Greetings all!
I'm going to start with a lot of explanation and then a basic question at the end.
I have an application that is incredibly disk intensive. It looks up combinations of items from a fairly complex database. Disk based queries were just too slow, so I decided to move the database (it's static... for lookups only) to a RAM drive.
VERY dramatic difference. Somewhere around 40 times faster.
So... the machine I use has a Q6600 on a P35-DS3L mobo with 8GB DDR2 6400 RAM with 7Pro for the OS. Previously, I'd not really overclocked this particular machine much. I had it at around 2.8GHz, but hadn't really looked at the overclocking in any detail. I upped it to 3.2GHz and everything was nice and stable.
SO... doing some investigating, I realized that I was running the memory at half the speed, or at least asynchronously. I changed that so that it was running synchronously and with that alone got about a 14-15% speed increase.
The program, unfortunately, is single threaded, so it only uses one core, but it doesn't max that one core out, so it still isn't CPU bound, it appears to still be bottlenecked soley by data access speed.
My question is this. Could I switch to a set of DDR2 1066 modules and run the memory in a mode that ups the memory to that speed while not stressing or overclocking the rest of the machine farther than it already is? Heck... I'd even be willing to reduce the OC on the CPU if I could get even faster I/O with the memory.
Right now, I'm cranking work out at about 45 times faster than a stock machine in the normal configuration (spinning disks) does, but if I could crank the memory from 800 to 1066 and get another 33%, that would have me around 60 times the speed. What took an hour done in a minute!
So... is this doable? Can I simply plug in DDR2 1066 sticks, change the ratios and get the memory to run at those speeds? Would I be able to do that and keep the memory synchronous with the CPU?
Thanks in advance for any and all replies,
Joe
I'm going to start with a lot of explanation and then a basic question at the end.
I have an application that is incredibly disk intensive. It looks up combinations of items from a fairly complex database. Disk based queries were just too slow, so I decided to move the database (it's static... for lookups only) to a RAM drive.
VERY dramatic difference. Somewhere around 40 times faster.
So... the machine I use has a Q6600 on a P35-DS3L mobo with 8GB DDR2 6400 RAM with 7Pro for the OS. Previously, I'd not really overclocked this particular machine much. I had it at around 2.8GHz, but hadn't really looked at the overclocking in any detail. I upped it to 3.2GHz and everything was nice and stable.
SO... doing some investigating, I realized that I was running the memory at half the speed, or at least asynchronously. I changed that so that it was running synchronously and with that alone got about a 14-15% speed increase.
The program, unfortunately, is single threaded, so it only uses one core, but it doesn't max that one core out, so it still isn't CPU bound, it appears to still be bottlenecked soley by data access speed.
My question is this. Could I switch to a set of DDR2 1066 modules and run the memory in a mode that ups the memory to that speed while not stressing or overclocking the rest of the machine farther than it already is? Heck... I'd even be willing to reduce the OC on the CPU if I could get even faster I/O with the memory.
Right now, I'm cranking work out at about 45 times faster than a stock machine in the normal configuration (spinning disks) does, but if I could crank the memory from 800 to 1066 and get another 33%, that would have me around 60 times the speed. What took an hour done in a minute!
So... is this doable? Can I simply plug in DDR2 1066 sticks, change the ratios and get the memory to run at those speeds? Would I be able to do that and keep the memory synchronous with the CPU?
Thanks in advance for any and all replies,
Joe
Last edited: