What's better for RAM?

Dougiebabe2003 · 19 Aug 2007 at 02:09

Hi,

Specs:

E6600
Asus P5B
2GB 800MHZ CellShock

Im just thinking, I know that ram can get close to 1066mhz at 5-5-5-15 timings and wondering while I'm O/C if it's better for me to have the ram running faster or at a 1:1 divider and slower with tighter timings?

Surely the extra speed will help the e6600 better than timings?

I tested in everest on a 939 X2 3800+ with ddr ram @ 2-2-2-5 runing at 416mhz and it's latency was 48ns, e6600 @ 5-5-5-15 timings was 58ns

I belive that the 939 X2 liked the lower timings ove the faster FSB and just thought that being able to shove data into the e6600 would be better than the latency to get at it...?

Thebug · 19 Aug 2007 at 05:42

i tried various ways and found it ran best at 1:1 so cpu @ 401 mem @ 802.

Big.Wayne · 19 Aug 2007 at 06:10

Dougiebabe2003 said:
I belive that the 939 X2 liked the lower timings ove the faster FSB and just thought that being able to shove data into the e6600 would be better than the latency to get at it...?

I think the Athlon 64 chips like tighter timings because the memory controller is built into the chip itself whereas on the INTEL's the data still travels via the Northbridge.

I read that INTEL are gonna integrate the memory controller on future CPU's so kudos to AMD for getting the tech in early.

Memory bandwidth is application sensitive I believe, i.e some programs or games thrive on it whereas some other applications/games don't get much benefit.

In your case I would leave the memory on 1:1 until I had totally sussed out what the CPU and mobo/chipset could do, once thats done you can just play around with different settings for the RAM and run a few benches/games and see what u feel is best. . .

In short I don't think the Conroe is memory bandwidth starved even running at PC2-6400 speeds, perhaps Quad core may change this?

MumboJumbo · 19 Aug 2007 at 11:49

I saw a lot of graphs on this somewhere. About 30/40 in total probly.

The faster timings benefit one thing and the faster clock speed another.

They were using balistic ram i beleive and they had it at somthing like:

5-5-5-15 1000 mhz.

4-4-4-12 1000 mhz

5-5-5-15 1050mhz

The 4-4-4-12 always outperformed the 5-5-5-15 at 1000mhz. But half the graphs showed that 1050 at looser timings where better and the other half showed titer timings where better.

But overal youll never notice the difference in real world use. Just benchmarks.

weescott · 19 Aug 2007 at 15:10

900mhz and tight timings (1T) is the sweet spot. Above that it's just about raw mhz without the timing gain.

Dougiebabe2003 · 19 Aug 2007 at 18:35

K, Well I gave it a go and at the minuite have the following overclock...

FSB: 306x9
Voltage(In CPUZ): 1.240

CPU @ 2.75
RAM @ 1020mhz 5-5-5-15 2T

I found it not to be great at around 300 but read that I should keep going and it got better. I did have the FSB @ 310x9 which gave me almost 2.8 but after leaving it doing Seti@Home I found it had crashed. 306x9 seems to be ok though... may have been the ram... May contuine on with the ram lowered on speed.

What benifit do I gain by having the divider @ 1:1, if I done that now I would have 612Mhz ram which would give good timings but surely be slower than faster ram with loser timings?

Cheers to everyone for replying!

Dougiebabe2003 · 19 Aug 2007 at 20:53

K, I have now lowered the ram speed to find out my max CPU overclock & have found something weird...

Core 1 will do fine at (in bios) 1.25v @ 3.1 but Core 0 is still moaning, kicking up errors in orthos AND performing worse than the other?!?

Is this normall? When I say performance Core 1 was performing iterations in Orthos quite a bit quicker than the other...noticeable diffrence in performance let's just say that.

Any help?

Edit to add that Core 0 is moaning now 1.3v and temps are hitting 62 each core, 64 overall. Temps are taken in everest (nothing else will work with me in vista)