1T VS. 2T on Nforce 680i

[Dumo]

Thanks Franc

Heres quick comparo @ 3-3-3-autos/704 and 4-4-4-autos/1006 with Team Xtreem 6400C3....





3D...

[ryaan]

so 1t is SLIGHTLY better.

[Dumo]

Probly so

It needs more testing with higher cpu mhz. like 5Ghz and up for 3Ds.

Also if some ddr2 5400 with default cas3 can run 1T with this board....it would be nice

[irev210]

nice review. I am having one hell of a time with quad core and this board.


This board definitely is strange, it needs some bios help.

[fhpchris]

Why are the runs so slow?

[NickS]

680i's ram performance is just sad.

Nice effort + comparison though.

[fhpchris]

I know, I do not know what Kunaak was smoking when he said that 975 and 680i have similar chipset latencies...

[zakelwe]

Why are the runs so slow?

There is derfinitely something up with 680i and Superpi on the longer runs, I was running a 32M the other day and I was minutes behind the Intel 975 at the end for same settings. In 3d though it is pretty close. Weird.

Regards

Andy

[Dumo]

Yep, its a little slow..

[bachus_anonym]

Thanks for starting this thread, Dumo

I hope you don't mind me adding some of my results? I did not want to create a new thread for the same analysis, so I joined the fun here

Enjoy

Testbed onfiguration
Intel Core 2 Duo E6600, cooled w/ Intel stock cooler
eVGA 122-CK-NF68-AR (nForce 680i SLI), BIOS P20 official
Corsair TWIN2X2048-8888C4DF 2x1024MB
EVGA 7800GTX KO @ 490+40/1390, ForceWare 84.21
Windows XP SP2 with all current updates.

--- DDR2-858 4-3-3-3.0-1T (Tras-Trp-Trcd-Cas-Cmd) ---
SPi32M - 13:35................... 3DMark 2001 - 48695......... 3DMark 03 - 22166............. AM3 - 21056 CPU


Everest Ultimate 3.5 - more


--- DDR2-858 4-3-3-3.0-2T (Tras-Trp-Trcd-Cas-Cmd) ---
SPi32M - 14:12................... 3DMark 2001 - 48435......... 3DMark 03 - 22133............. AM3 - 20299 CPU


Everest Ultimate 3.5 - more


... and for comparison, higher memory clocks @ 2T

--- DDR2-938 4-3-3-3.0-2T (Tras-Trp-Trcd-Cas-Cmd) ---
SPi32M - 13:59................... 3DMark 2001 - 48556......... 3DMark 03 - 22139............. AM3 - 20646 CPU


Everest Ultimate 3.5 - more


--- DDR2-1206 4-3-4-4.0-2T (Tras-Trp-Trcd-Cas-Cmd) ---
SPi32M - 13:32................... 3DMark 2001 - 48683......... 3DMark 03 - 22189............. AM3 - 21272 CPU


Everest Ultimate 3.5 - more


--- DDR2-1306 4-4-4-5.0-2T (Tras-Trp-Trcd-Cas-Cmd) ---
SPi32M - 13:27................... 3DMark 2001 - 48737......... 3DMark 03 - 22171............. AM3 - 21409 CPU


Everest Ultimate 3.5 - more



So, 1T is quite fun, isn't it? Although I can boot to Windows and run e.g. SPi 1M @ DDR2-900 5-3-3-3.0-1T, it was not very stable for most other stress benchmarks. Also, getting even DDR2-858 8-3-3-3.0-1T to be ORTHOS or HCI Memtest stable, borders a miracle

I will do some more testing as I hope to somehow get DDR2-900 5-3-3-3.0-1T to be at least SPi32M stable? So far, no joy But I really like the kick 1T gives in this particular benchmark!

[Dumo]

Yep, 1T is fun ..Really nice clock on those 8888 Bachus ...Reminds me of 6400C3 which imo is a 'free for all fsb/Dram' memory.

[OBR]

My little test ... on eVGA nF680i SLI + Corsair XMS2 8500C5 + C2D@3600MHz, difference between 1T and 2T is +- zero ...



Dominators are on the way ...

[EnJoY]

Looks like the next challenge is to get 1000MHz 1T with 4-4-4 timings or less, and you'll have the ultimate performance that even DDR1350+ can't match.

I like this lower latency stuff, it's very Athlon-ish.

[TaPaKaH]

question:
How many Mhz in mem OC (same volt, same timing) do you lose setting 2T -> 1T on average ???

[EnJoY]

question:
How many Mhz in mem OC (same volt, same timing) do you lose setting 2T -> 1T on average ???

From the results I've seen, usually around 300MHz or more lost to 1T.

[bingo13]

My little test ... on eVGA nF680i SLI + Corsair XMS2 8500C5 + C2D@3600MHz, difference between 1T and 2T is +- zero ...


Dominators are on the way ...

The synthetic benchmarks will not show that big of a difference, it shows up in applications.

[guess2098]

well..... becasue you guy need to try 4-3-3 @ 1200
mine HZ runs 4-3-3 @ 1200 2.5v

[lopri]

800MHz/3-3-3-1T was faster than 1100MHz/4-4-4-2T from my testing. I couldn't run many tests, however, because 1T wasn't stable enough.

[lopri]

Why are the runs so slow?

http://img292.imageshack.us/my.php?i...dr8001tjx0.png

Once you get 1T working, it's not slower than 975X.

[EnJoY]

well..... becasue you guy need to try 4-3-3 @ 1200
mine HZ runs 4-3-3 @ 1200 2.5v

Well that's some nuts G.Skill you've got there Denny. But hardly what the average forum goer posesses.

[Lestat]

well..... becasue you guy need to try 4-3-3 @ 1200
mine HZ runs 4-3-3 @ 1200 2.5v

4-3-3 ?????? oh come on now man thats pretty outrageous for D9GMH
i never doubt you guess2098 but thats insane for gskill HZ's
unless you have the PC8000 HZ's then thats D9GKX

[guess2098]

i think all D9 can do 4-3-3 from 1000 to its max (GCT, GMH, GKX)
mine HZ is 1000 GKX tho

[Dumo]

4-3-3 1t