Nebuchadnezzar
Legend
The Exynos 5420 is also supposed to have some dedicated SRAM for GPU and memory interface. Given that the SRAM for the A7 is also in direct vicinity of those blocks, they might be correlated in function.
Next-Gen iPhone & iPhone Nano Speculation
- Thread starter Arun
- Start date
A nice-sized chunk of SRAM would be useful to a lot of the processor sub-systems, though I'm not sure any of the included designed cores would call for it (maybe the ISP for the camera processing.) Dedicating it as the workspace for matching the fingerprint data pulled from storage would seem wasteful, but I suppose it's the kind of flexibility Apple would have when they build a device from top to bottom. I'm not sold on the theory right now, though.
Their labeling of the GPU blocks as cores is a bit annoying. Rogue clusters don't have all the necessary functionality to operate as independent GPUs, so that "shared digital logic" is the GPU core as much as any other part is.
Their labeling of the GPU blocks as cores is a bit annoying. Rogue clusters don't have all the necessary functionality to operate as independent GPUs, so that "shared digital logic" is the GPU core as much as any other part is.
anexanhume
Veteran
Associated with the frame buffer compression logic?
Technically it should call all of it one core and the highlighted parts the clusters
I find it interesting how spread out the SDRAM interfaces are, on the whole.
GPU is taking around 30% more die area (normalised for chip process) than the GPU in the A6.
Pretty impressive for a x3.7 improvement in GL2.7 and nearly x2 fillrate.
In fact its around 20% smaller than the A6x GPU, and still comfortably outperforms it.
All we need now is a nice power comparison test by anandtech.
Pretty impressive for a x3.7 improvement in GL2.7 and nearly x2 fillrate.
In fact its around 20% smaller than the A6x GPU, and still comfortably outperforms it.
All we need now is a nice power comparison test by anandtech.
Nebuchadnezzar
Legend
Now that you mention it that is probably the best guess until now, the 5420 touts this new MIC memory compression. The A7 having an equivalent is very probable.
The transistor density is likely way higher on the A7 GPU vs. the A6X GPU. And the performance difference between the two GPU's is neglible in GFXBench 2.5, although there is ~ 30% performance improvement in GFXBench 2.7 IIRC (albeit still at framerates that are not yet smooth and playable). Still a very impressive SoC all things considered.
On a side note, something was holding the A6 GPU back in GFXBench 2.7 because it's framerate was only 1/3 that of the A6X GPU.
Associated with the frame buffer compression logic?
Technically it should call all of it one core and the highlighted parts the clusters
I find it interesting how spread out the SDRAM interfaces are, on the whole.
Just to confirm, we are looking at a dual-channel memory configuration, that has been spread out around the SoC, rather than a quad-channel setup?
Here's a really good article on the advantages and disadvantages of the 64bit support in the A7 and iOS7: http://www.mikeash.com/pyblog/friday-qa-2013-09-27-arm64-and-you.html
Of particular interest is the way Apple leverages unused bits in 64bit pointers to accelerate ObjC reference count management, which results in serious performance upgrades for apps that do a lot of object allocation/deallocations.
Of particular interest is the way Apple leverages unused bits in 64bit pointers to accelerate ObjC reference count management, which results in serious performance upgrades for apps that do a lot of object allocation/deallocations.
That's not that much of a disadvantage in apple's ecosystem... Figures released recently indicate that a week after iOS7 launched its webshare of internet traffic was at ~57% IIRC. After ONE WEEK... What's android's uptake after one week?
Historically, within a few months the new OS will have propagated to 90+ percent of users.
Historically, within a few months the new OS will have propagated to 90+ percent of users.
That's not that much of a disadvantage in apple's ecosystem... Figures released recently indicate that a week after iOS7 launched its webshare of internet traffic was at ~57% IIRC. After ONE WEEK... What's android's uptake after one week?
Historically, within a few months the new OS will have propagated to 90+ percent of users.
Does upgrading to iOS 7 make the A5 & A6s in all those iPad 2,3 & 4, iPhone 4,4s & 5 transform into 64-bit SoCs?
No.
But if you'd read the article, then you'd know that iOS7 supports fat binaries that support both.
Non-sequitur.
anexanhume
Veteran
Yes, apple said this compile option would be available soon.
ltcommander.data
Regular
https://developer.apple.com/library...roduction.html#//apple_ref/doc/uid/TP40013501
To clarify, you can already develop 32-bit/64-bit fat binaries for iOS 7. What's coming, supposedly next month, is building 32-bit/64-bit fat binaries supporting both iOS 6 and 7.
I don't know how accurate this guys measurements are for the baytrail CPU and GPU IP, but assuming they are accurate, they are interesting comparative comments regarding A7.
http://investorshub.advfn.com/boards/read_msg.aspx?message_id=92266257
I'm assuming he has identified the large block uppermost in the baytrail die shot as being the gpu.
He gets Baytail GPU @18.4mm2 against A7 GPU @14mm2 (normalised for 22nm).
Thats a 25%+ smaller gpu (or baytrail being 30% larger, take your pick), with better performance, working within the power budget of a handheld device, versus one working within a tablet power budget.
Factually he states that baytrail would have xtra die space due to supporting DX11 which is correct, but also implies that the A7 isn't es3.0 compliant, which is wrong.
http://investorshub.advfn.com/boards/read_msg.aspx?message_id=92266257
I'm assuming he has identified the large block uppermost in the baytrail die shot as being the gpu.
He gets Baytail GPU @18.4mm2 against A7 GPU @14mm2 (normalised for 22nm).
Thats a 25%+ smaller gpu (or baytrail being 30% larger, take your pick), with better performance, working within the power budget of a handheld device, versus one working within a tablet power budget.
Factually he states that baytrail would have xtra die space due to supporting DX11 which is correct, but also implies that the A7 isn't es3.0 compliant, which is wrong.
Last edited by a moderator:
ltcommander.data
Regular
http://appleinsider.com/articles/13...g-out-its-ios-gpu-development-team-in-orlando
Apple continues to hire for their GPU design team. It took two years after they started branding their own SoC before they came up with a truly custom CPU design in the A6. Now that they've started branding their own GPU in the A7, perhaps we'll be seeing the results of their work in the A9.
Apple continues to hire for their GPU design team. It took two years after they started branding their own SoC before they came up with a truly custom CPU design in the A6. Now that they've started branding their own GPU in the A7, perhaps we'll be seeing the results of their work in the A9.
ltcommander.data
Regular
https://developer.apple.com/library...html#//apple_ref/doc/uid/TP40013599-CH106-SW1
The anti-climatic Apple A7 GPU.