You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I want to go straight to the core, the A9X chip. What a beast!!!
You must definitely have a look at the Ars Techinca review
Let’s go over the basics first: Geekbench reports that the A9X is a dual-core chip running at about 2.25GHz. The A8X used three CPU cores to boost performance, but the A9’s “Twister” CPU architecture and the big boosts in clock speed that Apple is squeezing out of it (up from 1.84GHz in the iPad Air 2) both apparently made that third core unnecessary. These seem to be the same CPU cores we saw in the new iPhones earlier this year, just running 300-or-so MHz faster. That speed increase is reflected in the Geekbench and browser-based CPU tests. Also note that while the iPad Pro doesn’t look all that much faster than the iPad Air 2 in multi-core CPU tests, the fact that it’s doing that work with two cores means the single-core scores are drastically higher.
The A9X can’t quite get up to the level of a modern U-series Core i5 based on Broadwell or Skylake (see the 2015 MacBook Air and Surface Pro 4 results), but it’s roughly on the same level as a Core i5 from 2013 or so and it’s well ahead of Core M. And despite the fact that it lacks a fan, the A9X shows little sign of throttling in the Geekbench thermal test, which bodes well for the iPad Pro’s ability to run professional-caliber apps for extended periods of time.
Things are even more impressive on the GPU side, where the OpenGL version of the GFXBench test shows the A9X beating not just every previous iDevice, but every Intel GPU up to and including the Intel Iris Pro 5200 in the 15-inch MacBook Pro and the Intel HD 520 in the Surface Pro 4. Once we see Iris and Iris Pro chips from the Skylake family, Intel may be on top again, but those aren’t due out until early next year, and they only ship in the fastest of Apple’s products.
Well, I definitely like it, it is a masterpiece in terms of hardware specs. This Apple SoC is a serious threat for the Intel/AMD duo. To date it is probably the best chip ever in terms of performance per watt. Gorgeous. Just thinking about a quad core SoC, what a monster it would be.
Would I buy it? To be honest, I do not know. As a tablet replacement, it is too expensive (and probably also too big for my taste), as a laptop alternative, I must admit that I´m not very comfortable with this iOS yet. Accessories, a must if we plan to use it as a work machine, are very pricey too.
The A9X is insanely fast. The CPU is the same speed as a Skylake Core i5 in single thread performance. The GPU beats the Iris Pro 5200 and Nvidia Tegra X1 hands down.
I just wonder if they will put the same SoC in the iPad air3, as the screen size is much smaller they don't need such a powerful GPU.
Anyone have a view of what the TDP is? Several reports I have read are saying closer to 10W, which may explain why we haven't seen a iPad Air 3 announced with it in.
Also does anyone have a view why in geekbench, the A9x scaling between single core and multi-core is only 1.70x, whereas on the latest core-M it is 1.95x?
Lefty, from what I have read, the GPU scores are not comparable as mobile benchmarks are at FP16 and the PC benchmarks at FP32, hence A9X graphics performance is not anywhere near core level. An expert on here can correct me if I have been misinformed on this one.
Either way it is an incredible achievement by Apple (and Imagination too).
Also does anyone have a view why in geekbench, the A9x scaling between single core and multi-core is only 1.70x, whereas on the latest core-M it is 1.95x?
Lefty, from what I have read, the GPU scores are not comparable as mobile benchmarks are at FP16 and the PC benchmarks at FP32, hence A9X graphics performance is not anywhere near core level. An expert on here can correct me if I have been misinformed on this one.
Yes, I'd say you've being mis-informed. Shaders writen for mobile GPUs use lowp for some operations (but only about 10 -20%) and a GPU supporting FP16 will run these shaders better. However, GPUs not supporting FP16 are simply less capable. So, it's the same as saying that a benchmark "doesn't count" simply because it uses a lot of tessellation and therefore favours GPUs with good tessellation, or a benchmark uses a lot of memory bandwidth and therefore favours GPUs with lots of bandwidth.
I also point out that you will start to see FP16 implemented in PC GPUs soon (in fact the Tegra X1 supports FP16).
Well, I'm sad to say my iPad Pro has not arrived yet. The cover arrived last Monday though. Here is the delivery trail which is a bit funny:
Location
|-
| Louisville, KY, United States
| 11/20/2015
| 7:35 A.M.
| Saturday delivery has been requested. / Delivery will be rescheduled for Saturday.
|-
|
| 11/20/2015
| 7:30 A.M.
| We left the package in our facility. This may cause a delay. We will deliver as quickly as possible. / Delivery will be rescheduled for Saturday.
|-
|
| 11/20/2015
| 3:10 A.M.
| Import Scan
|-
|
| 11/20/2015
| 1:18 A.M.
| Arrival Scan
|-
| Anchorage, AK, United States
| 11/19/2015
| 3:30 P.M.
| Departure Scan
|-
|
| 11/19/2015
| 12:11 P.M.
| Arrival Scan
|-
| Chek Lap Kok, Hong Kong
| 11/19/2015
| 8:29 P.M.
| Departure Scan
|-
| Chek Lap Kok, Hong Kong
| 11/18/2015
| 9:31 P.M.
| Your package is in transit. We're updating plans to schedule your delivery. / The package will be forwarded to a UPS facility in the destination city.
|-
|
| 11/18/2015
| 9:27 P.M.
| Your package is in transit. We're updating plans to schedule your delivery.
|-
| Louisville, KY, United States
| 11/15/2015
| 6:07 A.M.
| Arrival Scan
|-
| Anchorage, AK, United States
| 11/14/2015
| 8:43 P.M.
| Departure Scan
|-
|
| 11/14/2015
| 6:14 P.M.
| Arrival Scan
|-
| Chek Lap Kok, Hong Kong
| 11/15/2015
| 2:03 A.M.
| Departure Scan
|-
| Shenzhen, China
| 11/13/2015
| 11:45 P.M.
| Departure Scan
|-
|
| 11/13/2015
| 4:00 P.M.
| Origin Scan
|-
| China
| 11/13/2015
| 4:31 A.M.
| Order Processed: Ready for UPS
This package is racking up serious frequent flyer miles!
I finally got my iPad Pro and am using it now. Setup was easy and I have not experienced any technical problems yet. One thing I can tell you is that it is BIG!!! Seriously big!!! And heavy!!! As compared to the iPad 2. I cannot imagine traveling with this thing or even taking it to the gym. My biggest fear is that this big old screen is going to break!
As for speed, the iPad Pro is blazingly fast as compared to the iPad 2. Clearly the speed is outpacing my WiFi connection and I have very fast WiFi at home.
I'd be curious to see how it compares to the Core-M based 2015 MacBook. I know the MacBook is super light, and my experience with it shows that the performance is pretty good.
One wonders if Apple will ever put at A9 into a full feature laptop? They must be thinking about the idea....
I'd be curious to see how it compares to the Core-M based 2015 MacBook. I know the MacBook is super light, and my experience with it shows that the performance is pretty good.
One wonders if Apple will ever put at A9 into a full feature laptop? They must be thinking about the idea....
It depends how you define what a "laptop" is. With a 12.9 inches screen, a detachable keyboard, 32GB/128G Storage, 4GB RAM, I can call iPad Pro is a full feature laptop or a notebook already.
I'd be curious to see how it compares to the Core-M based 2015 MacBook. I know the MacBook is super light, and my experience with it shows that the performance is pretty good.
One wonders if Apple will ever put at A9 into a full feature laptop? They must be thinking about the idea....
Would it have everything that is need to run a Mac OS laptop in terms of ability to interface with peripherals and external storage
? Or are we still comparing something that is still essentially a mobile chip? Also any view on cost of producing it. It seems pretty large and any views on what the yield would be and whether there would be binning?
Maybe not the same Mac OS, but I do not see any problem there. Chrome OS can run on ARM, Windows RT ran on ARM.
The bigger the chip the lower the yield, but again, it is the same for the x86 SoCs. Not yet spotted the A9X size on the net, but it should be most likely around 150 mm2. That would make the total chip count per wafer around 400. About cost, I must guess, but I would say in 65-85$ range. Yield per wafer, at this stage, 60-65%. A mature yield for such a node and chip size, should be around 80%.
About the binning, sure, but mainly in terms of voltage pre-settings to be applied at the SoC (chips can have slightly different std-by currents for example). Some repairability might also be possible (by fusing), but that´s pretty much it.
While Intel and AMD can downgrade and sell a chip at a lower frequency or with a fewer working cores (if one or more are dead), since they have more "models" available on the market, the A9X is instead a kind of unique. Downgrade the A9X into an A9, the only option possible (fewer GPU clusters, less SRAM, lower CPU frequency), doesn´t sound plausible to me.
Thanks astilo, very useful. I am just trying to form a view of whether Apple would actually go ahead and launch an A-Series laptop running Mac OS. We can be sure that Apple doesn't pay anything near list prices for its Intel chips, so it is the relationship between an A-series performance and all-in cost that will determine what they decide to do. And of course Intel's potential reaction - they may well cut Apple off, destroying its higher end business, or cease giving them good prices. Plus of course the fact that a lot of MacBook users also run Windows. Maybe what we will see over time is that iOS morphs more towards Mac OS.
On the binning, would that mean that Apple would decide on a frequency target and then anything unable to operate above this would be tossed (or held back for a future lower performance product) and chips above the target would be all adjusted through the voltage settings to stay at the frequency?
On the binning, would that mean that Apple would decide on a frequency target and then anything unable to operate above this would be tossed (or held back for a future lower performance product) and chips above the target would be all adjusted through the voltage settings to stay at the frequency?
Yes and no. Outside a well specified and designed operative range, a part is of course marked as bad. Anyway, inside the normal distribution, they keep everything and apply a different binning.
I found an article on anandtech from another semiwiki blogger that is extremely good to explain that.
Andrei Frumusanu]Process variations during manufacturing can lead silicon to have different electrical characteristics, leading to cases where a product would no longer be able to function properly under its target specifications. What does very commonly happen though is that vendors try to increase voltages to compensate for such process variations, overcoming problematic manufacturing issues in this way. Chipsets are tested at the factory for their characteristics and each chip is then permanently marked with the information by burning it to on-chip fuses. For the Exynos chipsets these bins are called ASV groups. The groups with the higher voltages represent bins with “slow-” or “cold” silicon, meaning process variations cause transistors to not to be fully able to reach the design frequency without having to raise VDD from the nominal targets. One advantage of cold chips is that their static leakage is reduced over other bins. On the other spectrum we have “fast” or “hot” silicon with lower threshold voltages that are able to hit the desired clock-rate at a lower VDD. In contrast to cold silicon, hot silicon has much more static leakage due to the lower Vt. Power consumption in today’s large SoCs is mostly determined by the dynamic leakage (gate to drain, drain-induced barrier lowering, etc) of a chip as it overshadows static leakage currents which can be mitigated by power-gating mechanisms. FinFET also comes into play as it helps to dramatically reduce static leakage compared to planar technologies. It is thus certainly almost always more advantageous to have a “hot / fast” bin which is able to reach lower operating voltages.
As seen in the graphic, the range between a worst-case and best-case for the Exynos 7420 can be as high as 150mV which represents up to 32% more dynamic power on the highest frequency of the A57 cores. Luckily, one should not have to worry too much about the bin in one’s device as shipped units follow a Poisson distribution pattern where the vast majority of chipsets fall at or around the lambda of ASV10-ASV11. <asv6, which="" doesn’t="" mean="" they="" don’t="" exist,="" but="" may="" be="" very="" rare.="" the="" device="" we’ve="" tested="" power="" on="" in="" this="" article="" came="" with="" a="" chipset="" graded="" asv10="" (highlighted="" green="" graph)="" cpu="" clusters="" and="" memory="" controller="" asv11="" gpu,="" by="" way="" also="" points="" out="" to="" fact="" that="" main="" soc="" blocks="" are="" individually="" characterized="" necessarily="" fall="" same="" grading="" bin="" category.<="" span=""></asv6,>
And of course Intel's potential reaction - they may well cut Apple off, destroying its higher end business, or cease giving them good prices. Plus of course the fact that a lot of MacBook users also run Windows. Maybe what we will see over time is that iOS morphs more towards Mac OS.
With the flat, slow, or sometimes even decreasing Windows desktop and laptop computer market, Intel can't afford to cut Apple off from their customer list when Apple decides to use more their own SoC/processors. As matter of fact, PC vendors like Asus, Acer, Lenovo, and HP all use both AMD and Intel processors on their products for many years already.
I posted my iPad Pro delivery tracking above. It looks like mine was recalled and a new one shipped. I have read the same from others. I wonder how many made a similar journey and why exactly was it recalled?
