I understand this is somewhat OT but we have some smart guys here so I
am throwing it out here.
I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
I have bought several sets of 2g sticks trying to get to 8g and I
can't get any to work right. The last batch are exactly the same
Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
swears were tested and OK for this machine (DDR2 667mz). I put them in
BIOS says they are there and OK.
Windows 7 hangs at the first "flag" screen for several minutes.
Eventually it will come up but it is too slow to do anything
productive. I tried every combination of any 2, in the proper slots.
Same deal. One chip seems OK (any one)
Is this a hardware problem or a windows problem?
I haven't tried sticking an XP drive in here but I do have one from
this machine.
I also have a quad core X3230 chip I haven't tried yet. "Break one
thing at a time" is what I always say but I am wondering if it would
affect my RAM problem.

gfretwell@aol.com wrote:
> I understand this is somewhat OT but we have some smart guys here so I
> am throwing it out here.
> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
> I have bought several sets of 2g sticks trying to get to 8g and I
> can't get any to work right. The last batch are exactly the same
> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
> swears were tested and OK for this machine (DDR2 667mz). I put them in
> BIOS says they are there and OK.
> Windows 7 hangs at the first "flag" screen for several minutes.
> Eventually it will come up but it is too slow to do anything
> productive. I tried every combination of any 2, in the proper slots.
> Same deal. One chip seems OK (any one)
> Is this a hardware problem or a windows problem?
> I haven't tried sticking an XP drive in here but I do have one from
> this machine.
> I also have a quad core X3230 chip I haven't tried yet. "Break one
> thing at a time" is what I always say but I am wondering if it would
> affect my RAM problem.

"- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
in Dual Channel Interleaved Mode, the largest memory capacity possible
is 8 GB.

(16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
* 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
* 1G/1024M * 1 M/(K*K) = 8 GB)

- Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"

The 965 address space is 64GB, and it looks like the memory is constrained
by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
per slot, sets the upper memory limit.

And this means that the entire 8GB can be used, without reservation.
There were some other chipsets, where you plugged in 8GB of sticks,
and maybe ~7GB was "Free". The decoding was eating some of it.

With your chipset, you could plug in 8GB of RAM and an 8GB video card,
and it's supposed to accept that. And even if the BIOS double mapped
the PCIe space for the video card (8GB cached, 8GB uncached) there
would still be enough space (because a total of 24GB of addresses would
be defined out of 64GB max).

Notice how the tech writer skillfully slips in that pish about
DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
a real limit, but the BIOS might well have all of that encoded in the
programming logic of the "Auto" settings for RAM. This was likely
a loading related reasoning, when the 1Gbit/chips didn't exist yet,
and they were predicting trouble. The thing is, as time progressed,
the trend was for bus loading to be a "chip count" issue, rather
than "what is inside the chip" issue. One reason the industry settled
on input capacitance standard values, was so DIMM stick designers
would stop putting caps across the inputs to "equalize" things.

DDR2-800 and DDR2-667, should both have bus loading issues, but
this is typically handled on Intel, via Command Rate 2 on everything.
This is "half-rate address presentation", which generally doesn't
do a great deal of harm to the performance.

The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
there are 16 chips on the module, arranged as two ranks of 8 chips.
8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
It all checks out. It is a "low density module". If the module had
used 256Mx4 chips (a server type of chip), then no, Intel hates that.
And Intel notes the hatred for x4 chips, in every MCH datasheet.
Only x8 and x16 chips are allowed, nobody uses x16, and Intel
does not accept x4.

There were some VIA chipsets that accepted a full loading of x4 chips.
As well as one special chipset, that could accept server DIMMs or
consumer DIMMs (established by one brave home experimenter just
shoving them in and trying them). But Intel on the other hand,
has their practices fine tuned, and there is only one
practical formulation, which is what you bought - a DIMM with
sixteen x8-wide chips on it.

I couldn't tell you why it's not working. If the BIOS memory
settings are on auto, unless the SPD is grossly mis-programmed,
the BIOS should compute the correct CAS and memory clock values,
which you can verify with CPUZ. If one stick will work, then
you can

1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
2) Boot OS.
3) Use CPUZ (cpuid.com) to verify what settings were used.

Sometimes the numbers shown in the BIOS, do not match
the values in CPUZ. I could, for example, enter CAS2
on one board, and CPUZ would show CAS3. There are occasionally
arithmetic mistakes in the BIOS, but it doesn't happen all that
often. This means that an "Auto" setting could actually
do something naughty.

Check and make sure memory hoisting is enabled. This
allows 8GB of RAM, to consist of 3GB of RAM below the
4GB mark, and 5GB of remaining RAM above the 4GB mark.
The 5GB chunk is "hoisted". Not all BIOS use this kind
of terminology for it.

Windows 7 has a memory license, the following values are
for the x64 version. The 32 bit OS only supports 4GB.
I had to buy Pro for the Test Machine, to get past 16GB.
Windows will ignore the excess RAM.

Windows 7 Professional 192 GB
Windows 7 Home Premium 16 GB
Windows 7 Home Basic 8 GB
Windows 7 Starter 2 GB

Paul

On Mon, 12 Jul 2021 22:46:30 -0400, Paul <nospam@needed.invalid>
wrote:

>gfretwell@aol.com wrote:
>> I understand this is somewhat OT but we have some smart guys here so I
>> am throwing it out here.
>> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
>> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
>> I have bought several sets of 2g sticks trying to get to 8g and I
>> can't get any to work right. The last batch are exactly the same
>> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
>> swears were tested and OK for this machine (DDR2 667mz). I put them in
>> BIOS says they are there and OK.
>> Windows 7 hangs at the first "flag" screen for several minutes.
>> Eventually it will come up but it is too slow to do anything
>> productive. I tried every combination of any 2, in the proper slots.
>> Same deal. One chip seems OK (any one)
>> Is this a hardware problem or a windows problem?
>> I haven't tried sticking an XP drive in here but I do have one from
>> this machine.
>> I also have a quad core X3230 chip I haven't tried yet. "Break one
>> thing at a time" is what I always say but I am wondering if it would
>> affect my RAM problem.
>
>"- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
> in Dual Channel Interleaved Mode, the largest memory capacity possible
> is 8 GB.
>
> (16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
> * 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
> * 1G/1024M * 1 M/(K*K) = 8 GB)
>
> - Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"
>
>The 965 address space is 64GB, and it looks like the memory is constrained
>by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
>per slot, sets the upper memory limit.
>
>And this means that the entire 8GB can be used, without reservation.
>There were some other chipsets, where you plugged in 8GB of sticks,
>and maybe ~7GB was "Free". The decoding was eating some of it.
>
>With your chipset, you could plug in 8GB of RAM and an 8GB video card,
>and it's supposed to accept that. And even if the BIOS double mapped
>the PCIe space for the video card (8GB cached, 8GB uncached) there
>would still be enough space (because a total of 24GB of addresses would
>be defined out of 64GB max).
>
>Notice how the tech writer skillfully slips in that pish about
>DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
>a real limit, but the BIOS might well have all of that encoded in the
>programming logic of the "Auto" settings for RAM. This was likely
>a loading related reasoning, when the 1Gbit/chips didn't exist yet,
>and they were predicting trouble. The thing is, as time progressed,
>the trend was for bus loading to be a "chip count" issue, rather
>than "what is inside the chip" issue. One reason the industry settled
>on input capacitance standard values, was so DIMM stick designers
>would stop putting caps across the inputs to "equalize" things.
>
>DDR2-800 and DDR2-667, should both have bus loading issues, but
>this is typically handled on Intel, via Command Rate 2 on everything.
>This is "half-rate address presentation", which generally doesn't
>do a great deal of harm to the performance.
>
>The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
>there are 16 chips on the module, arranged as two ranks of 8 chips.
>8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
>It all checks out. It is a "low density module". If the module had
>used 256Mx4 chips (a server type of chip), then no, Intel hates that.
>And Intel notes the hatred for x4 chips, in every MCH datasheet.
>Only x8 and x16 chips are allowed, nobody uses x16, and Intel
>does not accept x4.
>
>There were some VIA chipsets that accepted a full loading of x4 chips.
>As well as one special chipset, that could accept server DIMMs or
>consumer DIMMs (established by one brave home experimenter just
>shoving them in and trying them). But Intel on the other hand,
>has their practices fine tuned, and there is only one
>practical formulation, which is what you bought - a DIMM with
>sixteen x8-wide chips on it.
>
>I couldn't tell you why it's not working. If the BIOS memory
>settings are on auto, unless the SPD is grossly mis-programmed,
>the BIOS should compute the correct CAS and memory clock values,
>which you can verify with CPUZ. If one stick will work, then
>you can
>
>1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
>2) Boot OS.
>3) Use CPUZ (cpuid.com) to verify what settings were used.
>
>Sometimes the numbers shown in the BIOS, do not match
>the values in CPUZ. I could, for example, enter CAS2
>on one board, and CPUZ would show CAS3. There are occasionally
>arithmetic mistakes in the BIOS, but it doesn't happen all that
>often. This means that an "Auto" setting could actually
>do something naughty.
>
>Check and make sure memory hoisting is enabled. This
>allows 8GB of RAM, to consist of 3GB of RAM below the
>4GB mark, and 5GB of remaining RAM above the 4GB mark.
>The 5GB chunk is "hoisted". Not all BIOS use this kind
>of terminology for it.
>
>Windows 7 has a memory license, the following values are
>for the x64 version. The 32 bit OS only supports 4GB.
>I had to buy Pro for the Test Machine, to get past 16GB.
>Windows will ignore the excess RAM.
>
>Windows 7 Professional 192 GB
>Windows 7 Home Premium 16 GB
>Windows 7 Home Basic 8 GB
>Windows 7 Starter 2 GB
>
> Paul

This is 64 bit w/7 pro so that should be OK unless something didn't
load right. This is an approved configuration according to the Intel
web site. I reset the BIOS back to defaults. Somebody lied ore I have
some kind of board problem. I beat up the sites Google pointed me to
with no joy. I think I will try the other CPU next and hope the bug is
in the CPU.

gfretwell@aol.com wrote:
> On Mon, 12 Jul 2021 22:46:30 -0400, Paul <nospam@needed.invalid>
> wrote:
>
>> gfretwell@aol.com wrote:
>>> I understand this is somewhat OT but we have some smart guys here so I
>>> am throwing it out here.
>>> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
>>> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
>>> I have bought several sets of 2g sticks trying to get to 8g and I
>>> can't get any to work right. The last batch are exactly the same
>>> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
>>> swears were tested and OK for this machine (DDR2 667mz). I put them in
>>> BIOS says they are there and OK.
>>> Windows 7 hangs at the first "flag" screen for several minutes.
>>> Eventually it will come up but it is too slow to do anything
>>> productive. I tried every combination of any 2, in the proper slots.
>>> Same deal. One chip seems OK (any one)
>>> Is this a hardware problem or a windows problem?
>>> I haven't tried sticking an XP drive in here but I do have one from
>>> this machine.
>>> I also have a quad core X3230 chip I haven't tried yet. "Break one
>>> thing at a time" is what I always say but I am wondering if it would
>>> affect my RAM problem.
>> "- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
>> in Dual Channel Interleaved Mode, the largest memory capacity possible
>> is 8 GB.
>>
>> (16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
>> * 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
>> * 1G/1024M * 1 M/(K*K) = 8 GB)
>>
>> - Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"
>>
>> The 965 address space is 64GB, and it looks like the memory is constrained
>> by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
>> per slot, sets the upper memory limit.
>>
>> And this means that the entire 8GB can be used, without reservation.
>> There were some other chipsets, where you plugged in 8GB of sticks,
>> and maybe ~7GB was "Free". The decoding was eating some of it.
>>
>> With your chipset, you could plug in 8GB of RAM and an 8GB video card,
>> and it's supposed to accept that. And even if the BIOS double mapped
>> the PCIe space for the video card (8GB cached, 8GB uncached) there
>> would still be enough space (because a total of 24GB of addresses would
>> be defined out of 64GB max).
>>
>> Notice how the tech writer skillfully slips in that pish about
>> DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
>> a real limit, but the BIOS might well have all of that encoded in the
>> programming logic of the "Auto" settings for RAM. This was likely
>> a loading related reasoning, when the 1Gbit/chips didn't exist yet,
>> and they were predicting trouble. The thing is, as time progressed,
>> the trend was for bus loading to be a "chip count" issue, rather
>> than "what is inside the chip" issue. One reason the industry settled
>> on input capacitance standard values, was so DIMM stick designers
>> would stop putting caps across the inputs to "equalize" things.
>>
>> DDR2-800 and DDR2-667, should both have bus loading issues, but
>> this is typically handled on Intel, via Command Rate 2 on everything.
>> This is "half-rate address presentation", which generally doesn't
>> do a great deal of harm to the performance.
>>
>> The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
>> there are 16 chips on the module, arranged as two ranks of 8 chips.
>> 8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
>> It all checks out. It is a "low density module". If the module had
>> used 256Mx4 chips (a server type of chip), then no, Intel hates that.
>> And Intel notes the hatred for x4 chips, in every MCH datasheet.
>> Only x8 and x16 chips are allowed, nobody uses x16, and Intel
>> does not accept x4.
>>
>> There were some VIA chipsets that accepted a full loading of x4 chips.
>> As well as one special chipset, that could accept server DIMMs or
>> consumer DIMMs (established by one brave home experimenter just
>> shoving them in and trying them). But Intel on the other hand,
>> has their practices fine tuned, and there is only one
>> practical formulation, which is what you bought - a DIMM with
>> sixteen x8-wide chips on it.
>>
>> I couldn't tell you why it's not working. If the BIOS memory
>> settings are on auto, unless the SPD is grossly mis-programmed,
>> the BIOS should compute the correct CAS and memory clock values,
>> which you can verify with CPUZ. If one stick will work, then
>> you can
>>
>> 1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
>> 2) Boot OS.
>> 3) Use CPUZ (cpuid.com) to verify what settings were used.
>>
>> Sometimes the numbers shown in the BIOS, do not match
>> the values in CPUZ. I could, for example, enter CAS2
>> on one board, and CPUZ would show CAS3. There are occasionally
>> arithmetic mistakes in the BIOS, but it doesn't happen all that
>> often. This means that an "Auto" setting could actually
>> do something naughty.
>>
>> Check and make sure memory hoisting is enabled. This
>> allows 8GB of RAM, to consist of 3GB of RAM below the
>> 4GB mark, and 5GB of remaining RAM above the 4GB mark.
>> The 5GB chunk is "hoisted". Not all BIOS use this kind
>> of terminology for it.
>>
>> Windows 7 has a memory license, the following values are
>> for the x64 version. The 32 bit OS only supports 4GB.
>> I had to buy Pro for the Test Machine, to get past 16GB.
>> Windows will ignore the excess RAM.
>>
>> Windows 7 Professional 192 GB
>> Windows 7 Home Premium 16 GB
>> Windows 7 Home Basic 8 GB
>> Windows 7 Starter 2 GB
>>
>> Paul
>
> This is 64 bit w/7 pro so that should be OK unless something didn't
> load right. This is an approved configuration according to the Intel
> web site. I reset the BIOS back to defaults. Somebody lied ore I have
> some kind of board problem. I beat up the sites Google pointed me to
> with no joy. I think I will try the other CPU next and hope the bug is
> in the CPU.

I'd stick with the 6600 until this is resolved. You need to prove
your platform out first, before sticking one of those "modified"
Xeons in the system :-) They're very tempting.

Have you had a chance to verify how the working single-stick
DIMM case reports to CPUZ ?

https://www.cpuid.com/softwares/cpu-z.html (portable ZIP)

Even running Memtest gives some bandwidth numbers on the left,
that give some idea whether the thing is within bounds or not.
With two sticks (dual channel), you should see a bit less than 4GB/sec.
With only a single channel working, it's going to be less.

http://memtest.org/ (scroll to 50% of web page)

If Memtest reports timings on the right, they're typically
incorrect. But the bandwidth on the left, is more believable,
and I looked at the code once upon a time, and it looked
reasonable (there is a cache flush first, before it measures
the RAM-based speed).

With Windows booted and CPUZ available, the timings listed in
there are accurate. I've not run into a problem there.
The proof, is whether any "unusual" numbers were listed and
the computer remained stable, that might suggest a problem.
For example, some Intel chipsets can't go below CAS4, due
to internal timing limitations (probably a bus turnaround issue).
Whereas the DDR2 on my VIA chipset ran all the time at CAS3
(DDR2-533 though, so who is counting). What I liked about
running the DDR2-533 speed, is I could do sixteen hours
of Prime95, no trouble at all. Never an error.

https://www.mersenne.org/download/

Paul

On Tue, 13 Jul 2021 01:00:43 -0400, Paul <nospam@needed.invalid>
wrote:

>gfretwell@aol.com wrote:
>> On Mon, 12 Jul 2021 22:46:30 -0400, Paul <nospam@needed.invalid>
>> wrote:
>>
>>> gfretwell@aol.com wrote:
>>>> I understand this is somewhat OT but we have some smart guys here so I
>>>> am throwing it out here.
>>>> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
>>>> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
>>>> I have bought several sets of 2g sticks trying to get to 8g and I
>>>> can't get any to work right. The last batch are exactly the same
>>>> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
>>>> swears were tested and OK for this machine (DDR2 667mz). I put them in
>>>> BIOS says they are there and OK.
>>>> Windows 7 hangs at the first "flag" screen for several minutes.
>>>> Eventually it will come up but it is too slow to do anything
>>>> productive. I tried every combination of any 2, in the proper slots.
>>>> Same deal. One chip seems OK (any one)
>>>> Is this a hardware problem or a windows problem?
>>>> I haven't tried sticking an XP drive in here but I do have one from
>>>> this machine.
>>>> I also have a quad core X3230 chip I haven't tried yet. "Break one
>>>> thing at a time" is what I always say but I am wondering if it would
>>>> affect my RAM problem.
>>> "- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
>>> in Dual Channel Interleaved Mode, the largest memory capacity possible
>>> is 8 GB.
>>>
>>> (16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
>>> * 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
>>> * 1G/1024M * 1 M/(K*K) = 8 GB)
>>>
>>> - Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"
>>>
>>> The 965 address space is 64GB, and it looks like the memory is constrained
>>> by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
>>> per slot, sets the upper memory limit.
>>>
>>> And this means that the entire 8GB can be used, without reservation.
>>> There were some other chipsets, where you plugged in 8GB of sticks,
>>> and maybe ~7GB was "Free". The decoding was eating some of it.
>>>
>>> With your chipset, you could plug in 8GB of RAM and an 8GB video card,
>>> and it's supposed to accept that. And even if the BIOS double mapped
>>> the PCIe space for the video card (8GB cached, 8GB uncached) there
>>> would still be enough space (because a total of 24GB of addresses would
>>> be defined out of 64GB max).
>>>
>>> Notice how the tech writer skillfully slips in that pish about
>>> DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
>>> a real limit, but the BIOS might well have all of that encoded in the
>>> programming logic of the "Auto" settings for RAM. This was likely
>>> a loading related reasoning, when the 1Gbit/chips didn't exist yet,
>>> and they were predicting trouble. The thing is, as time progressed,
>>> the trend was for bus loading to be a "chip count" issue, rather
>>> than "what is inside the chip" issue. One reason the industry settled
>>> on input capacitance standard values, was so DIMM stick designers
>>> would stop putting caps across the inputs to "equalize" things.
>>>
>>> DDR2-800 and DDR2-667, should both have bus loading issues, but
>>> this is typically handled on Intel, via Command Rate 2 on everything.
>>> This is "half-rate address presentation", which generally doesn't
>>> do a great deal of harm to the performance.
>>>
>>> The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
>>> there are 16 chips on the module, arranged as two ranks of 8 chips.
>>> 8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
>>> It all checks out. It is a "low density module". If the module had
>>> used 256Mx4 chips (a server type of chip), then no, Intel hates that.
>>> And Intel notes the hatred for x4 chips, in every MCH datasheet.
>>> Only x8 and x16 chips are allowed, nobody uses x16, and Intel
>>> does not accept x4.
>>>
>>> There were some VIA chipsets that accepted a full loading of x4 chips.
>>> As well as one special chipset, that could accept server DIMMs or
>>> consumer DIMMs (established by one brave home experimenter just
>>> shoving them in and trying them). But Intel on the other hand,
>>> has their practices fine tuned, and there is only one
>>> practical formulation, which is what you bought - a DIMM with
>>> sixteen x8-wide chips on it.
>>>
>>> I couldn't tell you why it's not working. If the BIOS memory
>>> settings are on auto, unless the SPD is grossly mis-programmed,
>>> the BIOS should compute the correct CAS and memory clock values,
>>> which you can verify with CPUZ. If one stick will work, then
>>> you can
>>>
>>> 1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
>>> 2) Boot OS.
>>> 3) Use CPUZ (cpuid.com) to verify what settings were used.
>>>
>>> Sometimes the numbers shown in the BIOS, do not match
>>> the values in CPUZ. I could, for example, enter CAS2
>>> on one board, and CPUZ would show CAS3. There are occasionally
>>> arithmetic mistakes in the BIOS, but it doesn't happen all that
>>> often. This means that an "Auto" setting could actually
>>> do something naughty.
>>>
>>> Check and make sure memory hoisting is enabled. This
>>> allows 8GB of RAM, to consist of 3GB of RAM below the
>>> 4GB mark, and 5GB of remaining RAM above the 4GB mark.
>>> The 5GB chunk is "hoisted". Not all BIOS use this kind
>>> of terminology for it.
>>>
>>> Windows 7 has a memory license, the following values are
>>> for the x64 version. The 32 bit OS only supports 4GB.
>>> I had to buy Pro for the Test Machine, to get past 16GB.
>>> Windows will ignore the excess RAM.
>>>
>>> Windows 7 Professional 192 GB
>>> Windows 7 Home Premium 16 GB
>>> Windows 7 Home Basic 8 GB
>>> Windows 7 Starter 2 GB
>>>
>>> Paul
>>
>> This is 64 bit w/7 pro so that should be OK unless something didn't
>> load right. This is an approved configuration according to the Intel
>> web site. I reset the BIOS back to defaults. Somebody lied ore I have
>> some kind of board problem. I beat up the sites Google pointed me to
>> with no joy. I think I will try the other CPU next and hope the bug is
>> in the CPU.
>
>I'd stick with the 6600 until this is resolved. You need to prove
>your platform out first, before sticking one of those "modified"
>Xeons in the system :-) They're very tempting.
>
>Have you had a chance to verify how the working single-stick
>DIMM case reports to CPUZ ?
>
>https://www.cpuid.com/softwares/cpu-z.html (portable ZIP)
>
>Even running Memtest gives some bandwidth numbers on the left,
>that give some idea whether the thing is within bounds or not.
>With two sticks (dual channel), you should see a bit less than 4GB/sec.
>With only a single channel working, it's going to be less.
>
>http://memtest.org/ (scroll to 50% of web page)
>
>If Memtest reports timings on the right, they're typically
>incorrect. But the bandwidth on the left, is more believable,
>and I looked at the code once upon a time, and it looked
>reasonable (there is a cache flush first, before it measures
>the RAM-based speed).
>
>With Windows booted and CPUZ available, the timings listed in
>there are accurate. I've not run into a problem there.
>The proof, is whether any "unusual" numbers were listed and
>the computer remained stable, that might suggest a problem.
>For example, some Intel chipsets can't go below CAS4, due
>to internal timing limitations (probably a bus turnaround issue).
>Whereas the DDR2 on my VIA chipset ran all the time at CAS3
>(DDR2-533 though, so who is counting). What I liked about
>running the DDR2-533 speed, is I could do sixteen hours
>of Prime95, no trouble at all. Never an error.
>
>https://www.mersenne.org/download/
>
> Paul

On 7/12/2021 9:25 PM, gfretwell@aol.com wrote:
> On Mon, 12 Jul 2021 22:46:30 -0400, Paul <nospam@needed.invalid>
> wrote:
>
>> gfretwell@aol.com wrote:
>>> I understand this is somewhat OT but we have some smart guys here so I
>>> am throwing it out here.
>>> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
>>> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
>>> I have bought several sets of 2g sticks trying to get to 8g and I
>>> can't get any to work right. The last batch are exactly the same
>>> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
>>> swears were tested and OK for this machine (DDR2 667mz). I put them in
>>> BIOS says they are there and OK.
>>> Windows 7 hangs at the first "flag" screen for several minutes.
>>> Eventually it will come up but it is too slow to do anything
>>> productive. I tried every combination of any 2, in the proper slots.
>>> Same deal. One chip seems OK (any one)
>>> Is this a hardware problem or a windows problem?
>>> I haven't tried sticking an XP drive in here but I do have one from
>>> this machine.
>>> I also have a quad core X3230 chip I haven't tried yet. "Break one
>>> thing at a time" is what I always say but I am wondering if it would
>>> affect my RAM problem.
>>
>> "- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
>> in Dual Channel Interleaved Mode, the largest memory capacity possible
>> is 8 GB.
>>
>> (16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
>> * 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
>> * 1G/1024M * 1 M/(K*K) = 8 GB)
>>
>> - Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"
>>
>> The 965 address space is 64GB, and it looks like the memory is constrained
>> by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
>> per slot, sets the upper memory limit.
>>
>> And this means that the entire 8GB can be used, without reservation.
>> There were some other chipsets, where you plugged in 8GB of sticks,
>> and maybe ~7GB was "Free". The decoding was eating some of it.
>>
>> With your chipset, you could plug in 8GB of RAM and an 8GB video card,
>> and it's supposed to accept that. And even if the BIOS double mapped
>> the PCIe space for the video card (8GB cached, 8GB uncached) there
>> would still be enough space (because a total of 24GB of addresses would
>> be defined out of 64GB max).
>>
>> Notice how the tech writer skillfully slips in that pish about
>> DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
>> a real limit, but the BIOS might well have all of that encoded in the
>> programming logic of the "Auto" settings for RAM. This was likely
>> a loading related reasoning, when the 1Gbit/chips didn't exist yet,
>> and they were predicting trouble. The thing is, as time progressed,
>> the trend was for bus loading to be a "chip count" issue, rather
>> than "what is inside the chip" issue. One reason the industry settled
>> on input capacitance standard values, was so DIMM stick designers
>> would stop putting caps across the inputs to "equalize" things.
>>
>> DDR2-800 and DDR2-667, should both have bus loading issues, but
>> this is typically handled on Intel, via Command Rate 2 on everything.
>> This is "half-rate address presentation", which generally doesn't
>> do a great deal of harm to the performance.
>>
>> The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
>> there are 16 chips on the module, arranged as two ranks of 8 chips.
>> 8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
>> It all checks out. It is a "low density module". If the module had
>> used 256Mx4 chips (a server type of chip), then no, Intel hates that.
>> And Intel notes the hatred for x4 chips, in every MCH datasheet.
>> Only x8 and x16 chips are allowed, nobody uses x16, and Intel
>> does not accept x4.
>>
>> There were some VIA chipsets that accepted a full loading of x4 chips.
>> As well as one special chipset, that could accept server DIMMs or
>> consumer DIMMs (established by one brave home experimenter just
>> shoving them in and trying them). But Intel on the other hand,
>> has their practices fine tuned, and there is only one
>> practical formulation, which is what you bought - a DIMM with
>> sixteen x8-wide chips on it.
>>
>> I couldn't tell you why it's not working. If the BIOS memory
>> settings are on auto, unless the SPD is grossly mis-programmed,
>> the BIOS should compute the correct CAS and memory clock values,
>> which you can verify with CPUZ. If one stick will work, then
>> you can
>>
>> 1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
>> 2) Boot OS.
>> 3) Use CPUZ (cpuid.com) to verify what settings were used.
>>
>> Sometimes the numbers shown in the BIOS, do not match
>> the values in CPUZ. I could, for example, enter CAS2
>> on one board, and CPUZ would show CAS3. There are occasionally
>> arithmetic mistakes in the BIOS, but it doesn't happen all that
>> often. This means that an "Auto" setting could actually
>> do something naughty.
>>
>> Check and make sure memory hoisting is enabled. This
>> allows 8GB of RAM, to consist of 3GB of RAM below the
>> 4GB mark, and 5GB of remaining RAM above the 4GB mark.
>> The 5GB chunk is "hoisted". Not all BIOS use this kind
>> of terminology for it.
>>
>> Windows 7 has a memory license, the following values are
>> for the x64 version. The 32 bit OS only supports 4GB.
>> I had to buy Pro for the Test Machine, to get past 16GB.
>> Windows will ignore the excess RAM.
>>
>> Windows 7 Professional 192 GB
>> Windows 7 Home Premium 16 GB
>> Windows 7 Home Basic 8 GB
>> Windows 7 Starter 2 GB
>>
>> Paul
>
> This is 64 bit w/7 pro so that should be OK unless something didn't
> load right. This is an approved configuration according to the Intel
> web site. I reset the BIOS back to defaults. Somebody lied ore I have
> some kind of board problem. I beat up the sites Google pointed me to
> with no joy. I think I will try the other CPU next and hope the bug is
> in the CPU.
>

Have you tried running Memtest86+ or some such self booting program with
various combination of the memory?

On Thu, 15 Jul 2021 15:32:36 -0700, Bob F <bobnospam@gmail.com> wrote:

>On 7/12/2021 9:25 PM, gfretwell@aol.com wrote:
>> On Mon, 12 Jul 2021 22:46:30 -0400, Paul <nospam@needed.invalid>
>> wrote:
>>
>>> gfretwell@aol.com wrote:
>>>> I understand this is somewhat OT but we have some smart guys here so I
>>>> am throwing it out here.
>>>> I have an Intel DQ965GF board with an Intel Core 2 6600 2.4GZ chip.
>>>> I have 3g on it 2 500m sticks and 2 1g sticks (800mz) That works OK
>>>> I have bought several sets of 2g sticks trying to get to 8g and I
>>>> can't get any to work right. The last batch are exactly the same
>>>> Crucial sticks (Crucial CT25664AA667.16FE/Micron MT47H128M8HQ) Intel
>>>> swears were tested and OK for this machine (DDR2 667mz). I put them in
>>>> BIOS says they are there and OK.
>>>> Windows 7 hangs at the first "flag" screen for several minutes.
>>>> Eventually it will come up but it is too slow to do anything
>>>> productive. I tried every combination of any 2, in the proper slots.
>>>> Same deal. One chip seems OK (any one)
>>>> Is this a hardware problem or a windows problem?
>>>> I haven't tried sticking an XP drive in here but I do have one from
>>>> this machine.
>>>> I also have a quad core X3230 chip I haven't tried yet. "Break one
>>>> thing at a time" is what I always say but I am wondering if it would
>>>> affect my RAM problem.
>>>
>>> "- By using 1 Gb technology (DDR2 800MHz 1 Gb technology is not supported)
>>> in Dual Channel Interleaved Mode, the largest memory capacity possible
>>> is 8 GB.
>>>
>>> (16K rows * 1K columns * 1 cell/(row * column) * 8 b/cell * 8 banks/device
>>> * 8 devices/DIMM-side * 4 DIMM-sides/channel * 2 channels * 1 B/8 b
>>> * 1G/1024M * 1 M/(K*K) = 8 GB)
>>>
>>> - Maximum DRAM address decode space is 8 GB (assuming 36-bit addressing)"
>>>
>>> The 965 address space is 64GB, and it looks like the memory is constrained
>>> by the DDR2 address bus limitations (1Gbit/chip max). The four slots times 2GB
>>> per slot, sets the upper memory limit.
>>>
>>> And this means that the entire 8GB can be used, without reservation.
>>> There were some other chipsets, where you plugged in 8GB of sticks,
>>> and maybe ~7GB was "Free". The decoding was eating some of it.
>>>
>>> With your chipset, you could plug in 8GB of RAM and an 8GB video card,
>>> and it's supposed to accept that. And even if the BIOS double mapped
>>> the PCIe space for the video card (8GB cached, 8GB uncached) there
>>> would still be enough space (because a total of 24GB of addresses would
>>> be defined out of 64GB max).
>>>
>>> Notice how the tech writer skillfully slips in that pish about
>>> DDR2-800 not working at the 1Gbit/chip level. This is unlikely to be
>>> a real limit, but the BIOS might well have all of that encoded in the
>>> programming logic of the "Auto" settings for RAM. This was likely
>>> a loading related reasoning, when the 1Gbit/chips didn't exist yet,
>>> and they were predicting trouble. The thing is, as time progressed,
>>> the trend was for bus loading to be a "chip count" issue, rather
>>> than "what is inside the chip" issue. One reason the industry settled
>>> on input capacitance standard values, was so DIMM stick designers
>>> would stop putting caps across the inputs to "equalize" things.
>>>
>>> DDR2-800 and DDR2-667, should both have bus loading issues, but
>>> this is typically handled on Intel, via Command Rate 2 on everything.
>>> This is "half-rate address presentation", which generally doesn't
>>> do a great deal of harm to the performance.
>>>
>>> The chip spec for your Micron part number is a 128Mx8. It's a by-8 chip,
>>> there are 16 chips on the module, arranged as two ranks of 8 chips.
>>> 8 chips times 8 bits wide is 64 bits, the standard width of a DIMM.
>>> It all checks out. It is a "low density module". If the module had
>>> used 256Mx4 chips (a server type of chip), then no, Intel hates that.
>>> And Intel notes the hatred for x4 chips, in every MCH datasheet.
>>> Only x8 and x16 chips are allowed, nobody uses x16, and Intel
>>> does not accept x4.
>>>
>>> There were some VIA chipsets that accepted a full loading of x4 chips.
>>> As well as one special chipset, that could accept server DIMMs or
>>> consumer DIMMs (established by one brave home experimenter just
>>> shoving them in and trying them). But Intel on the other hand,
>>> has their practices fine tuned, and there is only one
>>> practical formulation, which is what you bought - a DIMM with
>>> sixteen x8-wide chips on it.
>>>
>>> I couldn't tell you why it's not working. If the BIOS memory
>>> settings are on auto, unless the SPD is grossly mis-programmed,
>>> the BIOS should compute the correct CAS and memory clock values,
>>> which you can verify with CPUZ. If one stick will work, then
>>> you can
>>>
>>> 1) memtest the one 2GB stick first. We don't want to corrupt C: !!!
>>> 2) Boot OS.
>>> 3) Use CPUZ (cpuid.com) to verify what settings were used.
>>>
>>> Sometimes the numbers shown in the BIOS, do not match
>>> the values in CPUZ. I could, for example, enter CAS2
>>> on one board, and CPUZ would show CAS3. There are occasionally
>>> arithmetic mistakes in the BIOS, but it doesn't happen all that
>>> often. This means that an "Auto" setting could actually
>>> do something naughty.
>>>
>>> Check and make sure memory hoisting is enabled. This
>>> allows 8GB of RAM, to consist of 3GB of RAM below the
>>> 4GB mark, and 5GB of remaining RAM above the 4GB mark.
>>> The 5GB chunk is "hoisted". Not all BIOS use this kind
>>> of terminology for it.
>>>
>>> Windows 7 has a memory license, the following values are
>>> for the x64 version. The 32 bit OS only supports 4GB.
>>> I had to buy Pro for the Test Machine, to get past 16GB.
>>> Windows will ignore the excess RAM.
>>>
>>> Windows 7 Professional 192 GB
>>> Windows 7 Home Premium 16 GB
>>> Windows 7 Home Basic 8 GB
>>> Windows 7 Starter 2 GB
>>>
>>> Paul
>>
>> This is 64 bit w/7 pro so that should be OK unless something didn't
>> load right. This is an approved configuration according to the Intel
>> web site. I reset the BIOS back to defaults. Somebody lied ore I have
>> some kind of board problem. I beat up the sites Google pointed me to
>> with no joy. I think I will try the other CPU next and hope the bug is
>> in the CPU.
>>
>
>Have you tried running Memtest86+ or some such self booting program with
>various combination of the memory?
They seem to run OK, just slow