Q6600 optimal memory for performance?

[Bill]

I had a whitebox vendor build me a Quad Q6600 PC (4 GB RAM). Vista
Ultimate gives it a performance score ("Windows Experience Index") of
4.5, with the memory as the bottleneck:

Processor: 5.9
Memory: 4.5
Graphics 5.9
Gaming Grphics 5.9
Primary Hard Disk 5.4


1. Is this a reasonable performance for a Q6600?
2. Is my system a balanced design, or can I improve performance with
better memory? If so, how do I determine the right memory to upgrade
to? (I'm willing to experiment with overclocking although I've never
done it)

cpu-z shows the following parameters:

CPU:
====
Voltage: 1.23-1.26V
Core Speed 2400 MHz
Multiplier: x 9.0
Bus Speed: 266.7 MHz
Rated FSB: 1066.7 MHz

Mainboard:
==============
Manufacturer: ASUSTek
Model: P5N-E SLI
Chipset: NVIDIA 650i SLI SPP
BIOS: Phoenix ASUS P5N-E SLI ACPI Rev 0505 (03/05/2007)

Memory:
============
Frequency: 333.3 MHz
FSBRAM: 4:5
CAS# LAtency: 5.0 clocks
RAS to CAS: 4 clocks
RAS precharge: 4 clocks
Cycle Time (Tras) 31 clocks
Bank Cycle Time (Trc) 19 clocks
Command Rate: 2T

 

[kony]

I had a whitebox vendor build me a Quad Q6600 PC (4 GB RAM). Vista
Ultimate gives it a performance score ("Windows Experience Index") of
4.5, with the memory as the bottleneck:

Processor: 5.9
Memory: 4.5
Graphics 5.9
Gaming Grphics 5.9
Primary Hard Disk 5.4


1. Is this a reasonable performance for a Q6600?
2. Is my system a balanced design, or can I improve performance with
better memory? If so, how do I determine the right memory to upgrade
to? (I'm willing to experiment with overclocking although I've never
done it)

Forget about the Windows benchmark as it is established to
be not so applicable to real uses of systems, you need to
benchmark against your most demanding applications and
consider which subsystems are primary bottlenecks to those.

If your 4GB of memory is running stabily (hopefully
confirmed by at least 24 hours of testing with memtest86+) I
would leave it alone, any and all systems will have
"something" bottlenecking "some" uses, no matter what you
upgrade.

 

[Paul]

I had a whitebox vendor build me a Quad Q6600 PC (4 GB RAM). Vista
Ultimate gives it a performance score ("Windows Experience Index") of
4.5, with the memory as the bottleneck:

Processor: 5.9
Memory: 4.5
Graphics 5.9
Gaming Grphics 5.9
Primary Hard Disk 5.4


1. Is this a reasonable performance for a Q6600?
2. Is my system a balanced design, or can I improve performance with
better memory? If so, how do I determine the right memory to upgrade
to? (I'm willing to experiment with overclocking although I've never
done it)

cpu-z shows the following parameters:

CPU:
====
Voltage: 1.23-1.26V
Core Speed 2400 MHz
Multiplier: x 9.0
Bus Speed: 266.7 MHz
Rated FSB: 1066.7 MHz

Mainboard:
==============
Manufacturer: ASUSTek
Model: P5N-E SLI
Chipset: NVIDIA 650i SLI SPP
BIOS: Phoenix ASUS P5N-E SLI ACPI Rev 0505 (03/05/2007)

Memory:
============
Frequency: 333.3 MHz
FSBRAM: 4:5
CAS# LAtency: 5.0 clocks
RAS to CAS: 4 clocks
RAS precharge: 4 clocks
Cycle Time (Tras) 31 clocks
Bank Cycle Time (Trc) 19 clocks
Command Rate: 2T

You'd need to know who makes the memory and what is the part number,
to dig up more info on it.

Being a quad, it will help a little bit, to run the memory faster if
possible.

For stability testing, try something like this, while in Windows. Your
machine should be able to run this with no errors, for hours on end.
If it cannot, take it back to your builder. Orthos uses Prime95 code,
to test the CPU. The Prime95 code does some FFT calculations, which
are floating point. The answer is known, and the program can detect
arithmetic errors. If an error happens, it can be the RAM or it
could be the processor is not stable. In a lot of cases, fiddling
with the BIOS will fix it - you can do that, or your builder
may have a very good idea what needs to be tweaked.

http://sp2004.fre3.com/beta/beta2.htm

For some light reading material on Q6600, try here:

http://www.xtremesystems.org/forums/showthread.php?t=153096&highlight=q6600

"Which ram is better?"
http://www.xtremesystems.org/forums/search.php?searchid=4209772

Paul

 

[Bill]

You'd need to know who makes the memory and what is the part number,
to dig up more info on it.

Being a quad, it will help a little bit, to run the memory faster if
possible.

For stability testing, try something like this, while in Windows. Your
machine should be able to run this with no errors, for hours on end.
If it cannot, take it back to your builder. Orthos uses Prime95 code,
to test the CPU. The Prime95 code does some FFT calculations, which
are floating point. The answer is known, and the program can detect
arithmetic errors. If an error happens, it can be the RAM or it
could be the processor is not stable. In a lot of cases, fiddling
with the BIOS will fix it - you can do that, or your builder
may have a very good idea what needs to be tweaked.

http://sp2004.fre3.com/beta/beta2.htm

For some light reading material on Q6600, try here:

http://www.xtremesystems.org/forums/showthread.php?t=153096&highlight=q6600

"Which ram is better?"
http://www.xtremesystems.org/forums/search.php?searchid=4209772

The memory is Micron TEchnology 8HTF12864AY-667E1 PC2-5300 (333 MHz),
according to cpu-z

It's not on the P5N-E QVL:
http://dlsvr03.asus.com/pub/ASUS/mb/socket775/P5N-E SLI/P5N-E_SLI_QVL.pdf

I couldn't run Orthos on my Vista system. Something about WORKER.DLL
not found.

It seems to me that the best way to get a step-function improvement in
memory performance is to go to an approved DDR2-800 (from the current
DDR2-667). What else do I need to change in the system if I do this?

Thanks for your help!

Bill

 

[Paul]

The memory is Micron TEchnology 8HTF12864AY-667E1 PC2-5300 (333 MHz),
according to cpu-z

It's not on the P5N-E QVL:
http://dlsvr03.asus.com/pub/ASUS/mb/socket775/P5N-E SLI/P5N-E_SLI_QVL.pdf

I couldn't run Orthos on my Vista system. Something about WORKER.DLL
not found.

It seems to me that the best way to get a step-function improvement in
memory performance is to go to an approved DDR2-800 (from the current
DDR2-667). What else do I need to change in the system if I do this?

Thanks for your help!

Bill

Interesting. So another test (Orthos) doesn't work with Vista.

To a first order approximation, performance is proportional to processor
core speed. Usually memory is a secondary effect, and you can spend a lot
of money to get 10% more overall performance.

Your board has some "link/unlink" options. Normally, a board has fixed
simple integer ratios, between FSB and memory bus. I think the unlink
option, allows asynchronous operation, but I'm not really sure what shows
up as options in the BIOS screen.

In these tests, they show two conditions. One of the conditions is the one
you would be interested in. If you drop the CPU multiplier, you can then
raise the FSB. If done correctly, you would get close to the same core
clock rate as a result. Thus, your core is not overclocked (if you were
not interested in CPU overclocking). When the FSB is raised, that increases
the potential transfer rate at the FSB, easing any bottleneck which may be
there. In turn, that makes more room for your memory improvements.

P5N-E SLI overclocking result. DDR2-800 memory, using high or low multiplier.
http://www.anandtech.com/showdoc.aspx?i=2894&p=3

Another 650i overclocking result. DDR2-800 memory, using high or low multiplier.
http://www.anandtech.com/showdoc.aspx?i=2958&p=4

Now, one thing wrong with those tests, is there are no four stick configurations.
Anandtech has listed the test results of four stick configurations in some
past motherboard reviews. But I cannot find any four stick results for 650i.
In the past, there were few consequences from using four sticks, but I've
noticed an increased incidence of "you have to put the two sticks in the yellow
slots" type reports for recent motherboards, implying there are some
differences to be expected. In the past, only a minor timing change was needed.

Your motherboard has plenty of flexibility in setup. You could use DDR2-800
memory if you want. Perhaps a DDR2-1066 memory would also leave room for
improvement. But without having results in hand, to help predict what the
limits might be, I cannot guess what makes good, economic sense as a memory
purchase.

I see some mention of Micron D9 memory chips, as being capable of pretty
high memory clock speeds. Some of this stuff is pretty expensive.

http://forums.guru3d.com/showthread.php?t=206393
http://ramlist.ath.cx/ddr2/
http://www.shopatbuffalotech.com/cart/home.php?cat=42083

Paul

 

[Trimble Bracegirdle]

I was a bit puzzled that VISTA rated my 3 Gigs of DDR2 800 MHz Dual Channel
as the lowest value in there index thing on a E6600 running at 3.2 Gig Hz &
Geforce 8800 GTX.
The rating seems to say 'Memory Speed' rather than other aspects such as
amount .
(\__/)
(='.'=)
(")_(") mouse

 

[kony]

You need -at least- DDR2-800 buddy. Forget 667, it ain't gonna cut it.
Not to mention those latencies are horrible also. You can get a 4GB kit
of DDR2-800 running at CAS4-3-3-10, which is *very* good.

Sure, it'll be faster BUT until there is an actual use of
the system where the memory is a significant (per the user's
perception and needs) bottleneck, there is no particular
rush to throw out 4GB of memory just to replace with some a
little bit faster... such a change can come at any time and
ultimately, for the money it may not even be the biggest
enhancement to the performance of the apps needing a boost.