MSI Admits to FSB Overclocking

mmonninmmonnin Centreville, VA
edited June 2003 in Hardware
MSI Admits to FSB Overclocking @ Toms
The lab engineers made a curious discovery when taking a closer look at the 865 Neo 2 from MSI: the manufacturer had integrated an ingenious logic that increased the FSB clock from 6% to 8% when programs were run, which automatically led to overclocking the processor. The trick: with conventional tools, such as the popular WCPUID, Intel CPU Frequency Display, CPUZ and SiSoft Sandra 2003, the clock increase cannot be detected. Only a special tool was able to reveal the raised clock speed.
In the end, this was only made possible in that the board overclocked the 3 GHz process to over 3.2 GHz during operation due to the special logic.

Comments

  • mmonninmmonnin Centreville, VA
    edited June 2003
  • edited June 2003
    Yeah, read about this a week or 2 ago. Tom's has to make such a big deal out of it though; "special tools".:rolleyes:

    From my understanding of what I've read, when the proc gets loaded, the fsb speed overclocks 6-8% to give a little extra crunching power. I bet you could detect this simply by running wcpuid while you have that critter folding some Gromacs work, since the proc wil be 100% loaded.
  • mmonninmmonnin Centreville, VA
    edited June 2003
    Yeah I finally got around to reading their news letters they send out. Been a lil behind.

    Toms said it couldnt be detected by WCPUID or anything of the sort.

    That OCing does mess up the benchmarks tho.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    mmonnin said
    MSI Admits to FSB Overclocking @ Toms

    The lab engineers made a curious discovery when taking a closer look at the 865 Neo 2 from MSI: the manufacturer had integrated an ingenious logic that increased the FSB clock from 6% to 8% when programs were run, which automatically led to overclocking the processor. The trick: with conventional tools, such as the popular WCPUID, Intel CPU Frequency Display, CPUZ and SiSoft Sandra 2003, the clock increase cannot be detected. Only a special tool was able to reveal the raised clock speed.
    In the end, this was only made possible in that the board overclocked the 3 GHz process to over 3.2 GHz during operation due to the special logic.

    I don't know what the "special tool" was as I did not read the article, and maybe it isn't mentioned there either, but perhaps the "special tool" was just overclocking the board MANUALLY until they got the same results, and then just saying what they believed the FSB to be changed to?

    Who knows . . . I don't trust Tom's.
    Not that this isn't true, but until this is repeated by a couple other sites that I do think are trustworthy, then I don't give it much weight.
  • mmonninmmonnin Centreville, VA
    edited June 2003
    Some test show that it was OCed some, IE it outperformed in some tests where it was in the middle of the pack in others.

    If you clicked the link its on the first page.

    http://www6.tomshardware.com/cpu/20030522/images/thg_clock.exe
  • danball1976danball1976 Wichita Falls, TX
    edited June 2003
    I don't think that it matters if they do if it doesn't hurt anything. If it does some good, then its ok
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    I'm wondering how they'd know if the chip can handle the extra MHz.

    Do they just assume it can, or is there some super-test they devised that can determine whether the CPU can handle the extra MHz as the system POSTs?

    This Athlon XP I have here ($44 @ NewEgg) sucks at overclocking, weird thing is I can take it to 2GHz (maybe higher, haven't tried) but I have not been able to get it stable higher than 1.6GHz thus far. Seems odd to me though that it won't even stay stable higher than 1.6, but can go to 2GHz, albeit unstable. Makes me wonder if this Epox is flakey (wouldn't doubt it . . . never had a good Epox in the past, but for $85 I figured I'd give Epox one last try).
    danball1976 said
    I don't think that it matters if they do if it doesn't hurt anything. If it does some good, then its ok
  • ClutchClutch North Carolina New
    edited June 2003
    I'm assuming you have the 1700+ right a2j? What stepping do you have if so? What multiplier is it set to and what fsb?
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    DUT3C (wish it were DLT3C)

    I said in another post that it was an 0310, but I was mistaken, it's an 0312XPMW.

    I took it out today and put in another 0312XPMW DUT3C. Got Prime95 to run for 3 hours and 1 minute before it failed @ 1.8GHz using 1.7v VCore. That's a lot better than the other chip would do. I'll up the VCore a bit and see if I can make 1.8GHz stable . . . if so I'll shoot for 1.9 or 2GHz.

    The timing (as of this moment)
    FSB: 200MHz (DDR400)
    Multiplier: 9
    RAM: 200MHz (DDR400) 2-3-3-6
    VDD: 1.58
    VDimm: 2.63
    VCore: 1.7

    I'm going to change the VCore to 1.75 and see what happens.
    Clutch said
    I'm assuming you have the 1700+ right a2j? What stepping do you have if so? What multiplier is it set to and what fsb?
  • mmonninmmonnin Centreville, VA
    edited June 2003
    A chip is ALWAYS rated at a lower speed than it can go. If its rated at its max any little thing can cause it to not work. Chip manufactures want to make sure it will work and not die out so they rate it at something lower than it can go.
  • leishi85leishi85 Grand Rapids, MI Icrontian
    edited June 2003
    a2jfreak said
    DUT3C (wish it were DLT3C)

    I said in another post that it was an 0310, but I was mistaken, it's an 0312XPMW.

    I took it out today and put in another 0312XPMW DUT3C. Got Prime95 to run for 3 hours and 1 minute before it failed @ 1.8GHz using 1.7v VCore. That's a lot better than the other chip would do. I'll up the VCore a bit and see if I can make 1.8GHz stable . . . if so I'll shoot for 1.9 or 2GHz.

    The timing (as of this moment)
    FSB: 200MHz (DDR400)
    Multiplier: 9
    RAM: 200MHz (DDR400) 2-3-3-6
    VDD: 1.58
    VDimm: 2.63
    VCore: 1.7

    I'm going to change the VCore to 1.75 and see what happens.

    Clutch said
    I'm assuming you have the 1700+ right a2j? What stepping do you have if so? What multiplier is it set to and what fsb?

    dude at least up your vmem to 2.73 or something around that mark.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    Not true.
    Just look back to the 1.13GHz release of the Pentium !!!.
    mmonnin said
    A chip is ALWAYS rated at a lower speed than it can go. If its rated at its max any little thing can cause it to not work. Chip manufactures want to make sure it will work and not die out so they rate it at something lower than it can go.



    This is Corsair PC3200C2 and designed to run @ 2.6, so I think 2.63 is enough voltage, besides the system would crap out even using much lower FSB speeds, so I'm pretty sure RAM has nothing to do with it.
    leishi85 said

    dude at least up your vmem to 2.73 or something around that mark.




    I ran Prime95 using 1.75v this last time around (w/ the second DUT3C) and it crashed after 1h 45m or after 45m, I forget now. Since it actually performed worse with more volts, I backed it down to 1.675v to see if perhaps even 1.7v was a tad too much.

    If push comes to shove I guess I could get the DLT3C I purchased for my mom's machine . . . I know that chip is stable upto 2.3GHz @ 1.65v.
  • mmonninmmonnin Centreville, VA
    edited June 2003
    That was a mistake by intel and was quickly recalled. They were panicking cause they lost the battle to 1ghz and had to outdo AMD.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    Tried 1.675v and it bombed @ the 45m mark.

    Trying 1.6v now @ 1.5GHz (200x7.5) which is only about a 40MHz overclock (since the FSB is like 200.5 or thereabouts). We'll see if that is stable. I think it *should* be, but who knows. I'm growing weary of this. I may buy a DLT3C just so I can overclock and then buy a couple dirt cheap ECS mobos for these DUT3C cpus and turn them into folding rigs using the standard 1.47GHz. Heck, maybe I'll go w/ the Barton 2500+ . . . anyone know if the problem Bartons were having w/ nForce2 boards is fixed on the Epox 8RDA+?
    mmonnin said
    That was a mistake by intel and was quickly recalled. They were panicking cause they lost the battle to 1ghz and had to outdo AMD.
  • leishi85leishi85 Grand Rapids, MI Icrontian
    edited June 2003
    hmm, i said try to up the vmem, the voltage to your ram, not the vcore.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    I had already replied to you . . . look at my post just previous to the one you replied to.

    I just set the FSB to 133 and the multiplier to 12 and VCore to 1.65. Ran Prime95 for 6h 23m before I stopped it. I'll try a higher FSB and see how high it will go and be stable.
    leishi85 said
    hmm, i said try to up the vmem, the voltage to your ram, not the vcore.
  • NoFutureNoFuture In a 3D world...
    edited June 2003
    I'm not sure but I think DUT3C is a lot worse than DLT3C.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    They are . . . but I'm convinced it is the motherboard, or the Northbridge doesn't have adequate cooling (passing heatsink gets much more than warm to the touch).

    I tried the board @ 166MHz (which makes it use 167 on the dot) and a multiplier of 9.5, which kept the CPU right @ 1.6GHz (the same as 133x12, which ran for nearly 6.5 hours before I stopped it) but died 2 mins into Prime95. I'm running Prime95 again just to see how long it will stay, but since the MHz on the CPU is actually about a dozen MHz lower and the VCore is the same and the RAM is SERIOUSLY underclocked, I'm pretty sure it's the mobo.

    Stupid Epox sucks!

    On another note, I got the NF7-S back in for the system I was building for my mother (had to RMA the last one because of a couple of issues) and the board is not doing the same thing so all seems well. Those NF7-Ss (Rev2.0) boards are SWEET!!!!!!
    NoFuture said
    I'm not sure but I think DUT3C is a lot worse than DLT3C.
  • a2jfreaka2jfreak Houston, TX Member
    edited June 2003
    This is a 1700+ so it's supposed to run @ 1.47GHz (actually 1466/1467MHz, but I doubt 3/4MHz kills the thing) but yet it still craps out @ 1.47GHz when using higher than 133MHz FSB. I'm pretty sure it's not the RAM, but I'll check that next and use some Corsair that I *KNOW* is good for at least 200MHz, so anything lower than 200MHz should be fair game.

    I'm fairly well convinced it's the board and not the RAM at this point, but I'll find out.

    Here's a pic of CPUZ and Prime95.

    Ignore CPUZ's VCore reading as it reports the VDD not the VCore. The VCore was the default 1.6 for this run.

    Memory timings: Cas 2.0, 3-3-6 @ 1.63volts
Sign In or Register to comment.