LIttle memory comparison

1.) Newegg has:

http://www.newegg.com/Product/Produc...82E16820148488
Crucial Ballistix 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory
Model BLT2KIT2G3D1608DT1TX0
5 out of 5 eggs (16) | Write a Review

DDR3 1600 (PC3 12800)
Timing 8-8-8-24
Cas Latency 8
Voltage 1.5V
$16.99

2.) Installed in my desktop is:

http://www.gskill.com/products.php?index=222
System Desktop
System Type DDR3

Intel P55

CAS Latency 9-9-9-24-2N
Capacity 2GB (2GBx1)
Speed DDR3-1600 (PC3 12800)
Test Voltage 1.5 Volts
Height 40 mm / 1.58 inch
Registered/Unbuffered Unbuffered
Error Checking Non-ECC
Type 240-pin DIMM
Warranty Lifetime
Features Intel XMP (Extreme Memory Profile) Ready

I could use to add 4gb of mem. to my Asus P7H55D-M EVO mobo.

If I add the Crucial 2x2gb (slots are open), will the Crucial be reasonably
expected to deliver less thru-put than the old Ripjaws?

Which should be faster according to the specs??

Thx,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:
1.) Newegg has:

http://www.newegg.com/Product/Produc...82E16820148488
Crucial Ballistix 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory
Model BLT2KIT2G3D1608DT1TX0
5 out of 5 eggs (16) | Write a Review

DDR3 1600 (PC3 12800)
Timing 8-8-8-24
Cas Latency 8
Voltage 1.5V
$16.99

2.) Installed in my desktop is:

http://www.gskill.com/products.php?index=222
System Desktop
System Type DDR3

Intel P55

CAS Latency 9-9-9-24-2N
Capacity 2GB (2GBx1)
Speed DDR3-1600 (PC3 12800)
Test Voltage 1.5 Volts
Height 40 mm / 1.58 inch
Registered/Unbuffered Unbuffered
Error Checking Non-ECC
Type 240-pin DIMM
Warranty Lifetime
Features Intel XMP (Extreme Memory Profile) Ready

I could use to add 4gb of mem. to my Asus P7H55D-M EVO mobo.

If I add the Crucial 2x2gb (slots are open), will the Crucial be reasonably
expected to deliver less thru-put than the old Ripjaws?

Which should be faster according to the specs??

Thx,
P

You currently have a single stick of RAM, 2GB total ?

A single stick of RAM, runs in single channel mode.

Normally, you would purchase a matched pair, and run
in dual channel mode. If you really only have a single
stick installed, it runs at half the memory bandwidth
of a two stick configuration.

The original stick has XMP, with an additional profile.
That allows the stick to communicate with an XMP BIOS,
and the XMP BIOS then sets DDR3-1600 automatically.
Without XMP enabled, it might run DDR3-1333, and then
the user has to enter the BIOS and fix it up themselves.

I'm not an expert on XMP. I was reading the other
day, that XMP not only records a speed setting, it
also records a recommended stick configuration as well.
If the XMP is intended for two sticks, that's all
it works for. If you install four sticks of RAM,
and the XMP profile says two, then XMP should be
disabled. That's because, with slightly more bus
loading, some different numbers might be required.
Like any SPD chip information table, there are
limited slots available in the table, to specify
configurations with associated timing numbers.

So, buying options a

1) Buy a second 2GB stick of XMP RAM, Ripjaws series.
System runs DDR3-1600 CAS9, at 1.5V. Since both
sticks match, XMP profile is for two sticks, BIOS
sets it all up automatically for you. The same XMP
profile works with one or two sticks, but not with
three or four sticks.

2) Buy a second 2GB stick of non-XMP RAM. Any brand.
Ideally, at least DDR3-1600 CAS9, so the new stick
doesn't slow down the old stick (i.e. don't buy a
stick of DDR3-1066, then come back and complain the
setup seems slow). The purpose of doing this config,
is so we can run dual channel, for twice the memory
bandwidth. System will run DDR3-1333 in Auto mode.
The user hand-cranks the settings to DDR3-1600 CAS9

Now, we consider your crazy plan. Your plan is to set up
three sticks of RAM. What three stick configuration would
work ?

3a) You decide to combine a single 2GB XMP stick, with
two 2GB non-XMP sticks (total 6GB). For a number of
reasons, XMP will be disabled. You adjust the sticks
manually. It is not possible to populate equal quantities
of RAM on each channel. Of the total 6GB space, 4GB runs dual
channel, at full speed, while the upper 2GB address space
runs at half the memory bandwidth, for a slight change
in performance. I've run a system, with this characteristic
(address space dependent memory speed), and you can't
really notice a difference. Doing it this way, installing
3x2GB, is only going to bother guys like me, your
Friendly Memory Guy.

3b) To compensate for the channel imbalance, so the entire
memory space is dual channel and consistent, you
currently own 2GB, and buy a 2GB stick as well as a
4GB stick, for a total of 8GB in the form of three sticks.
You stick 2x2GB on one channel, you stick the 4GB DIMM on
the other channel. Total RAM is 8GB. Each channel has 4GB.
As long as the processor supports "Intel Flex Memory",
it works as predicted. Since Intel switched to putting
the memory controller right on the processor, the
quality of Intel documentation has dropped to the point,
that I can no longer give a personal guarantee of anything!
I think all the processors have Flex Memory (at least, the
dual channel processors), but if called on it, I can't point
to a line in the processor spec to prove it.

Another plan would be:

4) Buy the new 2x2GB non-XMP kit. It will run every bit as
fast as the old stick. Put the old stick back in its
anti-static bag. Total installed RAM is now 4GB. Speed
is DDR3-1600 CAS8. The difference of one in CAS, is pretty
close to invisible. Since the setup is non-XMP, you enter
the BIOS and set up the memory speed manually. If left
to auto, your DDR3-1600 two stick kit could default to
DDR3-1333. So you'll need to adjust it yourself.

The disadvantage of the story in (4), is you've retired 2GB
of RAM. You're doing this, on the theory that it's not possible
to reasonably match the original stick (plan (1) above). By buying
the matched two stick kit, you're compensating for not doing
it right the first time.

Using the Intel Flex Memory feature, is certainly a possibility,
nothing wrong with it. But it is a nuisance to have to buy
a single 2GB stick and a single 4GB stick, to try to
build up a channel-matched configuration.

Only (1), were we aiming for an XMP capable system. Due to
that limitation of XMP, where the XMP profile in the SPD
states the number of sticks (either half of the slots used
or all of the slots used), (1) is the only plan with a
good chance of having a totally automated setup. If XMP
was enabled in the BIOS, the BIOS supported XMP (check
motherboard specs), then the BIOS would set the speed
to DDR3-1600. If you're reasonably clever though, you
can dial in the speed yourself, a ton of people have
done that before you.

On the face of it, there is little difference between
DDR3-1600 CAS8 versus CAS9. The CAS8 is faster, but the
slightly reduced latency, could only be measured with a
synthetic benchmark (or a stopwatch). It's not like
you'll be able to watch five videos in Firefox, when
before you could only watch four videos.

Going to a mismatched two stick configuration,
a severely mismatched three stick configuration (3x2GB),
a mismatched three stick with channel balance (1x4GB+2x2GB),
or even you going out and buying three 2Gb sticks to add
to the system (for 4x2gB config), all will require
some twiddling in the BIOS.

Just to please me, you could buy two 2x2GB kits, populate
all four slots, then put the Ripjaws back in the anti-static
bag. Now, you'd have "completely matched" RAM, and for a total
project cost of $34 :-) (Note: I don't own stock in
Crucial or Micron, or any other memory company for that
matter.)

I hope you have a 64 bit OS, to support more than 4GB of
RAM, in an intelligent way... A Linux 32 bit OS with PAE
mode, can support more than 4GB, but not many people do
that (run Linux, percent market share). Windows 32 bit
systems are artificially limited to 4GB. If all you install
is a 2x2GB configuration, then, the OS choice isn't an issue.
If you install 3x2GB, 4x2GB, 2x2GB+1x4GB, then you have to
think about whether the OS will use the RAM or not.

Paul

On Sat, 08 Dec 2012 15:39:21 -0500, Paul wrote:

You currently have a single stick of RAM, 2GB total ?

No.

The mobo supports Intel XMP.

The mobo now has 2 dimms with 2gb each:
2.) Installed in my desktop is:
...
Capacity 2GB (2GBx1)

Near as I can tell, my 2x(2gb) modules support dual-channel and XMP on my board.

Being as I can't find anything on XMP support for the Crucial Ballistix, safe to
assume it would run slower and limit my collective thruput with all 8gb installed?

Aside from XMP, what bothers me is comparison between timings like 8-8-8-24 and
9-9-9-24-2N. Explanation in context of mem thruput?

Thx,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:
On Sat, 08 Dec 2012 15:39:21 -0500, Paul wrote:

You currently have a single stick of RAM, 2GB total ?

No.

The mobo supports Intel XMP.

The mobo now has 2 dimms with 2gb each:
2.) Installed in my desktop is:
...
Capacity 2GB (2GBx1)

Near as I can tell, my 2x(2gb) modules support dual-channel and XMP on my board.

Being as I can't find anything on XMP support for the Crucial Ballistix, safe to
assume it would run slower and limit my collective thruput with all 8gb installed?

Aside from XMP, what bothers me is comparison between timings like 8-8-8-24 and
9-9-9-24-2N. Explanation in context of mem thruput?

Thx,
P


OK, that simplifies things considerably.

The modules in any case, can only run as fast as their top
rated speed. In your example, one set is DDR3-1600 CAS8
and the other is DDR3-1600 CAS9. When they're mixed in
the same PC, all sticks share the slowest speed setting
of DDR3-1600 CAS9.

CAS9, is the delay from requesting the data, until the
first data of the burst is returned. Thus CAS9, you're
waiting one clock cycle longer to get your data than CAS8.
The CAS9 module is "slower". CAS12 would be even slower.

The difference is relatively small. To mirror this back
to DDR days, DDR3-1600 CAS8 = DDR2-800 CAS4 = DDR400 CAS2.
They would have the same latency. So if they were
DDR400 modules, one would be CAS2 and the other
would be "CAS 2.25". The practical difference should
be tiny.

When four DIMMs are present, XMP will be disabled.
That would be my guess, as it's unlikely anybody
includes an XMP profile for a 4 DIMM setup in their
XMP data on the DIMM. This really isn't that important,
but just means that the user has to enter the BIOS
and verify that the memory is set to DDR3-1600 and
the slower of the two CAS values (CAS9).

I expect the two configurations (original 2 DIMM config,
or proposed 4 DIMM config), you're not going to be
able to detect a difference in speed. The difference
will be that small.

Paul

On Sat, 08 Dec 2012 19:22:43 -0500, Paul wrote:

OK, that simplifies things considerably.

The modules in any case, can only run as fast as their top
rated speed. In your example, one set is DDR3-1600 CAS8
and the other is DDR3-1600 CAS9. When they're mixed in
the same PC, all sticks share the slowest speed setting
of DDR3-1600 CAS9.

CAS9, is the delay from requesting the data, until the
first data of the burst is returned. Thus CAS9, you're
waiting one clock cycle longer to get your data than CAS8.
The CAS9 module is "slower". CAS12 would be even slower.

The difference is relatively small. To mirror this back
to DDR days, DDR3-1600 CAS8 = DDR2-800 CAS4 = DDR400 CAS2.
They would have the same latency. So if they were
DDR400 modules, one would be CAS2 and the other
would be "CAS 2.25". The practical difference should
be tiny.

When four DIMMs are present, XMP will be disabled.

That's consistent with the note in the mobo manual:

"To obtain best performance ... install only 1 dimm
on each mem channel."

That would be my guess, as it's unlikely anybody
includes an XMP profile for a 4 DIMM setup in their
XMP data on the DIMM.

I'm less certain about this:
http://en.wikipedia.org/wiki/Serial_...le_.28X MP.29

This really isn't that important,
but just means that the user has to enter the BIOS
and verify that the memory is set to DDR3-1600 and
the slower of the two CAS values (CAS9).

I expect the two configurations (original 2 DIMM config,
or proposed 4 DIMM config), you're not going to be
able to detect a difference in speed. The difference
will be that small.

Still, the specs suggest that mem thruput per dimm would be
somewhat diminished with 4 dimms installed and XMP disabled.

That clarifies the XMP situation considerably. I thought XMP
might be full functional with either 2 or 4 dimms. Evidently
not the case.

Thanks,
P

"Law Without Equity Is No Law At All. It Is A Form Of Jungle Rule."

Puddin' Man wrote:
On Sat, 08 Dec 2012 19:22:43 -0500, Paul wrote:

OK, that simplifies things considerably.

The modules in any case, can only run as fast as their top
rated speed. In your example, one set is DDR3-1600 CAS8
and the other is DDR3-1600 CAS9. When they're mixed in
the same PC, all sticks share the slowest speed setting
of DDR3-1600 CAS9.

CAS9, is the delay from requesting the data, until the
first data of the burst is returned. Thus CAS9, you're
waiting one clock cycle longer to get your data than CAS8.
The CAS9 module is "slower". CAS12 would be even slower.

The difference is relatively small. To mirror this back
to DDR days, DDR3-1600 CAS8 = DDR2-800 CAS4 = DDR400 CAS2.
They would have the same latency. So if they were
DDR400 modules, one would be CAS2 and the other
would be "CAS 2.25". The practical difference should
be tiny.

When four DIMMs are present, XMP will be disabled.

That's consistent with the note in the mobo manual:

"To obtain best performance ... install only 1 dimm
on each mem channel."

That would be my guess, as it's unlikely anybody
includes an XMP profile for a 4 DIMM setup in their
XMP data on the DIMM.

I'm less certain about this:
http://en.wikipedia.org/wiki/Serial_...le_.28X MP.29

This really isn't that important,
but just means that the user has to enter the BIOS
and verify that the memory is set to DDR3-1600 and
the slower of the two CAS values (CAS9).

I expect the two configurations (original 2 DIMM config,
or proposed 4 DIMM config), you're not going to be
able to detect a difference in speed. The difference
will be that small.

Still, the specs suggest that mem thruput per dimm would be
somewhat diminished with 4 dimms installed and XMP disabled.

That clarifies the XMP situation considerably. I thought XMP
might be full functional with either 2 or 4 dimms. Evidently
not the case.

Thanks,
P

I'm just repeating what I've read here. I don't have any
XMP capable motherboards, or XMP capable DIMMs, so have no
first hand knowledge. I've read that while in theory, XMP
might offer both 2 DIMM and 4 DIMM timing info, a typical
DIMM is likely to only support the 2 DIMM setting. Users
were claiming it was disabled, and required them to set
things up manually themselves.

There's nothing magic about XMP. It's an attempt to set
timing and DIMM voltage automatically. If you have all the
parameters, say via using CPU-Z or a similar utility, or
even manually grabbing the SPD table with CPU-Z and analyzing
it yourself, you can use the parameters to set up the memory
yourself. It should be every bit as fast and reliable.
XMP support, just takes the pain out of setting it up.
A motivated end-user, can do the settings themselves
just as well as the automation can.

The memory settings in the BIOS, support Auto and Manual
settings at every level. You as the user, can change
just the ones you want, and have the computer work
out the other values. For example, in your case, I'd
take the memory off "totally Auto", then, only adjust
the DDR3-1600 (800MHz) clock setting. Allow the BIOS
to compute the CAS for itself. Leave CAS on Auto.
(Should be CAS9). Then, boot memtest86+ and test
the memory. *Do Not* boot into Windows as your first
test case. You can easily corrupt your OS (registry file damage),
by the usage of poorly adjusted memory settings. Even
a guy I suggested this to, did the memtest86+ test,
the test was clean, and his Registry still got
corrupted. But, if you didn't run memtest86+ first,
it would just improve the odds of suffering corruption.
Memtest86+ runs, are to weed out the truly horrible
settings. Note that, it's not always your fault - I
have one motherboard, where you set a certain CAS
value, and the BIOS loads an entirely different value
into the DIMM. A bad bug. So sometimes, it's the
fault of Asus...

If I'm particularly suspicious (I've heard the BIOS
has that bug), I do things in this order.

1) Set up BIOS.
2) Boot memtest86+ floppy or CD. Do one or more
error free passes. Return to the BIOS if something
is still amiss.
3) Next test, is boot a Linux LiveCD, like Ubuntu.
If the memory is particularly corrupt, the icons
in Ubuntu will start disappearing from the desktop,
and eventually you'll get a BSOD.
4) You can run the Linux version of Prime95 from
http://www.mersenne.org/freesoft/ . The integrity
test that does, of your memory, is a bit more thorough
than memtest86+. Allow it to run overnight, only
go to step 5 if the test is completely error free
(no execution threads stop before the morning).
5) Now, boot Windows. Ideally, have a backup of C:,
just in case.

If you're the kind of person who just "jumps to step 5",
at least have a backup of C: ready from before you did
the memory change, to restore to known-good conditions.

HTH,
Paul