( ESNUG 588 Item 26 ) --------------------------------------------- [10/30/20]

Subject: Naveed on SiFive's reaction to ARM's reactions to SiFive's actions
                     Virtual DAC'20 Troublemakers Panel


   Cooley: Naveed.


   Naveed: Yes, sir.


   Cooley: On last year's panel, you said SiFive was making big inroads 
           against ARM.  You said you had 70% of the cores that ARM had
           in the portfolio in only 3 years.  (See ESNUG 588 #7.)
     
           Since then ARM has expanded; they're not sitting down.  They're
           expanded to have stuff going directly against the RISC-V 
           portfolio like Neon and things like that.

           How are you catching up?  Or are you?


   Naveed: I think the interesting thing about SiFive is that we are a very
           interesting start-up in the sense that we are a company which has
           all the customers that you'd like to have.  We have all of them.  

           We have close to 180 customers.  Every large company.  And I 
           always make a joke, everybody but Broadcom and Marvell is using
           us, and this is a fact.  

           The reason it happened is because they all started replacing 
           their embedded cores.  So, we got inroads.  
           Our approach in all the customers is what we call land and 
           expand.  We are landed.  Literally everybody who is anybody in 
           the business -- we already have them as a customer.  All of them
           basically told us a year ago,

           "Guys, all these cores are good.  We need high-end cores".  

           So, we have dramatically shifted our SiFive engineering to go
           into very high-end cores.
           That's what we are doing now.  Because that's what people want.  
           You know we were in lower end, middle end cores.  But now, all
           our new offerings that we come out with next year, 2021, is
           going to be high end.  

           The second thing that has happened with RISC-V is that we have
           been joined by other people now.  Like Alibaba has a complete 
           unit which is targeting pretty high-end cores.  And I think you
           will see a couple of other companies come.
      
           ARM may have a lot of R&D and things like that.  We at SiFive
           are putting a lot on the high-end, but I think there are other
           people as well because of Open Source nature a lot of attraction
           in China and things like that.  

           I think in the next year in 2021 you will see higher and higher
           end cores come, which is the mainstay of ARM cores.  I think
           that's what you will see starting probably April of 2021.


   Cooley: April next year?  What's specific in April -- what do you mean?


   Naveed: We at SiFive will be releasing our pretty high-end core in
           April 2021.
      
   Cooley: How's that in contrast to...  I'm trying to think...  Apple just
           put out a press release, that they're using ARM for their 5 nm 
           chip that's used in all their Mac chips.  That's a big thing.  
           How do you compete against that?


   Naveed: Well I think the biggest barrier in the high-end cores is 
           actually not the core.  That is important, but I think it's 
           the software stack.  


   Cooley: Oh yeah, it's huge.  It's huge.


   Naveed: So that [SiFive's software stack] is where a lot of the work
           is going on.  A lot of RISC-V companies have banded together
           and are working together.  A lot of investment is going on;
           and all over the world.
         "The biggest barrier in the high-end cores is actually not the core. That is important, but I think it's the software stack."

- Naveed Sherwani of SiFive
           I think that's what is going to limit.  It's not us putting out
           a core.  It is that particular company committing to building 
           the software stack and I think that is going to be the limiting 
           factor.  Us having a core is just the first step.  

           We have a long road ahead.  I don't want to give the impression
           that next year April 2012 core comes out and then by June 2021
           everybody will have shifted to our SiFive core.  That's not the
           case.  The reason for that is the software stack -- that's going
           to take a few years to catch up.  The availability of our high
           end core is the first step.


   Cooley: Okay, one of the other questions I had was that ARM responded 
           directly to you guys -- SiFive and then the RISC-V market 
           altogether.  They [ARM] basically cut their prices in half.
      
           I saw when they did their Flexible Access program.  I was shocked
           at the price ARM was saying.  It was $75,000 a year for a 
           single chip, $200,000 for unlimited number of chips.  And you 
           don't even have to do a contract until you tape out.  I was like
           Oh My God they're cutting their prices radically.  Doesn't that
           kind of damage your SiFive profitability?


   Naveed: Well, first of all we have always said that we are all for what
           we call the Democracy of Silicon.  Day One we said that one 
           very important part of our DNA is to make sure that people all
           over the world can participate in the silicon revolution.

           We are very happy that other people who make processor cores are
           lowering their prices because it actually helps the cause of the
           Democracy of the Silicon.  That's very good.  

           We - I'm talking about SiFive, not RISC-V -- we have always said
           that in the long run, we want to enable people who are doing 
           silicon.  And we at SiFive want to be making money just not
           in the IP business but in the silicon business.

           In other words when ARM puts out a core, they charge a bigger
           number - $10 million, $20 million, $50 million -- and then they
           overcharge a 3% royalty.  They are trying to change it to a
           much higher percentage now.

           But our approach at SiFive was we'll do that, and we will help
           build silicon.  If you help build silicon, then you get the
           ASIC partner.  We have a different way of making money.  We
           are not shying away from ASIC.  We are going to do that, and
           we have invested in building silicon and I think that's how
           we want to make money in addition to selling our cores.  So that
           is not impacted by them [ARM] lowering their prices.


   Cooley: So just for numbers, because I like numbers, approximately how
           many customers do you at SiFive have?  Companies?

   Naveed: I think we have...  I think our design wins are close to 200.  
           Unique numbers, it depends on how many people are repeat 
           customers.  
     [ Editor's Note: Last year SiFive had 102 design wins, so they've
       close to doubled their design win footprint in 12 months.  - John ]


   Cooley: Well I was going to go to that.  How many are repeat and how
           many are unique?


   Naveed: We are now *almost* in the second cycle of most customers.  Most 
           customers picked us up 3 years ago, 2 years ago, and now we are
           in the second cycle.  So, I would say that unique customers are
           probably in the range of 100.  


   Cooley: And then how many are repeat?


   Naveed: I think...  that's why...  probably I would say...


   Cooley: Are you saying you have 100 repeat customers?


   Naveed: No.  I think total unique customers for SiFive are in the
           range of 100 or 110.  I would say at least 20% repeat.


   Cooley: Okay, alright, cool.  Thank you.

     [ Editor's Note: That 20% repeat rate was on January 2020.  Now
       in October 2020, it's at a 33% repeat customer rate as more
       of the 3 year licensing contracts come up for renewal.  - John ]

        ----    ----    ----    ----    ----    ----   ----

Related Articles:

    Anirudh on his Clarity/AWR/EMX end run around ANSS/SNPS in 3D IC
    Mo on Movellus customizable PLLs/DLLs are Netflix, not Blockbuster
    Naveed on SiFive's reaction to ARM's reactions to SiFive's actions

    Sawicki on his "free" AFS-XT being 5x-10x faster than Spectre-X
    Costello on Metrics cloud in 4 hours vs. SNPS/CDNS/MENT's 5 days
    Prakash on functional static sign-off to grow 2X in 5 years, $300M

Join    Index    Next->Item






   
 Sign up for the DeepChip newsletter.
Email
 Read what EDA tool users really think.












Feedback About Wiretaps ESNUGs SIGN UP! Downloads Trip Reports Advertise

"Relax. This is a discussion. Anything said here is just one engineer's opinion. Email in your dissenting letter and it'll be published, too."
This Web Site Is Modified Every 2-3 Days
Copyright 1991-2025 John Cooley.  All Rights Reserved.
| Contact John Cooley | Webmaster | Legal | Feedback Form |

   !!!     "It's not a BUG,
  /o o\  /  it's a FEATURE!"
 (  >  )
  \ - / 
  _] [_     (jcooley 1991)