New 28nm A72 Exynos SoCs in 2016.

[ShintaiDK]

https://benchlife.info/samsung-exynos-7650-7880-will-take-28nm-to-fight-snapdragon-10192015/

Only the 7890 will be build using Samsungs 14nm Finfet.

 

[Arachnotronic]

Basically Snapdragon 620 competitor. Samsung might be looking to oust Qualcomm from more of its phones.

 

[NTMBK]

Not surprising, given FinFET costs. Hopefully some new low-cost, higher performance node comes along for these budget chips- some posters over on S|A are awfully excited about 22FDX, but I'm dubious given GloFo's track record. (EDIT: And the cost of SOI wafers.)

 

[Arachnotronic]

Not surprising, given FinFET costs. Hopefully some new low-cost, higher performance node comes along for these budget chips- some posters over on S|A are awfully excited about 22FDX, but I'm dubious given GloFo's track record. (EDIT: And the cost of SOI wafers.)

Global Foundries is all hat, no cattle.

 

[mrmt]

https://benchlife.info/samsung-exynos-7650-7880-will-take-28nm-to-fight-snapdragon-10192015/

Only the 7890 will be build using Samsungs 14nm Finfet.

It seems that 14 is becoming the new 13 number of the foundry industry and Samsung is experiencing a debacle of their own with their 14nm node. Let's see whether TSMC did everything right with their 16nm.

 

[mrmt]

some posters over on S|A are awfully excited about 22FDX, but I'm dubious given GloFo's track record. (EDIT: And the cost of SOI wafers.)

Can't blame them. They were excited about Bulldozer.

In any case, it's too much exciting for a node that not a single partner of the former Common Alliance wanted to license.

 

[NTMBK]

It seems that 14 is becoming the new 13 number of the foundry industry and Samsung is experiencing a debacle of their own with their 14nm node. Let's see whether TSMC did everything right with their 16nm.

Oh yeah, they just shipped how many iPhone processors? Total debacle

 

[mrmt]

Oh yeah, they just shipped how many iPhone processors? Total debacle

They couldn't fulfill their entire iPhone quota and they seem to be unable to manufacture low cost products with their 14nm. Seems that their 14nm was a bit rushed.

 

[ShintaiDK]

The cost perspective is the most interesting part.

 

[dark zero]

They couldn't fulfill their entire iPhone quota and they seem to be unable to manufacture low cost products with their 14nm. Seems that their 14nm was a bit rushed.

Just Like Intel and Broadwell end even Skylake

 

[khon]

Even if they can't get their 14nm FinFET technology to work, why not make these new SoCs with their 20nm technology ? 28nm seems really outdated at this point.

 

[Snafuh]

They couldn't fulfill their entire iPhone quota and they seem to be unable to manufacture low cost products with their 14nm. Seems that their 14nm was a bit rushed.

What was their iPhone quota and how many chips did they actually ship? I'm interested in seeing numbers about this.
In the meantime Samsung shipped ~30 million Galaxy S6 with 14nm chips since April.

 

[krumme]

Kirin 950
A72 2.4ghz 16nm ff
http://www.gsmarena.com/huawei_mate_8_leak_reveals_a_beastly_specs_sheet-news-14552.php

 

[ShintaiDK]

Even if they can't get their 14nm FinFET technology to work, why not make these new SoCs with their 20nm technology ? 28nm seems really outdated at this point.

Cost.

 

[Andrei.]

It seems that 14 is becoming the new 13 number of the foundry industry and Samsung is experiencing a debacle of their own with their 14nm node. Let's see whether TSMC did everything right with their 16nm.

There's no debacle with 14nm.

Anyway if these SKUs are correct then it's clear who and what Samsung is targeting. They only need to push the vendors for adoption.

 

[Tuna-Fish]

some posters over on S|A are awfully excited about 22FDX, but I'm dubious given GloFo's track record. (EDIT: And the cost of SOI wafers.)

22FDX is more of an IBM node than it is a GloFo node, so there's not really need to be quite as gloomy of it as there is of GloFo usually.

However, 22FDX is not cheap in that way. The selling point of the node is that masks and design are much cheaper than with the finfet nodes. This is a huge advantage if you only need to do a few thousand chips but need near top-end node performance. However, the per-wafer cost will be higher than with it's main competitors, so it's not a good fit for a low-cost high-volume product.

 

[redzo]

Basically Snapdragon 620 competitor. Samsung might be looking to oust Qualcomm from more of its phones.

As of now, the non-Samsung mobile device market is being dominated by MT and Qualcomm. Exynos scored its own(Samsung) device wins. Samsung deployed Exynos in its own devices and that is it.

If Samsung looses a tiny bit of mobile market share, Exynos is screwed.

 

[NTMBK]

22FDX is more of an IBM node than it is a GloFo node, so there's not really need to be quite as gloomy of it as there is of GloFo usually.

However, 22FDX is not cheap in that way. The selling point of the node is that masks and design are much cheaper than with the finfet nodes. This is a huge advantage if you only need to do a few thousand chips but need near top-end node performance. However, the per-wafer cost will be higher than with it's main competitors, so it's not a good fit for a low-cost high-volume product.

I thought that per wafer costs were reduced by avoiding multi patterning, and hence having fewer manufacturing steps? (Counter balanced by adding SOI.) Though I could well be wrong.

 

[geoxile]

What phones are going to use these? I assume these will be launching concurrently with a new flagship SoC using Mongoose cores.