what is better? low ddr2 freq CL3 or high ddr2 freq CL4?

[guess2098]

Date: 8/18/2006; Time: 11:10PM
Air Temp : 75F ; location: Southern California



CPU: INTEL X6800 @ 1.575v
MOBO: ASUS P5WGD2 PR0 FULLY LOADED @ 0004
RAM: GSKILL 800 2GBHZ @ 2.8V DDR2 1092 CL4 vs DDR2 910 CL3
COOLING: TQ120
VC: XFX 7950GX2 XXX
PSU: PC P&C 1KW

CL3



CL4



looks like high freq of DDR2 is more important

[Micutzu]

Nice work and nice memory sticks there; i think some Lobby High comparo would've made the things more interesting .

[eva2000]

nice what vdimm for both runs 2.8v ?

[guess2098]

nice what vdimm for both runs 2.8v ?

yeah 2.8v for both... just lazy to adjust for 1092 one hehe

[andre X_X]

3dmark01 run would be nice to see....

[guess2098]

3dmark01 run would be nice to see....

can run 3d on 4G air lol

conroe sux~~~~~ lmao

[andre X_X]

do you think CL4 will perform better on 3dmark01?? just like spi32m ???

[SlackerXL]

Very interesting question...
you'll definetly hear many different opinions ..
but i think the benchmarks speak for themselves in this
extensive article over at x-Bit Labs

it looks as anything past 1000 (regardless timmings)
is unbeatable in almost every benchmark..
even compared to lower freq DDR2 3-2-2-8 memory

[heydo6]

great work

[nn_step]

Short version is For AMD Lower latency (that would be the lower speed in this case)
and for Intel, the Highier the Bandwidth the better (aka the Highier speed is always better)

[Aerou]

Short version is For AMD Lower latency (that would be the lower speed in this case)
and for Intel, the Highier the Bandwidth the better (aka the Highier speed is always better)

.. the same story going over and over ...

[fhpchris]

Short version is For AMD Lower latency (that would be the lower speed in this case)

This is wrong BTW.

Both AFI and Bachus said that max @ cas 4 is faster than max @ cas 3, and both run 2.8v+ on ddr2...

I think cas 4 is just better than cas 3 period...

[3oh6]

yeah boys...AM2 is the same as intel here. MHz > timings for the most part...especially in SuperPi. at some point i believe CAS 5 will also trump CAS 4 but don't have the hardware to test it with CAS 4 running so high already.

[nn_step]

This is wrong BTW.

Both AFI and Bachus said that max @ cas 4 is faster than max @ cas 3, and both run 2.8v+ on ddr2...

I think cas 4 is just better than cas 3 period...

lets think about this logically.
DDR2-800 @ 4-4-4-12 has similar latency to DDR-400 @ 2-2-2-6
DDR2-667 @ 3-3-3-9 has similar latency to DDR-444 @ 2-2-2-6
DDR2-1000 @ 4-4-4-12 has similar latency to DDR-500 @ 2-2-2-6

Thus it is the lower latency not the extra bandwidth for AMD

[fhpchris]

NN_STEP, you need to learn how to do some math.

Both of the people I listed are running over 460 cas 3 and 570 cas 4.

460 @ cas 3,3,3,x = 306 @ cas 2
570 @ cas 4,4,3,x = 285 @ cas 2

Yet, Cas 4-4-4 and 4-4-3 are faster at these speeds.

These are real world results, not your supid math...

[jmax_oc]

NN_STEP, you need to learn how to do some math.

Both of the people I listed are running over 460 cas 3 and 570 cas 4.

460 @ cas 3,3,3,x = 306 @ cas 2
570 @ cas 4,4,3,x = 285 @ cas 2

Yet, Cas 4-4-4 and 4-4-3 are faster at these speeds.

These are real world results, not your supid math...

With your words, 460MHz 3-3-3-x are faster than 570 4-4-x-x.
It's completely wrong.

[bachus_anonym]

Thanks for you comparison, Denny On AM2, situation is very similar...

I have done a little comparison earlier, for the purpose of some older thread, but I also did a fresh and quick test on my AM2 setup. I run 275x10=2750MHz @ DDR2-916 8-3-3-3.0 and DDR2-1100 8-4-4-4.0. FYI, my mainboard is volt-modded so actual voltage for DDR2 is 2.51v, measured with DMM. Therefore, ignore what AiBooster says - it simply shows what setting was chosen in BIOS.

DDR2-916 8-3-3-3.0

SPi32M --> 26:12.984
3DMark2001 & 3DMark03 ---> 30465 & 18514
3DMark06 ---> 4427


DDR2-1100 8-4-4-4.0

SPi32M --> 25:58.219
3DMark2001 & 3DMark03 ---> 30777 & 18517
3DMark06 ---> 4426

As you can see, as long as CPU speed is the same and we're trying to max out mems at particular dividers, higher memory clocks and bandwidth at CAS4 gives faster SPi32M calculation and 3DMark2001 score. On the other hand, as suspected, 3DMark03 & 3DMark06 don't care about it at all - results are the same...

In other words, there will always be a point where higher bandwidth catches up with low latency... Just a matter of figuring out the difference...

[coolaler]

high FSB and high frequency is better

[fhpchris]

Thanks for you comparison, Denny On AM2, situation is very similar...

I have done a little comparison earlier, for the purpose of some older thread, but I also did a fresh and quick test on my AM2 setup. I run 275x10=2750MHz @ DDR2-916 8-3-3-3.0 and DDR2-1100 8-4-4-4.0. FYI, my mainboard is volt-modded so actual voltage for DDR2 is 2.51v, measured with DMM. Therefore, ignore what AiBooster says - it simply shows what setting was chosen in BIOS.

DDR2-916 8-3-3-3.0

SPi32M --> 26:12.984
3DMark2001 & 3DMark03 ---> 30465 & 18514
3DMark06 ---> 4427


DDR2-1100 8-4-4-4.0

SPi32M --> 25:58.219
3DMark2001 & 3DMark03 ---> 30777 & 18517
3DMark06 ---> 4426

As you can see, as long as CPU speed is the same and we're trying to max out mems at particular dividers, higher memory clocks and bandwidth at CAS4 gives faster SPi32M calculation and 3DMark2001 score. On the other hand, as suspected, 3DMark03 & 3DMark06 don't care about it at all - results are the same...

In other words, there will always be a point where higher bandwidth catches up with low latency... Just a matter of figuring out the difference...

Thank you BA

Even on AM2 there is a point where lower latency just does not help as much as bandwith does...

I am impressed that even on single core the bandwith does matter.

[elmor]

Any comparison for 1:1, 4:5 and 2:3 for pi/3d?

[guess2098]

Any comparison for 1:1, 4:5 and 2:3 for pi/3d?

1:1 just.... to slow... DDR2 728 3-3-3?

[M.Beier]

1M I assume lower timmings is better, due faster accesstime, 32M as you've shown high mhz is better..

Can you do a 1M compare as well??

Coolaler, ras to cas 4