I saw an antutu score on the M9 for 51,727 after the throttling update was pushed. The Zenfone 2 got a score of 50,401 LOL. Qualcomm really screwed their customers with the 810.
I know this guy has industry insights, but it's got to be a touch embarrassing to write something like this. The S810 in the G Flex 2 and the M9 is underwhelming - not just from a throttling perspective but also efficiency.
The paragraph posted above is especially hilarious now that the S6 is in reviewer's hands and the 7420 looks as promised.
Good luck on that 16nm silicon. Same as it having a A57 cluster.Huawei revealed some details about their new Kirin 930 SoC. They are using Cortex A53e + Cortex A53 in a little.LITTLE configuration. It was previously rumoured that they would be using Cortex A57 + A53 instead (Kirin 920 packs Cortex A15 + A7). They also gave some hints about Cortex A72 performance, power and die size in one of their slides.
With 8 Cortex A53 cores and a Mali T628MP4 this chip should be very small @ TSCM 16nm FF. That leaves Samsung and Qualcomm as the only companies with Cortex A57-based phone SoCs (unless I'm missing something?).
Good luck on that 16nm silicon. Same as it having a A57 cluster.
The MediaPad X2 with the 930 was already on display during MWC and I played with it for a minute (Didn't know it had the 930 at the time else I would have looked more into it), you would imagine HiSilicon would be shouting from the rooftops if they have a 16nm SoC not only in production, but already in a product. Look at the die size in those slides. I'll go out buy a hat and then eat it if this thing is anything other than 28HPM. The chip has been on GeekBench since January and it scores like only 900-ish on single-thread.
The incorrect media hype around this thing is insane.
Good luck on that 16nm silicon. Same as it having a A57 cluster.
The MediaPad X2 with the 930 was already on display during MWC and I played with it for a minute (Didn't know it had the 930 at the time else I would have looked more into it), you would imagine HiSilicon would be shouting from the rooftops if they have a 16nm SoC not only in production, but already in a product. Look at the die size in those slides. I'll go out buy a hat and then eat it if this thing is anything other than 28HPM. The chip has been on GeekBench since January and it scores like only 900-ish on single-thread.
The incorrect media hype around this thing is insane.
Google Translate said:Came out of the mongoose jjirasi (Mongoose), known as the custom core of Samsung
Geek Bench 3 single score 2240 points
The layout of this is quite unusual – normally the memory marking (SEC 507 etc.) is in lines of text above the APU marking (7420 etc.), not in a diagonally opposed block.
Which leads me into the speculation that maybe the 7420 is out of GLOBALFOUNDRIES, rather than a Samsung fab in Korea or Texas. Could ALB be short for Albany (NY)? Is the G in the lot code short for GLOBALFOUNDRIES? That all seems rather unlikely, but if Samsung wants to switch on the volume quickly in anticipation of huge volumes for the S6, what better way than to use three fabs? They did sound very confident in their last quarterly analyst call, saying that they expect 14-nm to be 30% of the LSI line capacity by year end. And there are lots of rumours about Qualcomm using the Samsung 14-nm process..
..
The functional die size is ~78 mm2, which compares well with the 118.3 mm2 of the Snapdragon chip used in the Galaxy S5.
Cross-posting this new information here. (Originally from Mobile & Gadget forum)
Finally some concrete information on the guts.
Chipworks: Inside the new Samsung Galaxy S6
Exynos 7420 is only ~78mm² (!) and the chips are possibly manufactured by Global Foundries..(?!)
Qualcomm's next-generation application processor (AP), the Snapdragon 820, is expected to outsource to Samsung Electronics using a 14nm node, which the Korea-based foundry house has demonstrated as a proven process as showcased by the performance and power consumption of its14nm Exynos 7420 CPU developed in house, according to Digitimes Research.
Since Samsung has also been aggressively striving for orders with competitive pricing, other players such as Nvidia, AMD and MediaTek are believed to have a high chance of shifting part of their orders to Samsung, which in turn could affect the global mobile AP market in 2015 and 2016, Digitimes Research commented.
Qualcomm was forced to accelerate the roadmap of its mobile AP following a mishap of the Snapdragon 810 and is also being pushed to fabricate its Snapdragon 820 via Samsung's 14nm process as the development of a similar advanced 16nm process by Taiwan Semiconductor Manufacture Company (TSMC) is currently lagging Samsung's 14nm node by about one quarter, Digitimes Research believes. [...]
Some info about Samsung's custom ARM core. Codenamed Jirasi (Moongose) it scores 2240 @ Geekbench single-core operating at 2.3GHz.
Here's hoping this monster arrives earlier than the Galaxy S7 (perhaps with the Galaxy Note 5).
Some comparisons:
- Exynos 7420: ~1400-1500
- A8X (iPad Air 2): ~1800
- Tegra K1 Denver (Nexus 9): ~2000-2100
Ps: Mongooses eat snakes, gotcha Qualcomm (Krait, etc.)?
http://samgukji.net/won/link/?item_no=783753&ckattempt=2
Also I just updated my Exynos Galaxy Note 4 to Lollipop (OTA) and I will post some S810 and Exynos 7420 comparisons soon. At first glance CPU scores have improved a bit across the board (compared to KitKat).
About ~71% the size of 20nm Exynos 5433 with a larger GPU (Mali T760MP8 vs T760MP6). This leaves room for some big custom CPU cores (and GPUs) for the Galaxy Note 5 and Galaxy S7.
2 or 3 Mongoose/Jirasi cores + Mali T880MP6/MP8 (coupled with 4GB LPDDR4) would make one very fast SoC later this year.
Related news/rumour:
Samsung expected to see increasing orders on 14nm process
www.digitimes.com/news/a20150402PD204.html
How is that 71% possible with a larger gpu? Newer gf/ss tech..hdl??? Density should not improve like that for 14nm.
How is that 71% possible with a larger gpu? Newer gf/ss tech..hdl??? Density should not improve like that for 14nm.
Samsung's 20nm process uses 80nm 2D interconnect while the new 14nm
process uses 64nm 2D interconnect. The total Quad A57 Cluster plus L2
seems to be less than 9 mm2 down from 14 mm2 in the Exynos 5433.
It's still hidden under the top-layer. Next week we'll get to know more
from ChipWorks.
So, literally what Intel did with its 22nm process? 90nm gate pitch and 80nm minimum metal pitch?
You're saying ChipWorks measured the underlying layer?Samsung's 20nm process uses 80nm 2D interconnect while the new 14nm
process uses 64nm 2D interconnect. The total Quad A57 Cluster plus L2
seems to be less than 9 mm2 down from 14 mm2 in the Exynos 5433.
It's still hidden under the top-layer. Next week we'll get to know more
from ChipWorks.
ChipWorks has probably already given a bevel polish to the die so that they can look at the transistors in plan view.You're saying ChipWorks measured the underlying layer?
Our guys in the lab made their usual exceptional effort in enabling us to see what the process looks like within a few hours of getting the phone in-house, we have a decapsulated part and a cross-sectional sample under the microscope.
The Exynos 7420 uses 11 layers of metal, as you can see from the die seal cross-section above. Now lets look at the transistors:
You're looking at the cross-section of the die. Like if you'd cut (vertically) a brain in its two halves (hemispheres). The lowest layer, where you see the contacts and gates (not sure about the fins), is where the transistors are. Everything above that is the interconnect (BEOL): the copper wires that transmit the current.I have no idea what I am looking at so I will leave it up to more knowledgeable folks here.
I don't have them yet. I'll get back to you on that. I think David is wrong in his assumption though, I'm sure the CPUs are hitting the 650mV range at the low end. The GPU on the unit I played with at MWC had this voltage curve;Andrei, can you provide any insight on the CPU voltage ranges for Exynos 7420? The topic came up on RWT.
It is a full node shrink. The confusion was on the 5433 process being not the 20nm that we thought. The 7420 does not share the same BEOL. The theoretical shrink is around 44%, not 15%.@witeken: Thank you for the explanation. Not that I understood everything you wanted to teach, but I feel like I know more than before.
Anyhow, isn't 113mm² -> 78mm² in the realm of full node shrink, somewhat justifying Samsung's claim of 20nm -> 14nm? (78 ÷ 113 ≈ 0.69)
It is a full node shrink. The confusion was on the 5433 process being not the 20nm that we thought. The 7420 does not share the same BEOL. The theoretical shrink is around 44%, not 15%.
Around 0.69 yes.I am assuming the 44% shrink means the same as 70% the size of previous node? (1 ÷ 1.44 = 0.69) But I thought you said in an AT article that Samsung's 14nm is a half node sharing the same interconnects of its 20nm. (I could be wrong since I am not well versed in this area)