MRAM Update from IEDM and Zoom Call

 

Mark Webb, MKW VENTURES CONSULTING at IEDM, San Francisco Hilton Dec 11-15^th

^{ZOOM CALL Sunday and Monday evening}

^{QUICK update MRAM}

^MRAM:

^{Embedded MRAM is officially in the "It works, we implemented it, and now we need to optimize it for each application". This is true maturity working with each customer for what they need and what they will pay for.} 

1) Replace embedded flash (This should have happened long ago). e-Flash at any geometry is neither optimal for performance nor optimal for cost. e-Flash (which i worked on years ago) is a "its all we got, so we are stuck" technology. MRAM and possibly RRAM to the rescue

2) Replace/Augment Last Level Cache. Its not really cheaper or better than SRAM for pure cache yet... but it is within optimization timeframe (2 years). It needs to be about 30% smaller than current, and it needs to give up some retention for Bit error rate (BER) and speed .... but it is a possible option depending on applications

Discrete MRAM has made some progress with STT at Everspin but it is still a small market until we can show significant improvement over  NOR and DRAM/SRAM

We have a presentation and Zoom call covering new emerging memories.

**** Sunday 12/12 from 6PM to 630PM Pacific time AND

**** Monday 12/13 from 5:30-6PM Pacific time

**** ZOOM Meeting ID: 811 2666 3619     ZOOM Passcode  481708

We can also meet 1:1 or in a small group during the conference. Just text.

Topics:

* Emerging Memory Product Lifecycle: What steps are needed to deliver production memories and where does each memory stand? How can we tell whether a technology is 15 years from success or 2 years.  Note: Memories take 10+ years to mature, 90% don’t make it

* Revenue and scenarios for future for Optane/3DXPoint. Where are we today and what are the three possible scenarios for 5 years from now.

**the scenarios are named “I am so Proud”, “Groundhog Day”, and “I Won” **

* Revenue and projections for MRAM. Embedded and discrete.

* ReRAM status and challenges

* Status of other memories

 

Mark Webb

WWW.MKWVENTURES.COM

https://mkwventures-semi.blogspot.com/

https://mkwventures-nand.blogspot.com/

Emerging Memory Presentation at IEDM and Zoom Info

 

Mark Webb, MKW VENTURES CONSULTING at IEDM, San Francisco Hilton Dec 11-15^th

 

We will be at International Electron Device Meeting  Saturday 12/11 to Wed 12/15

We will be attending sessions and short courses

We have a presentation and Zoom call covering new emerging memories.

**** Sunday 12/12 from 6PM to 630PM Pacific time AND

**** Monday 12/13 from 5:30-6PM Pacific time

**** ZOOM Meeting ID: 811 2666 3619     ZOOM Passcode  481708

We can also meet 1:1 or in a small group during the conference. Just text.

Topics:

* Emerging Memory Product Lifecycle: What steps are needed to deliver production memories and where does each memory stand? How can we tell whether a technology is 15 years from success or 2 years.  Note: Memories take 10+ years to mature, 90% don’t make it

* Revenue and scenarios for future for Optane/3DXPoint. Where are we today and what are the three possible scenarios for 5 years from now.

**the scenarios are named “I am so Proud”, “Groundhog Day”, and “I Won” **

* Revenue and projections for MRAM. Embedded and discrete.

* ReRAM status and challenges

* Status of other memories

 

Mark Webb

WWW.MKWVENTURES.COM

https://mkwventures-semi.blogspot.com/

https://mkwventures-nand.blogspot.com/

Emerging Memories.... whats next?

 

New Memory Technologies …. Next stop? …. The Trough? The Moon? Other?

We have a report and will have an upcoming Zoom call similar to my Emerging memory update from Flash memory summit. This will be timed in parallel with IEDM so we can review latest info. We will identify where each memory stands in the New Memory Product Lifecycle (there are 10 different stages that go from concept to mass production) and projected revenue for each technology.

A quick update and preview:

1)     DRAM and NAND are ramping as fast as anytime in the past 5 years. No Brick wall coming, No replacement is planned. I am convinced DRAM and NAND will be around when I am long gone.

2)     Only two “new memories” have significant volume in bits

a.      3D Xpoint/Optane. More bits sold than all other new memories combined.

b.      Toggle and STT-MRAM: discrete Sales growing, embedded available from multiple foundries

3)     ReRam: lots of work, lots of issues identified. Similar performance to 3D XPoint, No revenue for densities over 1Mbit.

4)     Promising technologies: SOT MRAM, FE-RAM/FET… still promising.

The Challenges and Predictions

MRAM is ideal for embedded NVM and is steadily growing for discrete. What next?

3D XPoint is available to anyone who wants it today. But it is not growing, not game changing yet, not protecting Intel’s datacenter market yet. Is Intel still spending money on it? Is Intel adding capacity? Why did Micron abandon it? We have answers to where 3D Xpoint will be in 1,3,10 years.

Other Technologies are still path to eventually detailed Mbit+ array characterization

So, Personally, I am in the “Trough of Disillusionment” along with many Emerging memories IMHO. Now I am predicting what the Enlightenment will bring which we will present next week.

Mark Webb

www.mkwventures.com

Intel Breaks Ground in Arizona on New Fabs

 

Intel Broke ground on 2 new fabs in Arizona last week. They plan to open fabs in 2024 and spend $20B over the next 3 years to complete them.

https://www.reuters.com/technology/intel-breaks-ground-20-bln-arizona-plants-us-chip-factory-race-heats-up-2021-09-24/

Intel is making big announcements on Fab capacity to grow the company and its foundry operations. Intel's CEO Pat Gelsinger has done a great job of boldly stating where he wants Intel to be.

Unfortunately, Intel may find reality is a little different. We think Intel will not complete and ramp these fabs as planned and that the capex will be high but short of what Intel has announced. This isn't Pat's fault. It is just that ramping a foundry business is a 10 year plan, not a 3 year plan even if you have great press conferences. 

Actual Capex will average 19.5B per year including all of the announcements in 2021. This is 4B higher than previous 4 year average.

Intel Fab 52/62 will start at reduced capacity and most likely with delayed openings. We expect 25K wafers per month in 2025 timeframe.

We have actual loadings, capex, costs and technologies for each year and comparisons of Intel vs TSMC technologies and costs. We also have effects of government incentives and what Intel is hoping for going forward.

Call us for more info a bullet summary with numbers.

Mark Webb

www.mkwventures.com