C2D and DDR2 667

[450R]

I learneded how to link!

Do the results seem screwy to anyone else? Every other review I've seen has shown that latency is almost a non-factor for C2D. But MadShrimps is showing DDR2 400/533 with low latencies beating DDR2 667 C3/C4 in some tests. They're basically saying DDR2 667 will perform WORSE than 533. What's the deal?

 

[Shimmishim]

learn how to link!!!!

j/k

http://www.madshrimps.be/?action=getarticle&number=1&artpage=1962&articID=472

 

[acegazda]

Then why did everyone who got a conroe go out and buy ddr2-800 and cause the ridiculous fluctuation that we are suffering ATM? They could have just got some cheap ddr2-667 kit. And no, it seems like the ocing potential of the lower speed ddr2 modules would make up the gap if nothing else.

 

[myocardia]

What you're forgetting is that people are buying PC2-6400 so they can overclock, and run at even higher than 800 Mhz speeds. That test had no overclocking, it was simply running the RAM faster than the processor's fsb, which hardly ever improves performance (much).

 

[lopri]

Better Review

Best practical scenario

Non-OC: DDR2-533 with tightest timings
OC: DDR2-800 / 3-3-3

So it's basically 1:1 (FSB:RAM) with the tightest timings. As long as you don't run your memory slower than FSB, the performance difference is negligible at large.

 

[Shimmishim]

another interesting article for those that use the Asus P5B

http://www.bleedinedge.com/forum/showthread.php?t=23803

 

[imported_Buckster]

Interesting - as my P5-BD doesn't behave like that article

can do 3.19 at 357 or something OK at 1:1 - so my CAS5 memory rated at CAS4 at 667 is overclocked ...

but if I set 401 * 8 in the bios - with RAM set to 1:1 - at CAS5 - so thats 800MHz CAS5 - exactly what its rated as.

as soon as I save settings in the bios the pc hangs - you can't even get out of the bios !

my friends setup does exactly the same.

mobo limitation or CPU ? I can't see how it could be the CPU as it can run slightly lower -stable, even if 3.2 wasn't stable in Orthos I'd expect it to at least boot up properly.

Mark.

 

[hans007]

the only reason to run higher than ddr2-533 is latency. the voltage doesnt even matter.


technically ddr2-600 at cl3, would be just as fast as ddr2-1000 at cl 5.

3 cycles at 600mhz, is the same as 5 1000mhz cycles.


the bandwidth doesnt matter since its fsb limited anyway. it would only matter for integrated video.

 

[gobucks]

for some reason, when i ran a 4:5 divider to make my system run at DDR2-1000 CAS 5, it was substantially faster than running the memory at a 1:1 divider at DDR2-800 CAS 4. It seems to me that at least when overclocked C2D seems to really like the extra memory bandwidth.

 

[cmdrdredd]

Originally posted by: gobucks
for some reason, when i ran a 4:5 divider to make my system run at DDR2-1000 CAS 5, it was substantially faster than running the memory at a 1:1 divider at DDR2-800 CAS 4. It seems to me that at least when overclocked C2D seems to really like the extra memory bandwidth.

It does. Having faster channels to fetch from memory always helps, but the question is how much?

 

[Polish3d]

Originally posted by: gobucks
for some reason, when i ran a 4:5 divider to make my system run at DDR2-1000 CAS 5, it was substantially faster than running the memory at a 1:1 divider at DDR2-800 CAS 4. It seems to me that at least when overclocked C2D seems to really like the extra memory bandwidth.

That's how I'm running mine too. How did you measure the speed differene?

 

[lopri]

The timedemo in Doom 3 is an excellent tool for memory testing. (Run it without AA) Very sensitive to memory configuration and results are extremely repeatative.