Archived from groups: alt.comp.periphs.mainboard.asus (
More info?)
On Mon, 05 Sep 2005 03:33:25 GMT, nospam@needed.com (Paul) wrote:
>In article <6v5mh1tp2a1up4s85g7i9j1jtf4f665340@4ax.com>, curiosity <c@.t> wrote:
>
>> this cpu will run at 2.17Ghz with the fsb at 333 but my 2100 ram
>> requires 266fsb. I'm currently running it at 333 but it will crash
>> unpredictably within the hour or day depending on mood.
>>
>> I've tried to set the bios using 'manual' for the cpu and a suitable
>> multiplier on an fsb setting of 266 but the cpu insists on running at
>> 1.74Ghz whatever the multiplier.
>>
>> A google search suggests this is a problem other a7v333 users have
>> experienced.
>>
>> Does anyone have a workaround?
>
>I can see a couple possibilities. You could do some endless
>research on whether a locked multiplier processor can be
>unlocked, what bridges control it, and so on. Chopping up
>bridges on a processor doesn't sound like a lot of fun.
>
>Or, you could try to stabilize the RAM. There are a couple
>of things you could try for the RAM.
>
>The board has two jumpers that aren't documented in the
>manual. They are mentioned here, and elsewhere.
>
>http://forums.pcper.com/printthread.php?t=113092&pp=40
>
> "Just above the AGP slot there are 2 undocumented jumepers
> J1 and J2, with them it is possible to adjust the DIMM/RAM
> voltage. The settings are:
>
> JP1 [1-2] - JP2 [1-2] = 2.50volt
> JP1 [2-3] - JP2 [1-2] = 2.65volt
> JP1 [1-2] - JP2 [2-3] = 2.75volt
> JP1 [2-3] - JP2 [2-3] = 2.80volt "
>
>According to this article, the default JP1/JP2 settings are
>already overvolting the RAM, so perhaps no further tweaking
>is needed. 2.75V should be enough.
>
>http://www.lostcircuits.com/motherboard/asus_a7v333/5.shtml
>
> JP1 JP2
> x x i = jumper in lower position
> x x ! = jumper in upper position
> x x
>
> "ii 2.59 - 2.63V (same with one or both jumpers removed completely)
> i! 2.78 - 2.85V (default setting, out of spec according to JEDEC guidelines)
> !i 2.88 - 2.96V
> !! 2.95 - 3.06V "
>
>One article I read a while back, mentioned increasing the Trcd
>setting, as a way to help overclock a memory. Have a look in
>the BIOS for that setting.
>
>Changing the memory clock, means the timings have to change
>too. If set to Auto, the BIOS will compute the necessary
>scaling and rounding up, needed to figure out the new timing
>numbers. You can use a copy of CPUZ from cpuid.com , to see
>what the BIOS is using for timings. Write down the numbers,
>then go into the BIOS, select a Manual mode for the timing,
>and enter those numbers manually. Increasing Trcd by 1, is
>supposed to open up more head room, in case the RAM cannot
>really handle a 25% overclock.
>
>The following is how you would calculate new timing numbers
>manually:
>
>Say the RAM is 2-2-2-5 at DDR266. The clock rate in that case
>is 133MHz, and the clock period is 7.5ns (the inverse of 133).
>For CAS, the first number, 2*7.5=15ns. At the new speed of
>DDR333, the clock is 166Mhz and the clock period is 6ns.
>The internal timing number of 15ns, divided by the new clock
>period of 6ns, is 2.5. For CAS, you round to the next nearest
>half-integer value - in this case, 2.5 happens to be right on
>the line, so you get to use 2.5. If the division operation
>had returned 2.6, you would round up to CAS 3.0 . You
>do a similar calculation for the other timing numbers too,
>except the other ones round to the nearest integer. A
>division that returned 2.2 would round to the next full
>integer, which is 3. So, the 2-2-2-5 would become 2.5-3-3-7
>at DDR333. If the RAM did not like the overclock, the
>second number, Trcd, would be increased, as in 2.5-4-3-7.
>(There is a limit to how far these numbers can be pushed,
>as a hardware register somewhere has to have room for the
>number you just computed.)
>
>Test the memory, by using a copy of memtest86+ from
>memtest.org . Memtest86+ can boot the computer, without
>the use of an OS. Do at least a couple of full passes error
>free, with memtest86+, before considering booting back
>into Windows. Then, use Prime95 (torture test) from mersenne.org,
>as a final test that the overclocked memory timings are
>working properly. Again, no errors are acceptable.
>
>If you simply cannot stabilize the memory, either consider
>a future filled with processor bridge cutting, or buy some
>fast-enough cheap DDR ram before the price goes out of sight.
>DDR400 is backward compatible with DDR333, DDR266, DDR200, so
>can be used as a "universal donor".
>
>HTH,
> Paul
thanks Paul, I couldn't possibly have asked for a more comprehensive
answer! I have a mixture of reasonable and indifferent ram here so I
think - as interesting as it would be to put it collectively through
its paces - this might end up being a thankless task with
unsatisfactory results.
I'm a complete beginner on this subject but I see that DDR400 is still
around at reasonable prices - is it expected then that prices could
rise soon?
I'm not one of nature's overclockers - I run a modest audio
workstation in a small room and noise is ananthema so I prefer to run
with a quiet fan at low temps.
Much appreciate the response though - many thanks.