BeSang Inc
BeSang Inc
  • 17
  • 32 092

Comments

  • FPGA KING
    FPGA KING 8 months ago

    잘 봤습니다, 응원합니다.

  • Peter Kersting
    Peter Kersting Year ago

    adoption rate is real.....not only for 3D NAND, but all new technologies...once hype is over, money and efficiency talk

  • Sacco Belmonte
    Sacco Belmonte Year ago

    So, it's manufacturer adoption of the 3D Super NAND design a reality at the end of 2017? It all looks very impressive on paper.

  • 송영진
    송영진 Year ago

    thanks for your good presentation!.

  • klam77
    klam77 Year ago

    New Dram is nullfied by New OS.......a long running game! end of moore compound the problem

  • Chris Welch
    Chris Welch 2 years ago

    Very, very tasty analysis for a nerd like me

  • Michael Schmitz
    Michael Schmitz 2 years ago

    As I understood the big manufacturers never intended to stack the layers up to 128, 256 etc. but connecting smaller stacks together what in the video is called 3D Super-NAND. 3D NAND already uses that technology.

    • James Clarity
      James Clarity 10 months ago

      Clearly, the future is here!!

    • Tanj Bennett
      Tanj Bennett 2 years ago

      BeSang does not propose to use the multilayer vertical stacks. If you look at their web page you will see their layers are a 2D NAND with CMOS-under-cells, allowing cheap CMOS process for the base, combined with optimized high density vertical channel (like the "wire channels" proposed for 7nm) for the NAND cells, built in a separate add on epitaxial layer. The epitaxials can be added up to a relatively modest count by adding vias connecting down to the base CMOS. I'm not an expert to say if it is everything claimed, but it certainly differs from (3D) V-NAND. besang.com/structure.html

    • BeSang Inc
      BeSang Inc 2 years ago

      3D NAND has poor cell density. So, it should add more layers to be cost competitive. Big manufacturers are making "string stacking" which combines multiple single string to implement high number of cell stacking. For example, "36-layer + 36-layer" to implement 72-layer by SK hynix and "48-layer + 48-layer" to implement 96-layer by SanDisk.

  • PLOBEXRIME
    PLOBEXRIME 2 years ago

    So standard NAND is safer to store data if compared to 3D NAND (not 3D Super-NAND) ?

    • BeSang Inc
      BeSang Inc 2 years ago

      3D NAND is better for now. When 3D Super-NAND comes to the market, I will certainly recommend 3D Super-NAND.

    • PLOBEXRIME
      PLOBEXRIME 2 years ago

      Thanks for answer, but my question was, which is safer to store data ? I mean less probability of data loss/failure due to cells interference. If bigger distance means less interference, then 3D Super-NAND is best. But is planar NAND (15nm) still better than 3D NAND (13nm) ? BTW great video !

    • BeSang Inc
      BeSang Inc 2 years ago

      3D NAND is better than planar NAND because it could control more gate charges. 3D Super-NAND is much better than planar and other 3D NAND.

  • Ferdinand Aggenbach
    Ferdinand Aggenbach 2 years ago

    Best analysis of 3D XPoint on RU-clip to date!

  • James Han
    James Han 2 years ago

    Thank you Besang!!! 3D xpoint doesn't have much information in market.. and got some clue to understand the technology.

  • Mike
    Mike 3 years ago

    How do you see the HybriDimm from Netlist competing in this area? They announced at the Flash Memory Summit and it appears to be aimed at 3D Xpoint. insidehpc.com/2016/08/netlist-hybridimm-memory-unifies-dram-nand/

    • BeSang Inc
      BeSang Inc 3 years ago

      I think Netlist figures out smart way to allocate data between DRAM and NAND at the DIMM slot in order to increase performance of system. Though, it will be relatively slow storage class memory. 3D XPoint will work as main memory at NVDIMM, which will replace DRAM. HybriDimm cannot replace DRAM simply because NAND is too slow. So, it is fair to say that HybriDimm is fast NAND and 3D XPoint is main memory which will replace DRAM. HybriDimm is not a 3D XPoint killer.

  • Tobi Tobey
    Tobi Tobey 3 years ago

    Thank you, Mr. Lee. This presentation is an eye-opener to the issues with 3d NAND production and cost. I learned much.

  • Abe Dillon
    Abe Dillon 3 years ago

    It's confusing that you switch between comparing 3D XPoint to MLC and SLC NAND. How is it "fair" to claim it's only 10x more durable than NAND and 5x more expensive than NAND when you're talking about two very different implementations of NAND? EDIT: Also, when comparing to DRAM you leave out the difference that DRAM is volatile. It's not just 2x bits/dollar vs. DRAM, it's instant-on computers. It would be much better if you started with your comparison to 3D Super-NOR. You spend 11 minutes on a bogus comparison to NAND and DRAM then 1 minute introducing the interesting info about 3D Super-NOR. Not a great strategy.

    • BeSang Inc
      BeSang Inc 3 years ago

      Thank you for your input. The confusion starts from the claims of 3D XPoint. It says 1,000x endurance of NAND. In this case, I am saying that it should be fair to compare 3D XPoint with SLC NAND because endurance strongly depends on storage level of cell. Regarding the cost, it is reasonable to compare MLC NAND with 3D XPoint because Optane should mainly compete with SSD with MLC NAND or SSD with 3D NAND in the SSD market. Endurance is not a big issue for general use of SSD with MLC NAND. For fair comparison, the performance between Optane and SSD with MLC NAND also should be measured at the same PCIe interface. Intel/Micron mention that 3D XPoint is not low cost memory. I agree that 3D XPoint is non-volatile which is better than DRAM when there is no power supply. However, price advantage of 3D XPoint seems small compared to DRAM for main memory application and SSD with MLC NAND for storage class memory application. Unlike other NAND-based NVDIMM, DDR4 NVDIMM is an unique feature of 3D XPOint and NOR flash memory. When we talk about 1/2 price of DRAM, it will come from 2nd generation of 3D XPoint which has 4-layer of stacked memory cells. Even though 3D XPoint has advantages as a DDR4 NVDIMM, I think cost might be the biggest issue for wide adoption of 3D XPoint as NVDIMM. I did not share details of 3D Super-NOR in this tutorial. Sorry about that. There will be another chance to introduce 3D Super-NOR in detail at later time.

  • OnBoard
    OnBoard 3 years ago

    I'll take 2TB please (or 2x2TB if it's really 100$). Now somebody just needs to make it as a drive.

  • Im1CrazyCow
    Im1CrazyCow 3 years ago

    Heck I want stock in y'all because I can see how this is going to Break the back of other Memory manufactures when it hits, If you get some investment capital why not just make them your self and just dominate the Chip memory market. I will do my best to get this info up to one of the top You Tubers and on their weekly Wan Show. So glad I have the first post as I see the massive potential. Cow }:-o)

  • thoughtbox
    thoughtbox 3 years ago

    I really like your videos!! Keep making them.

  • pasta514
    pasta514 3 years ago

    Why does Intel call it 'optane'? because it is very hard to OBTAIN.