Explanations of DFI NF3-250GB Bios settings/Ramtimings

[joedeman]

Nice Gundamit, I cant wait for the OCZ Powerstream. I did some Uwackme burn-in over the weekend and I was able to take 2-2-2-5 timings to 240 (both sticks in there 2x512) with no errors in memtest #5 for 3 hours and passed 5 hours of prime. Thats on a crappy 3.2 vdimm that under load goes to 3.08!!!!

 

[gundamit]

Joedeman, I should mention I've been running single sticks while testing. I realize you're going for a 2x512 set-up. I was going to get around to it myself after I converted this rig from my bencher to my gamer, but you piqued my curiosity. Both my BH-5 HyperX sticks (from the Outpost deal last month) match performance wise, so I just slapped them both in. I'm actually surprised they booted at 250x10 3.4v. Got into the OS but got a file corruption message as I tried to load a program. Going to run off some Memtesting before I go back in.

I'm sure these sticks will be monsters when I go dual channel Q1 '05.

 

[joedeman]

I have always been doing 2 sticks cuz I kinda lazy but doing the UwackMe burn-in with both of the sticks at the same time seems to work for me. Believe me, I have tried to punish these sticks with all I have running memtest #5 for 36 hours straight. When I first got this stuff I could not boot at 10x250 at all, Nothing but beeps. After doin the burn-in stuff I easily boot into 10x250 with just a little lesser timings. I am sure these outpost sticks have been used so it will not take us as long to hammer them to where we want them.

 

[stowenrat]

Just had to thank you Jess1313, I am using you recommended settings and they work great.

 

[joedeman]

JESS, Need some help man, the memory is just killin memtest, no errors at number 5 test at 255 at 2-2-2-5 with 3.3 vdimm, prolly can go higher but now I am just trying to get the system stable becuase at 10x250 1:1, I cant pass prime for 5 mins. Vcore is at 1.6, chipset is 1.7, I know that the CO revisions like I have kinda suck but I would have to say I should be able to get 2500 or 2600 with this chip. I have a VENUS12 on the CPU which does a great job. I think I can do more vcore but I would like to keep that a low as possible. Right now my cpu idles at 26C and load does 39-40C. Any suggestions to take the cpu at little higher? It is definitely prime stable at 2400 but 2450 or 2500 it aint. Please help. Thanks in advance. Joe

 

[gundamit]

J Vcore is at 1.6

Try 1.5v x 110% in bios = 1.65v. Check temps, but should be okay as long as you stay below 50c full load. If you don't get better stability, you just might have an unlucky CPU.

 

[joedeman]

Thanks Gun, Ill try that tonite. Yea , I have seen that the CO rev. have not been doing that great overall but oh well, I aint buying another chip until the 939s have become mature in a year or so and DFI has a board for that.

 

[jmke]

C0 version does not do well with high FSB speeds unfortunately

 

[pazza316]

Two Sticks of BH5 Question

Hi All,

I am using 2 sticks of 256mb BH5 what settings should I use with two sticks to allow the FSB of the board to go over 270 stable. The memory is running at 243 with the use of a divider.

 

[bwanaaa]

thanks!

Jess1313 thank you very much for your post of BH-5/BH-6 prefered timings. With a single stick of mushkin222 lvl2, memtest passed up to 255mhz. Putting the divider to slow the ram let me crank the cpu to >2.5ghz stable. But at 1:1, i could not exceed 10x240-2-2-2. i tried different voltages etc , nothing helped, i gave it up until i found this thread. So many changes are recommended from auto, i can only imagine how many weeks it took you to get them.

Now I am prime stable at 10x250. Thats 300mhz gained but more importantly, a ton of bandwidth that makes things go much more smoothly.

Thank you.

 

[OldDarux]

Help with DFI LanParty UT nF3 250GB overclocking

Hi All,

My system is running ok, but I'm not sure if my settings are ok. Is there anything else I can change? Can someone help me optimize it?
I have:
Athlon 64 3200+, DFI LanParty UT nF3 250GB, 1GB OCZ (3-4-4-8) PC3200.

Here is what I have now:

FSB Bus Frequency: 244
AGP Bus Frequency: 67
LDT/FSB Frequency Ratio: 3
CPU/FSB Frequency Ratio: 10
CPU VID Control: 1.425
CPU VID Special Control: 110%
Chipset Voltage Control: 1.7
AGP Voltage Control: 1.6
DRAM Voltage Control: 2.8

DRAM Frequency Set: 180 (9/10)
CPC: Disabled
tCL: 3
tRCD: 4
tRAS: 10
tRP: 4
tRC: 15
tRFC: 17
tRRD: 00
tWR: 3
tWTR: 2
tRTW: 4
tREF: 4708 166 (1.95us)
tWCL: Auto
DQS Skew: Increase
DQS Skew Value: 64
DRAM Drive Strength: Level 1
Max Async Latency: 8
Read Preamble Time: 6
Idle Cycle Limit: 32
Dynamic Counter: Auto
R/W Queue Bypass: 16
Bypass Max: 4
32 Byte Granulation: Enable (4burst)


Thank you in advance for your help
DJ

 

[sizzla]

Help me with TCCDs

Hi Folks!


First of all, sorry for my bad English, school was long ago and i did not practice very much ;-)


Thats why i did not read the whole thread ( would last hours for me ) , so i thought i will just ask u guys..... ;-)

I got the following System:

AMD 3400+

DFI UT 250GB

Gainward 6800 Ultra

Corsair XMS XL with TCCDs ( 1 GB )


but i think, my rig is much too slow....its not working as fst as others with nearly the same setup!


I did try the Setting from the first page, but there were none for TCCDs.....can somebody please give me his setting for TCCD if they work fast?

atm i Got:

3400+@2640 Mhz

FSB/HTT is 240 the RAM running 1:1 ( also 200Mhz in Genie Settings )

Timings:

Standard ones: 2-3-3-6

Row Cycle Time: 7

Row Refresh ( Trfc ) : 11

RAS to RAS : 0

Write Recovery (Twr ) : 3

Write to Read delay ( Twtr ) : 2

Read to Write Dealay (Trtw ) : 4

Refresh Rate: 200 Mhz 3,9 ns

Read Write Queue Bypass: 16x

Cotroller Mode: 64bit

ECC : off

1T: enabled

Baypass MAx: 7x

Max Async Latency: 7ns

Read Preamble: 5ns

Idle Cycle Limit : 16 clcks

Dynamic Counter : enabled

MEMCLK Freq : 200

Any ideas how i can improve the Performance?

 

[ToastedToad]

in order that you listed them:

2.5,3,3,10
8
16
5
3
2
5
2592
8x
64 bit
off
enabled
7x
8ns
5ns
16clcks
enabled
200

 

[Tony420]

I have been using jess's timings with my 2x512 bh-5 and I have found that the suggested Read to Write delay(tWTR)=1 limits FSB alot. With 1x512 at 245fsb it would be fine, then add another 512 and 10thousand errors+...no voltage or anything would get rid of the errors untill I changed the (tWTR) to 2. Then the errorst magically dissapeared. I am now at 250fsb stable and about to test higher

I have also noticed not setting the DQS skew value and leaving it at 0 improves overclock

 

[ken1307]

hi everyone, my 1st post here
well, Ive been followed Jess1313's guide ( cheers m8 ), rite now Im stuck at 240@2-2-2-10 with 3.1v, its an BH-6 modules so Im pretty satisfied. However, just one question here, why you guys only use test #5 of Memtest? My ram can run that test #5 for hours without any errors but it fails like instantly at test #6
Cheers

 

[Clevor]

Jess1313, I have Crucial Ballistix memory. Do you have maximum stable achivable timings for Micron?

I have this DFI board running for all of one night. Figured out how to change VCORE and get CAS settings, and to keep it simple, I run 200(1:1) on the ram and just up the fsb on the 3400+ Clawhammer I got.

I tested 2x512 MB Ballistix PC4000 (you heard me right), and they rock on this board! They do:

238, 2.5-2-2-5 @ 2.6 VDIMM
244, same @ 2.7
248, same @ 2.8
250, same @ 2.9
252, same @ 3.0

I am running 250x10, 2.5-2-2-5 rock stable now. VDIMM shows as 2.87 in Windows. 1T of course as I have a CO step. SuperPI 1M time is 34 sec, 16M is 14'15".

I am running 4xLDT although I've completed benches OK at 5x. I understand even 4x is a little high here.

I did a comparison bench at 230 of 2.5-2-2-5 versus BH-5 at 2-3-2-5 and the Ballistix beats it across the board in the four 'Marks, except 2003.

I have two sets of Ballistix PC3200, early and late sets that are even faster than the PC4000, so I can't imagine how they will top out. Main thing to notice is at 250 fsb, I don't need more than 2.9 VDIMM . The ram hits a wall at 252; don't know if it's the ram or fact COs won't do high fsb. More VDIMM don't help above 252. Ram doesn't like 2-2-2-5 down low nor will it get higher at 2.5-3-3-6.

On a P4 rig, the ram did 225, 2-2-2-5 and up to 260 at 2.5-3-3-6, but I'm not complaining the ram won't run anything but 2.5-2-2-5 on this board .

 

[Clevor]

I don't understand this HTT thing. You are supposed to find max HTT, max CPU size, and max fsb on the ram, and use some tables to get all three working together. But a high HTT doesn't provide a performance increase so what's the big deal? Maybe that you have more latitude with combinations to hit your sweet spot???

Is high HTT due to the CPU? I have a 0411 CO step 3400+ Claw with 1 MB L2. I tested max HTT and quit at 335 since that seemed high enough. I could bench 16M SuperPI and 3DMark2001 here - at 3xLDT.

I settled for 320HTTx9, 2560 mhz, and 240 on the ram. But I got 1 sec less in SuperPI 1M versus a straight up 250x10, 250HTT, 200(1:1) on the ram. It's looking to me to be old school stuff: higher fsb on the ram, the faster the rig (even with a smaller CPU!).

I know HTT can be seen in CPU-Z, but is there any software that can accurately show you the fsb of your ram in Windows when you are running all the HTTxmulti combinations?

 

[masterwoot]

So when's this awesome, unbelievable thread going to get sticky status?????

 

[WinXP]

This need to come back up, with the useful info in it. Maybe someone will be getting one for Xmas an need the info.

 

[OnDborder]

Hey Jess, good job.
Did you happen to come across some settings for the tccd chips?