poachedeggs wrote:
>> poachedeggs wrote:
>>> I thought this would be overlooked if I tagged it onto the rest of my
>>> thread from a fortnight or so back:
>>> http://groups.google.com/group/alt.comp.hardware.homebuilt/browse_thr ...
>>> Advice from the previous wizard would be appreciated. :-)
>>> When I run Speccy or CPU-Z now, the Jedec readings run from 400mhz
>>> through 333 mhz to 266 mhz instead of 400, 266 and 200 and the
>>> apparently faulty 400, 400, 200 that showed for the unmatched sticks
>>> that were replaced.
>>> With the 400mhz, Jedec #3, set of numbers the Cas Latency is 6, as is
>>> Ras to Cas and Cas precharge. The previous 5,5,5, 18, 23 string of
>>> numbers I'd learnt to key into my BIOS now apply to Jedec #2, 333
>>> mhz. Does this suggest I have RAM with a Cas Latency of 6 now instead
>>> of 5, and that I should change the BIOS to 6,6,6, 18, 24 as per the
>>> apparently appropriate Jedec 3 readings in Speccy and CPU-Z? Or can I
>>> leave these BIOS entries as 5,5,5, 18, 23, in which case am I
>>> currently 'overclocking' this RAM, given that in the SPD readings
>>> 333mhz pertains to 5,5,5, 18, 23? All seems stable though. I hope
>>> what I'm saying is as clear as a CPU-Z dump.
>>> I am peeved slightly at the thought of possibly having a lower CL now,
>>> if I have, but this has gone on a bit too long for me and I'll settle
>>> with it. Just like to know if I should alter the BIOS string of
>>> numbers mentioned here for stability's sake or leave it.
>>> Many thanks.
>> Sounds like your memory is CAS6.
>>
>> First, you start by checking the paper specification for the RAM
>> you got. The paper specification overrides the SPD, on enthusiast
>> RAM. If they promise it supports DDR2-800 CAS4 for example, then
>> they probably warrant that to be the case, and you'd return the
>> product if it didn't meet the paper specification they promised.
>>
>> The SPD can contain the wrong information on purpose. "To encourage
>> the computer to be able to start when the memory is first plugged
>> in." But a customer who buys RAM, doesn't want the SPD table
>> to be so corrupt, as to cause the BIOS to malfunction.
>>
>> If the packaging for the product, says it is DDR2-800 CAS5, then
>> you'd assume the SPD was set that way (looser than the value printed
>> on the package) so the memory can be started OK.
>>
>> As it is, if you're seeing 6-6-6-18-24 in the SPD table for
>> DDR2-800 (400MHz), then you could have been given a higher
>> latency CAS6 memory product as a replacement.
>>
>> The way the JEDEC tables work, is they define three consecutive
>> CAS points. Think of it as X, X+1, X+2 CAS entries in the table.
>> The frequency values do not have to be consecutive, as the frequency
>> value is the point at which the new CAS value applies.
>>
>> The only thing that would not make sense, is a table with two 400MHz
>> entries, because that implies one table value is redefining the other.
>> You could have CAS4,5,6 and frequencies 266, 333, 400, or you
>> could have 266, 400, 533. But you shouldn't have 266, 400, 400, as
>> that doesn't make any sense (one of the table entries
>> is redundant in that case).
>>
>> If we take CAS4,5,6 and 266, 400, 533 as an example, it means to use
>> CAS6 between 533 and 400. CAS5 between 400 and 266. CAS4 at 266
>> or a slower frequency. In the case of the "missing frequency", like
>> the 333 value, we know that 333 is between 400 and 266, and it would be
>> CAS5 at that frequency. The table is defining ranges of values.
>>
>> If you use the BIOS auto setting, and somehow specify the RAM should
>> run at DDR2-800, then the BIOS will extract the values from the SPD
>> and use them. If the BIOS is brain dead (and some are), and you check
>> with CPUZ and see the values are not correct, you'd set them manually.
>>
>> If the paper specification for the memory, has a tighter set of values
>> listed, then you'd set them up manually, apply the Vdimm required for
>> those settings, and do memtest. If the memory doesn't meet its paper
>> spec, then you'd return it.
>>
>> *******
>>
>> So the first question I'd have to ask you, is what is the new part number
>> for the memory, and do you have a piece of paper stating what
>> timing parameters and recommended voltage that it uses. If you have
>> absolutely no documentation at all, then you have to rely on the SPD
>> as your best means of setting up the memory. And then it would be
>> CAS6 DDR2-800.
>>
>> Paul
>
> Absolutely brilliant, marvellously committed and generous response -
> thanks. Yes, no paper and not even any note on the stick itself
> saying if it's CL5 or CL6. I imagine and hope that as it seems 667
> mhz RAM still seems the average given in non-custom machines, laptops
> etc that with the RAM still being 800mhz it doesn't make too much
> difference about the higher latency, especially as I'm not a gamer.
>
> I think if I've followed you right the thing to do now is see how it
> goes with auto settings and then if it goes a bit cranky to use the
> manual settings for CL6. To recap, that double entry for 400mhz
> pertained to the sticks I sent back, or rather one of them. The new
> ones check out fine as a pair.
>
> I don't know what the 'part number' is, but this is what's on the
> sticks' container: Integral IN2T2GNXNFX and then DDR2 2GB 800mhz Non-
> ECC DIMM and finally 44-80-40 D09, if anything meaningful can be
> extracted from that, I imagine the longer letter/number combo is just
> a serial number anyway. This is basic £35 per stick memory, probably
> not what people call 'enthusiast' memory, at a guess.
>
> Thanks again.
Yup, that "Integral IN2T2GNXNFX" is the part number, as you can see
here. It is DDR2-800 CAS6 memory. Go ahead and use it that way.
The thing is, if you do the arithmetic, CAS5 at DDR2-667 is going to be
roughly the same as CAS6 at DDR2-800, so there is no latency difference
measured in nanoseconds, between the two of them. But there is a
bandwidth difference, and so you'd select DDR2-800 as the operating
point, since it has higher bandwidth. The BIOS will likely select that
setting as well, barring an inability to find a divider to enable that speed.
http://www.misco.co.uk/applications/SearchTools/item-details.asp?EdpNo=339169&Tab=2&NoMapp=0
CAS5 * 800/667 = 5.997, so CAS5 @ DDR2-667 has same latency as CAS6 @
DDR2-800
Set it to DDR2-800, if it isn't there already, and start your testing.
Paul
XML Sitemap
>>> I thought this would be overlooked if I tagged it onto the rest of my
>>> thread from a fortnight or so back:
>>> http://groups.google.com/group/alt.comp.hardware.homebuilt/browse_thr ...
>>> Advice from the previous wizard would be appreciated. :-)
>>> When I run Speccy or CPU-Z now, the Jedec readings run from 400mhz
>>> through 333 mhz to 266 mhz instead of 400, 266 and 200 and the
>>> apparently faulty 400, 400, 200 that showed for the unmatched sticks
>>> that were replaced.
>>> With the 400mhz, Jedec #3, set of numbers the Cas Latency is 6, as is
>>> Ras to Cas and Cas precharge. The previous 5,5,5, 18, 23 string of
>>> numbers I'd learnt to key into my BIOS now apply to Jedec #2, 333
>>> mhz. Does this suggest I have RAM with a Cas Latency of 6 now instead
>>> of 5, and that I should change the BIOS to 6,6,6, 18, 24 as per the
>>> apparently appropriate Jedec 3 readings in Speccy and CPU-Z? Or can I
>>> leave these BIOS entries as 5,5,5, 18, 23, in which case am I
>>> currently 'overclocking' this RAM, given that in the SPD readings
>>> 333mhz pertains to 5,5,5, 18, 23? All seems stable though. I hope
>>> what I'm saying is as clear as a CPU-Z dump.
>>> I am peeved slightly at the thought of possibly having a lower CL now,
>>> if I have, but this has gone on a bit too long for me and I'll settle
>>> with it. Just like to know if I should alter the BIOS string of
>>> numbers mentioned here for stability's sake or leave it.
>>> Many thanks.
>> Sounds like your memory is CAS6.
>>
>> First, you start by checking the paper specification for the RAM
>> you got. The paper specification overrides the SPD, on enthusiast
>> RAM. If they promise it supports DDR2-800 CAS4 for example, then
>> they probably warrant that to be the case, and you'd return the
>> product if it didn't meet the paper specification they promised.
>>
>> The SPD can contain the wrong information on purpose. "To encourage
>> the computer to be able to start when the memory is first plugged
>> in." But a customer who buys RAM, doesn't want the SPD table
>> to be so corrupt, as to cause the BIOS to malfunction.
>>
>> If the packaging for the product, says it is DDR2-800 CAS5, then
>> you'd assume the SPD was set that way (looser than the value printed
>> on the package) so the memory can be started OK.
>>
>> As it is, if you're seeing 6-6-6-18-24 in the SPD table for
>> DDR2-800 (400MHz), then you could have been given a higher
>> latency CAS6 memory product as a replacement.
>>
>> The way the JEDEC tables work, is they define three consecutive
>> CAS points. Think of it as X, X+1, X+2 CAS entries in the table.
>> The frequency values do not have to be consecutive, as the frequency
>> value is the point at which the new CAS value applies.
>>
>> The only thing that would not make sense, is a table with two 400MHz
>> entries, because that implies one table value is redefining the other.
>> You could have CAS4,5,6 and frequencies 266, 333, 400, or you
>> could have 266, 400, 533. But you shouldn't have 266, 400, 400, as
>> that doesn't make any sense (one of the table entries
>> is redundant in that case).
>>
>> If we take CAS4,5,6 and 266, 400, 533 as an example, it means to use
>> CAS6 between 533 and 400. CAS5 between 400 and 266. CAS4 at 266
>> or a slower frequency. In the case of the "missing frequency", like
>> the 333 value, we know that 333 is between 400 and 266, and it would be
>> CAS5 at that frequency. The table is defining ranges of values.
>>
>> If you use the BIOS auto setting, and somehow specify the RAM should
>> run at DDR2-800, then the BIOS will extract the values from the SPD
>> and use them. If the BIOS is brain dead (and some are), and you check
>> with CPUZ and see the values are not correct, you'd set them manually.
>>
>> If the paper specification for the memory, has a tighter set of values
>> listed, then you'd set them up manually, apply the Vdimm required for
>> those settings, and do memtest. If the memory doesn't meet its paper
>> spec, then you'd return it.
>>
>> *******
>>
>> So the first question I'd have to ask you, is what is the new part number
>> for the memory, and do you have a piece of paper stating what
>> timing parameters and recommended voltage that it uses. If you have
>> absolutely no documentation at all, then you have to rely on the SPD
>> as your best means of setting up the memory. And then it would be
>> CAS6 DDR2-800.
>>
>> Paul
>
> Absolutely brilliant, marvellously committed and generous response -
> thanks. Yes, no paper and not even any note on the stick itself
> saying if it's CL5 or CL6. I imagine and hope that as it seems 667
> mhz RAM still seems the average given in non-custom machines, laptops
> etc that with the RAM still being 800mhz it doesn't make too much
> difference about the higher latency, especially as I'm not a gamer.
>
> I think if I've followed you right the thing to do now is see how it
> goes with auto settings and then if it goes a bit cranky to use the
> manual settings for CL6. To recap, that double entry for 400mhz
> pertained to the sticks I sent back, or rather one of them. The new
> ones check out fine as a pair.
>
> I don't know what the 'part number' is, but this is what's on the
> sticks' container: Integral IN2T2GNXNFX and then DDR2 2GB 800mhz Non-
> ECC DIMM and finally 44-80-40 D09, if anything meaningful can be
> extracted from that, I imagine the longer letter/number combo is just
> a serial number anyway. This is basic £35 per stick memory, probably
> not what people call 'enthusiast' memory, at a guess.
>
> Thanks again.