Opinions and advice on Overclock please...

Discussion in 'Abit' started by tHatDudeUK, Sep 14, 2003.

  1. tHatDudeUK

    tHatDudeUK Guest

    Hardware

    XP2200+ (Thoroughbred A [I think])
    Coolermaster Aero 7 lite
    Samsung DDR333 512mb rated CAS 2.5
    Thermaltake Aluminum active DDR cooler
    Abit NF7-S v2.0
    Antec 550 watt PSU

    Current Bios settings and benchmarks

    219 FSB X 9 (= 1971Mhz) (If I remember correctly I already tried a
    multiplier of 9.5 but no joy, enjoyed many CMOS resets, hehe.)

    Processor set at 1.725 volts (I think)

    Based on above memory speed = DDR438

    Memory 8 - 3 - 3 - CAS 3.0

    Memory Voltage 2.9 volts.

    Memory bandwidth Benchmark SiSoft Sandra 3150/2834 mb/s (compared with
    2852/2601 mb/s at 206x9.5 9-4-4-2.5)

    (This is a far cry from being stuck at 133mhz FSB on my Gigabyte 7VAXP!)

    Will this overclock be likely to melt my stick of RAM or any other piece of
    hardware? How can I really tell it's stable?

    Would I be better off lowering my FSB and sticking up my RAM timings to get
    better performance? (The SiSoft sandra benchmark does little to indicate
    anything about response times, just throughput.)

    Where should I stick my tiny Coolermaster aluminum chip heatsinks? In the
    draw or on my southbridge etc?

    Any other thoughts or recommendations?

    TIA

    tHatDudeUK
     
    tHatDudeUK, Sep 14, 2003
    #1
    1. Advertisements

  2. tHatDudeUK

    Ben Pope Guest

    Thats probably quite good for a Tbred A, and crap for a Tbred B.
    Thats quite high.
    I've not had any experience with melting RAM or running it much over spec...
    I suggest looking up the data sheet for the memory (if it's samsung thats
    easy). I would suggest that 2.9V is pushing the RAM pretty hard, does it
    have heat spreaders?

    Prime95 pulls out cpu errors for me. And memtest86 seems pretty thorough.

    If you can run both of those tests for an hour, you're probably reasonably
    stable.
    Throughput is probably still a reasonable measure - the greater throughput
    you can get, the quicker it is. Make sure FSB and RAM are in synch though.
    Well... if you are going to be running your FSB at 220 or so, it might be an
    idea. Touch the chip to see if it gets hot. Careful though, some of the
    components on the A7N8X reach up to about 95°C (I think thats a voltage
    regulator). Southbridge should be considerably less :)

    Ben
     
    Ben Pope, Sep 14, 2003
    #2
    1. Advertisements

  3. tHatDudeUK

    tHatDudeUK Guest

    That's what I thought. I'll try 2.8 but it seems to work ok as it is.
    A thermaltake aluminum heatspreader with an active fan cooler. It seems
    quite cool to the touch, especially in comparison with when it was running
    below spec and directly touching the chips. I'm using a Coolermaster
    ATCS-201-SX case and it seems to have pretty good airflow.
    though.

    For some weird reason my Abit board has dividers for 3/3, 4/4 and 6/6. I
    have set it at 4/4 although I can't quite understand how any of these 3
    settings will make much difference.
    Still enough to make me take my finger away :)
     
    tHatDudeUK, Sep 15, 2003
    #3
  4. tHatDudeUK

    Ben Pope Guest

    Use the lowest you can get away with. I'm using 2.7V on some Corsair
    XMS3200LL and I can get 5-2-2-2 at 220MHz.
    Active cooling on your memory is going some... :)
    Looks like somebody missed the point and got confused as we all know that
    3/3 = 4/4 = 6/6 = 1
    I've not tried touching mine. If it's hot to the touch then a heatsink may
    be of use.

    Ben
     
    Ben Pope, Sep 15, 2003
    #4
  5. Not keen on overclocking. How often do you need the extra speed?
     
    Johannes H Andersen, Sep 15, 2003
    #5
  6. tHatDudeUK

    Skid Guest

    Every time I boot up ;>)

    Seriously, why should anyone settle for less than the most performance their
    hardware can deliver? My P4 2.4g is just as stable at 3.26, and a hell of a
    lot quicker at everything I do.
     
    Skid, Sep 15, 2003
    #6
  7. But it's unknown territory. You only know that it works so far, instead of
    that it works because the manufacturer assures that it works. Your computer
    might not have encountered the whole spectrum of possible conditions.
     
    Johannes H Andersen, Sep 15, 2003
    #7
  8. tHatDudeUK

    Ben Pope Guest

    Manufacturers assurances are not guarantees that they will work.
    Manufacturers cannot test every possible condition either.

    Ben
     
    Ben Pope, Sep 15, 2003
    #8
  9. Agreed, but they'll probably have a far better idea.
     
    Johannes H Andersen, Sep 15, 2003
    #9
  10. tHatDudeUK

    Ben Pope Guest

    And the difference is the risk the overclocker is willing to take.

    If you can get through memtest86 and Prime95 torture, you're not gonna be
    far off stable and reliable enough for most tasks..

    Ben
     
    Ben Pope, Sep 15, 2003
    #10
  11. tHatDudeUK

    Skid Guest

    I've torture-tested this machine extensively with MemTest, Prime95, Sandra's
    burn-in utility, 3DMark 2003 and evey game and benchmark I can throw at it.
    I found the upward limit of stability and backed down a notch until it was
    solid as a rock. It passes all tests with flying colors.

    If it makes you feel better to run at default settings, by all means do so.

    It makes me feel better to know that I have fine-tuned my rig for maximum
    performance at minimum cost.

    My approach takes a lot more skill and a little more luck. On the other
    hand, it's an intellectual challenge with an economic payoff.

    Your approach allows Intel and AMD to continue charging vastly higher prices
    for cpus based on exactly the same core because they set different default
    multipliers. At least AMD has the decency to sell them unlocked so you can
    experiment more easily to find the chips true capacity.

    This is an old argument, and it never really gets settled.

    But given the choice between a solid 2.4 ghz system and a solid 3.26 ghz
    system with the same quality components for the same price, I'm going to
    pick the faster one every time. That's the fun of overclocking.

    I also have a Radeon 9500 that is modded and overclocked to perform as well
    as a 9700 Pro. That's a huge jump in performance at zero cost.

    If you prefer life in the slow lane, sobeit. Me, I can't drive 55 ;>)
     
    Skid, Sep 15, 2003
    #11
  12. tHatDudeUK

    tHatDudeUK Guest

    Which is why I got rid of the crap Gigabyte 7VAXP with AC'97 audio and no
    5th multiplier (hence stuck at 133mhz FSB). Replace with Abit NF7-S v2.0 and
    all is nice and fast and the overclocking options in the BIOS are just too
    tempting to not fiddle with!

    All the other hardware is the same. Just goes to show that a motherboard
    upgrade alone can be worth it in the correct circumstances.
     
    tHatDudeUK, Sep 15, 2003
    #12
  13. tHatDudeUK

    tHatDudeUK Guest

    Of how to make money! Most systems pre-built these days will harp on about
    DDR RAM and won't give the speed (probably because it's only DDR266),
    they'll harp on about the size of the hard disk, but forget about the speed.
    It's all a case of cheapest goes and sell it at the highest possible going
    price.

    Apparently DDR400 was unofficial (probably still is?) for a while and not an
    officially recognised standard although every motherboard manufacturer got
    in on the game!
     
    tHatDudeUK, Sep 15, 2003
    #13
  14. tHatDudeUK

    tHatDudeUK Guest

    Also possibly by the time you destroyed your hardware by overclocking it
    would only be worth hardly anything anyway what with computers devaluing
    faster than cars and all...
     
    tHatDudeUK, Sep 15, 2003
    #14
  15. Jens C. Hansen [Odense], Sep 15, 2003
    #15
  16. tHatDudeUK

    tHatDudeUK Guest

    Unfortunately this has led to fake CPU's entering the market... I wonder if
    AMD will react by changing this fact?!
    Woahh. Well, I'll probably be stuck with my Ti4200 for a while until the GFX
    market starts to change and prices drastically drop (IE. 9800 pro for under
    £180). I think my card just about manages 4600 speeds.
     
    tHatDudeUK, Sep 15, 2003
    #16
  17. I'll choose a computer that works over a faster one - any time ;>)
     
    Johannes H Andersen, Sep 15, 2003
    #17
  18. tHatDudeUK

    Skid Guest

    Sure. Me too. But mine works quite nicely, thanks. In fact, that overclocked
    fsb means it works faster than any stock Pentium 4 system. My faux 3.2 at
    12x270 is faster than a genuine 3.2 at 16x200 because it has more bandwidth
    to play with and faster memory feeding it.

    It's running at stock voltage, and now that I've replaced the stock hsf it
    runs cooler than it did before it was overclocked. I've tested it with every
    conceivable piece of burn-in and cpu/memory checking software I can find,
    and it's as stable as it was at 2.4g -- just quicker.

    I'm not one of those "push it until it fries" or "damn the spontaneous
    reboots, just let me break the record in 3DMark" kind of guys. I wouldn't
    settle for anything less than 100 percent stability.

    But as both Intel and AMD increase their yields and improve their fabs,
    there is a lot more headroom in this generation of chips. It's an individual
    choice whether to take advantage of it or not.
     
    Skid, Sep 15, 2003
    #18
    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.