DDR2 1066 unlinked VS. DDR2 800 1:1 (on E6420)

alpaca007

Junior Member
Jul 8, 2007
3
0
0
I want 2 gigs on my eVGA 680i with C2D E6420.
I am trying to decide between running DDR2 1066 (5-5-4-12) that is unlinked to FSB since I will be overclocking, or using DDR2 800 (4-4-3-5) with 1:1 ratio and the CPU overclocked to 8x400 = 3.2 ghz.

However, I may end up with a more modest overclock of 7x400 = 2.8 ghz, also using DDR2 800 in 1:1. Another question: can I even do this? Can the multiplier on the E6420 be lowered?

Thank you.

Any suggestions would be great.
 

JustaGeek

Platinum Member
Jan 27, 2007
2,827
0
71
I believe you can do it, if this feature in enabled in the MB's BIOS. You can do it on P5N-E SLI with E6600, should not be different for E6420.

IMHO, it is better to run DDR2 800 linked to FSB at 400MHz as opposed to "funny" unlinked memory dividers for the 1066MHz frequency. Also, with increased latencies for the 1066, the performance difference (SANDRA, Everest) will be minimal, if any.
 

bfdd

Lifer
Feb 3, 2007
13,312
1
0
Originally posted by: JustaGeek
I believe you can do it, if this feature in enabled in the MB's BIOS. You can do it on P5N-E SLI with E6600, should not be different for E6420.

IMHO, it is better to run DDR2 800 linked to FSB at 400MHz as opposed to "funny" unlinked memory dividers for the 1066MHz frequency. Also, with increased latencies for the 1066, the performance difference (SANDRA, Everest) will be minimal, if any.

Memory benchmarks always showed unlinked 1066mhz running faster for me than 1:1 800 on my setup when I used to run Corsair XMS2 PC-8500.
 

JustaGeek

Platinum Member
Jan 27, 2007
2,827
0
71
Originally posted by: bfdd
Originally posted by: JustaGeek
I believe you can do it, if this feature in enabled in the MB's BIOS. You can do it on P5N-E SLI with E6600, should not be different for E6420.

IMHO, it is better to run DDR2 800 linked to FSB at 400MHz as opposed to "funny" unlinked memory dividers for the 1066MHz frequency. Also, with increased latencies for the 1066, the performance difference (SANDRA, Everest) will be minimal, if any.

Memory benchmarks always showed unlinked 1066mhz running faster for me than 1:1 800 on my setup when I used to run Corsair XMS2 PC-8500.

Could you give me a comparison in hard numbers, like 7000MB/s versus 6500MB/s ?

My intention is not to dispute your experience, just for my personal reference... We were discussing this issue on the P5N-E SLI forum. What I mean is - same FSB, different memory frequency...

Thank you.
 

bfdd

Lifer
Feb 3, 2007
13,312
1
0
If the kit actually still worked and I didn't have 2 of them break on me I would be glad to re run benchmarks for you. But I don't have it working anymore and have the ram I have now. It was definitely faster for me in my experience.
 

JustaGeek

Platinum Member
Jan 27, 2007
2,827
0
71
Thanks, JAG87.

Another words, DDR2 800 with 4-4-4-12 performs slightly better that DDR2 1000 with 5-5-5-15, or the difference is negligible.

That was exactly my experience - without a headache of higher voltage reguired for DDR2 1000, and obviously more heat generated inside the case.

Look at the numbers in my sig for DDR2 800 with FSB 325MHz, Unlinked divider of 16:13. SANDRA ~6500MB/s, Everest ~8900MB/s, Everest latency of ~55. I think that with FSB of 400MHz, linked 1:1, the numbers would only be better!

But I don't want to go this high on FSB with stock HSF...