Announcement

Collapse
No announcement yet.

parhelia voltage mod finished

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #91
    Originally posted by Greebe
    But you insist that the internal supply voltages are sloppy and that sir is incorrect.
    If it's so tightly set, sending more current to the chip shouldn't do much -or did you mean something more elaborate?

    /me ponders on the value of pi as a toastpan.

    Comment


    • #92
      If it's so tightly set, sending more current to the chip shouldn't do much -or did you mean something more elaborate?
      Current is akin to mass as Voltage is akin to velocity, greater the mass the greater the power. Remove virtually all current and you have static electricity.

      Which brings me to the only point I'm trying to make here is to clarify the error ridden comments Fender is making... especially in light that he's also used what all he thinks he knows and bashes others with comments like...
      I think you need some courses in electrical engineering
      Regardless of the happy face spin he's putting on it, it's still an insult simply because he doesn't know what he's talking about. Comparasions given are not only grossly generalized, but simply do not apply. The one thing I dispise most is someone using anothers gullibility to fill their head with crap... "if you can't dazzle them with brilliance, baffle them with BS" ego trip just doesn't get it.

      To emphesize this point further, claiming to OC a 486DX33 to 66mHz is impossible. No 486MB made could do 66mHz fsb. The only options available were 25(486DX(SX)25, DX50 and the DX2/50), 33 (DX(SX)33 & DX2/66) and at the high end 40 to support some of the late generation AMD cpu's at the time. With the sole exception being the Diamond FastBus VLB which was the first MB designed with OCing in mind, adjustable in 1 mHz steps and even then it's top end was just a tad over 40mHz.
      "Be who you are and say what you feel, because those who mind don't matter, and those who matter don't mind." -- Dr. Seuss

      "Always do good. It will gratify some and astonish the rest." ~Mark Twain

      Comment


      • #93
        Originally posted by Greebe
        Current is akin to mass as Voltage is akin to velocity, greater the mass the greater the power. Remove virtually all current and you have static electricity.

        Which brings me to the only point I'm trying to make here is to clarify the error ridden comments Fender is making... especially in light that he's also used what all he thinks he knows and bashes others with comments like...

        Regardless of the happy face spin he's putting on it, it's still an insult simply because he doesn't know what he's talking about. Comparasions given are not only grossly generalized, but simply do not apply. The one thing I dispise most is someone using anothers gullibility to fill their head with crap... "if you can't dazzle them with brilliance, baffle them with BS" ego trip just doesn't get it.

        To emphesize this point further, claiming to OC a 486DX33 to 66mHz is impossible. No 486MB made could do 66mHz fsb. The only options available were 25(486DX(SX)25, DX50 and the DX2/50), 33 (DX(SX)33 & DX2/66) and at the high end 40 to support some of the late generation AMD cpu's at the time. With the sole exception being the Diamond FastBus VLB which was the first MB designed with OCing in mind, adjustable in 1 mHz steps and even then it's top end was just a tad over 40mHz.
        oh lol, I should have known there would be people like this. sorry for the other person about "you should..." but i think himself didn't take this as bad...

        you say I'm wrong about voltages but don't give any clue about the truth. if it were so wrong, you could easily beat it.

        and for the last thing, lol, why wouldn't my little 33MHz go up to 66MHz ??? you enumerate sx / dx / dx2 models and so what ? the only thing wich could have limited me would have been the mobo, and many mobos accepted both 33 and 66MHz 486... I never claimed to have put the fsb to 66 (I would have liked but it couldn't get upper 50MHz (and not 40 lol, make a simple google search and you'll see many 486@fsb50). I only changed the multiplier (it was already possible too) to 2x

        oh well, keep on claiming whatever official/commercial crap you want about o/cing many people 'round the world know how to do it and what it implies (yeah, sometimes it makes it last shorter, we know that)

        Comment


        • #94
          Fender, I'm going to try to explain this once and only once.

          There are people around here that know certain things about Matrox products. They also have things called NDA's which I'm sure you already know are Non Disclosure Aggrements.

          So yes at times they can say 'No, it won't work' but cannot answer why. You get used to it, and you also learn to in general take their advice because usually it is good.

          Now please go read the survival guide in the FAQ files
          Juu nin to iro


          English doesn't borrow from other languages. It follows them down dark alleys, knocks them over, and goes through their pockets for loose grammar.

          Comment


          • #95
            Originally posted by Sasq
            Fender, I'm going to try to explain this once and only once.

            There are people around here that know certain things about Matrox products. They also have things called NDA's which I'm sure you already know are Non Disclosure Aggrements.

            So yes at times they can say 'No, it won't work' but cannot answer why. You get used to it, and you also learn to in general take their advice because usually it is good.

            Now please go read the survival guide in the FAQ files
            that's ok for this part, it was just that he was trying to say false things to make people believe i lied about what i said, which i can't really appreciate.
            and despite the respect i can have for those people, you have to admit that there's always a slight difference btw officiel statement and truth, it's the same with every board/proc/chip maker in the world. a long time ago, intel claimed the same thing about their processors, even if they don't bother doin' it anymore now everybody knows. it has always been this way ...
            I can't blame people for that, but you have to take into account that some people out there don't take official statements for granted and in many cases, it appears they are right not to do it. I don't see why it should be different for matrox (even if i really appreciate this brand, as I remember my good ol' mystique lol).

            I'm ok to shut up about this if you want, but sorry, I'll never take any "official statement" for completely granted, that's all.

            have a good night

            Comment


            • #96
              Considering the number of Retail Parhelia's that have issues running at 220MHz, and the fact that the Parhelia 256MB card doesn't even ship at 220MHz, i would not be too quick to attempt to push my card to that high of a clock speed.

              About clocking the memory down... they could be using higher timings on the memory, thus lowering the safe max from 300MHz. if you look at most memory it can be spec'd out at one speed (say 133MHz) at a certainly latency (lets say, 3 cycles), but if you were to decrease the latency (say to 2 cycles) it would also decrease the rated speed (say to 100MHz). this is a pretty common practice in the memory industry. Clock speed is not everything.

              That being said I highly doubt that overclocking a Parhelia is gonna result in that much of a performance increase at this time.

              RE the 486's - yes, several did ship at a 50MHz bus. I believe AMD was the only one to, Cyrix might have as well. they only shipped two parts at this speed - the 486DX-50 and the 486DX2-100. As far as overclocking goes, i personally highly doubt that a 486DX33 running at 66MHz would have continued to work for all that long. My own experiences involved taking a DX5-133 (or a 5x86-133, depending on who you talk to) up to 160MHZ and it """died""" after about 5 minutes of use. That being said it didn't acctually die, but have you ever seen a healthy system (or processor) cause INT0 errors on boot up?

              Comment


              • #97
                Oh that's ok Dan, have this urge to disect this one up

                oh lol, I should have known there would be people like this. sorry for the other person about "you should..." but i think himself didn't take this as bad...
                I see you've had the same reaction elsewhere

                you say I'm wrong about voltages but don't give any clue about the truth. if it were so wrong, you could easily beat it.
                You are comparing 1 semester basic DC electronics to complex switching power supply design. That in itself proves my point.

                To extend what I am refering to are the actual components (IC's, MOSFET's, etc) that supply the voltage rails to the GPU and memory

                (any fool who is halfway competent in electronics would have realised at this point they're goose is cooked. Because all you have todo is look up the bloody parts and measure the values with a decient DMM . It's not like those parts are inside info only, Doh!)

                (links are to PDF data sheets, files may be large)

                First up we have an IRU3037A, an 8 pin synchronous PWM controller.

                Notice the Fb Voltage Line Regulation, Max is only 0.35%! The typical application specs are virtually identical to what Parhelia has in most respect to what we are discussing here, Line regulation and tolerance (not current)

                BTW R3 and R5 make up the feedback network which you guys are modding with your pencil trick.
                [b]This also is a basic DC voltage divider circuit[b] and functions as Fender describes... but two bloody resistors are a far cry from what all we're looking at here!

                A switch mode design of this type has not only accuracy in it's regulation, but both high noise rejection and isolation from the source which provides excelllent control when the source voltage varies.

                These 3037's drive several different MOSFET's (a type of high input impedence (read:high isolation) power transistor) and their selection is decided by operating frequency voltage and output current characteristics (primarily) If interested in their operating characteristics, here are their white papers...


                IRF7460 SMPS MOSFET

                FDS6676 30V N-Channel PowerTrench® MOSFET

                Si4920DY Dual N-Channel 30-V (D-S) MOSFET

                Their outputs supply 1.8, 2.5 and 3.3vdc

                BTW the "soft start" feature of this controller IC is for using it as a DC motor controller limiting it's output in a slower ramp up fashion, so the motor isn't slammed with a massive rush of power (max current when at rest) which would otherwise blow the motor or irreputably damage it or the output transistors.

                and for the last thing, lol, why wouldn't my little 33MHz go up to 66MHz ??? you enumerate sx / dx / dx2 models and so what ? the only thing wich could have limited me would have been the mobo, and many mobos accepted both 33 and 66MHz 486...
                The 66mHz cpu's made were internally clock doubled, meaning the FSB ran at 33 and the core was doubled. Intel multiplier locked their cpu's starting with the 386. BUSTED

                I never claimed to have put the fsb to 66 (I would have liked but it couldn't get upper 50MHz (and not 40 lol, make a simple google search and you'll see many 486@fsb50). I only changed the multiplier (it was already possible too) to 2x
                Pardon me, I did leave out 50 (twas very tired atm). But that is neather here nor there. As above it's impossible to change the multiplier on an i486. (The d00d's shoot'n blanks here)

                BTW, why can't you mention what MB was used? Oh I know! It doesn't exist!

                oh well, keep on claiming whatever official/commercial crap you want about o/cing many people 'round the world know how to do it and what it implies (yeah, sometimes it makes it last shorter, we know that)
                If they continued to take your ill advice sir, we would have several very pissed off members here with dead P's!!!
                "Be who you are and say what you feel, because those who mind don't matter, and those who matter don't mind." -- Dr. Seuss

                "Always do good. It will gratify some and astonish the rest." ~Mark Twain

                Comment


                • #98
                  hehe...I think you could change the multiplier on AMD 486's

                  and their were 50 486 mhz intels, I used to have one at a place I used to work, it was in a ceramic package and it had no HS

                  Was quite fast for what was but they were very rare, only produced with the first round of intel 486's, they seemed to stop production of them completely when the dx2's appeared.
                  As I recall the intel 50mhz parts had no multiplier adjustemet, only AMD chips did.
                  Last edited by Marshmallowman; 29 January 2003, 20:13.

                  Comment


                  • #99
                    Greebe,
                    First off, thanks for the time to explain some of your knowledge of the P...
                    on to the Q...

                    So by lowering the Resistance on R3 and R5, the Core voltage is being raised and what else is being done??

                    Sorry for my novice Questions..


                    On another note, Sasq mentioned
                    So yes at times they can say 'No, it won't work' but cannot answer why. You get used to it, and you also learn to in general take their advice because usually it is good.
                    This does happen often on this board, while at times it isn't an issue and completely understandable [such as NDA's], while other times this kind of phenomina verges on Big Brother looking over your shoulder..
                    While the advice might be well intentioned and correct it just raises more questions... like what are the intentions of such advice etc..
                    Kinda like when Mom answers your questions with "Because I said so"

                    thanks
                    Craig
                    1.3 Taulatin @1600 - Watercooled, DangerDen waterblock, Enhiem 1046 pump, 8x6x2 HeaterCore Radiator - Asus TUSL2C - 256 MB Corsair PC150 - G400 DH 32b SGR - IBM 20Gb 75GXP HDD - InWin A500

                    Comment


                    • Lets use the "Typical application" reference here to get an idea of what is going on with the pencil trick.

                      Note: This is not Parhelia's circuit, for educational purposes only!

                      By penciling over R5 you are adding resistance in parallel and thus it's overall resistance would be reduced. This causes the Fb voltage divider network ratio to change and thus increased output voltage... it's not a pretty solution but it works (in a dirty way)

                      Better would be to add a potentiometer into the circuit with a lower value resistor replacing R5 in series with each other (in this example). Lets say a 500ohm resistor and a 1kohm pot you would limit how low/high the output voltage would be without alot of trial and error going on during the testing phase (and would be far more permanent that the pencil trick, albiet more work).

                      Now internally (I don't have a block diagram), but can see that they are feeding Fb with a sample of the output which is compared (and compensated) to a set precision voltage reference source of 1.25vdc. This output then is fed into a PWM (Pulse Width Modulation) circuit which runs at 400kHz. This output is then split 50/50 to drive the MOSFET's and filtered by L2 and C7, a low pass filter.

                      (now it's time for this jelly donut to hit the hay
                      "Be who you are and say what you feel, because those who mind don't matter, and those who matter don't mind." -- Dr. Seuss

                      "Always do good. It will gratify some and astonish the rest." ~Mark Twain

                      Comment


                      • Originally posted by Greebe
                        [B]Oh that's ok Dan, have this urge to disect this one up



                        I see you've had the same reaction elsewhere



                        You are comparing 1 semester basic DC electronics to complex switching power supply design. That in itself proves my point.

                        To extend what I am refering to are the actual components (IC's, MOSFET's, etc) that supply the voltage rails to the GPU and memory

                        (any fool who is halfway competent in electronics would have realised at this point they're goose is cooked. Because all you have todo is look up the bloody parts and measure the values with a decient DMM . It's not like those parts are inside info only, Doh!)

                        (links are to PDF data sheets, files may be large)

                        First up we have an IRU3037A, an 8 pin synchronous PWM controller.

                        Notice the Fb Voltage Line Regulation, Max is only 0.35%! The typical application specs are virtually identical to what Parhelia has in most respect to what we are discussing here, Line regulation and tolerance (not current)

                        BTW R3 and R5 make up the feedback network which you guys are modding with your pencil trick.
                        [b]This also is a basic DC voltage divider circuit and functions as Fender describes... but two bloody resistors are a far cry from what all we're looking at here!

                        A switch mode design of this type has not only accuracy in it's regulation, but both high noise rejection and isolation from the source which provides excelllent control when the source voltage varies.

                        These 3037's drive several different MOSFET's (a type of high input impedence (read:high isolation) power transistor) and their selection is decided by operating frequency voltage and output current characteristics (primarily) If interested in their operating characteristics, here are their white papers...


                        IRF7460 SMPS MOSFET

                        FDS6676 30V N-Channel PowerTrench® MOSFET

                        Si4920DY Dual N-Channel 30-V (D-S) MOSFET

                        Their outputs supply 1.8, 2.5 and 3.3vdc

                        BTW the "soft start" feature of this controller IC is for using it as a DC motor controller limiting it's output in a slower ramp up fashion, so the motor isn't slammed with a massive rush of power (max current when at rest) which would otherwise blow the motor or irreputably damage it or the output transistors.


                        The 66mHz cpu's made were internally clock doubled, meaning the FSB ran at 33 and the core was doubled. Intel multiplier locked their cpu's starting with the 386. BUSTED



                        Pardon me, I did leave out 50 (twas very tired atm). But that is neather here nor there. As above it's impossible to change the multiplier on an i486. (The d00d's shoot'n blanks here)

                        BTW, why can't you mention what MB was used? Oh I know! It doesn't exist!



                        If they continued to take your ill advice sir, we would have several very pissed off members here with dead P's!!!
                        As now, I'm more "enlightened" by you, I'm ok for the voltage (apologies here)

                        but heh, sorry, the first intel proc multiplier locked were the pentium II (and even on those, there was a trick to get over that)!!!

                        really, i can assure you ! 386, 486 and Pentium (first gen) had their multiplier unlocked !!


                        for the motherboard, I think it was an octek (does that brand still exist ?) but I cannot be sure as i had others unnamed (from full pcs) and at these times, I couldn't choose the mobos, so some could and some couldn't (most could, by jumpers though, if they were compatible with 33 and 66 procs)

                        so, I'm ok for the electrical part (even if i have seen many boards (sound, graphics or else) die because of an unstable power supply but it's a different chapter), but I can promise you this kind of overclock was possible, and almost easily even (at those times, procs were pretty cold, and if you had the chance to get a particularly good one, it could stand the 2x multiplier with a simple heatsink fan)

                        Comment


                        • I'm sorry but the first ones locked started with the i386 because Intel didn't want get a bad reputation from people recieving OC'd systems claiming to have a higher speed cpu than what was actually sold. Most of us at the time could see their point but instead of Intel moving to setup a different standard of cpu detection instead of measuring raw speed they locked the multiplier. This move proved to be bad news for us who actually had a clue on howto OC a system stable, so that pushed us more toward going with AMD due to extremely high prices and now hindering our efforts. It wasn't until the Diamond Fastbus VLB came out (the first MB that had OC'ing ie truely adjustable FSB). Once this new wave of MB's hit the market AMD's market share started to really take off. Later when AMD managed to start pumping out more power on the S7/SS7 MB's of the time (late i486's generation time frame) people were pushing their AMD DX4/80's way up(~120). This was the time when OCing started to hit the main stream. Later when the DX4/120 came out they were hitting 160+ in the face of the P66 and in a few areas the P90. During this time frame Intel cpu's had to be OC'd via the FSB and the AMD's with both the FSB and multiplier.
                          "Be who you are and say what you feel, because those who mind don't matter, and those who matter don't mind." -- Dr. Seuss

                          "Always do good. It will gratify some and astonish the rest." ~Mark Twain

                          Comment


                          • I too have seen PS's fail and take out componets, but that neither here nor there. That is unless you want to start another thread so we can look at all the possiblilties as to why this happens.
                            "Be who you are and say what you feel, because those who mind don't matter, and those who matter don't mind." -- Dr. Seuss

                            "Always do good. It will gratify some and astonish the rest." ~Mark Twain

                            Comment


                            • Originally posted by Greebe
                              I'm sorry but the first ones locked started with the i386 because Intel didn't want get a bad reputation from people recieving OC'd systems claiming to have a higher speed cpu than what was actually sold. Most of us at the time could see their point but instead of Intel moving to setup a different standard of cpu detection instead of measuring raw speed they locked the multiplier. This move proved to be bad news for us who actually had a clue on howto OC a system stable, so that pushed us more toward going with AMD due to extremely high prices and now hindering our efforts. It wasn't until the Diamond Fastbus VLB came out (the first MB that had OC'ing ie truely adjustable FSB). Once this new wave of MB's hit the market AMD's market share started to really take off. Later when AMD managed to start pumping out more power on the S7/SS7 MB's of the time (late i486's generation time frame) people were pushing their AMD DX4/80's way up(~120). This was the time when OCing started to hit the main stream. Later when the DX4/120 came out they were hitting 160+ in the face of the P66 and in a few areas the P90. During this time frame Intel cpu's had to be OC'd via the FSB and the AMD's with both the FSB and multiplier.
                              It looks very strange to me as I remember how I had the idea to overclock (I hadn't the net or BBS at this time) : I read the jumper configuration on my manual and there was a jumper for 1x and 2x and another for 25MHz 33MHz... I just tried to change the 1x to 2x and it worked for a few minutes (there wasn't any fan, just a little heatsink)
                              I mounted a fan and it worked as long as I wanted... (I don't pretend it could live long this way as I didn't keep this one more than 2 months after i discovered this) that's were I started...

                              and it was exactly the same when I got my P166 : just a jumper to move and it was a 200 (and it was definitely not fsb oc) so I can assure at least for 486 and Pentium, that it wasn't locked.
                              for the 386, I only had a word from someone saying it wasn't too, but didn't try as I have never had one)


                              About poor power supply, hmm well, if you want, but I don't really need it anymore as I'm now very careful about the quality of the supplies I buy as I had too many probs with "noname" supplies

                              Comment


                              • My P200MMX@250 was definitely unlocked as well...
                                Let us return to the moon, to stay!!!

                                Comment

                                Working...
                                X