having issues with TCCD on your DFI NF4 board?...heres some help

Ok like the title says this is for anyone having trouble getting their system stable that is running TCCD ram on their DFI NF4 board wether it be the ultra-d or the sli-d. Try these specific bios memory options. I compiled these timings from several posts made by DFI reps on other forums...enjoy. Mods i didnt put this in the memory section for a reason, this problem is almost specific to the DFI mobos so its not a general memory problem.

Gskill pc4400Le TCCD timings for DFI lanparty NF4 ultra-d/sli-d

@ DDR 400

Vddr: 2.7v
Tcl: 2
Trcd: 3
Tras: 6
Trp: 3
Trc: 9
Trfc: 12
Trrd: 2
Twr: 2
Twrt: 2
Trwt: 2
Tref: 4078 (166MHz 1.95us)
Twcl: 1

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

Gskill pc4400Le TCCD timings for DFI lanparty NF4 ultra-d/sli-d

@ DDR 550

Vddr: 2.7v
Tcl: 2.5
Trcd: 3
Tras: 6
Trp: 3
Trc: 9
Trfc: 12
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: 7ns
Read Preamble: 5ns
Idle Cycle: AUTO
Dyn. Counter: Enable
R/W Bypass: 16x
Bypass Max: 7x
32Bit Gran.: Disable

With these settings I got high performance and stability. The Tref (Refresh interval) can set to 4078 (166MHz 1.95us) for better performance.

If the settings mentioned above prove not to be stable, you can change the following:

- Max. Async. to 8ns
- Lower the refresh interval (Tref)
- Trrd to 3
- Idle Cycle at 128/256

When the memory works fine, then the problem must be the memory controller. You should disable the Command per Clock (CPC).

Gskill pc4400Le TCCD timings for DFI lanparty NF4 ultra-d/sli-d

@ DDR 600

Vddr: 2.7v
Tcl: 2.5
Trcd: 3
Tras: 6
Trp: 3
Trc: 9 (12)
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: 2 (4)
Max. Async: 8ns
Read Preamble: 5ns
Idle Cycle: AUTO
Dyn. Counter: Enable
R/W Bypass: 16x
Bypass Max: 7x
32Bit Gran.: Disable

The Tref can be tuned higher to e.g. 1.95us for better performance. Tref=3072 is 200MHz 3.9us refresh interval. You can also set Trfc to 12.

When those settings are not stable, you can change the following:

- Max. Async. to 8ns
- Lower the refresh interval (Tref)
- Trrd to 3
- Twr to 3
- Trwt to 3
- Idle Cycle at 128/256

Always try to run at CL2.5-3-3-6 1T to keep the performance at a high level. But if there is no other way to get this stable you can raise the Trcd to 4.

also keep in mind...
Test known stability increasing factors

- Data strength 2 is the best value for stability, I recommend not increasing or decreasing it. But always test which setting is best for you
- Drive strength 7 is the best value for stability, I recommend not increasing or decreasing it.
- Always set the bank interleave to enable when you are using two modules. This will increase the performance and stability.

Definitely a nice thread, which will at least give me a point of reference, as my Ultra-D and 3000+ Venice are getting here on wednesday, which I will be pairing up with my GSkill PC4400LE 431 for some real fun, new BIOS 510-3 is here just in time too.

cornelious0_0 said:

Definitely a nice thread, which will at least give me a point of reference, as my Ultra-D and 3000+ Venice are getting here on wednesday, which I will be pairing up with my GSkill PC4400LE 431 for some real fun, new BIOS 510-3 is here just in time too.

Click to expand...

dont use the 510-3 use the 510-1 its memory table is better suited for TCCD

cell_491 said:

dont use the 510-3 use the 510-1 its memory table is better suited for TCCD

Click to expand...

I've heard that even though 510-1 is more "suited" for TCCD, there have been many people getting simliar or better OC/performance results with 510-3...i dunno, I'll give both a shot, cus knowing me, I'll quickly lose track of the number of hours I'm sitting infront of my monitor tinkering wiht my new toys.

cell_491 said:

also keep in mind...
Test known stability increasing factors

- Data strength 2 is the best value for stability, I recommend not increasing or decreasing it. But always test which setting is best for you
- Drive strength 7 is the best value for stability, I recommend not increasing or decreasing it.
- Always set the bank interleave to enable when you are using two modules. This will increase the performance and stability.

Click to expand...

In regard to drive strength, odd numbered settings are considered "weak" and even are "normal". Drive strength should be weak for TCCD, normal for most everything else.

Not that you should be using TCCD on an A64 in the first place if there's a choice in the matter.

ashmedai said:

In regard to drive strength, odd numbered settings are considered "weak" and even are "normal". Drive strength should be weak for TCCD, normal for most everything else.

Not that you should be using TCCD on an A64 in the first place if there's a choice in the matter.

Click to expand...

i would like to think that the DFI technicians from where i got this info would actually know what they are talking about. Also TCCD is awesome for an A64 my shit hits ddr600 no problem at those specified timings at the low low vdimm of 2.7v. at 264 x11 my system gets 7559mbps memory bandwidth (sandra)

Did I say they didn't? I just explained why it was right. ^_^

Well at least without getting too technical. Anyway, G.Skill 4400 is nice stuff and all, but it's $240. Bandwidth and memory speed concerns belong on Intel, they're irrelevant on A64. You could be running 2-2-2 for half the price, but instead you have marginally worse performance for significantly more cost.

ashmedai said:

Did I say they didn't? I just explained why it was right. ^_^

Well at least without getting too technical. Anyway, G.Skill 4400 is nice stuff and all, but it's $240. Bandwidth and memory speed concerns belong on Intel, they're irrelevant on A64. You could be running 2-2-2 for half the price, but instead you have marginally worse performance for significantly more cost.

Click to expand...

i paid $180 for my stuff on ebay and besides vx and some BH-5 there isnt much ram that can do ddr600 @ 2.5-3-3-6

You're missing the point...it's not that the stuff isn't good or anything. It's that what it's good AT is misplaced for A64. It'd kick total ass on a P4. But on A64 you can get something that costs about half as much, and works better. Memory clock speed means very little to an A64, but it means the world to a P4.

ashmedai said:

Not that you should be using TCCD on an A64 in the first place if there's a choice in the matter.

Click to expand...

It's not too often that I blatently disrespect someone around here, but crap man.....please, do your reading ahead of time and learn the advantages, benefits, and pitfalls of different solutions, in this case...different types of DDR ram for an A64. It is true that you SHOULD be more interested in running tight timings on most AMD setups, but consider this...even the highest end memory out there aimed at tight timings (OCZ Gold VX and Mushkin's Redline come to mind) are only gonna (typically) take you to a 1:1 HTT of 260-270.....what if that isn't enough. I don't have the cash to drop on an FX-55, so I settled with a 3000+ Venice. Are you telling me that if I have the choice, I should go for ram that at the MOST (in a 1:1 scenario) is going to take me to about 270x9, for a clock of 2.43GHz??? Not quite, consider this then, what if my chip is capable, and I wanna push 'er farther then "just" 270HTT? Most available (newer/higher quality) TCCD is going to at LEAST take you to a 1:1 HTT of 300, which will/would have already taken my little 3000+ up to 2.7GHz. Depending on how much you DO want to drop on your ram, you could end up with some nice Mushkin, OCZ, or GSkill TCCD that'll be willing to scale much higher then 300MHz 1:1...giving someone with a lower cpu mulitplier a LOT more headroom to stretch their chips legs.

Dont get me wrong, I'm not completely downplaying all other options. If I had the cash, I WOULD be running an FX-55 right now, and therein lies the fun of this whole game. The FX-55 paired up with some Mushkin Redline or OCZ VX PC4000 will most likely get to a 1:1 HTT of about 260 or so...and guess what, then the unlocked multipliers on the FX actually come in handy, because you aren't limited to a stock multiplier. If you've got the cooling to handle the cpu clocks, having a 3800+ or faster A64 with some nice VX ram will pay off greatly, because that "measly" 260-270 HTT no longer is that much of a hinderance.

That being said, if an individual ends up with a particularly stellar A64 with a VERY capable memory controller, 1:1 HTT's of 330+ can be achieved...and paired with some nice cooling and a "lower end" chip.....you get not only stellar CPU clocks, but ridiculous memory bandwidth to boot.

So....depending on which way someone chooses to go with their cooling/cpu/mobo, their memory preference and options will change drastically, but to simply say "Not that you should be using TCCD on an A64 in the first place if there's a choice in the matter" is one of the most ignorant things you could have chosen to say in this particular matter.

It'd kick total ass on a P4

Click to expand...

Now I really know you haven't done a LOT of reading. Very often, a lot of TCCD won't scale to it's intended clocks in many Intel configs. Not to say that it wont happen, but TCCD was originally designed and released to go hand in hand WITH A64 users who are able to push their setups to higher and higher HTT's (with the ability to drop their cpu multipliers accordingly) and take advantage of the capability of the memory. I've seen plenty of happy Intel users when it comes to TCCD ram, but the truth is that more often then not, TCCD does NOT "kick total ass" on a P4.

Memory clock speed means very little to an A64

Click to expand...

I'm not gonna get really "into" this one...but I want to issue you a challenge. Head on over to www.xtremesystems.org and start up a thread in "Xtreme Bandwidth" saying that memory speed means very little to an A64, and how if given a choice, A64 users shouldn't even be looking at TCCD ram, tell me what kind of response you get. It's true that raw memory clocks and bandwidth are far more important to an Intel setup, but what you're essentially saying is that it doesn't matter at all to AMD. The gains might be traditional and typically far less, but they ARE still there, and do play a role in day to day, game, and benchmark performance.

I hope that you might have learned something here, whether it be related to how certain available memory plays with a particular A64 cpu, or about reading before you speak, either way...good day to you, class dismissed.

ashmedai said:

You're missing the point...it's not that the stuff isn't good or anything. It's that what it's good AT is misplaced for A64. It'd kick total ass on a P4. But on A64 you can get something that costs about half as much, and works better. Memory clock speed means very little to an A64, but it means the world to a P4.

Click to expand...

if thats right then the jump from 5839 mbps to 7559mbps memory bandwidth when i take the memory and fsb to 264mhz must have been because the bandwidth genie lives inside my chipset...and dont ell me that crap about memory ratios not affecting bandwidth because thats a load of crap as well at 264 @ 1:1 i get 7559mbps at 264 @ 2/3 I get like 6000 sumthing

I have to admit, I only read that up to the point where you advocated running 1:1 HTT.

Sadly it’s a rather common fallacy – but that doesn’t change the fact that it’s complete nonsense. A64 works differently than Intel, and not everything from one can be applied to the other. If you want to be arrogant and insulting, feel free to do so – but please do me the dignity of bothering to research the topic enough to get the facts straight first instead of parroting someone else’s errors and assuming they’re true.

cell_491 said:

if thats right then the jump from 5839 mbps to 7559mbps memory bandwidth when i take the memory and fsb to 264mhz must have been because the bandwidth genie lives inside my chipset...and dont ell me that crap about memory ratios not affecting bandwidth because thats a load of crap as well at 264 @ 1:1 i get 7559mbps at 264 @ 2/3 I get like 6000 sumthing

Click to expand...

I honestly don't know where he was/is getting his "information" from...but it might be best to just leave well enough alone, as I can see this very easily turning into a screaming match, something which I know I'm above.

I'm willing to settle things via PM's if he finds a way to argue anything I said above, but always remember:

ashmedai said:

If you want to be arrogant and insulting, feel free to do so – but please do me the dignity of bothering to research the topic enough to get the facts straight first instead of parroting someone else’s errors and assuming they’re true.

Click to expand...

Are you actually trying to tell ME to do my research and get my facts straight?!?!

If that's the way you feel then that's your call, I still have no idea where you're trying to go with this in trying to prove me wrong, or how you're going to argue with what I've already said, but feel free to try.

And remember, stop on over at XS and see what they have to say, I think you'll find that you'll get put in your place quicker then you thought possible.

cell_491 said:

if thats right then the jump from 5839 mbps to 7559mbps memory bandwidth when i take the memory and fsb to 264mhz must have been because the bandwidth genie lives inside my chipset...and dont ell me that crap about memory ratios not affecting bandwidth because thats a load of crap as well at 264 @ 1:1 i get 7559mbps at 264 @ 2/3 I get like 6000 sumthing

Click to expand...

Yes, increasing clock speed increases bandwidth. That's more or less what its purpose is. However outside of a memory bandwidth benchmark, it tends to have a very minimal effect. This is the good old "increasing the bandwidth when there's already more than enough" problem. Good for marketing though, they made a nice chunk of money off of moving from AGP 4x to AGP 8x, and from AGP to PCI-E...and that's just recently.

Since the memory ratio is the variable that affects what divider the memory is ran off of the CPU with, it'd kind of redundant to state that it affects bandwidth or clock speed. That's like saying turning on the light makes the room get brighter. Anyone that knows what it is, knows that's what it does. But again - you're increasing the bandwidth, when it already has more than enough. This isn't Intel, it's an A64. Intel's the one with bandwidth problems.

And to reiterate...anyone who brings up 1:1 (an Intel-related convention) in the context of A64 needs to do some more research first instead of just parroting everyone else's misdenomers.

This thread is awsome.

Kudos x lots.

cornelious0_0 said:

Are you actually trying to tell ME to do my research and get my facts straight?!?!

Click to expand...

Trying, no. If you read what you quoted, it rather clearly says that I think you need to do so. However it sounds like you're into "yell and act like there's no possibility what I believe to be true is wrong" instead of "question what I know in hopes of finding something new to learn". A popular stance, not a good one, but one that's not suprising from the way you were already acting.

Anyway given your attitude there's not really anything constructive I can say as it will fall on deaf ears. But in case I'm wrong, a quickie:

A64 calculates memory speed directly off the CPU. The actual divider is calculated by dividing the CPU multiplier by the memory ratio, and rounding the result up to the nearest whole number. For example, with the 11x multi from a SD 3700+, and a ratio setting of 200 (which I believe to be the one you're refering to, even if with the wrong terminology) you would have a memory divider of 11 (11/1 rounded up = 11). Running the memory at a 1:1 with it's source clock - the CPU speed - would be well beyond the capacity of DDR.

ashmedai said:

And to reiterate...anyone who brings up 1:1 (an Intel-related convention) in the context of A64 needs to do some more research first instead of just parroting everyone else's misdenomers.

Click to expand...

Believe me bro, I'm fully aware that memory ratios have not nearly the same "hit" on overall performance with A64's as they do on an Intel rig, but the difference is still there. I'm trying to moreso relate this to specific memory here also, if you're running TCCD in the first place (which believe it or not, a lot of people DO) the point of the memory is abnormally high 1:1 clock speeds. Also, once you get into more hardcore/serious overclocking, it's more of a challenge on a motherboard or ondie memory controller to be running at 1:1 as opposed to 5:4, and while the benefits of running in sync isn't as huge for AMD as it is for Intel, it still is the better way to go, and there's enough available product (different types/brands of memory) available right now that it doesn't really make sense to be running anything other then 1:1.

If you're hell-bent on trying to prove me wrong I think you need to sit back for a second and really think about what you're trying to prove or disprove, because you're really not getting anywhere right now.

I will not be discussing anything further with you in particular in this thread related to this debate (if you can call it that) as I don't see the point. If you feel like you really need to you are free to PM me and we can try to sort out what you're getting at that way...or you can drop it and leave well enough alone...your call.

And to reiterate...anyone who brings up 1:1 (an Intel-related convention) in the context of A64 needs to do some more research first instead of just parroting everyone else's misdenomers.

Click to expand...

ok 1:1 is an intel related convention? i thought it was called a ratio wich is a math term... And ratios are important because they can tell you what speed you memory is running at by using that math stuff, you know 264x(2/3)=175ish multiply that by two and you get 350mhz ram or ddr350. Also im not sure about you but i do insane overclocks just for the sake of benchmarks (i leave the not so insane ones for day to day stuff) and i can tell you that my cpu at 275x10 gets almost 200 points higher in 3dmark than it does at 250x11 (both 2750mhz core)

you know what dude...I refuse to argue with you on a thread that was made in order to help people if you respond again with anything off topic ill make sure you get "moderated"

cell_491 said:

you know what dude...I refuse to argue with you on a thread that was made in order to help people if you respond again with anything off topic ill make sure you get "moderated"

Click to expand...

Good call bro, good to hear I'm not the only one who wants to keep the peace around here.

You know, insulting people just because they say something different from you is a bad thing. Going from that to threats is just sending my opinion of you down the tube.

ashmedai said:

You know, insulting people just because they say something different from you is a bad thing. Going from that to threats is just sending my opinion of you down the tube.

Click to expand...

i never insulted you... however you did insult me and frankly i dont care what you think of me

cell_491 said:

and frankly i dont care what you think of me

Click to expand...

There is the key, not taking anything said on the forums personally. Someone might insult or put down your input or opinion, but they're (typically) not attacking you as a person. Trust me, after being around here over two years now and posting as much as I do, I've been through my fair share of flamefests and arguements....but in the end you figure out/realise that it's all for nothing.

The bottom line is that if someone says something you disagree with or see as incorrect, post you opinion in a polite manner and discuss from there, or PM them if you think things might get a little "heated".

This is nothing personal, I just honestly do NOT think you fully know enough about what you're trying to argue in reference to TCCD memory, not the whole dram ratio thing, I'm talking about your comments on TCCD mostly...and I stand behind my previous statements.

If you disagree with what I've said then that's fine, but you're not doing any good by cluttering up threads trying to argue something like that, especially when the original intent of this thread was to help people.....sry things ended up this way...nothing else to say I guess but to finish this (if need be) via PM's or get ahold of the mods.

I would personally prefer to have all posts related to this little conflict deleted from the thread so we can get back ontopic, and it's sad that it has come to that point.

cornelious0_0 said:

I would personally prefer to have all posts related to this little conflict deleted from the thread so we can get back ontopic, and it's sad that it has come to that point.

Click to expand...

I totally agree i want this thread to help people

Slightly more ontopic...my Ultra-D and 3000+ show up in two days, and as soon as I get my PC4400LE 431 GSkill installed I'll start working away and try out both the 510-1 and 510-3 BIOS' and let you guys know how I do for some more results to compare.