Aggressive memory settings questions

  • Thread starter Thread starter Howie
  • Start date Start date
H

Howie

A7N8X-Deluxe board running 3 sticks of 256mb Infineon PC3200 DDR400
RAM CL2.5...

I swapped one of the sticks at my local computer shop for one stick of
Samsung PC3200 CL3 512mb. The logic was twofold: 1) Wanted to
increase system RAM to 1gig and 2) Take advantage of the Dual Channel
capability of this board (256+256 in slots 1 & 2, 512 in slot 3).

Got home, installed the 512mb stick, everything booted up fine, 1gig
of RAM in place. I then ran Sandra to analyze the memory and it
reported the new stick was CL2.5 up to 166MHz and CL3 up to 200MHz.
While I didn't feel this was anything to be concerned about in the
grand scheme of things, I did take note of the fact my 256mb sticks
were rated at CL2.5 at 200MHz and the new 512mb stick CL3.

Then, quite ignorantly, I decided to play a bit with the memory
settings. I selected "Aggressive" from the BIOS, saved, and received
an audible report the "System Failed Memory Test". I quickly realized
I could not simply reboot and get into the BIOS to change this setting
back, so I removed the new stick of memory, leaving the two 256mb
sticks in Slots 1 & 2. I was somewhat surprised to receive the same
error warning, so I removed the stick from Slot 2, tried again, and no
dice.

Before resorting to flashing a new BIOS, on a whim I removed the
remaining 256mb stick and replaced it with the 512mb stick-- Slot 1.
I say "a whim" as it was rated at CL3 and I recall the "Aggressive"
settings being in the neighborhood of CL2.5 in the BIOS memory setting
menu.

I was shocked to hear the computer POST and I was able to get into the
BIOS, change the setting, reinstall the 1gig of memory, and proceed as
if nothing happened. I considered myself lucky as I didn't really
feel like fooling with a reflash!

Anyway, a couple of questions that I have from that episode...

1) Does it make sense that with the "Aggressive" settings, my machine
would boot with memory rated "slower" (rated CL3 memory installed
allowed it to boot where rated CL2.5 failed)?

2) Does the possible answer lie in the logic that my CL3 rated Samsung
memory is faster than advertised and the CL2.5 rated Infineon memory
is perhaps "overrated"?

3) I believe, dual channel or otherwise, my system will default to the
slowest rated RAM (CL3). Is this correct?

4) Even with CL2.5 rated RAM, if the BIOS memory settings call for
CL3, I presume the memory will "throttle back" to the CL3 speed (it
appears the "Auto" setting calls for this CL3 speed)?

5) It has been suggested in this group that the benefits of
dual-channel capability are miniscule, at best, versus single channel.
What discernable performance difference will one find with memory
running at CL2.5 versus CL3? I should have asked this question prior
to fiddling with the setting as it might be one of those imperceptible
gains noticeable only on benchmark testing.

Thanks in advance for your answers, input, and counsel!

--Howie
 
Howie said:
A7N8X-Deluxe board running 3 sticks of 256mb Infineon PC3200 DDR400
RAM CL2.5...

I swapped one of the sticks at my local computer shop for one stick of
Samsung PC3200 CL3 512mb. The logic was twofold: 1) Wanted to
increase system RAM to 1gig and 2) Take advantage of the Dual Channel
capability of this board (256+256 in slots 1 & 2, 512 in slot 3).

Got home, installed the 512mb stick, everything booted up fine, 1gig
of RAM in place. I then ran Sandra to analyze the memory and it
reported the new stick was CL2.5 up to 166MHz and CL3 up to 200MHz.
While I didn't feel this was anything to be concerned about in the
grand scheme of things, I did take note of the fact my 256mb sticks
were rated at CL2.5 at 200MHz and the new 512mb stick CL3.

Then, quite ignorantly, I decided to play a bit with the memory
settings. I selected "Aggressive" from the BIOS, saved, and received
an audible report the "System Failed Memory Test". I quickly realized
I could not simply reboot and get into the BIOS to change this setting
back, so I removed the new stick of memory, leaving the two 256mb
sticks in Slots 1 & 2. I was somewhat surprised to receive the same
error warning, so I removed the stick from Slot 2, tried again, and no
dice.

Before resorting to flashing a new BIOS, on a whim I removed the
remaining 256mb stick and replaced it with the 512mb stick-- Slot 1.
I say "a whim" as it was rated at CL3 and I recall the "Aggressive"
settings being in the neighborhood of CL2.5 in the BIOS memory setting
menu.

I was shocked to hear the computer POST and I was able to get into the
BIOS, change the setting, reinstall the 1gig of memory, and proceed as
if nothing happened. I considered myself lucky as I didn't really
feel like fooling with a reflash!

Anyway, a couple of questions that I have from that episode...

1) Does it make sense that with the "Aggressive" settings, my machine
would boot with memory rated "slower" (rated CL3 memory installed
allowed it to boot where rated CL2.5 failed)?

2) Does the possible answer lie in the logic that my CL3 rated Samsung
memory is faster than advertised and the CL2.5 rated Infineon memory
is perhaps "overrated"?

3) I believe, dual channel or otherwise, my system will default to the
slowest rated RAM (CL3). Is this correct?

4) Even with CL2.5 rated RAM, if the BIOS memory settings call for
CL3, I presume the memory will "throttle back" to the CL3 speed (it
appears the "Auto" setting calls for this CL3 speed)?

5) It has been suggested in this group that the benefits of
dual-channel capability are miniscule, at best, versus single channel.
What discernable performance difference will one find with memory
running at CL2.5 versus CL3? I should have asked this question prior
to fiddling with the setting as it might be one of those imperceptible
gains noticeable only on benchmark testing.

Thanks in advance for your answers, input, and counsel!

--Howie
Click to expand...

1) On some boards, the more aggressive memory settings (like "Turbo")
select a CAS2 setting. There are now a couple of memory supplies who
have silicon that is likely to do CAS2, no matter how it is binned.
The latest revision of Samsung TCCD and the Micron chips used in
Ballistix are examples, and between them, there are seven brands
of CAS2 memory using those chips. If the same chips are used to
make ordinary DIMMs, there is a good chance they too will run at
enhanced timings.
2) You are possibly looking at different generations of silicon dice,
so the answer is yes.
3) Yes, the slowest DIMM determines the timings used.
4) Yes. The CAS value is actually written into the DRAM chip at POST,
as part of the DRAM initialization. There is a register in the chip,
where the CAS value is written. If the BIOS writes too aggressive a
value into the DIMM, then obviously the machine will crash big time.
Selecting "turbo" would do that. When "By SPD" is used, the BIOS
reads the acceptable timing from the SPD chip on the DIMM, to figure
out what CAS value to write into the DDR memory chips. This special
write operation is required, as the DDR chips and the Northbridge
have to agree on the delay to expect, before data will be delivered.
5) The easiest way to test the difference, is to prepare a memtest86
boot floppy (memtest.org). It only takes a few seconds to boot
(less than booting Windows), and puts a bandwidth number in the
upper left corner of the screen. You can compare memory bandwidth
numbers that way. I think you will be quite surprised by the
answer. A second way to do it, is to get a Knoppix boot CD, which
has a copy of memtest86 built in, and memtest86 can be run from
the boot prompt. It too boots in a matter of seconds, and gives
the same info. Knoppix is a large download (700MB), so it is best
to borrow a copy from someone and burn your own. I like Knoppix,
because you can disconnect the hard drive, and experiment with
overclocking, without having to worry about corrupting your
hard drive. You can download and run Prime95 (Linux version,
from mersenne.org) while using Knoppix. Once you pass memtest and
Prime95, then it is safe to boot Windows. You can watch the Knoppix
boot screen, and if you see "segmentation fault", or the boot
process freezes, those are good indications that your core
frequency is too high.

HTH,
Paul
 
(e-mail address removed) (Paul) mused:

....Big snip...

:>
:>HTH,
:> Paul


It sure did...thanks for taking the time to enlighten.

--Howie
 
Hi;
After the bios flash. your front side bus reverted to a safe setting lower than 200: either 133
or 166 so your DDR memory (2 x FSB) became DDR266 or 333)==> acceptable for any of your sticks
at CL2.
It always revert to a SAFE SETTING and you have to adjust the FSB to the CPU's or end up w.
lower perf.
 
Except there was not a bios flash...


"Mikey" <savagem(-DELETE-)@sympatico.ca> mused:

:>Hi;
:>After the bios flash. your front side bus reverted to a safe setting lower than 200: either 133
:>or 166 so your DDR memory (2 x FSB) became DDR266 or 333)==> acceptable for any of your sticks
:>at CL2.
:>It always revert to a SAFE SETTING and you have to adjust the FSB to the CPU's or end up w.
:>lower perf.
:>
:>
:>> In article <[email protected]>, Howie
:>>
:>>> A7N8X-Deluxe board running 3 sticks of 256mb Infineon PC3200 DDR400
:>>> RAM CL2.5...
:>>>
:>>> I swapped one of the sticks at my local computer shop for one stick of
:>>> Samsung PC3200 CL3 512mb. The logic was twofold: 1) Wanted to
:>>> increase system RAM to 1gig and 2) Take advantage of the Dual Channel
:>>> capability of this board (256+256 in slots 1 & 2, 512 in slot 3).
:>>>
:>>> Got home, installed the 512mb stick, everything booted up fine, 1gig
:>>> of RAM in place. I then ran Sandra to analyze the memory and it
:>>> reported the new stick was CL2.5 up to 166MHz and CL3 up to 200MHz.
:>>> While I didn't feel this was anything to be concerned about in the
:>>> grand scheme of things, I did take note of the fact my 256mb sticks
:>>> were rated at CL2.5 at 200MHz and the new 512mb stick CL3.
:>>>
:>>> Then, quite ignorantly, I decided to play a bit with the memory
:>>> settings. I selected "Aggressive" from the BIOS, saved, and received
:>>> an audible report the "System Failed Memory Test". I quickly realized
:>>> I could not simply reboot and get into the BIOS to change this setting
:>>> back, so I removed the new stick of memory, leaving the two 256mb
:>>> sticks in Slots 1 & 2. I was somewhat surprised to receive the same
:>>> error warning, so I removed the stick from Slot 2, tried again, and no
:>>> dice.
:>>>
:>>> Before resorting to flashing a new BIOS, on a whim I removed the
:>>> remaining 256mb stick and replaced it with the 512mb stick-- Slot 1.
:>>> I say "a whim" as it was rated at CL3 and I recall the "Aggressive"
:>>> settings being in the neighborhood of CL2.5 in the BIOS memory setting
:>>> menu.
:>>>
:>>> I was shocked to hear the computer POST and I was able to get into the
:>>> BIOS, change the setting, reinstall the 1gig of memory, and proceed as
:>>> if nothing happened. I considered myself lucky as I didn't really
:>>> feel like fooling with a reflash!
:>>>
:>>> Anyway, a couple of questions that I have from that episode...
:>>>
:>>> 1) Does it make sense that with the "Aggressive" settings, my machine
:>>> would boot with memory rated "slower" (rated CL3 memory installed
:>>> allowed it to boot where rated CL2.5 failed)?
:>>>
:>>> 2) Does the possible answer lie in the logic that my CL3 rated Samsung
:>>> memory is faster than advertised and the CL2.5 rated Infineon memory
:>>> is perhaps "overrated"?
:>>>
:>>> 3) I believe, dual channel or otherwise, my system will default to the
:>>> slowest rated RAM (CL3). Is this correct?
:>>>
:>>> 4) Even with CL2.5 rated RAM, if the BIOS memory settings call for
:>>> CL3, I presume the memory will "throttle back" to the CL3 speed (it
:>>> appears the "Auto" setting calls for this CL3 speed)?
:>>>
:>>> 5) It has been suggested in this group that the benefits of
:>>> dual-channel capability are miniscule, at best, versus single channel.
:>>> What discernable performance difference will one find with memory
:>>> running at CL2.5 versus CL3? I should have asked this question prior
:>>> to fiddling with the setting as it might be one of those imperceptible
:>>> gains noticeable only on benchmark testing.
:>>>
:>>> Thanks in advance for your answers, input, and counsel!
:>>>
:>>> --Howie
:>>
:>> 1) On some boards, the more aggressive memory settings (like "Turbo")
:>> select a CAS2 setting. There are now a couple of memory supplies who
:>> have silicon that is likely to do CAS2, no matter how it is binned.
:>> The latest revision of Samsung TCCD and the Micron chips used in
:>> Ballistix are examples, and between them, there are seven brands
:>> of CAS2 memory using those chips. If the same chips are used to
:>> make ordinary DIMMs, there is a good chance they too will run at
:>> enhanced timings.
:>> 2) You are possibly looking at different generations of silicon dice,
:>> so the answer is yes.
:>> 3) Yes, the slowest DIMM determines the timings used.
:>> 4) Yes. The CAS value is actually written into the DRAM chip at POST,
:>> as part of the DRAM initialization. There is a register in the chip,
:>> where the CAS value is written. If the BIOS writes too aggressive a
:>> value into the DIMM, then obviously the machine will crash big time.
:>> Selecting "turbo" would do that. When "By SPD" is used, the BIOS
:>> reads the acceptable timing from the SPD chip on the DIMM, to figure
:>> out what CAS value to write into the DDR memory chips. This special
:>> write operation is required, as the DDR chips and the Northbridge
:>> have to agree on the delay to expect, before data will be delivered.
:>> 5) The easiest way to test the difference, is to prepare a memtest86
:>> boot floppy (memtest.org). It only takes a few seconds to boot
:>> (less than booting Windows), and puts a bandwidth number in the
:>> upper left corner of the screen. You can compare memory bandwidth
:>> numbers that way. I think you will be quite surprised by the
:>> answer. A second way to do it, is to get a Knoppix boot CD, which
:>> has a copy of memtest86 built in, and memtest86 can be run from
:>> the boot prompt. It too boots in a matter of seconds, and gives
:>> the same info. Knoppix is a large download (700MB), so it is best
:>> to borrow a copy from someone and burn your own. I like Knoppix,
:>> because you can disconnect the hard drive, and experiment with
:>> overclocking, without having to worry about corrupting your
:>> hard drive. You can download and run Prime95 (Linux version,
:>> from mersenne.org) while using Knoppix. Once you pass memtest and
:>> Prime95, then it is safe to boot Windows. You can watch the Knoppix
:>> boot screen, and if you see "segmentation fault", or the boot
:>> process freezes, those are good indications that your core
:>> frequency is too high.
:>>
:>> HTH,
:>> Paul
:>
 
Back
Top