Very important note to TCCD users !

Folks,

As it it these days, every MoBo maker has decided to include fairly detailed memory timings in the BIOS. To complement that, we now have various chip specific BIOS's being released. Case to point: Modded BIOS's for the DFI NF4 series.

Most people say Weak Drive strength's are for TCCD and strong ones are for BH-5. I have been stuggling of late to eliminate memtest errors running 1:1 DDR600. A few minutes ago, I decided to try BH-5 drive strengths i.e Drive strength of 8 and Data Drive Strength of 1. Guess what? From 194 memtest #8 errors I dropped to zero errors on two passes! O.K its just a couple of passes, but I thought it would be nice to share this info.

Moral of the story:
Do your own testing every module is different! And, don't shy away from adapting BH-5 timings on TCCD. It may work

Hope this gives all those frustrated at not being able to run DDR600 on their TCCD some hope. Detailed test results (memtest only), tonight.



S-N

I look forward to seeing if your findings pan out. This could chang the way people go about adjusting timings.

The timings you can/will run will be very dependant on your CPU as well. Every core will want different timings to run the same speeds

Steve,

My 3200 Venice can easily hit 2.85GHz at 1.7V. However, I've clocked down to 2.7. A definite improvement over my 3500 NC. Just for preliminary testing purposes, my RAM is at 2.5-4-4-10-1T,2.9V.

So are you saying loose timings may not work and some CPU's actually prefer tighter timings?

Thanks,

S-N

I was able to get my XL's/TCCD to DDR580 with 2.5-4-4-8 and 3V, memtest #5 30 passes/0 errors (didn't do anymore and on #8 15 passes 0 errors (also didn't do more than that).

P.S. Super Nade I will try upping the Drive Strenght and see if I can do DDR600, but I will need to lower my CPU clocks since I can't go higher than 2650MHz , wish I had one of the LBBLE's.

The basic cas-trcd-trp-tras timings will be the same from CPU to CPU, but the other timings that the A64's have can change from CPU to CPU.

question for a memory noob? what does drive strength do? I've tried googling it, but can't find anything...

That's one thing I like about AMD64 processors and nVidia nForce4, all the controls you have in memory. Intel doen't give you much except the basics CAS x-x-x-x. Heck even my daughter's old Shuttle SK41G AMD socket A SFF can adjust the t command in BIOS.

Audioaficionado said:

That's one thing I like about AMD64 processors and nVidia nForce4, all the controls you have in memory. Intel doen't give you much except the basics CAS x-x-x-x. Heck even my daughter's old Shuttle SK41G AMD socket A SFF can adjust the t command in BIOS.

Click to expand...

my dfi nforce 2 board has quite a bit...i just dont know what they do

lowfat said:

my dfi nforce 2 board has quite a bit...i just dont know what they do

Click to expand...

Have you read this thread from the DFI forum?
http://dfi-street.com/forum/showthread.php?t=11397
It's for the NF4 boards, but there's probably a lot that hasn't changed.

Updates :

I still have errors but they were reduced by a huge amount. From 200+ errors on the first pass alone, I have 291 errors over 6 hours 47min. Nothing spectacular, but it still offers a hint that I may be right.

I have to tweak the timings a bit more.

Edit*
Tightening the timings (trp 4->3) and setting Drive strength to Weak (Another BH-5 setting) gave me 50% less errors!

Another update:

Reduced errors from 279 in 6 hours 23 min to 93 in 9 hours 47 min. This is turning out quite well indeed!

lowfat

what does drive strength do? I've tried googling it, but can't find anything...

Click to expand...

Check this out

http://www.dfi-street.com/forum/showthread.php?t=11397&page=1

Sometimes called driving strength. This feature allows you to control the memory data bus' signal strength. Increasing the drive strength of the memory bus can increase stability during overclocking. DRAM drive strength refers to the signal strength of the memory data line. A higher number means a stronger signal and is generally recommended for an overclocked module to improve stability.

Click to expand...

EMC2, http://www.xtremesystems.org/forums/showthread.php?t=65168

There are two drive strengths... what are typically called DRAM Drive Strength (or sometimes simply Drive Strength ) and Data Drive Strength.

They control the drive strength of two seperate groups of signals to/from the memory.

Drive strength itself has nothing to do per say with the actual voltage levels of the signals. What it controls is the slew rate of the signals - i.e. how much current is available to drive the signals from one state to the other (high-to-low or low-to-high).

Too weak of a drive strength limits your max OC because it takes longer for the signals to transition (change) from one state to another.

Too strong of a drive strength can cause ringing and signal reflections, which means it will take longer before the signals stabilize after a transition, which again limits your max OC.

There are also other factors that come into play (like termination resistor values, the structure of the output drivers in the memory chips, etc.) that determine the best drive strength for a particular setup.

In regards to what the numbers indicate in regards to slewrates - varies by MB... but for the DFI nF4, Bigtoe's info in the thread you linked tells you the answer in regards to which numbers are stronger and which are weaker drive strengths.

Click to expand...

this post has ****ed me off big time!

now all i can do is think of all the tccd ive run on my a64 systems that i might have been able to milk more out of!

interesting results though.... just wish they didnt send my mind flashing back to all the tccd ive had and sold

Ok I tried upping the Drive Strenght the other day but I forgot to post back, anyways it wouldn't work and it can't even get past the boot screen.

I guess not all TCCD are the same

I have a neat anecdote here. I have two sets of BH-5 memory, Mushkin and Corsair.

Playing with both last night, using the same motherboard, I found that each set likes different DRAM Drive Strength/Data Drive Strength values. My Mushkin practically *needs* 7/2 to run stably, the Corsair prefers 3/1.

With the same chip type in both sticks, I find that quite peculiar. It tells me that either the PCB, or the memory chip batch, has a larger effect on optimum Drive Strength values than the memory IC type.

thats interesting. i know mushkin was touting its pcb as one of its main selling points when i picked up my kit of the lvl 2 3500 black a couple years ago. i know when i was studying up on tccd the diffrent pcb used was a factor in what i purchased (although admitedly it was from what others were recomending/experiencing than my own in depth detective work)

which one of the kit performs better in the end?

Both made it to 270 MHz with memtest86+ stability, the mushkin given 3.7V in the BIOS (3.78V measured with a multimeter), the Corsair given 3.6V in the BIOS (3.68V measured with a multimeter).

However, past 270 MHz the Corsair seems to do better, I've got it running at 274 MHz right now, trying to get it stable. The Mushkin memory seems to be held back from stability above 270 MHz by two adresses, which are the only two that start erroring when I scale the memory higher. I scaled it to 280 MHz using 3.9V (measured with a multimeter), the adresses at 465.5 megs and 510 megs were the only two that errored, other than that the sticks ran clean for half an hour. The timings I ended up with for each stick were also slightly different, the Corsair memory saw a stability gain from a CAS 1.5 timing, and an additional stability gain from a tRAS timing of 3 - I ended up running it at the timings in my sig, 1.5-2-2-3-7-13. The Mushkin ran best at CAS 2, with a tRAS of 6, a tRC 8, and a tRFC of 16. And, as mentioned above, the two sets preferred different Drive Strength values. The Corsair memory also benefits in stability from an Increased DQS of 55, the Mushkin did not play well with any DQS values, Increased or Decreased. Other than that, the "optimal" timings for balanced stability and performance were pretty much identical between the sets. Neither would run a TRWT of 1.

To summarize, the Mushkin memory sees its gains peak at a higher overvoltage than the Corsair memory (~0.1V higher), but doesn't go as far due to what is likely a single memory IC that isn't "perfect". The Mushkin memory has 3052E stepped BH-5 chips, the Corsair has 310WE, and 322WF stepped BH-5 memory chips. The Corsair memory is a PC3500 spec part, the Mushkin PC3200. Both have had their heatspreaders completely removed.

You have to do what you have to to get TCCD Memtest clean in #5/6/8 - no matter how odd the settings. With TCCD I shoot for 300, 2.5-3-3-6 here, most recently with 2x512 Vitesta PC600.

But that's only the first step. Assuming the ram is even 3D stable here, to adequately clear Five-Mark (all four Marks and Aquamark), each may need tweaking in the Trrd-Trwt settings. I found I can get by 3DMark2005 at 2-2-2-2, but need 3-2-2-2 to get by the other Marks. It's usually the opposite with Winnies. To get by 30' of the IL-2 and SS2nd Encounter demo required 3-2-2-3.

Depends on the ram too. As I recall my PQI TCCD benched Five-Mark here at 2-2-2-2 without a hiccup. I know I didn't have to mess with the settings.

Don't be ashamed of loose settings too. Even at 3-2-3-4 I think I got a 28.131 SuperPI 1M time at 290x10, 2.5-3-3-6. Those are the numbers that count . It helps to be running a San Diego. BH-5 at 264x11, 2-2-2-5 netted me 27.962 at same clock, so that ain't bad at all for TCCD at 2.7 VDIMM .

Well well well

Its nice to see someone take this thread seriously. I figured it was dead an buried like my thread on the Athlon64 mem-controller.

hawtrawkr:
Mate, just because someone posts screenies of DDR600, doesn't mean they run it 24/7. What I am looking for is 24/7 stability. Which means 0 memtest errors and 0 Super Prime/Pi errors. It has been a very painful process thus far. I changed one parameter (tRCD) and got 1213 test 8 errors over 413 passes (13 hrs 27 min ). This I believe is due to temprature. I leave the AC off when I'm at school and my room gets to almost 90F. You are a pretty experienced and thorough fellow, why don't you run some tests and we can compare notes

RedDragonXXX:
You cannot just up the drive strength and call it a day . Possibly, all other timings have to be changed. Its not easy and you will have to spend weeks on it. Since I have no idea what correlations exist between the Drive strength(s) and other parameters, I have to resort to crude trial and error.

felinusz my friend
You have brought up another point (which deserves a thread of its own), namely, the relation between drive strength and PCB's. Rephrasing, one may ask, what timings needed to be changed if you move frome one PCB to another. (Nate or infinitevalence, please look into this when you revisit your JEDEC v/s BP tests). Another point you mention deserves more thought this apparent weak stick phenomena. Is it just me or has anybody else noticed that there is always one weak stick in a dual channel kit. How does this happen? What is exposing this weakness? Can this be fixed in the BIOS. I speak from experience because the "weak" stick on my DFI-NF4 works perfectly fine in my ASUS A8V ! Surely there is more substance to this matter ?

Clevor:
I can't boot into windows with CAS 3. Sometimes I can't even Boot into memtest (using DFI NF4 5/10-1 BIOS). Why is that?

I hope more people see this and run some tests.

More testing wiill be done tonight