Apple A10 Fusion is ** Quad-core big.LITTLE **

[Andrei.]

The intriguing part I previously mentioned is that "New Caches" on the bottom right. If this is part of the front-end, I wonder if that could be some form of micro-op (or even trace) cache, but it would be oddly placed between L1I and L2. Or perhaps Apple went crazy and implemented a TAGE/ITTAGE larger than the I-cache

One of the sets of small boxes above the I-cache is most likely the tags.

Don't know if it's actually part of the front-end or just happens to be right next to it, even the hi-res die shot doesn't really tell. But it should be clearly part of the core and Linley's core area is therefore inaccurate.

 

[Nothingness]

Don't know if it's actually part of the front-end or just happens to be right next to it, even the hi-res die shot doesn't really tell. But it should be clearly part of the core and Linley's core area is therefore inaccurate.

Linley partitioning of Big/{Little,Small}/L2 (figure 3) is the same as yours, and I found no statement that the "New caches" structure has been excluded from the Big core area, so it should be right. Might have missed something...

 

[Andrei.]

Linley partitioning of Big/{Little,Small}/L2 (figure 3) is the same as yours, and I found no statement that the "New caches" structure has been excluded from the Big core area, so it should be right. Might have missed something...

His core area estimate is missing the exact difference from that new cache block.

 

[slashbinslashbash]

One more addition.
I think the reason you and I so often butt heads is that for the most part I am asking "why" questions, and you (and people like you) are providing "what" answers.

For example, if you look at this latest dialog stream, MY concerns have to do with WHY one would structure a memory controller in a certain way as compared to various alternatives. Your answers, while giving various correct facts, do not answer that question. And this is the constant pattern I see between us.

It is reasonable for you to say that you do not know the answer to my "why" questions; it is NOT reasonable for you to insist that people are answering my questions and I am just being dogmatic in ignoring the answers. They are NOT answering the questions. They are providing various facts (sometimes relevant, sometimes of dubious relevance) but I am usually not asking for specific facts, I am asking, as I said, for a more abstract, conceptual theory of WHY these facts and not others.

I am completely unqualified to talk about any of the technical details that you and Andrei have been discussing. However, I do have to question why you think that you have a claim to Andrei's time and effort, in demanding that he explain XYZ to you when clearly he knows what he is talking about and clearly you are coming from a place of lesser knowledge of the subject. It is all well and good to want to learn, but unless you are paying him to teach you then you can't expect to simply bully him into replying by claiming that he's wrong and that he must reply in order to prove himself right.

Learn some manners, is what I'm saying. Or take some paid classes from someone where it's their job to answer your questions.

 

[Nothingness]

Slightly off-topic: it looks like Intel modem is performing significantly worse than the Qualcomm one.

http://cellularinsights.com/iphone7

 

[Nothingness]

A10 core area versus the competition

 

[Lodix]

A10 core area versus the competition

Thanks for the link

 

[witeken]

That Broadwell Core, though .

 

[witeken]

Not bad, TSMC. Not bad.

"Posted 2 years ago"

Mark Liu CEO of TSMC stated that 10nm technology will be available to customers in late 2016 followed by expanded volume production in early 2017 which is when revenue from the first 10nm wafer sales are expected to roll in.

Quite a comeback. I honestly did not think that any article that far in the past about TSMC would be anywhere near this accurate and precise up to the quarter. (I was just reading some Wikipedia Tick-Tock and stuff and they linked to there.)

TSMC had faced serious challenges with its 16nm FinFET manufacturing process. The process faced three major setbacks amounting to nine months of delay.

Hey, I know another company that also had 9 months of delay, although the so called delay they're talking about didn't materialize actually ("pushing mass production deep into 2016"). Intel: prolong manufacturing nodes to 3 years. TSMC: do a 1.6x shrink within 1.5 years after a 1.9x shrink. Quite the comeback indeed after their horrendous 20/16nm node.

I once saw someone wondering why not everyone's on equal footing because they use the same equipment (IIRC). Intel's big push with Tick-Tock has always been their rapid node sucession within 2 years, with or without a full 2x shrink combined with their top notch rapid yield learning and Copy Exactly! With the yield learning having slowed down and 14nm not being a sole exception to the rule, it indeed seems the playing field in terms of density is leveling somewhat. But I don't see how anyone could achieve a major advantage .

http://wccftech.com/tsmc-promises-10nm-production-2016-7nm-2017/

The first five top comments, though . Definitely shame on me.

(I wonder if the choice of LELELE (Samsung) vs. SADP/SATP/SAQP (Intel) has anything to do with this.)

 

[nvgpu]

http://www.realworldtech.com/apple-custom-gpu/

After years of recruiting graphics architects, Apple has designed its own custom GPU, which is already shipping in the A8, A9, and A10 processors that power the iPhone 6, 6S, and 7. Based on publicly available evidence it is clear that the shader core in Apple’s GPU is architecturally very different from Imagination Technologies PowerVR line. This further implies that Apple wrote its own Metal and OpenGL ES compiler for its GPUs and almost certainly wrote the entire driver as well. The GPU in Apple’s processors still retains some fixed-function hardware from PowerVR, but the shader core and the driver are now developed internally.

 

[Thope]

does anybody know if A10 fusion deep dive is coming.

 

[Eug]

I think my phone is overheating.

I had a video file that was h.265 HEVC 1080p to play on my trip on the plane. My laptop is a Core 2 Duo and can't play that at all. Slideshow.

So, I put it on my iPhone 7 Plus instead. Playing the video, the phone in a plastic Spigen case over time got quite warm and then after about 20 minutes it started to get bad stuttering, which I interpreted as CPU throttling. If I stopped the video and then started playing it soon after, I'd get the same behaviour again in less than a minute. However, if I waited say 3 minutes, then I could play the video for several minutes before it started stuttering again. So it really does seem like throttling.

Mind you, with similar files before with an open back case, I didn't get that behaviour. Perhaps the files were a bit different, but I think the case used had a lot to do with it.

Overall though, 20 minutes of software based 1080p HEVC h.265 playback in an insulating case seems pretty decent. Maybe next time I'll try this same file with the case off to see what happens.

I do need a new laptop though. Can't wait for that Kaby Lake MacBook 12".

 

[Eug]

I think my phone is overheating.

I had a video file that was h.265 HEVC 1080p to play on my trip on the plane. My laptop is a Core 2 Duo and can't play that at all. Slideshow.

So, I put it on my iPhone 7 Plus instead. Playing the video, the phone in a plastic Spigen case over time got quite warm and then after about 20 minutes it started to get bad stuttering, which I interpreted as CPU throttling. If I stopped the video and then started playing it soon after, I'd get the same behaviour again in less than a minute. However, if I waited say 3 minutes, then I could play the video for several minutes before it started stuttering again. So it really does seem like throttling.

Mind you, with similar files before with an open back case, I didn't get that behaviour. Perhaps the files were a bit different, but I think the case used had a lot to do with it.

Overall though, 20 minutes of software based 1080p HEVC h.265 playback in an insulating case seems pretty decent. Maybe next time I'll try this same file with the case off to see what happens.

I do need a new laptop though. Can't wait for that Kaby Lake MacBook 12".

With the case off, the phone lasted just under 30 minutes before stuttering again. So, yeah, I think it's throttling under the heat. At about 15 minutes, the phone was only moderately warm, but by about 25 minutes it was very warm. And a few minutes later is when it throttled.

 

[sxr7171]

I think my phone is overheating.

I had a video file that was h.265 HEVC 1080p to play on my trip on the plane. My laptop is a Core 2 Duo and can't play that at all. Slideshow.

So, I put it on my iPhone 7 Plus instead. Playing the video, the phone in a plastic Spigen case over time got quite warm and then after about 20 minutes it started to get bad stuttering, which I interpreted as CPU throttling. If I stopped the video and then started playing it soon after, I'd get the same behaviour again in less than a minute. However, if I waited say 3 minutes, then I could play the video for several minutes before it started stuttering again. So it really does seem like throttling.

Mind you, with similar files before with an open back case, I didn't get that behaviour. Perhaps the files were a bit different, but I think the case used had a lot to do with it.

Overall though, 20 minutes of software based 1080p HEVC h.265 playback in an insulating case seems pretty decent. Maybe next time I'll try this same file with the case off to see what happens.

I do need a new laptop though. Can't wait for that Kaby Lake MacBook 12".

When that happens I request ice water and dip the phone into it. It will get the processor going again. I even have benchmarks showing that it will slow down to a third when throttling compared to being under ice.

 

[IntelUser2000]

I once saw someone wondering why not everyone's on equal footing because they use the same equipment (IIRC). Intel's big push with Tick-Tock has always been their rapid node sucession within 2 years, with or without a full 2x shrink combined with their top notch rapid yield learning and Copy Exactly! With the yield learning having slowed down and 14nm not being a sole exception to the rule, it indeed seems the playing field in terms of density is leveling somewhat. But I don't see how anyone could achieve a major advantage .

I think its quite obvious why Intel is losing the process advantage. Sure, same equipment, but every cook that has the same equipment don't cook the same quality food. Also, you can use somewhat different tools and achieve competitive result.

"Mobile Effect" is why I believe Intel is losing its edge. No, not just being unable to penetrate mobile markets directly with mobile-focused chips. The world's attention has turned away from Intel, to mobile guys. The top mobile players are Apple and Samsung. If you are looking for work in an area related to chip design and manufacturing, would you want to go to a company with its future not being sure?

Intel kept their process edge because they were the darlings of the tech industry. Top guys went there. We don't get to see and hear what kind of decisions guys at Intel made when they found out they were mere followers in the mobile market, then having to quit. They modified their 22nm process to fit with the mobile strategy. Do you think process guys will be happy about what's going on? Do you think they don't get the blame?

Apple having LPDDR4-3200 is amazing! Intel used to be the first to implement latest RAM standards in the tech industry. Their laptops are still stuck to LPDDR3(because LP models offer advantage over non LP ones which is why they don't use DDR4), and high end desktops reach 3000 speeds only with enthusiast models. Now with A10 they have LPDDR4 with 4266 speeds. Fastest Skylake desktops barely reach that speed, and a puny mobile chip in a 5W envelope does that with an advanced PoP stack.

Technology-wise, their chips are shameful compared to what Apple is doing. What we expected Intel, no, dreamed of, is all being done at Apple.*

Because companies are made up of its people, and we tend to see it instead as a single giant entity.

*If Apple can do this I'd have expected Intel to bring out Core M that can sustain 2GHz speeds and feature LPDDR4-4266 with 2 slice(48 EU) graphics, and small/power efficient enough to be used in an 8-inch Windows Tablet. Which is what Apple is going to do with A10X. Heck, that's the bare minimum. To be substantially better it should have had top tier 15W U performance CPU and Skull Canyon NUC Iris Pro graphics performance and still power efficient/small enough to be used in a 8-inch Tablet.

 

[Nothingness]

The world's attention has turned away from Intel, to mobile guys. The top mobile players are Apple and Samsung. If you are looking for work in an area related to chip design and manufacturing, would you want to go to a company with its future not being sure?

Apple doesn't look like a nice place to work: pay is good but the paranoiac environment seems to be a pain. Intel is so big that unless you work in the Architecture Group or one of the design teams that work on Core/Xeon, you're too far from what is really interesting. Samsung might be an option, but the way Koreans work might be an obstacle to many.

In fact, there's a better company to work for if you're into chip design: ARM

Also I don't see how you can predict Intel demise. Their Core/Xeon line is still untouchable by anyone and I don't see this dramatically change for many years.

 

[witeken]

Apple doesn't look like a nice place to work: pay is good but the paranoiac environment seems to be a pain.

Intel is so big that unless you work in the Architecture Group or one of the design teams that work on Core/Xeon, you're too far from what is really interesting.

Samsung might be an option, but the way Koreans work might be an obstacle to many.

In fact, there's a better company to work for if you're into chip design: ARM

Those prejudices .

But in any case I think if you want to work as an engineer, there are other companies too. For example if you live in the Netherlands, there's ASML.

 

[imported_ats]

Apple having LPDDR4-3200 is amazing! Intel used to be the first to implement latest RAM standards in the tech industry. Their laptops are still stuck to LPDDR3(because LP models offer advantage over non LP ones which is why they don't use DDR4), and high end desktops reach 3000 speeds only with enthusiast models. Now with A10 they have LPDDR4 with 4266 speeds. Fastest Skylake desktops barely reach that speed, and a puny mobile chip in a 5W envelope does that with an advanced PoP stack.

You too can be limited to 4GB of ram. Have fun. Do realize that everything has trade offs. What? You thought 4266 LPDDR4 didn't have downsides..