ARM Sparkles

ARM had Q4 revenue of  $262m up 21% on Q4 2011 and profits up 16% at $80m. Full year revenues were $913m for a profit of $276m. Q4 operating margin was 46.5% and full year margin was 45.6%.

“Customers are developing products to meet the needs of the post PC era and are driving demand for ARM’s most advanced technology,” says ARM CEO Warren East, “2013 brings exciting opportunities and challenges as ARM enters competitive new markets where we are well positioned to succeed.”

Q1 revenues are expected to be flat.

36 processor licenses were signed; two 64bit v8 architecture licenses were signed; six 64bit v8 processor ordinary  licenses were signed; three A15 processor licenses were signed.; seven Mali graphics processor licenses were signed and five physical IP licenses were signed.

The licence order backlog is up 25% which is good for future revenues.

Royalties were up 21%. License fees were up 26%.  Physical IP licenses were up 36% at $15.4m.



  1. SecretEuroPatentAgentMan

    Interesting. The ARM carpet seems bulging with embarrassments of the past.
    Did you keep any documentation about Piccolo?

  2. The same happened to their first 64bit sortie from 2003. I was talking to the head of their v8 programme last year and he didn’t even know it had existed.

  3. SecretEuroPatentAgentMan

    Realisation of the problems with Piccolo must have set in quickly. I believe I saw an article about it early 1996 and we sent off for more information, so we are talking just weeks.
    Today ARM has eradicated most traces of it ever having existed so they must have been really ashamed of the design. I haven’t even seen a detailed description of ISA or instruction set.

  4. Well unless you count an ICL1907 where I wrote a printer handler in octal, I began on a HP MC2 processor, but then went through all ths usual 6800, Z80, 6502, 68000 and so on.
    I really don’t know why Intel keep the old instructions – no compiler uses an INC mem instruction nowadays for example.
    You were lucky ARM didn’t come back – a long time ago I reviewed an EU project using Piccolo and called it a dog’s dinner which the ARM guy agreed it was 🙂
    Hexagon looks very good for what it is, but I don’t believe it does speculative calculations, something I had a AMD2900 bit-slice machine doing in the 80s.
    I really think we need to look at 32 instruction wide words (512-1024 bits) nowadays as with on-chip cache this width is not an issue, and then use feedforward validation to get the maximum parallelisation out of any code, and there are of course various high performance computing projects working on this avenue. I produced a prototype Verilog model using a cut down ARM7TDMI instruction set about a decade ago which hammered through some serial code at an average of 10 times what the standard chip did, though of course this is very peaky and certainly not low power (almost 100 times the single core though this could have been reduced). Nowadays even Intel seems more worried about power consumption than peak processing power.

  5. SecretEuroPatentAgentMan

    My last x86 job was also a graphics system but not for games. Back in the day I heard people say that if you started on 6502 or 6809 you would prefer 68K, ARM and the like. Those starting on Z80 would be happy with x86.
    In my case I did indeed start on 6502, a processor I still consider elegant. What was your start in the world of assembly programming?
    Anyway I agree that a break is needed; some kind of flag day where one could drop old stuff. The x87 instructions take up valuable real estate in the instruction code charts.
    If VLIW is yor preference I’d like your view on Hexagon:
    I wanted to try this myself and requested a license for 4stack but never got a reply. That was a back up plan for ARM Piccolo where ARM never returned to us when we flagged our interest (or more correctly, a dire need for it).

  6. In the 90s I produced a complete games graphic subsystem with video playback in x86 assembly and it seemed fine to me. And the space for translation is now down to

  7. My profound apologies Scunnerous. I have sent your entire Comment to the Head of Movable Type IT. Last week I got the following reply from him when I complained about the ‘too many comments’ error message: ” 
    There is a throttle on comments which I’d expect you to be affected by occasionally – and that’s set to no more than one comment per minute.  I can easily see you racing through at that rate, but then again, spam robots will go faster than that as well, and we do try to slow them down.  The setting is changeable across the whole installation, but for various reasons, including the spam issue that’s a change that would have to be requested formally by an IT Account Manager and me managed through change control.
    “As far as users seeing the effect after days or weeks of non-use, it’s possible that if they are on networks where there are many users sharing an IP address, a colleague may have just also commented, but I suppose that’s probably less likely for a modest traffic blog like yours.  In that case I can’t offer any explanation – only to make sure they have cleared their browser’s cache – all the usual stuff we tell people to do when we don’t have a more concrete answer.  If they are making a comment  then immediately editing it or making another, then the throttle will kick in as expected – it’s the _present 60 second slot_  that counts, no consideration is given to previous inactivity.”
    Hopefully this conveys more to you than it did to me, Scunnerous, but I think it means you have to wait for 60 seconds and try again. Having been caught by this many times I always copy all my comments before posting and so, if I get caught by the ‘too many…’ message, it is not too onerous to try re-posting it again . . .and again. I am so sorry about this. As I said I’ve complained again this evening since getting your Comment and hopefully they’ll come up with something this time.

  8. SecretEuroPatentAgentMan

    Well it is the ISA that I am concerned with. The programmer visible x86 ISA is extreme CISC with enormous sets of prefix codes. Sure it is RISC inside, but the translation is not free and the ISA has a real cost in terms of space.
    I did embedded programming in x86 assembly code and it was painful.
    ARM is far more pleasant but I would have expected 30 years research would have given us more. 64-bit ARM does bring some changes but from what I can see so far much is cosmetics and additions. There is a fear of deleting old baggage, interestingly it was AMD that used the hedge trimmers on the x86 instruction set, not Intel.

  9. I posted a comment here which first failed on the “too many posts” and then, on the 2nd attempt, was apparently “held for moderation”, even though I was signed in with Movable Type. It seems to now be lost and I did not save the text locally so I cannot recover it.

  10. Apart from some instruction codings, little is left of the original Intel or ARM processors in the latest Core7 or Cortex A15 processors.

  11. It can’t be lack of resources SEPAM because Wilson and Furber did their architecture on a shoestring. Maybe Arm have done such a good job of upgrading and extending it that there isn’t a need for a new one – and it’s becoming so pervasive – phones, tablets, consumer, computers, microcontrollers that everyone’s used to using it.

  12. SecretEuroPatentAgentMan

    Intel architecture is from the 70’s.
    ARM architecture is from the 80’s.
    Why nothing new coming after 30 years of research?

Leave a Reply

Your email address will not be published. Required fields are marked *