Is Pseudo-Sync the same as "asynchronous mode"?

Discussion in 'IBM' started by Anon, Sep 12, 2004.

  1. Anon

    Anon Guest

    I believe that pseudo-sync is synchronized thus not asynchronized.

    However, is pseudo-sync also called 'asynchronous mode'?

    i've put some evidence that it is called 'asynchronous mode' at the
    end of the post in 2 chunks each marked <extract>...</extract>

    The evidence that led me to my belief that pseudo-sync is synchronized
    and not asynchronized is below, all the way down, up to the 2
    <extract>s.

    a disclaimer is at the end

    Based on the references below i've concluded that async really means
    separate clock, not derived. However, since the introduction of
    pseudo-sync / pseudo-synchronous memory, some manufacturers have
    called
    pseudo-sync 'async' (without the quotes) which is wrong(even with
    quotes is wrong), as you can't suddenly change and completely
    contradict the traditional meaning of the word. And there is
    absolutely
    no way for that meaning of async to change, since it's de facto (been
    in common usage for years) and de jure (set in law, coined many moons
    ago no doubt).

    I will follow with further proof that pseudo-sync is sync (and thus
    can't be async because async means not sync - so it's either one or
    the
    other)
    I went to my BIOS and set the clock for my memory bus to 100MHz,
    (note,
    my RAM is 266 DDR-SDRAM i.e. 133*2 MHz.
    I ran Si Sandra and it told me that my RAM was at 266MHz = (2*133MHz)
    multiplier of 4/3 x
    So that means that my RAM which is DDR-SDRAM (the S standing for
    synchronous) was operating at (4/3) times my FSB, that is a
    pseudo-synchronous rate.
    Pseudo-synchronous (keith and frank agree) is when the said clock's
    speed is derived by a non-integer multiplier or a divider(divisor?)
    that is 1/[insert non-integer here] like 1/2.5 (often stated as 2/5)
    So my multiplier of 4/3 (note it's top heavy, thus a multiplier not a
    divider) of 1.33x is pseudo-sychronous, and my RAM is DDR-SDRAM
    SD= ***synchronous*** dynamic.
    And my DDR-SDRAM was taking my FSB=100 as input, multiplying that by
    4/3 to get 133MHz and then the RAM was writing/reading data onto the
    memory bus at DDR thus 133*2=266MHz So it took the 100, multiplied
    it, then DDR'd it.

    note: yeah, i know with FSB=100MHz and RAM at 133MHz, i'm not making
    use of the extra speed that i'm running my RAM at. But my processor is
    1.3GHz and since i've got an AMD Athlon, I guess my multiplier is
    locked at 13. Thus If I increase my FSB to 133(for max RAM
    efficiency),
    it would overclock my processor. I will probably unlock the
    processor's multiplier (big hassle and risky) then set my FSB to
    133MHz
    and my multiplier to 10x, I will be overclocking my processor by
    30GHz(no big deal!). My AGP @ 1/2x and PCI @ 1/4x should be ok.

    References - these threads explain async, sync, and pseudo-sync
    [1]
    Re: Asus SP97-V question (9th december 2000)
    http://tinyurl.com/4n3mx
    note: pseudo-sync is explained here but all the way to the end, keith
    and frank use different definitions.

    (please see disclaimer, this is my understanding - which think is
    correct - of what keith and frank said. I don't intend to put words
    into their mouths)


    Keith says pseudo-sync is a case of sync
    Frank says it's a case of async

    Keith's definition from years of experience and based on the
    traditional definition of async which i tyhink he would argue has not
    changed and can't refer to a derived clock. asynchrous clocks are 2
    separate clocks, not derived from each orther,

    no multipliers or dividers
    "Asynchronous clocks need not have
    any frequency relationship and therefor have no phase
    relationship"
    "They are *ASYNCHRONOUS*. There is no phase alignment."

    (so that is why keith says pseudo-synchronous / pseuso-sync is not
    async -
    he says it is sync as it uses a derived clock like x1 or x 2/5 or 4/3

    (i'm not sure if he used the term derived clock)

    Frank's definition is from manuals and glossaries, he defines
    pseudo-sync as a case of async.
    in the 37th post of this thread "Subject: Re: Recommendations on
    cheap
    AT motherboard" frank gives a lot of evidence from his
    manuals.

    [2]
    Re: Cyric MII 333 problems (from message 12 in thread) 21st Jan 1999
    http://tinyurl.com/5v4nl
    note: pseudo-sync explained brilliantly by keith.

    [3]
    Re: Recommendations on cheap AT motherboard (12 October 2001)
    http://tinyurl.com/4qpxt
    note : keith and frank still disagree on whether pseudo-sync is
    async.
    Frank gives lots of evidence from manuals in the 37th post of the
    thread.


    Extracts giving evidence that pseudo-sync = 'async mode'
    extract from an overclockers website article
    www.overclockers.com/tips1039/index02.asp
    <extract>
    "You should now have noted that a FSB of 133 MHz and one of 166 MHz
    both produce the same memory speed of 333 MHz for DDR memory as a
    result of the multiplier change in value at 166 MHz. The difference
    between the two cases being the first is in an asynchronous mode of
    operation while the second is in a synchronous mode. "

    My thinking, is that he's saying
    a)FSB 133 RAM (333=166*2) Mult=1.25
    b)FSB 166 RAM (333=166*2) Mult=1

    he calls 'a' async (from the threads mentioned below, I know that b
    is
    sync and a is pseudo-sync as pseudo-sync is mult by an non-integer or
    by 1/[non-integer] )
    Thus, since b is pseudo-sync and this guy called it 'asynchronous
    mode'
    I guess that pseudo-sync is the same as asynchronous mode.
    </extract>

    and more evidence that pseudo-sync is 'asynchronous mode'
    http://forums.amd.com/index.php?s=76b91b820399d27c78b4ccdd5528cb7a&show
    topic=18694&st=0&#entry191509
    or http://tinyurl.com/4mq2u
    <extract from="AMD forum">
    "On the other hand, if you want to keep the RAM frequency in 266Mhz
    (by
    SPD) or change to 333Mhz, you'll be forced to run in asynchronous mode
    (FSB:DRAM ratio= 4:6 or 4:5), "
    </extract>
    notice also, that he's talking about DDR RAM (the guy mentions it
    elsewhere. Also, I think 266MHz in our times smacks of DDR 133*2).


    DISCLAIMER
    disclaimer: my references make mention of keith and frank's
    discussion.
    Since their discussion was very long, I summarised my understanding of
    their statements. I do not mean to speak for them, I have done my
    best
    to state their points and where they differ as accurately as possible,
    to the best of my understanding. Just take my comments on their
    comments as a general guide at best, and if you want to know their
    exact comments, I included references to the threads they spoke in,
    and
    I included many quotes which I hope I haven't taken out of context.
    I am not an engineer nor do I have any practical experience with PCBs
    or chips. I just started reading about this stuff last week. So please
    don't consider my words authoritative!! Anything i've stated that has
    been derived largely from my brain , i've been careful to precede with
    'i guess' or 'i think'(note the small i), unless i've backed it up
    with strong
    evidence. Since i've been composing in the wee hours and am posting
    it at 5am, there may be the odd silly mistake, but since i've repeated
    myself so much in this post so as to make my points as strongly as
    possible, the rest of the post will show the silly mistake to be a
    slip up and not a misunderstanding. hopefully, any misunderstandings
    will glare out from the post (along with the eccentricity). I'm being
    careful 'cos i'm stating other peoples' statements a lot on a subject
    i started reading last week as a hobby, and at 5am i'm dizzy and more
    prone to slip ups.
     
    Anon, Sep 12, 2004
    #1
    1. Advertisements

  2. Anon

    Franc Zabkar Guest

    I hate to revisit that old pile of bile, but my take on the issue is
    as follows.

    Synchronous mode is when two clocks are in phase alignment in such a
    way that each edge of the slower clock coincides with an edge of the
    faster clock. This occurs, for example, when a single oscillator is
    used to generate two frequencies which are integer multiples of each
    other.

    Asynchronous mode is when there is no phase alignment between the two
    clocks. This would be the case if the two clocks, even at the same
    frequency, were generated from two different free-running oscillators.

    Some vendors such as Asus (a motherboard manufacturer) define a
    pseudo-synchronous mode where two clocks are derived from the same
    oscillator but where not every edge of the slower clock aligns with an
    edge of the faster clock. This occurs when the frequencies are not
    integer multiples, but integer ratios, eg 2:5. The two clocks in such
    an example would still be synchronised because the pattern repeats at
    every 5 cycles of the faster clock.

    The confusion arises when other vendors (including manufacturers of
    clock generator ICs) choose to define async as anything which is not
    sync. Hence, as far as they are concerned, pseudo-sync is async.

    The scenario which was the basis of the old thread involved a single
    14MHz crystal oscillator and a single clock generator IC. This would
    have made the chip's FSB/PCI clock outputs pseudo-synchronous, as
    defined by Asus. However, the IC manufacturer's datasheet made no such
    distinction and referred to it simply as async.

    IIRC, Keith claimed that he was in possession of two versions of the
    same Asus motherboard which used both clocking schemes, but when I
    challenged him to identify the respective clock chip(s) he was
    strangely silent. So the matter was never conclusively resolved.


    - Franc Zabkar
     
    Franc Zabkar, Sep 13, 2004
    #2
    1. Advertisements

  3. Anon

    keith Guest

    You would be correct.
    No. Some manufacturers did an asynchronous chipset (eg. SiS). Via's were
    pseudo-sync, which meant a non-integral, but synchronous multiplier
    between the processor bus and the PCI bus.

    Yes.

    <snip much>

    You are correct.
     
    keith, Sep 14, 2004
    #3
  4. Anon

    keith Guest

    Yes, I found the evidence, and there were indeed different versions of
    the SP97(?). I simply found it uninteresting to argue with someone who
    was *so* wrong. BTW, the board I had in my drawer indeed did have only
    one oscillator. ...so you were right there. OTOH, I had the records from
    *hundreds* of boards (from every manufacturer - including the SP97) that
    said there were both. The SiS chipsets could do either. Via was indeed
    pseudo-sync (non-integral synchronous) only.
     
    keith, Sep 14, 2004
    #4
  5. Anon

    Franc Zabkar Guest

    You avoided several opportunities to present the "evidence" which you
    claimed was at your fingertips. Instead you chose to fabricate a
    preposterous lie because you were not man enough to admit that you
    were wrong.
    Of course I was. The photographs on Asus's website were evidence
    enough.
    BS. IIRC, there were only two versions of the SP97, one with both
    SIMMs and DIMMs, the other with SIMMs only. Both used a single
    oscillator and clock generator.
    At that time many users discovered that a combination of AMD K6 CPU
    and SiS5597/5598 chipset was unstable when using a pseudo-synchronous
    PCI/FSB clock in the ratio of 2:5. This problem was well documented in
    various forums.
    I doubt it. Show me the evidence.


    - Franc Zabkar
     
    Franc Zabkar, Sep 14, 2004
    #5
  6. Anon

    Anon Guest

    there are a number of sites that seem to refer to 1 clock chipsets as
    being able to run in 'asynchronous mode'

    http://www.hardwarepage.nl/viaapollomvp4.html
    and http://www6.tomshardware.com/motherboard/19980731/socket7-02.html
    "VIA's MVP3 chipset allows you to use a synchronous or asynchronous
    mode for your memory"


    If async means 2 clocks, - and given the quote above - then maybe
    'asynchronous mode' is different from 'real' async. If that were the
    case, then maybe 'async mode' is a pseudonym for pseudo-sync
     
    Anon, Sep 15, 2004
    #6
  7. Anon

    Anon Guest

    there are a number of sites that seem to refer to 1 clock chipsets as
    being able to run in 'asynchronous mode'

    http://www.hardwarepage.nl/viaapollomvp4.html
    and http://www6.tomshardware.com/motherboard/19980731/socket7-02.html
    "VIA's MVP3 chipset allows you to use a synchronous or asynchronous
    mode for your memory"


    If async means 2 clocks, - and given the quote above - then maybe
    'asynchronous mode' is different from 'real' async. If that were the
    case, then maybe 'async mode' is a pseudonym for pseudo-sync
     
    Anon, Sep 15, 2004
    #7
  8. There's nothing about asynchronous that requires ANY clock. It simply means
    the events are not 'synchronous' with each other. I.E. They are not
    necessarily occurring in a fixed time relationship with each other; they
    happen when they happen.
     
    David Maynard, Sep 15, 2004
    #8
  9. Anon

    Anon Guest

    ok, i'll just correct something i said. when i said "async means 2
    clocks" I meant 2 oscillators. But I take what you say, so even with
    2 oscillators it can be synced if 1 oscillator happens to be aet the
    same beat as another, or a multiple.
    The thing is though, that I think the VIA MVP3 chipset's clocks use 1
    oscillator with multipliers and dividors, thus there is a fixed time
    relationship.

    Coming to think of it. Every rational number can be put in the form
    A/B where A and B are integers. Thus, given 2 clocks each with its
    own oscillator, then whatever speeds the 2 clocks are, there will
    always be a fixed timing relationship between them. So I don't see
    how your definition of sync can be right. There's always a
    non-integeral timing relationship i.e. an A/B relationship and thus
    by that definition of sync, there is no async!
     
    Anon, Sep 15, 2004
    #9
  10. Anon

    Franc Zabkar Guest

    Oscillators are not perfect. Frequencies drift with temperature, for
    example.


    - Franc Zabkar
     
    Franc Zabkar, Sep 15, 2004
    #10
  11. Same thing in this context.
    No, if they are sync'd to each other then they're synchronous.

    You're confusing the analog world with the digital world and there's no
    such thing as two clocks that 'just happen' to be at the same frequency (an
    analog measurement), or multiple, of each other. There is always some
    deviation, unless they are synchronized.
    There is a fixed relationship between the derived clocks, of course, but
    you're confusing the memory types. The asynchronous memory they're talking
    about does not operated by any of those clocks. It is asynchronous.

    A data request is made and 'some time' later the data is ready. The only
    thing you know about when the data will be ready is that it should be less
    than the specified read delay; typically 60ns for good FPM SIMMS. But it is
    not ready in 'sync' with any clock. It's ready when it's ready: asynchronously
    In the first place, I said asynchronous doesn't require ANY clocks but you
    keep insisting on defining it with clocks.

    Second, you're confusing idealized math with the real world. By idealized I
    mean you're ignoring all the other things that comprise the real world
    (analog) signals; the first of which I already mentioned: that there's no
    such thing as two clocks that 'just happen' to be at the same frequency.
    The next is there's no such thing as a 'pure' and constant time domain.
    Clocks do not run at 'X' frequency, nor at it forever. They run at X -+
    some error with Y jitter and Z drift over time. So, if you want two clocks
    to maintain a particular relationship over time then you have to do
    something to cause them to stay in that relationship. That is precisely
    what synchronization means: the relationship remains constant over time.

    Your 'intuitive' extrapolation from rational numbers is true in that there
    is a ratio between them at point 'A' in time. But there's another one at
    point 'B', and another one at point 'C', and so on and so on, because the
    clocks are not perfect, error free, things that remain constant. And, to
    make matters worse, the error adds up over time.

    Put another way, your math presumes things that are not true unless
    intentionally caused to be so... e.g. by synchronizing the clocks.
     
    David Maynard, Sep 15, 2004
    #11
  12. Anybody as interested as you should have the VIA data sheets. Look up the
    MVP3 chipset (598.pdf), specifically the VT82C598 North Bridge. It
    supported FSB:pCI clock ratios of 2, 2.5 and 3 and that's err, all folks.
    FWIW there were many mbrds which were scripted for 2.5 jumpering but I
    never heard of one which actually implemented it - even when there was a
    jumper, instead of a couple of spare pads, it did not actually force a 2.5
    ratio. AIR a look at the clock chips used showed that it was just not
    supported - mbrd mfr cheaped out or possibly the the VIA 598 didn't
    actually work right that way.

    If you don't have the .pdf and are umm, nice (for a change :-}) I'll e-mail
    it to you.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
     
    George Macdonald, Sep 16, 2004
    #12
  13. Amazing the fiction you can find on WWW, presented as fact. The MVP3 did
    *NOT* support "asynchronous" (not a good term anyway since it means
    something different in other apps) clocking of any kind. The memory was
    clocked the same as the FSB; the AGP was clocked at 2x PCI and the FSB:pCI
    allowed ratios of 2, 2.5 and 3.
    Nope. Just read the data sheets - it's in there.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
     
    George Macdonald, Sep 16, 2004
    #13
  14. And what makes you think tomshardware is wrong?

    http://www.mtiusa.com/sy-5eh.htm

    That board uses the MPV3 chipset and supports FPM and EDO; both of which
    are asynchronous memory.
     
    David Maynard, Sep 16, 2004
    #14
  15. I forgot about the Soyo 5EHM (were there others - don't think so) and we
    have a couple in the office. Never used FPM/EDO with them though and it
    was a glaring waste of mbrd real estate... given the relative prices in
    hmm, 1998(?). Did any company make FPM/EDO PC memory SIMMs which would
    respond to a 100MHz FSB pass-through rate?... 40ns?

    Of course that was not the "asynchronous" being discussed here anyway..
    which was to do with the FSB:pCI ratio. I guess I fell into the trap the
    previous poster set for himself.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
     
    George Macdonald, Sep 16, 2004
    #15
  16. Anon

    Franc Zabkar Guest

    Unfortunately such internal documents are available only to a select
    few. Mere mortals don't qualify.
    I'm always nice ... to nice people. But I detest liars and pompous
    blowhards.

    Thanks for your offer of the .pdf file. I suspect the OP would
    appreciate a copy, too. I don't really need it myself, though, as I
    don't doubt your integrity, nor your knowledge. OTOH, I have proven
    Keith to be wrong on so many occasions, on even basic issues, that I'm
    loathe to accept anything he says without independent confirmation.

    May I suggest you upload the VIA datasheet to your own webspace and
    post a link to same.

    - Franc Zabkar
     
    Franc Zabkar, Sep 16, 2004
    #16
  17. That's a useless accusation, especially since they were right and you were
    wrong on the support of asynchronous memory.
    All of which, even if true, means nothing as it only takes one to prove the
    chipset supports it whether you think it was a 'good idea' or not.
    The clock rate is irrelevant to asynchronous memory. It's, ahem, asynchronous.

    To illustrate the point, how did you come up with 40ns in order to "respond
    to a 100MHz FSB pass-through rate" that's a 10ns clock period? You see,
    *synchronous* memory would need a 10ns access rate to keep up. Asynchronous
    can be anything because it's, ahem, asynchronous.

    And 60ns memory doesn't 'respond to' a 66MHz bus "pass-through rate"
    either. The effective data rate for 60ns memory would be, at best (ignoring
    all other delays), 16.7 Mhz. That's why PC66 SDRAM was 'better'.

    The term asynchronous doesn't depend on 'which clocks' or 'which bus' nor
    is there 'the one here'. Asynchronous is asynchronous.
     
    David Maynard, Sep 17, 2004
    #17
  18. Anon

    David Wang Guest

    Nothing is truly "asynchronous" in this context. The DRAM is considered
    "asynchronous" because it did not explicitly move from one state to
    another state in response to s clock input, rather, it changed states
    relative to the timing of some control signals, but those control
    signals were generated by a DRAM controller, which is itself
    "synchronous" and operated on a base clock frequency.

    "Asynchronous" PB EDO could have a 10ns "cycle rate" in the sense that
    it could burst out the data at the 10ns (100MHz) rate, if you so pushed
    the margins and picked the right parts, but I agree, it would have been
    tough.
    "60ns" memory is 60ns tRAC**, and that has nothing to do with the effective
    datarate, since the DRAM manufacturers were actually pipelining and
    spitting out data on xEDO parts just as they were on SDRAM parts.
    Asynchronous is asynchronous, except when it's not really asynchrnous.
    Even when it's "synchronous", it's not really "synchronous". What does
    "clock" mean when the strobe signals arrive the device interfaces at
    different times? It's all relative... To something... Usually an
    explicit clock or strobe, but could be implicitly relative to the clock.

    ** tRAC is Randomc access latency. Which is basically tRCD + tCAS.
    Micron's data sheet on obsolete FPM and EDO parts are still available
    online at www.micron.com.
     
    David Wang, Sep 17, 2004
    #18
  19. Anon

    chrisv Guest

    Guffaw.
     
    chrisv, Sep 17, 2004
    #19
  20. True now but back in the days of the MVP3, VIA had pretty good general
    availability of data sheets.

    I guess we can all be rubbed the wrong way but I've always found Keith a
    valuable source of in depth info. Have you still not figured where he
    works and what he does?
    I've never even bothered to use any "webspace" I might have from my ISP...
    and would be concerned that it's allocation might work against my e-mail
    quota.

    Rgds, George Macdonald

    "Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
     
    George Macdonald, Sep 18, 2004
    #20
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.