The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
So they designed all the easy stuff. When they design their own CPU or GPU cores then I'll be impressed.
Cores are, by far, the most important and difficult portion of an SoC to design. By comparison, a NPU is simple to create compared to a CPU.
If it is so easy, why is Apple behind?
Apple is often behind in many of the superficial benchmarks that people look at. They don't care if they have the most RAM because they've put in the hard work to make their OS more efficient and design their own chips right down to the core in order to further take advantage of that synergy all to allow for better efficiency.
That doesn't change the observation. If this is the 'easy stuff' (and let's be honest, modems and WiFi are key smartphone areas), why hasn't Apple corrected the situation?
I’d bet that 99% of the time, you wouldn’t see any difference in those areas at all. It’s not important. When almost no one can get much more than 30Mb/s from most any network, it doesn’t matter that in the edge cases, there’s a difference. So if one modem can get to 800Mb/s, and another to 500Mb/s, the real time speeds where the real download speed is 25Mb/s will be the same.
Cores are, by far, the most important and difficult portion of an SoC to design. By comparison, a NPU is simple to create compared to a CPU.
<snip>
If it is so easy, why is Apple behind?
Lol, wtf are you talking about???!!! Apple’s new NPU had 800% performance improvement over their 1st one. It’s ASTOUNDING. By far the most impressive neural chip on the planet. Pretending it’s “behind” anything is going to be about as difficult to convince people of as telling people Michael Phelps needs swimming lessons. You're tripping.
Up until a couple of years ago, it was thought that 5nm would be so difficult, from a cost standpoint, a design and fabrication standpoint, and from the limiting factor of just pure physicstht said:
"HiSilicon advised it was planning to spend a minimum of $300 million on a system-on-chip design using 7-nanometer technology, illustrating the potentially high costs involved in working at that level." I have no idea what the author is talking about here. Why would it cost any more to *design* at 7nm than at 10nm? I'm pretty sure the high costs/complexity is mostly on the manufacturing side.
What do you mean? He said it has a high cost, doesn’t say that 10nm or larger weren’t also expensive to develop. I think he’s trying to make a point that shrinking your designs costs a massive amount of money.
Don’t understand the questions. You can think of the cost of design for a single transistor as a constant cost per transistor. The higher density chip plants enable more transistors per chip, thereby costing more to design the chip.
Those bigger NPU, GPU, CPU, ISP, SE, power management units don’t come for free. Every additional transistor will have some cost to it during design time. There are 50% more transistors in the A12 versus the A11 or something like that. And, I would bet those 50% more transistors cost more to put into a design than the prior increase in transistors from the A10 to the A11.
A number of years ago AMD said that the cost of a transistor on 14nm was more than on 22nm, and that it would just get worse as size went down. That’s true. Chip plant costs have doubled from 22nm to 10nm, and keep rising. It’s the major reason why Global Foundries dropped out.
a big problem is that computer sales have flattened. Previously, higher costs were mediated by higher sales, but that’s ending. A big question is whether the many more, smaller chips produced on a wafer, offset the lack of increased sales of much larger, much more expensive chips. The wafer size remains at 300 mm. The industry was preparing to move to 450 mm wafers years ago because of the supposed lowered costs per square mm, but it never happened, because those savings couldn’t be proven, as a result of the equipment costs for such were so much higher.
so we’re stuck.
Yeah, I agree with the folks here who are saying costs per transistor are only going up for the smaller nodes, with some increasingly complex design in the transistors and circuits themselves to account for quantum effects, increasing power leakage, thermals, etc. And, there probably will have to be software to correct for some of those effects too, which will be quite strange. I mean something like applying statistics to determine if a transistor did indeed change states, because in the quantum world, you never really know, and if software is being run a on a non-determinate machine, how the heck would we know the computations would be right in the first place to make the correction?
Samsung’s 10nm fab supposedly cost $17b. Intel’s still not working 10nm fab is probably $30b now, after what, 6 years of working on it? It’s like government sized capital investments now. If the USA gov’t was smart, billions of R&D dollars should have been awarded, should be continually awarded, to help get to 7nm, 5nm and 3nm nodes. There can only be one fab technique with Moore’s Second Law likely going exponential rather than the (approximate) linear doubling of costs every four years that we’ve seen.
And maybe 450 mm wafers at 10nm fab densities might be a real outcome if <7nm densities don’t pan out? The PC market is on a slow decline, but smartphones, servers, wearables, implantables, whatever, still has some cycles left in it.
Up until a couple of years ago, it was thought that 5nm would be so difficult, from a cost standpoint, a design standpoint, a fabrication standpoint and from the limiting factor of pure physics, that it could be impossible to accomplish. Now, it’s certainly believed , by the few manufacturers who are announcing their intentions, that I hope it will work out. But 3nm still seems impossible. The laws of physics alone seems to rule it out.
the future now is dependent on new technology, such as spintronics, molecular computing, and possibly carbon nanotubes. All of which are not expected to show any kind of real world promise before the middle of the next decade, and no commercial products before the end of the decade, or later, if progress slogs down. If those don’t get out the door, nobody knows what happens next other than a major hardware slowdown.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
So they designed all the easy stuff. When they design their own CPU or GPU cores then I'll be impressed.
Cores are, by far, the most important and difficult portion of an SoC to design. By comparison, a NPU is simple to create compared to a CPU.
If it is so easy, why is Apple behind?
Behind? Sorry, troll, it's Huawei that's behind.
Patiently waiting for your cherry-picked example of where Apple is behind while ignoring the major areas where they are far ahead.
Yes. Behind.
On the 'easy stuff'. Your words. Not mine. No need to cherry pick. You yourself jumped in with that claim.
I see you have no answer. May I suggest it isn't as easy as you imagine?
You understand why Huawei might know a bit more about those fields than Apple, right?
I would agree that Huawei has more cellular capability than Apple and it's partner, and has implemented into smartphones systems that exceed what carriers can provide, other than a few urban sites; hence, really not that big of a deal from a feature standpoint, at least at this time.
Everything else though, Apple is either equal to or better than Huawei's technology.
I think it is fine the Huawei teamed with Zeiss to implement the triple camera system, and credit where due, it's the first of its kind in volume production. Then again, Apple has a substantial in-house imaging team and IP to accomplish the same, and as it has been designing its own optics, optical stabilization, and focus systems, I'd say that they are doing fine.
Then we have to look at all of the processors and sensors that Apple has created for its wearables, because frankly, they lead in that as well. Given the number of letters in the Alphabet that Apple has already used for these, I'd say that they definitely have an advantage, over everybody, for the record.
In the meantime, I'm assuming that Huawei has an ARM architectural license, so we might see the first custom cores in 2019; great progress, but, not yet at the level of Apple or Samsung, and we still haven't seen the A12X yet.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
So they designed all the easy stuff. When they design their own CPU or GPU cores then I'll be impressed.
Cores are, by far, the most important and difficult portion of an SoC to design. By comparison, a NPU is simple to create compared to a CPU.
If it is so easy, why is Apple behind?
Behind? Sorry, troll, it's Huawei that's behind.
Patiently waiting for your cherry-picked example of where Apple is behind while ignoring the major areas where they are far ahead.
Yes. Behind.
On the 'easy stuff'. Your words. Not mine. No need to cherry pick. You yourself jumped in with that claim.
I see you have no answer. May I suggest it isn't as easy as you imagine?
You understand why Huawei might know a bit more about those fields than Apple, right?
Do you think repeating lies over and over will make them true?
The only thing Apple doesn't make is the cellular modem. They make their own WiFi, Bluetooth, ISP, DSP and NPU as well. Oh, wait, Apple also makes their NVMe storage controller (nobody else does this and even the 980 relies on the vastly inferior UFS). Let's see, what else.....oh yeah, Apple also makes their own inline hardware encryption module. Including the CPU and GPU that's 4 more items Apple designs that Huawei doesn't.
And you have the nerve to claim Huawei is ahead because they put a modem in their SoC?
And yes, a CPU and GPU are FAR MORE difficult to make than all those other listed components. Which is why we have countless companies that make all those very same components and only a couple that make CPU's or GPU's.
But you still didn't answer the question!
Always trying to control the narrative. Such arrogance. Why should I waste time answering anything you ask when you intentionally ignore points I’ve made?
Kuo writes the move to the 7-nanometer Kirin 980 helps Huawei narrow the gap between itself and Apple "in terms of user experience," as well as differentiating its products from other Android-based smartphone vendors.
How can it help them "narrow the gap" when they're still relying on an inferior operating system and App ecosystem?
In theory, Google will have the same advantage, given time, but I'm not sure they can exploit it and keep Android OS nominally "open".
From its inception, Android was only a response to an already existing iOS and iPhone. Google got it wrong from the start (and had to quickly and hectically correct mistake after that first Jobs presentation) and they continue getting it wrong every time and still need Apple’s “inspiration” to stay on the path. And that is for one simple reason - they do not have a coherent vision/plan....just like good old Sammy,
iOS wasn't even a rumor much less "a thing" when Google began developing Android as a smartphone OS.
Is that why Google completely missed the mark and had to start from scratch after Apple presented the iPhone 1? My bet is Google got someone at Apple to leak a bit of info on what Apple was doing. It wasnt a lot of info, as Apple was very secretive (for a good reason), so Google eng team had to invent a lot of things and they did....and missed the boat.
You aren’t by any chance referring to ‘the Mole?’
Eric the Mole who I assume you're referring to didn't arrive at Apple until late summer of 2006, August to be precise. He could hardly have "leaked" any Apple iPhone plans in late 2004/early 2005. The iPhone was all but done by the time he joined Apple.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
So they designed all the easy stuff. When they design their own CPU or GPU cores then I'll be impressed.
Cores are, by far, the most important and difficult portion of an SoC to design. By comparison, a NPU is simple to create compared to a CPU.
If it is so easy, why is Apple behind?
Apple is often behind in many of the superficial benchmarks that people look at. They don't care if they have the most RAM because they've put in the hard work to make their OS more efficient and design their own chips right down to the core in order to further take advantage of that synergy all to allow for better efficiency.
That doesn't change the observation. If this is the 'easy stuff' (and let's be honest, modems and WiFi are key smartphone areas), why hasn't Apple corrected the situation?
Because Apple goes after the hard parts first. A best in class SoC is harder than a modem and offers *much* more return on investment. Next.
"HiSilicon advised it was planning to spend a minimum of $300 million on a system-on-chip design using 7-nanometer technology, illustrating the potentially high costs involved in working at that level." I have no idea what the author is talking about here. Why would it cost any more to *design* at 7nm than at 10nm? I'm pretty sure the high costs/complexity is mostly on the manufacturing side.
Since wafer manufacturers are approaching the limits of fundamental physics with regard to circuit density, I will posit that wafer manufacturability is 100% contingent on the physical design of the integrated circuit. They’re probably having to experiment with a variety of design approaches using width, thickness, layering, different material properties and doping to achieve any kind of performance and reliability. It’s always been hard to miniaturize. But it’s harder than ever now and takes more resources to accomplish.
Being design enginner in that area (only originally and not active for years) I heard that story already in '90. Good fairy tale. Not to be bought. The real limit is temporary with tools.... as usually. Yeah phuysics limits... and then we went magnitudes smaller over decades.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
The 980 was benchmarked a few days ago, assuming that there wasn't a cheat anyway.
Huawei Mate 20 Armed With Kirin 980 Thrashes the Competition in the Latest Geekbench Run
Kirin 980 Manages to Outdo the Android Competition but Fails to Surpass the Scores of the A12 Bionic
I agree with you that it was a solid upgrade from the Kirin 970,
Considering that Samsung and Qualcomm are sampling their 7nm SOC's and will be shipping in a few months, this is one of the few opportunities that Huawei has to outshine the competition in performance, but only for a few months.
Make the best of it!
Apple has the A12X for the iPad Pro's still to announce, possibly a week or so after Huawei's Mate P20 announcement.
"HiSilicon advised it was planning to spend a minimum of $300 million on a system-on-chip design using 7-nanometer technology, illustrating the potentially high costs involved in working at that level." I have no idea what the author is talking about here. Why would it cost any more to *design* at 7nm than at 10nm? I'm pretty sure the high costs/complexity is mostly on the manufacturing side.
Since wafer manufacturers are approaching the limits of fundamental physics with regard to circuit density, I will posit that wafer manufacturability is 100% contingent on the physical design of the integrated circuit. They’re probably having to experiment with a variety of design approaches using width, thickness, layering, different material properties and doping to achieve any kind of performance and reliability. It’s always been hard to miniaturize. But it’s harder than ever now and takes more resources to accomplish.
Being design enginner in that area (only originally and not active for years) I heard that story already in '90. Good fairy tale. Not to be bought. The real limit is temporary with tools.... as usually. Yeah phuysics limits... and then we went magnitudes smaller over decades.
If it’s a fairy tale then please elaborate and share your wisdom. Enquiring minds want to know.
By the way, while I never did silicon, I used to do schematic capture and CCA layout back in the day. Different ballgame I know. But I’m not a complete ‘ijit’ with regard to this stuff.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
Oh look, our favorite knock-off cheerleader is here bragging about yet another chinese knockoff that doesn't even exist yet. "But it's better than Apple!!"
Can't wait until you find some other home to waste time on.
"HiSilicon advised it was planning to spend a minimum of $300 million on a system-on-chip design using 7-nanometer technology, illustrating the potentially high costs involved in working at that level." I have no idea what the author is talking about here. Why would it cost any more to *design* at 7nm than at 10nm? I'm pretty sure the high costs/complexity is mostly on the manufacturing side.
Since wafer manufacturers are approaching the limits of fundamental physics with regard to circuit density, I will posit that wafer manufacturability is 100% contingent on the physical design of the integrated circuit. They’re probably having to experiment with a variety of design approaches using width, thickness, layering, different material properties and doping to achieve any kind of performance and reliability. It’s always been hard to miniaturize. But it’s harder than ever now and takes more resources to accomplish.
Being design enginner in that area (only originally and not active for years) I heard that story already in '90. Good fairy tale. Not to be bought. The real limit is temporary with tools.... as usually. Yeah phuysics limits... and then we went magnitudes smaller over decades.
Well, back with Intel’s Netburst technology (really, just a lengthened instruction line) they thought they could get away with anything. 10, even 15GHz was going to happen. That crashed down at 90nm. But with clever design and different materials, we’ve gotten to where we are. But yes, indeed, we’re near the end. Quantum effects will assure that. No more cleverness can overcome the laws of physics. The dia. of the typical atom used in semiconductor work, is about 0.50 angstrom. That leaves very few atoms making up a line in a process size of 3nm. The problem is that now almost all of the electrons traveling that lead are at, or near the edges. Quantum effects assure that a lot of this will tunnel out, causing havoc.
just because most effects were mitigated earlier, doesn’t mean that they will still be able to be mitigated.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
We’ll se when they actually test it, but the truth is that if they continue doing what they’ve been doing, then those tests will be embarrassing, again.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
The 980 was benchmarked a few days ago, assuming that there wasn't a cheat anyway.
Huawei Mate 20 Armed With Kirin 980 Thrashes the Competition in the Latest Geekbench Run
Kirin 980 Manages to Outdo the Android Competition but Fails to Surpass the Scores of the A12 Bionic
I agree with you that it was a solid upgrade from the Kirin 970,
Considering that Samsung and Qualcomm are sampling their 7nm SOC's and will be shipping in a few months, this is one of the few opportunities that Huawei has to outshine the competition in performance, but only for a few months.
Make the best of it!
Apple has the A12X for the iPad Pro's still to announce, possibly a week or so after Huawei's Mate P20 announcement.
We’ll see how anandtech’s custom tests do. The interview with ‘Huawei on cheating is very interesting. How Huawei is handling this on the 980 will also be interesting to see, now that they’ve been outed. So the question is what do those tests mean? How real are they?
Kuo writes the move to the 7-nanometer Kirin 980 helps Huawei narrow the gap between itself and Apple "in terms of user experience," as well as differentiating its products from other Android-based smartphone vendors.
How can it help them "narrow the gap" when they're still relying on an inferior operating system and App ecosystem?
In theory, Google will have the same advantage, given time, but I'm not sure they can exploit it and keep Android OS nominally "open".
From its inception, Android was only a response to an already existing iOS and iPhone. Google got it wrong from the start (and had to quickly and hectically correct mistake after that first Jobs presentation) and they continue getting it wrong every time and still need Apple’s “inspiration” to stay on the path. And that is for one simple reason - they do not have a coherent vision/plan....just like good old Sammy,
iOS wasn't even a rumor much less "a thing" when Google began developing Android as a smartphone OS.
Is that why Google completely missed the mark and had to start from scratch after Apple presented the iPhone 1? My bet is Google got someone at Apple to leak a bit of info on what Apple was doing. It wasnt a lot of info, and as Apple was very secretive (for a good reason).
My guess is that you're a lousy guesser. Go back and do a little history searching.
Sorry, but Anton is correct. Android. As envisioned by Andy Rubin, and Google, after they bought Rubin’s company, was the Blackberry. This is so well understood, that reading anything to the contrary isn’t worth the time.
and yes, Google pretty much started over with the UI after the iPhone was shown.
LOL... You didn't even bother reading what he wrote before saying he was right did you? Go back and look at his post, #14. Anton said, quoting: " From its inception, Android was only a response to an already existing iOS and iPhone."
Read your history.... At Android's inception there was no iPhone. There was no iOS. But there was Microsoft.
The Kirin SoCs use standard ARM technology and cores. The performance is nothing to write home about, as the expression goes.
They also use designed in-house WiFi, modem, ISP and DSP etc. There is more to a SoC than cores.
Meh, their chips don’t have a great reputation. The 970 is behind the i35, much less the contemporary 845. The 980 will be behind Apple, Qualcomm and Samsung, as usual. Their AI section is a couple of years behind too. Not impressed.
its good that they try though. Someday they’ll get it right.
Strange conclusion.
Last year's Kirin 970 had better NPU performance and a Cat 18 modem. The best phone of the year on many fronts is still the P20 Pro - which has a Kirin 970. That's hardly 'meh' worthy. Apple, even with its new SoC is just catching up with the modem, but is still lagging.
The Kirin 980 has a Cat 19 modem and an ultra fast (I believe it's -or will be shortly- the fastest on a phone) wi-fi chipset. The co-processors, dual ISPs and DSP are sure to be no slouches either. That puts your 'still behind' comment into context.
You realise that the 970 was shipping around five months before the 845 and Huawei deliberately chose to launch without the latest ARM cores for questions of maturity and stability? Are you surprised that the 845 produces better numbers? That didn't stop them from releasing the phone of the year.
I don't know what 'don't have a great reputation' is supposed to mean.
Not strange at all, go look at the tests yourself. Oops! Be careful as most of them add to the results because of Huawei’s cheating. Even so, it not great.
its funny that you’re so sure about the performance of this chip that isn’t out yet, but by the specs, will not be a top performing device. Huawei can’t produce their own cores, so they use whatever ARM licenses directly. Don’t make pitiful excuses.
Excuses? Nope. This is a solid upgrade but don't take my word for it:
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
Oh look, our favorite knock-off cheerleader is here bragging about yet another chinese knockoff that doesn't even exist yet. "But it's better than Apple!!"
Can't wait until you find some other home to waste time on.
Rumours are rumours and should be taken for what they are. The Kirin 980 is real. The Cat 21 modem is real. The dual frequency GPS is real. The WiFi chip is real. The ISPs and DSPs are real.
I think most people actually accept this.
Incessantly calling everything a knock-off just shows how blind you are to reality but that is your problem. That reality is, and has been, staring you in the face for a while. You just refuse to accept it. Perhaps you simply can't accept it.
Comments
Apple is way ahead of everyone else in all the areas that matter. Huawei isn’t even in the running for third place. Maybe forth, or fifth.
Apple’s new NPU had 800% performance improvement over their 1st one. It’s ASTOUNDING. By far the most impressive neural chip on the planet.
Pretending it’s “behind” anything is going to be about as difficult to convince people of as telling people Michael Phelps needs swimming lessons.
You're tripping.
Up until a couple of years ago, it was thought that 5nm would be so difficult, from a cost standpoint, a design standpoint, a fabrication standpoint and from the limiting factor of pure physics, that it could be impossible to accomplish. Now, it’s certainly believed , by the few manufacturers who are announcing their intentions, that I hope it will work out. But 3nm still seems impossible. The laws of physics alone seems to rule it out.
the future now is dependent on new technology, such as spintronics, molecular computing, and possibly carbon nanotubes. All of which are not expected to show any kind of real world promise before the middle of the next decade, and no commercial products before the end of the decade, or later, if progress slogs down. If those don’t get out the door, nobody knows what happens next other than a major hardware slowdown.
Everything else though, Apple is either equal to or better than Huawei's technology.
I think it is fine the Huawei teamed with Zeiss to implement the triple camera system, and credit where due, it's the first of its kind in volume production. Then again, Apple has a substantial in-house imaging team and IP to accomplish the same, and as it has been designing its own optics, optical stabilization, and focus systems, I'd say that they are doing fine.
Then we have to look at all of the processors and sensors that Apple has created for its wearables, because frankly, they lead in that as well. Given the number of letters in the Alphabet that Apple has already used for these, I'd say that they definitely have an advantage, over everybody, for the record.
In the meantime, I'm assuming that Huawei has an ARM architectural license, so we might see the first custom cores in 2019; great progress, but, not yet at the level of Apple or Samsung, and we still haven't seen the A12X yet.
Always trying to control the narrative. Such arrogance. Why should I waste time answering anything you ask when you intentionally ignore points I’ve made?
”The Kirin 980 really does look like it’s on its way to be an exceptionally well balanced SoC with major improvements in every regard"
https://www.anandtech.com/show/13298/hisilicon-announces-the-kirin-980-first-a76-g76-on-7nm/2
This is the first SoC to use ARM's latest cores and GPU. I take it you are suggesting they are 'nothing to write home about'. The rest is largely Huawei produced, including the modem, ISP, DSP, etc
I believe the modem is actually Cat 21.
I'm not 'so sure' about the performance of the chip but from the specs alone, we can safely assume that the modem and WiFi will deliver. Do you doubt that things like dual frequency GPS won't offer any improvements?
As for the phone the SoC will debut on, even if only half the rumours are true, it will be anything but 'meh'.
https://wccftech.com/huawei-mate-20-kirin-980-geekbench-scores/
Huawei Mate 20 Armed With Kirin 980 Thrashes the Competition in the Latest Geekbench Run
Kirin 980 Manages to Outdo the Android Competition but Fails to Surpass the Scores of the A12 Bionic
I agree with you that it was a solid upgrade from the Kirin 970,
Considering that Samsung and Qualcomm are sampling their 7nm SOC's and will be shipping in a few months, this is one of the few opportunities that Huawei has to outshine the competition in performance, but only for a few months.
Make the best of it!
Apple has the A12X for the iPad Pro's still to announce, possibly a week or so after Huawei's Mate P20 announcement.
If it’s a fairy tale then please elaborate and share your wisdom. Enquiring minds want to know.
By the way, while I never did silicon, I used to do schematic capture and CCA layout back in the day. Different ballgame I know. But I’m not a complete ‘ijit’ with regard to this stuff.
Can't wait until you find some other home to waste time on.
just because most effects were mitigated earlier, doesn’t mean that they will still be able to be mitigated.
We’ll se when they actually test it, but the truth is that if they continue doing what they’ve been doing, then those tests will be embarrassing, again.
https://www.anandtech.com/show/13318/huawei-benchmark-cheating-headache
and:
https://www.anandtech.com/show/13285/huawei-gpu-turbo-investigation
So, heavy claims by Huawei in their presentation. Sure, it sounds impressive, as it does every year. But the reality is something very different.
Anton said, quoting: " From its inception, Android was only a response to an already existing iOS and iPhone."
Read your history.... At Android's inception there was no iPhone. There was no iOS. But there was Microsoft.
I think most people actually accept this.
Incessantly calling everything a knock-off just shows how blind you are to reality but that is your problem. That reality is, and has been, staring you in the face for a while. You just refuse to accept it. Perhaps you simply can't accept it.
Apple is behind in the areas I mentioned.
Why, oh WHY, do a tiny group of people have so many problems accepting that AI can even mention anything beyond Apple?
Why, oh WHY, do those same people have such a hard time accepting Apple might not be the best at something?
Why, oh WHY, can't they just accept the facts and live with them?
Why, oh WHY, must they ALWAYS go on the defensive?
Yes, why, oh WHY!
The new Pixels will be here soon. Expect a fair bit of non-Apple Pixel related news on AI.
Soon after that, the Mate 20 series will arrive. There might even be a word on that too.
Will you be praying for Pixel Insider too?
I fear I know the answer to that one already.