I already told you Vic... 3.6GHz is the best a Q6600 will do stable on air and normal cooling.. you can expect the same with the x48
Clocking memory is a art and a science. I cant sit here and go through an entire tuning session with you. I can however point you to some posts about memory clocking.
My goal is to PURCHASE memory/CPU/motherboard that will allow me to run a 450MHz FSB and therefore shift gears on the northbridge. Since the Q6600 has a locked multiplier of 9 you wont be able to do that. You would have to move to a quad with either a QX prefix that allows multiplier changes, or, a locked 8x multiplier which will not exceed its GHz ability or a 9 multiplier in which the quad can run 4GHz as demonstrated in this thread on a x48 at 450MHz
http://forums.avsim.net/dcboard.php?az=show_topic&forum=197&topic_id=43631&mode=...as you scroll through that thread you will see where I worked with simjunkie to get him in the 450 zone but he also has a Q9650 which is running 9x450 or just over 4GHz too and those procs will do that on air
Here is some info about memory tuning.. does not matter if its P35, x38 or x48
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=1read the entire article
Charts here:
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=9More here on settings
http://www.anandtech.com/mb/showdoc.aspx?i=3172&p=1Excellent source for all this tuning as Tony works for OCz and helps Asus and Gigabyte with BIOS designs
all these posts are about getting the most stable and highest perf result
http://www.ocztechnologyforum.com/forum/forumdisplay.php?f=122Success in that goes back to the memory purchase, the motherboard and the CPU ability
CPU speed is the primary factor in FSX along with keeping that OS clean however memory is also a player many do not understand.
One of the reasons I left DDR2 in March of 2007 for DDR3.
Consider this... the absolute best memory performance you can achieve no matter how high memory is clocked is 60% utilization of the memory bandwidth. Todays tech is limited to that 50-60% max ability.
OK, so what is better 50-60% of DDR2 800-1066 (400 to 533MHz) or 50-60% of DDR3 1600 - 1800 (800 to 900MHz)?
Thats just memory speed, the critical part in making use of that 50-60% is in what is known as 'latency'
In both cases the CAS and MCH latency of the memory play greatly into the use of that 50-60% and there are formulas for CAS/FSB/STRAP which drive the MCH latency changes on memory product. One of the reasons why the better memory can go from 150 bucks for 2 sticks running CAS 5-5-5 @ 1066 to 300+ bucks for 2 sticks running CAS 4-4-4 @ 1066-1150+. Those expensive sticks are only valuable to the person who knows how to use them correctly. Someone who simply plugs them into a system and does not tune them in properly is getting no perf increase over the CAS 5 1000-1066 product. Same with the 1066 products over the 800, etc.
As memory speed increases CAS latency can increase too and still run memory faster. There are a set of guidelines for CAS/SPEED in the MCH latency formula and an example of that formula in use can be seen here
http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=9Now, those charts are old and really do not take 'correctly' into account DDR3 speeds above 1500 with lower CAS because the article targets DDR2 more than DDR3 but those charts show how the lower CAS at higher memspeed affect MCH READ DELAY or tRD which is the primary key to getting the most out of that 50-60% bandwidth use.
Its like this.. you take DDR2 1066 @ 60% of the bandwidth being used and then add in the communication 'RATE' boost, or, speed the data making it to and from the memory to the CPU because the 'stop light' in the circuit is no longer 'RED' for a long period of time (or reduced latency) ... memory performance to the CPU speed is now significantly increased. You are only multiplying that effect when you move into DDR3 memory speeds
We EE's have known about this for years... Intel in the last few years just started allowing BIOS access to what really makes a CPU and system sing. If you are interested in the science behind it without going too far into the 'engineese' you may refer to this post
http://www.thetechrepository.com/showthread.php?t=195
Pages: 1
