tighter timings or higher async frequency?

[dragonfang]

I've been doing some overclocking on my new build, and im closing in on a sweetspot
Here's a little background info:
c2d e4300
AC Freezer 7 Pro
Gigabyte GA-965P-DS3 rev3.3 f12 bios
eVGA GeForce 7600GT
Corsair XMS2 PC6400C4 2x1gb 4-4-4-12
Corsair HX520

I've been following this guide
http://www.hardforum.com/showthread.php?t=1169366
and I'm pretty happy with my results so far-
I kept my ram voltage at 2.1v, overvolted the northbridge, pci-e, and the fsb by .1v
I raised the fsb from the default 200 to 335mhz while keeping the vcore at stock. so this yields 335x9= 3015mhz, with the corsair underclocked at 670mhz
I'm running the ram multiplier at sync 2x with tightened 3-3-3-12 timings
I ran memtest86+ overnight for 25 passes without any errors and orthos blend test for 5 minutes, and passed double instances of superpi32. Orthos has been running for a few hours on smallfft (to stress the cpu). Temps are a little warm with tat readings of 67C on load (70-72C on tat load), but it should be ok. So far everything looks great

now the dilemma is- should I keep the tighter timings at cl3, or should i run the ram asynchronously with a 2.5 multiplier and stock/looser timings, which would overclock the corsair at 837mhz? which setup would yield better performance? I researched this on google and a couple of sources say that the higher async frequency is only worth it to a certain extent or if the difference is big enough.

please give me some feedback
thanks

 

[Accord99]

From here:

http://www.xbitlabs.com/articl...uo-memory-guide_5.html

DDR2-800 CAS4 is slightly faster than DDR2-667 CAS3 across the board with the clockspeed/FSB of the CPU held constant.

 

[dragonfang]

Originally posted by: Accord99
From here:

http://www.xbitlabs.com/articl...uo-memory-guide_5.html

DDR2-800 CAS4 is slightly faster than DDR2-667 CAS3 across the board with the clockspeed/FSB of the CPU held constant.

thanks for the link
correct me if im wrong though, but the testbed they used for their tests was an x6800 that they overclocked to 400mhz fsb. you see the performance increase with the ddr800 vs the ddr600 because with ddr800, the ram is running synchronously with the cpu. however, i can't run my system at ddr800 since i'll need a lot more vcore to have 400mhz fsb. the tests show that if the ram is running faster than the cpu, the data transfers would be bottlenecked by the bandwidth of the fsb. and right now, if i choose to run it at ddr837, wouldn't my 335mhz fsb would be bottlenecking the extra frequency gain of the ram?

In our tests we overclocked the FSB to 400MHz and enjoyed a considerable performance growth on installing DDR2-800 SDRAM. This growth was much more tangible than what we had at a 266MHz FSB, other conditions being the same. However, the system didn't speed up on our using 1000MHz memory because the speed of CPU-memory data transfers was now limited by the bandwidth of the FSB. Still it is clear that if you overclock the FSB to frequencies above 400MHz, you will see performance gains on using memory faster than DDR2-800, too.

as I read what i wrote again, i find it very confusing. i hope it makes sense

 

[Accord99]

The main portion of the review was conducted with the X6800 at stock speeds. Only the portion starting at page 8 and onwards uses the overclocked X6800.

the tests show that if the ram is running faster than the cpu, the data transfers would be bottlenecked by the bandwidth of the fsb. and right now, if i choose to run it at ddr837, wouldn't my 335mhz fsb would be bottlenecking the extra frequency gain of the ram?

The FSB does bottleneck the memory somewhat but increasing the memory clockspeed still improves performance because the system is spending less time waiting for memory. A sweet spot for an E4300 might be to use the 8X multiplier and set the FSB to 400MHz and memory to DDR2-800.

 

[dragonfang]

thanks for clearing that up
i wish i can do 400x8, but i'd have to ramp up the vcore to get the same amount of speed as the one im getting with stock vcore

 

[jeffw2767602]

try 7x429 and run your ram at 858mhz

 

[firewolfsm]

Basically, find your max stable speed for the processor, and see what the lowest multiplier that can maintain that is. Even 6x5XX would be nice.

 

[darkfalz]

1:1 and tighter timings IMO

 

[imported_Shaq]

Originally posted by: Accord99
From here:

http://www.xbitlabs.com/articl...uo-memory-guide_5.html

DDR2-800 CAS4 is slightly faster than DDR2-667 CAS3 across the board with the clockspeed/FSB of the CPU held constant.

hmmm...the gist of it is at stock, higher memory clocks are better even at looser timings and when overclocked it is better to have tighter timings than higher memory clocks ie. 4-4-4-12 @ 800 is better than 5-5-5-15 @ 1000. Put another way, 1:1 is best when overclocking, and the tighter the timings the better.

Thanks for posting this as I was wondering about where to set my ram SPD, so I'll leave it at 1:1 where I've got it now.

 

[dragonfang]

Originally posted by: Shaq

Originally posted by: Accord99
From here:

http://www.xbitlabs.com/articl...uo-memory-guide_5.html

DDR2-800 CAS4 is slightly faster than DDR2-667 CAS3 across the board with the clockspeed/FSB of the CPU held constant.

hmmm...the gist of it is at stock, higher memory clocks are better even at looser timings and when overclocked it is better to have tighter timings than higher memory clocks ie. 4-4-4-12 @ 800 is better than 5-5-5-15 @ 1000. Put another way, 1:1 is best when overclocking, and the tighter the timings the better.

Thanks for posting this as I was wondering about where to set my ram SPD, so I'll leave it at 1:1 where I've got it now.

that was the way I interpreted it the first time i read the article
but this contradicts what accord99 just told me
can anyone clarify?

im going to try 7x429 after i'm done with the orthos and see if i can get past the strap

 

[dragonfang]

well, i couldn't get past the fsb hole, so i guess i'm going to stick to 335x9
i tried running the ram at 3-2-2-8 with the normal 2.1v and it passed memtest for more than enough passes
but when i booted into windows and opened cpu-z, it only shows up as 3-3-3-8. is this the right reading? i triple checked the bios and made sure it was 3-2-2-8, yet it gives me a looser timing on cpu-z. what gives?

i couldn't actually verify what timings were tested on the memtest since the chipset is unsupported

cpuz
oh and eveytime i set any cpu frequency above 266mhz in bios, it always POSTs 1mhz lower than it was actually set to. you can see it on the cpuz validation, i had it set to 335 in bios, but it shows up as 334 (not that it's that big of a deal). is this normal?

 

[The-Noid]

Higher frequency is better than tighter timings.

 

[rstrohkirch]

I gained a 4-5% improvement in CS:S/DoD:S going from 1:1 748mhz@4-3-4-12 to 1:1.25 935@5-5-5-12

Only real world test I cared about

 

[dragonfang]

yeah, i just confirmed this today. i actually noticed an average of 45 seconds of improvement on superpi 32mb (i did several tests on each setup) on the 2.5spd (4:5 with 4-4-4-12) setting over the tighter cas3 timings. this is good enough reason for me to have it async, however small the difference may be. thanks for all the feedback, i'll keep it at async

btw does anyone know why the fsb on POST & windows is always 1mhz lower than what i set n the bios?

 

[Accord99]

Originally posted by: dragonfang
btw does anyone know why the fsb on POST & windows is always 1mhz lower than what i set n the bios?

BIOS quirk perhaps, Asus sometime's likes to go 1MHz above the selected frequency (to look better on benchmarks no doubt).

 

[brencat]

Originally posted by: dragonfang
btw does anyone know why the fsb on POST & windows is always 1mhz lower than what i set n the bios?

Normal. Don't worry about it.

My almost 5 year old Asus A7N8X Deluxe does exactly the same thing.

 

[imported_Shaq]

Originally posted by: dragonfang
yeah, i just confirmed this today. i actually noticed an average of 45 seconds of improvement on superpi 32mb (i did several tests on each setup) on the 2.5spd (4:5 with 4-4-4-12) setting over the tighter cas3 timings. this is good enough reason for me to have it async, however small the difference may be. thanks for all the feedback, i'll keep it at async

btw does anyone know why the fsb on POST & windows is always 1mhz lower than what i set n the bios?

Remember that SuperPi is memory bandwith limited and the article showed that it did help with that application but others it didn't. It depends what applications you use most often. For games there was a slight decrease overall. I wonder if the Q6600 would be any different that the C2D they reviewed? It sounds like it would need more bandwith but I don't really want to run all the tests. If anyone has seen an article with the Q6600 and different memory settings please post it.