DD3 1333 vs DDR2 1066

Val

Limp Gawd
Joined
Nov 1, 2005
Messages
299
DDR2 has timing C5 and DDR3 C9. Although frequency > timings, this is a big difference.

Money no object, which one should i get?
 
How much memory total do you want to buy right now? Price aside, there is a MUCH better selection right now in DDR2 in both 2gb and 4gb kits. Also, if you want perfect, fast, and stable, keep in mind that the DDR3 platforms are VERY new and may still have bugs. Personally, I would not buy DDR3 for atleast another 60-90 days minimum so that the platform support is better.

But, if none of that matters, the latencies for DDR3 are not bad at all, they just seem that way. If you want to be cutting edge, go for it. The Gigabyte boards are supposed to be clocking DDR3 like mad.
 
Yeah, overclocking seems awsome. Close to 1500! :cool: Try that with DDR2 :D
 
As Mike says, the timings don't directly relate from DDR2 to DDR3, since the two technologies operate differently, so C9 timings don't mean the same as DDR2 running C9.

I just purchased an ASUS P5K Deluxe board (P35 chipset) that runs DDR2, so that should let you know how I feel about it! I should hit 1300mhz+ with the right DDR2 kit. ;)
 
And which kit is that??

I have some PC2-8500 Reaper that just showed up, that I will be testing and a backup kit of PC2-9600 Flex XLC coming soon as a backup. :) I have had some very good reports on the P35 chipset and the ASUS DDR2 board in particular.

I am out of town until Thursday, so on Friday I'll be setting the new rig up.
 
Why would you need that? Boards can't run stable at 600 FSB (other than P35 based boards).. and it boggles me why you would want to run an async upclock.. if you can even get it stable.
a) Because he works at OCZ.
b) Because he can.
c) Both of the above. :)
 
Well, with memory rated to run at higher frequencies, the timings can always be reduced. Or, a divider can just be used to make use of the higher frequency. What's wrong with running the memory frequency async to the CPU's FSB frequency?
 
Well, with memory rated to run at higher frequencies, the timings can always be reduced. Or, a divider can just be used to make use of the higher frequency. What's wrong with running the memory frequency async to the CPU's FSB frequency?

Doesn't stablize.

Intel memory controller takes more voltage to be running RAM so high when you get into the high FSBs. It's almost like doing two things at once - so it is always better to lock down 1:1.
 
Doesn't stablize.

Intel memory controller takes more voltage to be running RAM so high when you get into the high FSBs. It's almost like doing two things at once - so it is always better to lock down 1:1.

From the benchmarks ive seen, frequency eats games for breakfast. Big time. Why go 1:1 and reduce your framerate by like 10-15FPS?

Moreover, im running asynchro and im dead stable.

This 1:1 seems really usless and overhyped to me.
 
From the benchmarks ive seen, frequency eats games for breakfast. Big time. Why go 1:1 and reduce your framerate by like 10-15FPS?

Moreover, im running asynchro and im dead stable.

This 1:1 seems really usless and overhyped to me.

Entirely untrue. Not sure what benchmarks you've seen.

You will not notice any difference when running RAM higher than FSB in terms of real world performance. Especially in games.
 
It is a fact that running higher bandwidth gives better performance than running 1:1, but I also agree that the difference will not be noticable..... I dont think most of everything we buy gives noticable difference anyways cept maybe video cards for gaming......

For most consumer, an E4300, basic 965p mobo and 2g of value pc2 5300 is all thats required and everything higher than that wont make a noticable difference in real world performance....
 
Oops, sorry guys, you are right. The test i was thinking about also increased FSB with mem speed, so that would explain the frame rates. Indeed, memory speed has very little effect on game peformnce.

But what about multitasking? I.e. if i go alt+tab all the time? I noticed that this process happens a lot faster with faster memory? So there should be benefit in going at least 1:2? Like 1333/1333?
 
Some problems with those tests. They didn't show how much more of a difference a higher clocked CPU would bring instead of just overclocking the RAM.

Of course taking into account that if you ran 1:1 with those RAM speeds (not possible..) then you just might have the fastest system ever. :eek:
 
I'd put my 880Mhz DDR2 at 1:1 against the DDR-1067 test in that review any day if I had a 7900 on hand.

Even when clear performance gains can be obtained when running memory at higher frequencies...

Is your memory magic?
 
But, if the CPU clock remained constant, higher performance will be gained if the memory frequency was higher.

Doesn't matter.

Assuming your primary objective of overclocking is performance: once you reach higher FSBs, your memory controller will struggle to keep memory running 2:3, or 4:5, etc. Its speed would be set way too high. Your maximum performance will always be the highest CPU clock speed with the highest FSB possible running a 1:1 ratio.

Those benchmarks are flawed in that they do not show the difference when you overclock the processor.
 
Those benchmarks are misleading in that they do not show the difference when you overclock the processor.

Changed flawed to misleading.

Take this example:

My Conroe @ 3.6 GHz running 1:1 with crappy 800MHz value RAM will beat by a longshot a Conroe @ 2.93 GHz running whatever with any $500 PC-10000 1250 MHz RAM.

Of course we have not discussed chipset/FSB straps and memory latency.. but that's a whole different story.
 
higher mhz with the same latency is always better... DDR800 never beats DDR1066
 
My Conroe @ 3.6 GHz running 1:1 with crappy 800MHz value RAM will beat by a longshot a Conroe @ 2.93 GHz running whatever with any $500 PC-10000 1250 MHz RAM.

Of course it will but, if both were running at the same frequency and one had the memory set at a higher frequency than the other, then this one would perform better. I don't know why you've started changing variables. Doing so would make that article flawed; it's fine as it is.
 
Okay I give up. Go ahead and pay 500 dollars on Class X memory.

It's charts like that that persuade noobs into buying expensive ass RAM. I too was fooled by RAM figures.. on my merge from S939 to 775 I bought my 8500 Dominator kit and was expecting to get a huge OC while running my memory at a big OC also. Intel's memory controller just doesn't work that way. I learned the hard way: You can't have both without excessive voltmodding/tweaking. Memory is one of the factors that will have the least overall effect when overclocking a system.

Of course it will but, if both were running at the same frequency and one had the memory set at a higher frequency than the other, then this one would perform better.

Good luck on getting that stable. :)

I've noticed on my 680i, starting at the 1333 strap is very, very difficult to run async upclock on.
 
On the new p35 chipsets no-one is having trouble doing 1:2 in 1333/1333 configs from what i've read so far. In fact board is designed with this in mind.
 
On the new p35 chipsets no-one is having trouble doing 1:2 in 1333/1333 configs from what i've read so far. In fact board is designed with this in mind.

That's good to know. With the P5K Deluxe board i'm getting i think i might just get some Crucial Ballistix PC-8500 since they're the same price as the PC-6400 at the moment.
 
The only divider the P965 has trouble with is 4:5 and that doesnt take affect till around 420fsb..... The P35 has that covered but you still need rams that will do high frequency to take advantage of it..... The question still remains the same, is putting your hardware to that kind of stress worth it? ... theres no right or wrong answer cos its hardware and you can do what you want with them.....

I can run my computer right now 3600mhz 9x 400fsb at 2:3 divider ( ddr1200), Huge difference in bandwith than 1:1 but I dont think Ill notice the difference so why bother......
 
Back
Top