Apple A5X SoC

[dagamer34]

Poly edges are sub pixel in accuracy. To make edge aliasing non visible to the human eye the best approach is to make the pixels smaller then the eye can discern on an individual basis. If you do this, you eliminate visible poly edge aliasing, unless of course you can't display the image at the native resolution. If you are seeing edge aliasing, the game isn't running at the native resolution.



Which metric would you like me to use? Compute power where desktop GPUs have been in the TFLOP range for years? Geometric throughput? Which metric? No matter which you decide on, the A5X is going to be utterily humiliated to an absurd degree. Having monster specs for a SoC doesn't mean it is in the league of even the weakest current desktop GPU.



Those benches won't change, they are primitive throughput- perhaps I should say those benches shouldn't change because they are primitive throughput.



I was quite clear long before we had any of the details on what matters in a quality display. Contrast was at the top of my list, well before we had any reviews of the new iPad. The reviews are in, the new iPad loses to the dollar store quality Kindle Fire in the most important metric. You can take the word of someone who can't tell the difference between bilinear and trilinear at a glance- based on your comments you wouldn't be able to either.

As far as performance- the benched back me up, across the board actually. You can keep your devotion going, the performance data spells out exactly what I said. With both at their native resolution the new iPad has inferior performance to the iPad2 in anything GPU intensive.

Umm... you're comparing graphics cards that run in a 300W envelope to an SoC that runs at 4W envelope. That is a 75x difference in TDP. Any performance comparisons are absurd at that point...

 

[BenSkywalker]

Umm... you're comparing graphics cards that run in a 300W envelope to an SoC that runs at 4W envelope.

I am not the one who started the comparisons, some truly ignorant people in this thread were trying to compare SoC GPUs to desktop parts, the comparison is shockingly dumb on any level. Unlike the people trying to act like they are comparable in performance, I'm not trying to say they *should* be compared- I was pointing out that they shouldn't be.

Unforunately, due to the extremely poor design choices made the die sizes are getting comparable, that is something that needs to be gotten under control or Apple's upcoming SoCs are going to be pushing Fermi size......

 

[MrX8503]

I was quite clear long before we had any of the details on what matters in a quality display. Contrast was at the top of my list, well before we had any reviews of the new iPad. The reviews are in, the new iPad loses to the dollar store quality Kindle Fire in the most important metric. You can take the word of someone who can't tell the difference between bilinear and trilinear at a glance- based on your comments you wouldn't be able to either.

As far as performance- the benched back me up, across the board actually. You can keep your devotion going, the performance data spells out exactly what I said. With both at their native resolution the new iPad has inferior performance to the iPad2 in anything GPU intensive.

Because everyone knows that contrast is the #1 measurement for display quality, lol. Didn't you complain about saturation levels? Apple fixed that and on top of that the pixel density has been stunning across the board for every review I've read. Can you show me reviews of the Kindle Fire display being better than the iPad3?

You said that games would look bad on the new iPad....Well it turns out that the games don't look bad at all. There's a minor performance decrease because of the extra pixels (which I didn't disagree with), but you seem to lump less performance and terrible performance into one ball.

I am not the one who started the comparisons, some truly ignorant people in this thread were trying to compare SoC GPUs to desktop parts, the comparison is shockingly dumb on any level. Unlike the people trying to act like they are comparable in performance, I'm not trying to say they *should* be compared- I was pointing out that they shouldn't be.

Unforunately, due to the extremely poor design choices made the die sizes are getting comparable, that is something that needs to be gotten under control or Apple's upcoming SoCs are going to be pushing Fermi size......

It was shockingly dumb for you to compare the two on any level. I called out the poster that the two have no comparisons at all, but you say the A5X GPU sucks because it can't compete with a desktop counterpart. Really?

The MP4 GPU is a performance beast for a tablet and so was the A5 chip before it. Haters gonna hate.

 

[runawayprisoner]

I am not the one who started the comparisons, some truly ignorant people in this thread were trying to compare SoC GPUs to desktop parts, the comparison is shockingly dumb on any level. Unlike the people trying to act like they are comparable in performance, I'm not trying to say they *should* be compared- I was pointing out that they shouldn't be.

And who might that be? I have just read every single post in this thread and I couldn't find said posts comparing the A5X to desktop parts. You were the one that started the A5X vs desktop part comparison as far as I can see.

 

[smartpatrol]

This thread is hilarious. All the Android fanboys are whining that iPad has 4x the resolution but only double the GPU performance.

I can't wait until then next-gen Galaxy Tab comes out with a higher resolution than iPad and a weaker GPU. These very same people will be jizzing their pants over it.

 

[stormkroe]

And who might that be? I have just read every single post in this thread and I couldn't find said posts comparing the A5X to desktop parts. You were the one that started the A5X vs desktop part comparison as far as I can see.

Nope, it was puddle that brought IB into the comparison.

 

[runawayprisoner]

Nope, it was puddle that brought IB into the comparison.

Yeah, but I believe he mentioned mobile Ivy Bridge (for tablets and ultra portables), not desktop.

 

[dagamer34]

This thread is hilarious. All the Android fanboys are whining that iPad has 4x the resolution but only double the GPU performance.

I can't wait until then next-gen Galaxy Tab comes out with a higher resolution than iPad and a weaker GPU. These very same people will be jizzing their pants over it.

Well that pretty much guaranteed: 2560x1600 Super AMOLED display running on an Exynos 5250 with a dual-core ARM Cortex A15 and ARM Mali T604 GPU. But that should be assumed as fact by now.

Regardless, I am more annoyed by games that don't update for the new display, because it doesn't matter what is in the SoC if it's still using old image assets.

 

[runawayprisoner]

Well, looks like A5 in Apple TV is 32nm.

http://www.chipworks.com/en/technic...ple-a5-from-the-apple-tv-3-surprise-surprise/

So it would seem Apple did try 32nm and probably failed enough to churn out the Apple TV.

On a side note, if they can get it to work, I suspect they may try to fit A5X into the next iPhone...

 

[bearxor]

Well, looks like A5 in Apple TV is 32nm.

http://www.chipworks.com/en/technic...ple-a5-from-the-apple-tv-3-surprise-surprise/

So it would seem Apple did try 32nm and probably failed enough to churn out the Apple TV.

On a side note, if they can get it to work, I suspect they may try to fit A5X into the next iPhone...

And in case anyone missed it, new iPad 2's (iPad 2,4) have a 32nm A5 with both cores enabled.

This was obviously a test of 32nm for Apple/Samsung.

 

[ChAoTiCpInOy]

And in case anyone missed it, new iPad 2's (iPad 2,4) have a 32nm A5 with both cores enabled.

This was obviously a test of 32nm for Apple/Samsung.

Probably going to see a 32nm chip in the new iPhone this year.

 

[Eug]

Probably going to see a 32nm chip in the new iPhone this year.

Definitely, but the bigger question will be if it's an A5X or something else, like a quad-core A6. I'm thinking the former though.

 

[ChAoTiCpInOy]

Definitely, but the bigger question will be if it's an A5X or something else, like a quad-core A6. I'm thinking the former though.

I don't think it'll be the A5X because the iPhone doesn't need a 4 core GPU. Then again, Apple usually doesn't release chip updates in the iPhone. The iPad has always been where they release chip updates and the iPhone is where they release new software.

 

[alent1234]

I don't think it'll be the A5X because the iPhone doesn't need a 4 core GPU. Then again, Apple usually doesn't release chip updates in the iPhone. The iPad has always been where they release chip updates and the iPhone is where they release new software.

it will probably be A5X because apple likes to use the same parts in all products

 

[bearxor]

The only improvement the A5X brings that would be useful to a new iPhone is 1GB of RAM.

I'm putting my money on a August/September iPhone launch that has the A6 chip that is 32nm DC A15 with the 543MP2. The A6X will follow in January with higher clicked DC and the 543MP4. I just don't see anything on a roadmap that is practical for a GPU upgrade anytime soon. Especially since even the MP2 pretty much still pwns.

 

[ChAoTiCpInOy]

it will probably be A5X because apple likes to use the same parts in all products

Yeah. With that logic it'll probably be an A5X with 1GB of ram but the 2 extra cores in the GPU disabled. That makes sense (not being sarcastic).

 

[Eug]

The only improvement the A5X brings that would be useful to a new iPhone is 1GB of RAM.

I'm putting my money on a August/September iPhone launch that has the A6 chip that is 32nm DC A15 with the 543MP2. The A6X will follow in January with higher clicked DC and the 543MP4. I just don't see anything on a roadmap that is practical for a GPU upgrade anytime soon. Especially since even the MP2 pretty much still pwns.

Hmm... Dual-core Cortex A15 would be decent with 543MP2, esp. with twice the RAM, although I was kinda wondering about dual-core A7 + dual-core A15.

I still wonder if it might just be a 32 nm A5X though.

 

[ChAoTiCpInOy]

dual-core A7 + dual-core A15

What's that? Something like a Tegra 3?

 

[dagamer34]

Yeah. With that logic it'll probably be an A5X with 1GB of ram but the 2 extra cores in the GPU disabled. That makes sense (not being sarcastic).

That makes no sense. Disabled cores still take up space on the die. And 45nm production is an enough known quantity that they shouldn't be die harvesting that many chips just to fix production. And 32nm production will be in full swing by then.

The A5X as we know it is likely a on-off design that will only be used in the current iPad. We'll see a die shrink + CPU/GPU updates and/or clock speed increases in the iPhone this year.

 

[Eug]

What's that? Something like a Tegra 3?

Sorta but not quite. Tegra 3 is quad-core A9 on 40 nm, plus a lower power but not ultra low power 5th A9 core.

I was talking about ARM's so-called big.LITTLE configuration which pairs two ultra-low power A7 cores with two higher power but much higher performance A15 cores.

These A7 and A15 cores run the exact same instruction set. A7 is a lot slower than A15, but is fast enough for basic smartphone tasks, and runs a lot lower power than A15's minimum power requirement. For hardcore gaming or other heavy duty smartphone tasks, the OS can switch to A15.

ARM: big.LITTLE Processing

From my understanding, A15 is way faster than A9, and A7 should be way lower power than Tegra 3's low power A9.

Furthermore, A7 supposedly is faster clock-for-clock than A9, so A7 is no slouch.

Anand: ARM's Cortex A7: Bringing Cheaper Dual-Core & More Power Efficient High-End Devices

 

[smartpatrol]

Wow, I'd heard of A7 but I didn't realize it was supposed to be faster than A8. A15 with big.LITTLE will probably be the biggest leap we've ever seen in terms of performance and efficiency in a smartphone CPU.

32nm A15/A7, with 543MP4 (or something similar) and integrated LTE radio is my dream smartphone SOC. Wonder if it will happen this year. . .

 

[lothar]

I'll go against the grain here:
I doubt Apple will release an A15(or a hybrid form of it like A15/A7 big.little combination) in the next iPhone...It seems possible, but not likely. Most A15 designs(besides Qualcomm's Krait) are behind schedule and won't be ready in the June-October launch time frame(unless Apple delays the launch again like they did for 4 months last year).

If they're still keeping that June-October time frame, I'd expect either:
32nm dual core with "updated" graphics with/without LTE.
32nm quad core with the same iPhone 4S graphics with/without LTE

 

[dagamer34]

I'll go against the grain here:
I doubt Apple will release an A15(or a hybrid form of it like A15/A7 big.little combination) in the next iPhone...It seems possible, but not likely. Most A15 designs(besides Qualcomm's Krait) are behind schedule and won't be ready in the June-October launch time frame(unless Apple delays the launch again like they did for 4 months last year).

If they're still keeping that June-October time frame, I'd expect either:
32nm dual core with "updated" graphics with/without LTE.
32nm quad core with the same iPhone 4S graphics with/without LTE

Krait != A15. I don't know how many times I have to keep correcting this on the Internet. It has simiar performance but the core is 100% designed by Qualcomm. It's like saying a Core 2 Duo and a Phenom II are the same because they both run x86.

 

[badb0y]

It's probably going to be a die-shrunk A5X with the A6 following in the next iPad.

 

[lothar]

Krait != A15. I don't know how many times I have to keep correcting this on the Internet. It has simiar performance but the core is 100% designed by Qualcomm. It's like saying a Core 2 Duo and a Phenom II are the same because they both run x86.

That's all you could pick apart from my post?

"A15", "A15 like"; who gives a hoot? It still has A15 performance.
No need to get your panties in a wad.

In fact, I'll keep saying that just to annoy the Grammar Nazi police.