand
Conclusion: an A6 based mini will be as fast as the iPad 3 with the same power requirements as the iPhone 5 scaled to the larger 7.85" display. The increase in display size is offset by the larger space for the battery resulting in a similar 10 hour run time and size and weight as the current mini.
Your conclusion is erroneous. You've noted that the A6 has the same GPU performance as the A5X once it's been adjusted. This has no barring on the power consumption needed for all the components to power a 2048x1536 display on the iPad mini. You simply can't scale it the way you are attempting to by looking at one adjusted benchmark.
Did you even look at the memory bandwidth differences between the A6 and A6X chips? Did you consider the power saving from the 45nm A5 to the 32nm A6?
There is a reason Apple isn't using the A6 in the iPad 4 and it's the same reason why the iPad mini with a 2048x1536 display can't use the A6 chip and have the same performance as the iPad 4.... and that's before even considering the power usage and cost for those miniaturized display components and GPU.
Let me come at this another way. You have only considered a single adjusted GPU benchmark in your argument. Look at the GPU in the iPad 2 and iPad 3. Note that the GPU in the iPad 3 is much better than the GPU in the iPad 2. Yet, you're argument states that because it's better it should not have only not gotten a much larger battery that is heavier and caused the casing to be thicker, but made that overall device smaller. Why didn't that happen with a more power efficient GPU? Simple, the GPU had to push 4x as many pixels 'and the GPU was not the only component that was affected by YoY change. Per pixel power usage was reduced, but we're talking about 4x as many so unless the device as a whole has advanced to a point that it's using 1/4th the amount of power to push each pixel then you're going to use more more for a given duration and use.
Yet the benchmarks indicates it can and Anandtech states the performance of the two GPUs are "identical" when normalized for resolution.
So you are once again ignoring facts because they disagree with your opinion.
Show evidence that the A6 can't handle 2047x1536.
From AnandTech's detaild review of the iPad mini from the page titled A Retina mini? which was completed 3 months after the iPhone 5 review:
With the mini's display using a 1024 x 768 resolution, this option would give it a 7.85-inch 2048 x 1536 panel. That would be the same resolution as the iPad 3/4, but in a much smaller display giving it a pixel density of 326 PPI (vs ~263 for the iPad 3/4). Apple could do this, but it would then need to make all of the same changes it made in going to the iPad with Retina Display, primarily the introduction of a larger battery and much larger SoC. The bigger battery is needed to drive the more powerful backlight, and the X-series of SoCs is needed to actually render the UI and games at such a high resolution. Both of these things would increase the size and cost of the mini, which would make it distinctly un-mini.
Note: This does not contradict what you previously quoted.
Yet the benchmarks indicates it can and Anandtech states the performance of the two GPUs are "identical" when normalized for resolution.
I'm s… what? You're saying that this:
will work identically at the same resolution. Despite Anandtech'sbenchmarks showing the A6X outperforming the A6 in every single category, since it has a better GPU, despite the higher resolution. Where in your world the A6 would have to be doing better in every single category to then be able to bench identically at a higher resolution.
From AnandTech's detaild review of the iPad mini from the page titled A Retina mini? which was completed 3 months after the iPhone 5 review:
Note: This does not contradict what you previously quoted.
Look at the Egypt benchmarks in the review you linked and you'll see that the A6 is comparable to the A5X.
To get iPad 4 performance you need the A6X. To get iPad 3 performance you don't.
Some will argue that the iPad 3 was a little underpowered but that's neither here nor there. The benchmarks clearly show that the A6 is capable enough when compared to the iPad 3 as a baseline.
This is clearly intuitive given its a generation newer with a process node advantage, a higher clock and faster memory subsystem.
The primary architectural advantage that the A5X enjoys is wider (but slower) memory bandwidth and an extra GPU core.
Something the faster clock rate and better memory controller makes up for...again AS CAN BE SEEN IN THE BENCHMARK RESULTS. The A6 and A5X are within 10% of each other whether the A6 is ahead or the A5X is ahead. In the CPU benchmarks there's no contest. Plus the A6 can likely be clocked a little higher in a mini vs the phone.
An A6 iPad Mini would be on par with the iPad 3 in 3D performance and way better everywhere else.
will work identically at the same resolution. Despite Anandtech'sbenchmarks showing the A6X outperforming the A6 in every single category, since it has a better GPU, despite the higher resolution. Where in your world the A6 would have to be doing better in every single category to then be able to bench identically at a higher resolution.
Not the A6 and A6X but the A6 and A5X as is clearly stated in each post I've made.
But you aren't actually reading to understand but reading to "gotcha" so I guess that's par for the course.
Look at the Egypt benchmarks in the review you linked and you'll see that the A6 is comparable to the A5X.
To get iPad 4 performance you need the A6X. To get iPad 3 performance you don't.
Some will argue that the iPad 3 was a little underpowered but that's neither here nor there. The benchmarks clearly show that the A6 is capable enough when compared to the iPad 3 as a baseline.
This is clearly intuitive given its a generation newer with a process node advantage, a higher clock and faster memory subsystem.
The primary architectural advantage that the A5X enjoys is wider (but slower) memory bandwidth and an extra GPU core.
Something the faster clock rate and better memory controller makes up for...again AS CAN BE SEEN IN THE BENCHMARK RESULTS. The A6 and A5X are within 10% of each other whether the A6 is ahead or the A5X is ahead. In the CPU benchmarks there's no contest. Plus the A6 can likely be clocked a little higher in a mini vs the phone.
An A6 iPad Mini would be on par with the iPad 3 in 3D performance and way better everywhere else.
Again, you are only looking at adjusted performance, not what is needed to push 4x as many pixels. As the the Anand review notes you will need a the X chip. You need the increased memory bandwidth. There is really no way to get around this.
Again, you haven't once considered power usage. Anand's review did.
I don't get why you insist on interpreting a clear comment from AnandTech incorrectly but then ignore a very clear comment on the iPad mini review from AnandTech.
I'm not sure what your agenda here is. I get it that you want the iPad mini to be Retina sooner rather than later. Everyone except Apple's competitors want that to happen but it's simply not going to unless there are other technological changes that take place. Rogue 6 and IGZO look to be the most promising solutions on the horizon to allow that to happen, which means a 2013 release unless they still keep the iPad mini a year behind with the tech which then would mean 2014 at the earliest.
Again, you are only looking at adjusted performance, not what is needed to push 4x as many pixels. As the the Anand review notes you will need a the X chip. You need the increased memory bandwidth. There is really no way to get around this.
Adjusted performance levels the playing field to show what the GPUs can do relative to each other. The indications is that 3 cores running at higher speeds leads to a similar fill rate and 3D performance as 4 slower cores. Yes, ideally you want to bench the 543MP3 driving a 2048x1536 but you can't so the normalized benchmarks are the best you can do. Fill rate and triangle texture rate are very similar. The GLBenchmark show that the A6 as faster than the A5X.
While the memory interface is wider on the A5X I believe it is slower. In any case, the changes made for the A6X to be 128 bits wide can be back ported to the A6 but I covered all that already. It's probably easier just to move to the 554MP2 which will have a similar die size based on what I remember of the A6X if that is required. Again this is existing 2012 technology.
So there are at least three ways to get around this: increase (memory) clock speed or make the A6 128 bits wide or move to a A6X with two graphic cores designed out for lower power and footprint
Quote:
Again, you haven't once considered power usage. Anand's review did.
I don't get why you insist on interpreting a clear comment from AnandTech incorrectly but then ignore a very clear comment on the iPad mini review from AnandTech.
I'm not sure what your agenda here is. I get it that you want the iPad mini to be Retina sooner rather than later. Everyone except Apple's competitors want that to happen but it's simply not going to unless there are other technological changes that take place. Rogue 6 and IGZO look to be the most promising solutions on the horizon to allow that to happen, which means a 2013 release unless they still keep the iPad mini a year behind with the tech which then would mean 2014 at the earliest.
I don't see how that's misinterpreted. The two GPUs perform very similarly to each other. The other "clear comment" doesn't have much in the way of actual data behind it unlike the one where the benchmarks clearly show similar performance between the two GPUs.
Since TS seems to like this kind of chart:
Mobile SoC GPU Comparison
PowerVR SGX 543MP3
PowerVR SGX 543MP4
SIMD Name
USSE2
USSE2
# of SIMDs
12
16
MADs per SIMD
4
4
Total MADs
48
64
GFLOPS @ 200MHz
19.2 GFLOPS
25.6 GFLOPS
GFLOPS As Shipped by Apple/ASUS
25.5 GFLOPS
25.6 GFLOPS
The agenda is you're attempting to browbeat people into your position with incorrect assertions that the iPad mini can't go retina until the next gen GPUs arrive because the last and current gen lacks sufficient power. As seen here (25.5 GFLOPS vs 25.6GFLOPS) the iPhone 5 GPU is sufficiently capable. The 554MP2 even more so.
As far as power requirements go, yes it's going to be under higher load and draw more power and yes there's more panel to power and so forth but the Mini's battery is 3 times the capacity as the iPhone 5 at 16.7 Whr vs 5.45 Whr. That's not 4 times but given that the CPU, RAM, Flash, WiFi, 4G, GPS etc power load doesn't increase over the iPhone 5 it's probably a wash in terms of getting to that 10 hour mark. Especially with the die shrink and more advanced CPU architecture the power per watt numbers favor the A6 over the A5X in terms of achieving iPad 3 levels of user experience.
Neither Rogue 6 nor IGZO is required. What was required was sufficient iPhone 5 panels and A6 processors, which given that the iPhone 5 has been supply constrained until recently were clearly NOT available for the iPad Mini launch.
Apple will indeed keep the iPad Mini a year or more behind because the newest 2013 technology will go into the iPhone 6 and iPad 5. Given that iPhone 5 technology is sufficient to built a retina Mini this falls into line with expectations with a 2013 retina iPad Mini using 2011/2012 technology (2011 if they opt for the 4S panels vs the 5).
Whether iPad 3 level performance is sufficient for Apple or if the A6 drops sufficiently in price could be debated BUT it is clear that it's not IMPOSSIBLE as you keep claiming. It is certainly possible and perhaps even likely. What's your agenda for insisting it can't happen despite data indicating it can?
This is tiring. I can't be more clear. AnandTech can't be more clear. Good luck with your Retina iPad mini with an A6 sans X in March. I truly hope both AnandTech and I are wrong.
And it can't get an A6 if you want it to be Retina. It has to be at least an A6X, but that likely won't do because the A6X is shown to use too require too much power which is why I've been saying that it needs to be Rogue 6, which doesn't look to be available until the 2nd half of 2013.
In case we're not clear this is the statement discussed. As I've shown the A6 is capable IF you are willing to settle for iPad 3 level GPU performance. The iPad Mini does NOT require a A6X class GPU in the iPad 4 to go retina. it needs a A5X class GPU to go retina.
In case we're not clear this is the statement discussed. As I've shown the A6 is capable IF you are willing to settle for iPad 3 level GPU performance. The iPad Mini does NOT require a A6X class GPU in the iPad 4 to go retina. it needs a A5X class GPU to go retina.
Will quote AnandTech one last time...
The A6 has a narrower memory interface compared to the A5x (64-bits vs. 128-bits), but that makes sense given the much lower display resolution (0.7MP vs. 3.1MP).
[...]
The A5X features a 128-bit wide memory interface while the A6 retains the same 64-bit wide interface as the standard A5. In memory bandwidth limited situations, the A5X will still be quicker but it's quite likely that at the iPhone 5's native resolution we won't see that happen.
In order to drive its 2048 x 1536 Retina Display at reasonable frame rates, Apple needed much more memory bandwidth than the standard A5 SoC could provide. The solution was, at the time, the world's highest bandwidth memory controller for a mobile ARM based SoC. With four 32-bit LPDDR2 channels paired up with LPDDR2-800 DRAM, the iPad 3's A5X SoC was capable of a theoretical 12.8GB/s of memory bandwidth. That's not much by high-end PC standards, but unheard of in an ARM based mobile device.
[...]
The A5X/A6X place the memory interface blocks (and perhaps the controllers themselves?) adjacent to the GPU, while the A5/A6 more tightly integrate the CPU and memory controller. This highlights a pretty substantial difference in priority between the A5/A6 and A5X/A6X SoCs. The latter really do prioritize memory bandwidth delivery to the GPU, and for good reason. The Retina Display equipped iPads have over 4x the number of pixels as the iPhone 5.
I am absolutely done with this conversation. I can lead a horse to water but I can't make it drink. Keep ignoring the facts and cling to the one misinterpretation that suits your wishes. I really do hope this comes this year but it won't be based on your conclusion if it does.
will work identically at the same resolution. Despite Anandtech'sbenchmarks showing the A6X outperforming the A6 in every single category, since it has a better GPU, despite the higher resolution. Where in your world the A6 would have to be doing better in every single category to then be able to bench identically at a higher resolution.
Except the one where you replied to my reference to A6 and A6X without correcting me, meaning that's exactly what you meant.
Uhh... Reach much? I've been following the debate and I think solip and nht both make valid points- and I'm learning a great deal from both. And both actually have pretty solid evidence- from the same source no less. One thing I have definitely read is that nht has said 5X the entire time. You quoted the A6 vs the A6X. You were wrong. End of conversation. You seriously are going to say "because you didn't correct me once in one of my posts means everything else you've said in every post is null". Dude...
I mean really- the A6X is better at graphics rendering than the A6? Wow TS- thanks for the enlightenment!
I mean really- the A6X is better at graphics rendering than the A6? Wow TS- thanks for the enlightenment!
NO. He said this:
Conclusion: an A6 based mini will be as fast as the iPad 3 with the same power requirements as the iPhone 5 scaled to the larger 7.85" display.
And from there I began my string. So even just before this when I offered him a chance to correct if he had actually meant A5X-A6 the entire time and he didn't, it wouldn't have mattered, as that's NOT what he was saying.
If the A6 could have handled 2048x1536 to Apple's taste, there would not be an A6X.
This is tiring. I can't be more clear. AnandTech can't be more clear. Good luck with your Retina iPad mini with an A6 sans X in March. I truly hope both AnandTech and I are wrong.
The fact is that your initial assertion is incorrect. You do not need an A6X. Period. If this statement was not true then the iPad 3 would not exist. No other data is required to show this premise to be incorrect. If the premise is incorrect then the conclusion is always suspect even if all the other logic is correct.
March is debatable given there's no strong indication of a 6 month product cycle but a/the 2013 Mini being retina is a reasonable bet. It might be a tweaked current A6. It might be an A6 with 554MP2. It might even be a 32nm (Samsung) or even 28nm (TSMC) A5X which are also valid possibilities.
It's tiresome to you because you simply wont accept you were wrong and I didn't just roll over.
Hell, you still haven't even accepted that either the 543MP3 or the 554MP2 are sufficient to drive a 2048x1536 display.
And from there I began my string. So even just before this when I offered him a chance to correct if he had actually meant A5X-A6 the entire time and he didn't, it wouldn't have mattered, as that's NOT what he was saying.
You did nothing of the sort (offered a chance). It has ALWAYS been my position that the A6 is equivalent to the A5X and NEVER my position that the A6 and A6X are equivalent.
Anyone actually reading my posts would see that very clearly and you caught me in a "gotcha". Kudos since you seem to like that sort of thing. Oooh...I said something incorrect because I misread your post.
Quote:
If the A6 could have handled 2048x1536 to Apple's taste, there would not be an A6X.
The A6 offers iPad 3 level performance which was to Apple's taste in early 2012 and arguably could be Apple's taste again in 2013 for the iPad Mini. Nice movement of the goal posts though from Solipsism's "impossible" to merely "Apple's taste".
The fact is that your initial assertion is incorrect. You do not need an A6X. Period. If this statement was not true then the iPad 3 would not exist. No other data is required to show this premise to be incorrect. If the premise is incorrect then the conclusion is always suspect even if all the other logic is correct.
March is debatable given there's no strong indication of a 6 month product cycle but a/the 2013 Mini being retina is a reasonable bet. It might be a tweaked current A6. It might be an A6 with 554MP2. It might even be a 32nm (Samsung) or even 28nm (TSMC) A5X which are also valid possibilities.
It's tiresome to you because you simply wont accept you were wrong and I didn't just roll over.
Hell, you still haven't even accepted that either the 543MP3 or the 554MP2 are sufficient to drive a 2048x1536 display.
Ah, there's the problem. You think by answering my questions and responding to the comments I quoted from AT that it would mean you'd be rolling over. I gave you the benefit of the doubt in thinking you were being sincere, rational, and scientific in your approach to commenting on this forum. It certainly makes sense now why you've ignored all the various aspects I and AT have clearly laid out in this thread but I am honestly disappointed you'd let your hubris of admitting you didn't consider all aspects of the device get in the way of knowledge. But hey, you didn't "roll over" when the facts conflicted with your preconceived notions . Well done¡
I am absolutely done with this conversation. I can lead a horse to water but I can't make it drink. Keep ignoring the facts and cling to the one misinterpretation that suits your wishes. I really do hope this comes this year but it won't be based on your conclusion if it does.
Which fact am I ignoring? The one where a A5X is sufficient and not an A6X? You also skipped over the possibility of a 32nm or 28nm A5X if you believe the stock A6 cannot work.
Yes, the X series is architected to favor memory bandwidth the GPU over the CPU. The question isn't whether the A6 can keep up with the A6X but whether it can keep up with the A5X and how much tweaking must be done to do so. With 8.5GB/s raw the memory bandwidth in the A6 that's significantly higher than in the A5 but still shy of the 12.8GB/s for the A5X. Is that fast enough? Eh. If not then move to 128 bit wide on the A6 with the 543MP3. Done.
We already know that the A6X has been laid out this way...so your primary objection is that the A6X in the iPad 4 requires too much power. That's probably true but there are still many possible options. Such as a A6X with 543MP3 or a 554MP2. So rather than the A6X used in the iPad 4 it's one scaled down to A5X level GPU performance and much lower power requirements for the GPU.
Whether you want to call this a scaled down A6X or a scaled up A6 is up to you. Either way the fact remains it's a viable option for a 2013 iPad Mini and not "impossible" as you keep asserting.
Tell you what, I'll accept your belief that the A6 will not perform adequately for 3D games because under heavy load the GPU will be memory starved so Apple won't go that route for the Mini even though I believe that for most things it'll be fine.
That still leaves 3 other options on the table: 32nm A6X with 543MP3, 32nm A6X with 554MP2, 28nm A5X all with equal or better performance to the 45nm A5X in the iPad 3 with reduced power requirements.
Also note that while the iPad Mini is rated at 10 hours it seem to last 13 hours in battery life tests. Apple can still claim 10 hours for the retina mini but with less headroom or even, horrors, come in a little shy at 9.8 hours like the iPad 2. So going from the current 783 minutes (iPad Mini) to 590 minutes (iPad 2) give quite a bit of wiggle room for reaching that 10 hour marketing goal.
Comments
Your conclusion is erroneous. You've noted that the A6 has the same GPU performance as the A5X once it's been adjusted. This has no barring on the power consumption needed for all the components to power a 2048x1536 display on the iPad mini. You simply can't scale it the way you are attempting to by looking at one adjusted benchmark.
Did you even look at the memory bandwidth differences between the A6 and A6X chips? Did you consider the power saving from the 45nm A5 to the 32nm A6?
There is a reason Apple isn't using the A6 in the iPad 4 and it's the same reason why the iPad mini with a 2048x1536 display can't use the A6 chip and have the same performance as the iPad 4.... and that's before even considering the power usage and cost for those miniaturized display components and GPU.
Let me come at this another way. You have only considered a single adjusted GPU benchmark in your argument. Look at the GPU in the iPad 2 and iPad 3. Note that the GPU in the iPad 3 is much better than the GPU in the iPad 2. Yet, you're argument states that because it's better it should not have only not gotten a much larger battery that is heavier and caused the casing to be thicker, but made that overall device smaller. Why didn't that happen with a more power efficient GPU? Simple, the GPU had to push 4x as many pixels 'and the GPU was not the only component that was affected by YoY change. Per pixel power usage was reduced, but we're talking about 4x as many so unless the device as a whole has advanced to a point that it's using 1/4th the amount of power to push each pixel then you're going to use more more for a given duration and use.
Yet the benchmarks indicates it can and Anandtech states the performance of the two GPUs are "identical" when normalized for resolution.
So you are once again ignoring facts because they disagree with your opinion.
Show evidence that the A6 can't handle 2047x1536.
From AnandTech's detaild review of the iPad mini from the page titled A Retina mini? which was completed 3 months after the iPhone 5 review:
Note: This does not contradict what you previously quoted.
Originally Posted by nht
Yet the benchmarks indicates it can and Anandtech states the performance of the two GPUs are "identical" when normalized for resolution.
I'm s… what? You're saying that this:
will work identically at the same resolution. Despite Anandtech's benchmarks showing the A6X outperforming the A6 in every single category, since it has a better GPU, despite the higher resolution. Where in your world the A6 would have to be doing better in every single category to then be able to bench identically at a higher resolution.
Look at the Egypt benchmarks in the review you linked and you'll see that the A6 is comparable to the A5X.
To get iPad 4 performance you need the A6X. To get iPad 3 performance you don't.
Some will argue that the iPad 3 was a little underpowered but that's neither here nor there. The benchmarks clearly show that the A6 is capable enough when compared to the iPad 3 as a baseline.
This is clearly intuitive given its a generation newer with a process node advantage, a higher clock and faster memory subsystem.
The primary architectural advantage that the A5X enjoys is wider (but slower) memory bandwidth and an extra GPU core.
Something the faster clock rate and better memory controller makes up for...again AS CAN BE SEEN IN THE BENCHMARK RESULTS. The A6 and A5X are within 10% of each other whether the A6 is ahead or the A5X is ahead. In the CPU benchmarks there's no contest. Plus the A6 can likely be clocked a little higher in a mini vs the phone.
An A6 iPad Mini would be on par with the iPad 3 in 3D performance and way better everywhere else.
Not the A6 and A6X but the A6 and A5X as is clearly stated in each post I've made.
But you aren't actually reading to understand but reading to "gotcha" so I guess that's par for the course.
Originally Posted by nht
Not the A6 and A6X but the A6 and A5X as is clearly stated in each post I've made.
Except the one where you replied to my reference to A6 and A6X without correcting me, meaning that's exactly what you meant.
Again, you are only looking at adjusted performance, not what is needed to push 4x as many pixels. As the the Anand review notes you will need a the X chip. You need the increased memory bandwidth. There is really no way to get around this.
Again, you haven't once considered power usage. Anand's review did.
I don't get why you insist on interpreting a clear comment from AnandTech incorrectly but then ignore a very clear comment on the iPad mini review from AnandTech.
I'm not sure what your agenda here is. I get it that you want the iPad mini to be Retina sooner rather than later. Everyone except Apple's competitors want that to happen but it's simply not going to unless there are other technological changes that take place. Rogue 6 and IGZO look to be the most promising solutions on the horizon to allow that to happen, which means a 2013 release unless they still keep the iPad mini a year behind with the tech which then would mean 2014 at the earliest.
Quote:
Originally Posted by Tallest Skil
Except the one where you replied to my reference to A6 and A6X without correcting me, meaning that's exactly what you meant.
I was on my iPhone so I missed that. If you wish to use that lame excuse then whatever but clearly, if you read my posts it's very clear.
Quote:
Originally Posted by SolipsismX
Again, you are only looking at adjusted performance, not what is needed to push 4x as many pixels. As the the Anand review notes you will need a the X chip. You need the increased memory bandwidth. There is really no way to get around this.
Adjusted performance levels the playing field to show what the GPUs can do relative to each other. The indications is that 3 cores running at higher speeds leads to a similar fill rate and 3D performance as 4 slower cores. Yes, ideally you want to bench the 543MP3 driving a 2048x1536 but you can't so the normalized benchmarks are the best you can do. Fill rate and triangle texture rate are very similar. The GLBenchmark show that the A6 as faster than the A5X.
While the memory interface is wider on the A5X I believe it is slower. In any case, the changes made for the A6X to be 128 bits wide can be back ported to the A6 but I covered all that already. It's probably easier just to move to the 554MP2 which will have a similar die size based on what I remember of the A6X if that is required. Again this is existing 2012 technology.
So there are at least three ways to get around this: increase (memory) clock speed or make the A6 128 bits wide or move to a A6X with two graphic cores designed out for lower power and footprint
Quote:
Again, you haven't once considered power usage. Anand's review did.
I don't get why you insist on interpreting a clear comment from AnandTech incorrectly but then ignore a very clear comment on the iPad mini review from AnandTech.
I'm not sure what your agenda here is. I get it that you want the iPad mini to be Retina sooner rather than later. Everyone except Apple's competitors want that to happen but it's simply not going to unless there are other technological changes that take place. Rogue 6 and IGZO look to be the most promising solutions on the horizon to allow that to happen, which means a 2013 release unless they still keep the iPad mini a year behind with the tech which then would mean 2014 at the earliest.
I don't see how that's misinterpreted. The two GPUs perform very similarly to each other. The other "clear comment" doesn't have much in the way of actual data behind it unlike the one where the benchmarks clearly show similar performance between the two GPUs.
Since TS seems to like this kind of chart:
Mobile SoC GPU Comparison
PowerVR SGX 543MP3
PowerVR SGX 543MP4
SIMD Name
USSE2
USSE2
# of SIMDs
12
16
MADs per SIMD
4
4
Total MADs
48
64
GFLOPS @ 200MHz
19.2 GFLOPS
25.6 GFLOPS
GFLOPS As Shipped by Apple/ASUS
25.5 GFLOPS
25.6 GFLOPS
The agenda is you're attempting to browbeat people into your position with incorrect assertions that the iPad mini can't go retina until the next gen GPUs arrive because the last and current gen lacks sufficient power. As seen here (25.5 GFLOPS vs 25.6GFLOPS) the iPhone 5 GPU is sufficiently capable. The 554MP2 even more so.
As far as power requirements go, yes it's going to be under higher load and draw more power and yes there's more panel to power and so forth but the Mini's battery is 3 times the capacity as the iPhone 5 at 16.7 Whr vs 5.45 Whr. That's not 4 times but given that the CPU, RAM, Flash, WiFi, 4G, GPS etc power load doesn't increase over the iPhone 5 it's probably a wash in terms of getting to that 10 hour mark. Especially with the die shrink and more advanced CPU architecture the power per watt numbers favor the A6 over the A5X in terms of achieving iPad 3 levels of user experience.
Neither Rogue 6 nor IGZO is required. What was required was sufficient iPhone 5 panels and A6 processors, which given that the iPhone 5 has been supply constrained until recently were clearly NOT available for the iPad Mini launch.
Apple will indeed keep the iPad Mini a year or more behind because the newest 2013 technology will go into the iPhone 6 and iPad 5. Given that iPhone 5 technology is sufficient to built a retina Mini this falls into line with expectations with a 2013 retina iPad Mini using 2011/2012 technology (2011 if they opt for the 4S panels vs the 5).
Whether iPad 3 level performance is sufficient for Apple or if the A6 drops sufficiently in price could be debated BUT it is clear that it's not IMPOSSIBLE as you keep claiming. It is certainly possible and perhaps even likely. What's your agenda for insisting it can't happen despite data indicating it can?
Quote:
Originally Posted by SolipsismX
And it can't get an A6 if you want it to be Retina. It has to be at least an A6X, but that likely won't do because the A6X is shown to use too require too much power which is why I've been saying that it needs to be Rogue 6, which doesn't look to be available until the 2nd half of 2013.
In case we're not clear this is the statement discussed. As I've shown the A6 is capable IF you are willing to settle for iPad 3 level GPU performance. The iPad Mini does NOT require a A6X class GPU in the iPad 4 to go retina. it needs a A5X class GPU to go retina.
Will quote AnandTech one last time...
I am absolutely done with this conversation. I can lead a horse to water but I can't make it drink. Keep ignoring the facts and cling to the one misinterpretation that suits your wishes. I really do hope this comes this year but it won't be based on your conclusion if it does.
Uhh... Reach much? I've been following the debate and I think solip and nht both make valid points- and I'm learning a great deal from both. And both actually have pretty solid evidence- from the same source no less. One thing I have definitely read is that nht has said 5X the entire time. You quoted the A6 vs the A6X. You were wrong. End of conversation. You seriously are going to say "because you didn't correct me once in one of my posts means everything else you've said in every post is null". Dude...
I mean really- the A6X is better at graphics rendering than the A6? Wow TS- thanks for the enlightenment!
Originally Posted by Andysol
I mean really- the A6X is better at graphics rendering than the A6? Wow TS- thanks for the enlightenment!
NO. He said this:
Conclusion: an A6 based mini will be as fast as the iPad 3 with the same power requirements as the iPhone 5 scaled to the larger 7.85" display.
And from there I began my string. So even just before this when I offered him a chance to correct if he had actually meant A5X-A6 the entire time and he didn't, it wouldn't have mattered, as that's NOT what he was saying.
If the A6 could have handled 2048x1536 to Apple's taste, there would not be an A6X.
Quote:
Originally Posted by SolipsismX
This is tiring. I can't be more clear. AnandTech can't be more clear. Good luck with your Retina iPad mini with an A6 sans X in March. I truly hope both AnandTech and I are wrong.
The fact is that your initial assertion is incorrect. You do not need an A6X. Period. If this statement was not true then the iPad 3 would not exist. No other data is required to show this premise to be incorrect. If the premise is incorrect then the conclusion is always suspect even if all the other logic is correct.
March is debatable given there's no strong indication of a 6 month product cycle but a/the 2013 Mini being retina is a reasonable bet. It might be a tweaked current A6. It might be an A6 with 554MP2. It might even be a 32nm (Samsung) or even 28nm (TSMC) A5X which are also valid possibilities.
It's tiresome to you because you simply wont accept you were wrong and I didn't just roll over.
Hell, you still haven't even accepted that either the 543MP3 or the 554MP2 are sufficient to drive a 2048x1536 display.
Quote:
Originally Posted by Tallest Skil
And from there I began my string. So even just before this when I offered him a chance to correct if he had actually meant A5X-A6 the entire time and he didn't, it wouldn't have mattered, as that's NOT what he was saying.
You did nothing of the sort (offered a chance). It has ALWAYS been my position that the A6 is equivalent to the A5X and NEVER my position that the A6 and A6X are equivalent.
Anyone actually reading my posts would see that very clearly and you caught me in a "gotcha". Kudos since you seem to like that sort of thing. Oooh...I said something incorrect because I misread your post.
Quote:
If the A6 could have handled 2048x1536 to Apple's taste, there would not be an A6X.
The A6 offers iPad 3 level performance which was to Apple's taste in early 2012 and arguably could be Apple's taste again in 2013 for the iPad Mini. Nice movement of the goal posts though from Solipsism's "impossible" to merely "Apple's taste".
Originally Posted by nht
You did nothing of the sort (offered a chance).
You know what I mean.
Nice movement of the goal posts though from Solipsism's "impossible" to merely "Apple's taste".
Hardly.
Ah, there's the problem. You think by answering my questions and responding to the comments I quoted from AT that it would mean you'd be rolling over. I gave you the benefit of the doubt in thinking you were being sincere, rational, and scientific in your approach to commenting on this forum. It certainly makes sense now why you've ignored all the various aspects I and AT have clearly laid out in this thread but I am honestly disappointed you'd let your hubris of admitting you didn't consider all aspects of the device get in the way of knowledge. But hey, you didn't "roll over" when the facts conflicted with your preconceived notions . Well done¡
Quote:
Originally Posted by SolipsismX
I am absolutely done with this conversation. I can lead a horse to water but I can't make it drink. Keep ignoring the facts and cling to the one misinterpretation that suits your wishes. I really do hope this comes this year but it won't be based on your conclusion if it does.
Which fact am I ignoring? The one where a A5X is sufficient and not an A6X? You also skipped over the possibility of a 32nm or 28nm A5X if you believe the stock A6 cannot work.
Yes, the X series is architected to favor memory bandwidth the GPU over the CPU. The question isn't whether the A6 can keep up with the A6X but whether it can keep up with the A5X and how much tweaking must be done to do so. With 8.5GB/s raw the memory bandwidth in the A6 that's significantly higher than in the A5 but still shy of the 12.8GB/s for the A5X. Is that fast enough? Eh. If not then move to 128 bit wide on the A6 with the 543MP3. Done.
We already know that the A6X has been laid out this way...so your primary objection is that the A6X in the iPad 4 requires too much power. That's probably true but there are still many possible options. Such as a A6X with 543MP3 or a 554MP2. So rather than the A6X used in the iPad 4 it's one scaled down to A5X level GPU performance and much lower power requirements for the GPU.
Whether you want to call this a scaled down A6X or a scaled up A6 is up to you. Either way the fact remains it's a viable option for a 2013 iPad Mini and not "impossible" as you keep asserting.
Tell you what, I'll accept your belief that the A6 will not perform adequately for 3D games because under heavy load the GPU will be memory starved so Apple won't go that route for the Mini even though I believe that for most things it'll be fine.
That still leaves 3 other options on the table: 32nm A6X with 543MP3, 32nm A6X with 554MP2, 28nm A5X all with equal or better performance to the 45nm A5X in the iPad 3 with reduced power requirements.
Also note that while the iPad Mini is rated at 10 hours it seem to last 13 hours in battery life tests. Apple can still claim 10 hours for the retina mini but with less headroom or even, horrors, come in a little shy at 9.8 hours like the iPad 2. So going from the current 783 minutes (iPad Mini) to 590 minutes (iPad 2) give quite a bit of wiggle room for reaching that 10 hour marketing goal.
http://www.tuaw.com/2012/12/05/ipad-ipad-mini-win-battery-life-shootout-among-tablets/