Samsung Begins Mass Production of 10nm LPE SoCs

Subject: Mobile | October 17, 2016 - 07:01 AM |
Tagged: Samsung, 10nm

Earlier today, Samsung announced that mass production has started for system-on-a-chip (SoC) products on their first-generation 10nm process, which is called Low Power Early (10LPE). Chips produced from this node will begin to ship in devices starting early 2017. The press release claims that, for integrated circuits manufactured under the 10LPE process, die area could decrease up to 30%, with either an increase in performance of up to 27% or a decrease in power of up to 40%.

View Full Size

This is a little higher than the 10% increase in performance that AnandTech claimed in April. On the plus side, it was also expected that any design that was created for 10LPE could be migrated, pretty much without change, to the second-generation, Low Power Plus (10LPP) node. Jumping back to today's press release, Samsung claims that 10LPP will begin mass production in the second half of next year. So basically, early 10nm parts will launch in a couple of months, then a second wave will arrive the year after, using a more refined fabrication method.

Source: Samsung

Video News


October 17, 2016 | 10:46 AM - Posted by Anonymous (not verified)

Wow xbox scorpio might actually be 10nm by holiday 2017, maybe even lowcost hbm. 10nm mobile polaris/pascal??

October 18, 2016 | 11:59 AM - Posted by Anonymous (not verified)

This process is LPE, not LPP. OK for little mobile dies, not so good for the sort of jump dies APUs need, not for the low-margin-high-yield console market.

October 17, 2016 | 11:16 AM - Posted by Anonymous (not verified)

http://www.fool.com/investing/2016/10/01/intel-corporation-should-rename...

"From a marketing/public perception perspective, even if TSMC's 7-nanometer technology is "only" roughly equivalent to Intel's 10-nanometer technology, general perception may be that Intel is actually a full generation behind what TSMC is building."

Samsung & TSMC's 10nm process is just only reaching parity with Intel's 14nm process when it comes to density.

Don't be fooled by Samsung or TSMC's process names, they're not equal to industry leading Intel process.

October 17, 2016 | 11:34 AM - Posted by Anonymous (not verified)

Word. Were fucking light years ahead of everyone else. Hey anyone else on here in Hillsboro?

October 17, 2016 | 12:24 PM - Posted by Anonymous (not verified)

Who cares about the gate pitch its the gate size that gives most of the benefits! So Intel can cram a little more circuits onto a die with their process node, Intel will only use the space savings to get more dies per wafer while Samsung's customers will more often use any extra space for more features like better graphics. There is more competition in the ARM based devices market both custom ARM chips and Arm Holdings reference design chips. Samsung’s 10nm gates size will still have a power/other advantages no matter the gate pitch, and Intel is all about its margins so don’t look for Intel to be improving its SOC graphics anytime soon. Intel will never use its better graphics IP in its lower cost SKUs which is why AMD will have the better price/performance metric for available APU/SOC graphics at a much lower cost.

The only little advantage from Intel’s processes over the others processes is that Intel has a smaller gate pitch among little other features and you can be damn sure Intel will take most of the space saved and use it get more chips per wafer and not add any extra on die features! Intel will not be lowering its prices unless AMD’s Zen starts getting more market share. AMD will have much better Polaris based APU graphics so maybe Intel will be forced to improve its offerings but knowing Intel it will be costly!

October 17, 2016 | 02:36 PM - Posted by Anonymous (not verified)

A sad cry for help disguised as an attempt at insulting someone based on this posters personal failings.

October 17, 2016 | 05:45 PM - Posted by Anonymous (not verified)

No, it's reality for those that worship at the Intel altar to know that Intel will not give extra graphics to go along with the shrink atvantages it has. Intel will get more chips per wafer an milk that for all it's worth. This is Intel's plan to milk for margins at the cost of graphics.

AMD is the better choice for the better Graphics/Dollar metric. Zen APUs are coming and Intel will not do anything but keep its high end graphics on its most unused SKUs. Polaris graphics based APUs will be a whole level above what Intel will offer. The only crying will be Intel's stockholders when those dividends dry up!

October 17, 2016 | 07:12 PM - Posted by odizzido (not verified)

Personally I'd like intel just to drop their IGPUs on a bunch of processors and offer us a substantial discount for the significantly smaller die.....heh...haha....AHAHAHAHAHAHA yeah that's going to happen.

October 17, 2016 | 10:44 PM - Posted by Anonymous (not verified)

No Intel would not remove its graphics it will just keep the cash flowing to the laptop OEMs and the AMD laptop SKUs will be gimped with single channel only options. AMD really only needs to get a Zen/Polaris APU with a single stack of HBM2(4 or 8 GB) and the laptop OEMs will not be able to starve the APU's GPU of the necessary bandwidth. So even if the Zen/Polaris/HBM2 APU had a single channel to regular DIMM based DRAM it would not matter as the HBM2 would provide enough bandwidth from the included HBM2 stack for the GPU to feed from with any external DIMM based DRAM acting like a second tier RAM.

The single stack of HBM2 would act like a faster RAM Cache and any slower/single channel DIMM based DRAM would have its letency hidden by keeping gaming textures on the HBM2. It would not be very hard for AMD to configure any accesses to the HBM2 stack to be treated like a L4 cache with the most needed code/data/textures staged in the HBM2 and any memory transfers to and from DIMM based DRAM done in the background to keep the APU's graphics feeding from the HBM2 with its much greater effective bandwidth. Let's see Intel try and compete with an Zen/Polaris APU with its own stack of HBM2 feeding the APU's graphics cores, and also the Zen cores complex/s!

October 18, 2016 | 04:59 AM - Posted by Jann5s

I would love to see an APU with HBM2. I'm really hoping AMD will have the balls to create one. But then again, it would require software support, and we all know how well that worked out for bulldozer.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.