Karbosguide.com - Module 3d.5

About Cooling and Over clocking (continued)

The contents:

  • An example of overclocking
  • The SDRAM speed
  • Features of the Abit BX6 motherboard.
    		•
  • Next page
  • Previous page

    • An over-clocked Pentium II

    In the previous pages you can read about the theory behind over clocking. Here I describe a practical case.

    In April 1999 we needed a new workstation. It was to used for graphics work and sometimes video editing, so it had to be speedy. We decided to try some over clocking.

    Over clocking with Intel - earlier results

    Intel CPUs have always been good for over clocking.

    Back in 1997 we had a Pentium Pro designed for 200 MHz. It ran (and still runs) at 233 MHz without any problem at all.

    Later we got one of the first Pentium IIs. These processors were very friendly to over clocking, both the frequency of the system bus as well as the clock factor could be changed. A modest 233 MHz version ran (and still does) at 300 MHz.

    The Deschutes kernel of second generation Pentium II and Celeron was changed, so every CPU only could work with a specific clock factor. This means that you only can over-clock by increasing the bus frequency. This has been the situation with all later Intel processors.

    You see our over clocking results as described are not extreme. This has a reason; all our PCs function in a network and they are heavily used for various demanding tasks. So they have to be completely stable, which they also have been. Further over clocking would aggravate the inherent un-stability.

    The first attempt

    We started up with the cheapest solution. A 300 MHz Celeron should be doing fine at 450 MHz if the system bus was increased from 66 MHz to 100 MHz. We even added extra cooling, a fan placed above the SEC module:

    It never worked. But the motherboard was interesting, so we went for another approach.

    Pentium II and Abit BX6

    We then purchased a Pentium II-450 MHz. This processor was the clock factor 4.5 model of the Pentium II you could say.

    The motherboard was the newest version (2.0) of the pretty well-known Abit BX6. It is a BX-based board with is capable of delivering a lot of different frequencies. The clock multiplier goes up to factor 8, but since the Pentium II only works with clock factor 4.5, we had these options:

    Bus frequency
    (SDRAM speed)
    Clock
    factor
    Resulting
    CPU frequency
    L2 Cache
    speed
    66 MHz
    4.5
    300 MHz
    150 MHz
    75 MHz
    4.5
    338 MHz
    169 MHz
    83 MHz
    4.5
    375 MHz
    188 MHz
    100 MHz
    4.5
    450 MHz
    225 MHz
    112 MHz
    4.5
    504 MHz
    252 MHz
    117 MHz
    4.5
    527 MHz
    263 MHz
    124 MHz
    4.5
    558 MHz
    229 MHz
    129 MHz
    4.5
    581 MHz
    290 MHz
    133 MHz
    4.5
    599 MHz
    300 MHz
    138 MHz
    4.5
    621 MHz
    310 MHz
    143 MHz
    4.5
    644 MHz
    322 MHz
    148 MHz
    4.5
    666 MHz
    333 MHz
    153 MHz
    4.5
    689 MHz
    344 MHz

    Of course I could not expect my Pentium II to run at 689 MHz. The values are theoretical.

    When you increase the bus frequency it affects a lot of units within the PC. This is due to the architecture, where the system bus so to say is a local bus, with other attached buses and units working synchronously. Increasing the bus frequency influences:

  • The CPU clock frequency. Often Intel CPUs are capable of working at a higher frequency than what they are sold for. However, improved cooling is important.

  • The L2 Cache of the Pentium II module. It has an upper speed limit as all other RAM types do. Cooling is important for the L2 cache RAM chips.

  • The SDRAM speed. The RAM modules have to fast enough to cope with the increased bus frequency.

  • The PCI units. The graphics controller, EIDE controller and network controller all have to work at around 33 MHz, otherwise un-stability is the result (at least that is our experiences).

  • The AGP bus speed.

    Over clocking a PC is not that simple. All the mentioned units have to be tuned, so they work at right frequencies.

    Testing and trying

    One of the biggest problems is to control the speed of the PCI units. Our network (LAN) is a very good tool for testing this. I make a backup of all my documents (> 10.000 files) across the network from harddisk to harddisk, and if this works i am pretty sure that everything is all right with the new PC.

    With the Pentium II, I started increasing the bus frequency. Of course everything worked fine at 100 MHz. It should. 112 MHz was completely stable. 117 MHz as well, but at 124 MHz the problems came. Here you see the Soft Menu setting, which is an extremely nice feature of the BX6 board:

    The PC seemed to work at 558 MHz, but the file copy-test could not be performed. The PC froze. This probably was due to "slow" SDRAM. With better RAM it might have worked.

    SDRAM speeds

    Here is an theoretical calculation of the required SDRAM speed:

    Bus frequency
    SDRAM speed
    (Nano seconds)
    66 MHz
    15.02
    75 MHz
    13.33
    83 MHz
    12.00
    100 MHz
    10.00
    112 MHz
    8.93
    117 MHz
    8.55
    124 MHz
    8.03
    129 MHz
    7.75
    133 MHz
    7.52
    138 MHz
    7.25
    143 MHz
    6.99
    148 MHz
    6.76
    153 MHz
    6.54

    The RAM was of PC100 type. But this may be 10, 8 or 7 ns. In our case it was 8 ns, so the 124 MHz setting should have been working, it just didn't.

    Two versions of 117 MHz

    At 117 MHz I had two options. I could go for a PCI bus at 39 or 29 MHz. These values come out as one third or one quarter of the 117 MHz bus frequency. Unfortunately 39 MHz was too much for my PCI units:

    Soft Menu setting: PCI 1/3
    Soft Menu setting: PCI 1/4
    PCI frequency: 39 MHz
    PCI frequency: 29 MHz
    System stability: not good
    System stability: 100% all right

    So we ended up with a completely stable Pentium II system running at 527 MHz. That's absolutely OK.

    Features of the Abit BX6

    The Abit board seems pretty cool to me. The manual is OK but not overwhelming impressive. The board has 5 PCI slots which I like. But especially the Soft Menu II is great - a brilliant tool for over-clockers. You do not have to move a simple jumper on the BX6 board, so it is extremely simple to test your CPU and system at various frequencies.

    You also get thermistor to detect the CPU temperature:

    It is taped to the heat sink and connected to the motherboard.

    You get some software, among others this diagnostic tool:

    More over clocking?

    With better RAM we might tweak the full 689 MHz out of the Pentium II processor. Running with a bus frequency of 153 MHz, the PCI units have to work on 38,25 MHz which I very much doubt they can.

    My realistic guess would be that this configuration using 7 ns SDRAM might work:

    Bus
    frequency
    CPU
    frequency
    SDRAM
    speed
    PCI
    frequency
    138 MHz
    621 MHz
    7,25 ns
    34,5 MHz

  • Next page
  • Previous page
  • PC Architecture. More about CPUs

    Learn more

    Read more about the boot process and system bus in Module 2b

    Read more about I/O buses in module 2c

    Read more about the motherboard chip set in module 2d

    Read more about RAM in module 2e

    Read about EIDE in module 5b

    [Main page]
    [Contact]
    [Karbo's Dictionary]
    		 [The Software Guides]

    Copyright (c) 1996-2005 by Michael B. Karbo. www.Karbosguide.com.