TCCD DDR600+ DFI nF4 Ultra/SLI guide

[wai-wai]

gd guide~~!!
i love it >3<

[Crosis]

Hi there,
First thatks for that guide It's my first post here, so please forgive me my English Everybody are sharing with their experiance, so I'll do the same I've tried to reach stable 2,5-3-3-6, but I have no idea how to do it I've tried settigs that kakaroto posted here, I've tried also things he called "when those settings are not stable" and nothin' I have closed case, zalman F2 (low) on the window of case, 2 80mm Thermaltakes on the side and zalman 7000b (low). I'm trying to reach that score in "everyday conditions". Increasing vddr does nothin', better cooling the same. Now I'm testin' 2,5-4-3-6 in windows memtest. It's 800% without error so far. If that settings will be stable It will be ok, but I'm afraid that I would not reach highest frequency (310-315) with nice timings (2,5)

EDIT:
I've tried 2,5-4-3-6 on 310 and had errors in #5 so I increased Trp to 4 and #5 was passed... but errors in #8

[kakaroto]

hi kakaroto,

thanks for the posting and the screenshots.


i have more information coming after a weekend and today tweaking and benching and priming without end Btw. i have a winnie week 42 3500+, dfi ultra-d, 310p bios, the two OCZ 3200 rev2s...X-120 heatsink

.......................................
.......................................
.......................................

It MIGHT be that these highly tweaked settings are only good for CPUs at stock speed....eg. you can run your memory at those speeds and the CPU can handle the tight latency if its not overclocked....and if the CPU is already on the edge (like mine w/ 2600mhz)...it just cannot handle the latency ?

Your are totally right with that!, Memtest86++ test5 is far from 100% stable in windows (Prime95 long time stressing)
To get your memory fully stable, I think you must run the STD for some passes (~ 1 hour) and test 8 for a long time.

The memorycontroller is implemented in the core, but I think when the CPU is on the edge, this must have influence to the memorycontroller. But I haven't tested that so..

Hi there,
First thatks for that guide It's my first post here, so please forgive me my English Everybody are sharing with their experiance, so I'll do the same I've tried to reach stable 2,5-3-3-6, but I have no idea how to do it I've tried settigs that kakaroto posted here, I've tried also things he called "when those settings are not stable" and nothin' I have closed case, zalman F2 (low) on the window of case, 2 80mm Thermaltakes on the side and zalman 7000b (low). I'm trying to reach that score in "everyday conditions". Increasing vddr does nothin', better cooling the same. Now I'm testin' 2,5-4-3-6 in windows memtest. It's 800% without error so far. If that settings will be stable It will be ok, but I'm afraid that I would not reach highest frequency (310-315) with nice timings (2,5)

EDIT:
I've tried 2,5-4-3-6 on 310 and had errors in #5 so I increased Trp to 4 and #5 was passed... but errors in #8

First thanks everyone for using the guide, I'm happy it also works on most of the people.

What is your max stable with 2.5-3-3-6 1T @2.6~2.7v. If you get a speed working stable you can work further with that speed.

[flexy]

The memorycontroller is implemented in the core, but I think when the CPU is on the edge, this must have influence to the memorycontroller. But I haven't tested that so..

Probably old news...but tweaking various "memory" settings has quite a significant influence on how high you can overclock your CPU.

Maybe some people who run at a wall w/ their CPU should just relax some memory timings and might get a suprise because their CPU might overclock far better.

It might be old news for the Pros but i never saw it from that perspective.....eg. if you overclock CPU you think HTT, Core Voltage....but never really the memory timiings. Yesterday in some tests (OCCT, Prime) etc. i came close to 2700 with my winnie which had a hard time before reaching eg. 2610...

But..of course there are always trade-offs...if you increase latency and timings to get more mhz on the CPU you might lose a similliar amount on the memory (because the timings)...and..after all....the overall performance is the same.

[Ackbar]

I noticed that you placed Winchesters, Newcastles and Venice all in the "low to mid-range" of CPUs. While Clawhammers are in the high end, but you didn't include 3500+ Clawhammers in there. What are your thoughts (if any) on the quality of the memory controller of the 3500+ Clawhammers?

Great guide! I had no idea the number of Winchesters with a poor memory controller was so high!

6/10!!!

[WiCKeD]

Awesome! This is just what I was looking for!!!

One thing I would say is that TCCD gets hot at any voltage. Active cooling is definitely advisable.

[zeebs]

Props on a sick ram guide. Thanks for taking the time to make this, it should assist a lot of people in reaching their overclocking goals. I will let you know what it does for me when I get my tax return -> san diego/venice

[Sir Lancelot]

Hi there,

First of all, thank you for this guide!

But I have a problem here, those settings seem to work fine @ 1:1, but with the 180 divider, I cannot have my mems stable at 282MHz.


My rig is:
A64 3000+ @ 3100MHz, S&M 1.51 stable (Prommy cooled)
DFI NF4 Ultra-D
OCZ Powerstream 520W
ATI X850XT PE
4x 160 @ RAID0
and of course, 2x 512 GsKill 4800LA




Any help will be much appreciated



Thankx

[Crosis]

What is your max stable with 2.5-3-3-6 1T @2.6~2.7v.

Really? I haven't tried yet :P I've just tried 275 and 300.

If you get a speed working stable you can work further with that speed.

In Poland we have a small competition in SuperPi 16M, and I have some results to defeat :P So I'm searching for the best timings at ~315.

[Tutti-frutti]

Nice guide again
I have tested my Gskill TCCD too, with a lot of different settings.

My settings are almost the same as this guide now, but I have 2 different vallues:

Tref:
A value of 4708 will not result any error in Test #5 of memtest, I have 40 passes without an error. But when I run the STD test, 2-6 errors will pop up in test 3 and 4 each round. Even on a frequency of 260 mhz (haven't test lower than that).
When I changed this value to 3072 cycles, the errors are gone
This is a little bit strange, because it's faster and, in theory, it will result in less stability. I have talked about this with Kakaroto, and he saw the same problem. It seems TCCD loves Tref 3072c more than 4708c.

TRC:
The other differrent value is the TRC. Reducing this value to 8 (guide: 12) doesn't hit the stability of my system, but it gives a better bandwitdh (3099 MB/s --> 3297MB/s in memtest on ddr600)

So this will be my optimized memory configuration:

BIOS Settings

Vddr: 2.7-2.9v (2.9volt performs best here)
Tcl: 2.5
Trcd: 3
Tras: 6
Trp: 3
Trc: 8
Trfc: 14
Trrd: 2
Twr: 2
Twrt: 2
Trwt: 2
Tref: 3072(200MHz 3.9us)
Twcl: 1

Bank int.: Enable
Skew control: 255 +
Drive strength: 7
Data strength: 4
Max. Async: 8ns
Read Preamble: 5ns
Idle Cycle: AUTO
Dyn. Counter: Enable
R/W Bypass: 16x
Bypass Max: 7x
32Bit Gran.: Disable

[kakaroto]

Nice guide again
I have tested my Gskill TCCD too, with a lot of different settings.

My settings are almost the same as this guide now, but I have 2 different vallues:

Tref:
A value of 4708 will not result any error in Test #5 of memtest, I have 40 passes without an error. But when I run the STD test, 2-6 errors will pop up in test 3 and 4 each round. Even on a frequency of 260 mhz (haven't test lower than that).
When I changed this value to 3072 cycles, the errors are gone
This is a little bit strange, because it's faster and, in theory, it will result in less stability. I have talked about this with Kakaroto, and he saw the same problem. It seems TCCD loves Tref 3072c more than 4708c.

TRC:
The other differrent value is the TRC. Reducing this value to 8 (guide: 12) doesn't hit the stability of my system, but it gives a better bandwitdh (3099 MB/s --> 3297MB/s in memtest on ddr600)

So this will be my optimized memory configuration:

BIOS Settings

Vddr: 2.7-2.9v (2.9volt performs best here)
Tcl: 2.5
Trcd: 3
Tras: 6
Trp: 3
Trc: 8
Trfc: 14
Trrd: 2
Twr: 2
Twrt: 2
Trwt: 2
Tref: 3072(200MHz 3.9us)
Twcl: 1

Bank int.: Enable
Skew control: 255 +
Drive strength: 7
Data strength: 4
Max. Async: 8ns
Read Preamble: 5ns
Idle Cycle: AUTO
Dyn. Counter: Enable
R/W Bypass: 16x
Bypass Max: 7x
32Bit Gran.: Disable

Thanks Tutti-fruti, this finding is very nice

[ben805]

I thought the Bank Interleave need to be disable for stability?

[kakaroto]

I thought the Bank Interleave need to be disable for stability?

Interleaving allows banks with RAM to alternate their refresh and access cycles.
If 2x dimms (double sided) is used (2x 512), I recommend Interleaving mode. (4-way)
When you use single sided dimms (2x256), I recommend disable Interleaving mode (2-way) for best functionality.

[Sir Lancelot]

Kakaroto: and how about when using the 180 divider?...Mine works more or les fine @ 1:1, but with the 180 divider, it's a no go...



Thankx

[eva2000]

for BH-5 2x 512MB interleaving caused instability but for 2x 512MB TCCD they were more stable with interleaving

[fareastgq]

So if you use all the settings mentioned above in tuti fruti's post and your comp posts at 230 fsb 2x512, but reboots after it gets into windows and loads everything, how do you fix that? This might help some of u too, run your ram 2t, this helped me out, but I think I corrupted my inital os install so I have to reformat now :/ On a differnt note, anyone know what I can do for 1t settings and stopping the reboot? I was using 310p bios for this, I'm switching back to 2-17 and trying the settings now after a fresh install. I have seen you guys talk about ram timings, but what about the other voltages for the MB like LDT and chipset etc...here goes 1 long night of testing...

[mdzcpa]

Tref:
A value of 4708 will not result any error in Test #5 of memtest, I have 40 passes without an error. But when I run the STD test, 2-6 errors will pop up in test 3 and 4 each round. Even on a frequency of 260 mhz (haven't test lower than that).
When I changed this value to 3072 cycles, the errors are gone
This is a little bit strange, because it's faster and, in theory, it will result in less stability. I have talked about this with Kakaroto, and he saw the same problem. It seems TCCD loves Tref 3072c more than 4708c.

Same experiences here. I run all of my GSkill DFI combos with a Tref of 3072.

[Onepagebook]

Same experiences here. I run all of my GSkill DFI combos with a Tref of 3072.

When I run super pi, I set 100, 1.95
3072 can be normal using purpose.

[mdzcpa]

When I run super pi, I set 100, 1.95
3072 can be normal using purpose.

Thanks for the tip, I'll give that a try for benching

And, yes, the 3072 is for 24/7 use

[chew*]

Great guide thx I just purchased 2x256mb sticks of F1 LA and ill be using this guide to try t hit 8x300. very informative

[eva2000]

Thanks for the tip, I'll give that a try for benching

And, yes, the 3072 is for 24/7 use

my 2 sets of 2x 512MB Gskill PC4400LE 431 only hit 290mhz 2.5-4-3-7 1T at 2.9v (2.7 and 2.8v errors) on DFI NF4 SLI-D with 310 bios in memtest and 3072 was most stable, with new 325 beta bios 3120 TREF was more stable

[kakaroto]

Kakaroto: and how about when using the 180 divider?...Mine works more or les fine @ 1:1, but with the 180 divider, it's a no go...



Thankx

That a mobo bug I guess. I have troubles with high HTT with MP8 and 9.

[kiwi]

I've mainly used memtest and prime. Is STD some memory stability test? Link to download? Couldn't find it in google.

What about chipset and LDT voltage? Any suggested values? 1.3 and 1.7 is nice I guess.

[kakaroto]

I've mainly used memtest and prime. Is STD some memory stability test? Link to download? Couldn't find it in google.

What about chipset and LDT voltage? Any suggested values? 1.3 and 1.7 is nice I guess.

Noooo STD is a testing procedure in Memtest86++. You can select the test with "test selection"

Ik use 290MHz HTT with 1.2 LDT / 1.5v chipset.

[Jesse71]

Thanks for the great guide, I have just received my Gskill GBLA today

Tests for this week-end on my DFI Nf4 Ultra