NF7 rev2 VTT mod need it so bad...

Quote:

---

Originally posted by uwackme
Did he teach you the five fingered five point exploding heart trick?

---

:)
oh, btw uwackme, how did u do your vdimm mod?
i'm asking coz i had couple of people on my native forum do the vtt-vref mod and it didn't work... i'm still waiting for them to measure vref and vtt on the DIMM socket but... i'm suspecting that it only might work if one has vdimm mod done either "Dirty Punk's way" or "MickeyMouse's way" and not "MBfaqs' way". i might be wrong, though...
i remember u said once that vref tracked vdimm... but can't remember if that was before or after the cut-trace-vtt-mod.

Hey Bach....

I did the Vdimm the MBfaq's-way. Did the Vdd mod thier way too, this was before I discovered this forum. Vref and Vtt did NOT track.... until the Vtt mod using the cut-trace/pot method.

Quote:

---

Originally posted by uwackme
Did he teach you the five fingered five point exploding heart trick?

---

You damn skippy he did. Then I fed him poisoned fish and shaved off his beard. :p:

Quote:

---

Originally posted by uwackme
Hey Bach....

I did the Vdimm the MBfaq's-way. Did the Vdd mod thier way too, this was before I discovered this forum. Vref and Vtt did NOT track.... until the Vtt mod using the cut-trace/pot method.

---

that clears all up then...
i can't remember if vref tracked vdimm before i did mine vtt-mod but it looks like it did.
at 3.45vdimm, vtt is set with the pot to 1.725 but vref is 1.71v... so i' thinking that vref more-less tracks vdimm after DP's and MM's vdimm mods, but not MBFaqs' one. and doing vref-vtt mod (4th method) will be useless as vref is not tracking...

been testing some new...."maybe" BH5 sticks. Getting weird results. Decided to fiddle with Vtt/ref vs Vdimm.

On my CH5, with full 1Gig installed, I need to LOWER Vtt/ref to 1.545 instead of 1.65V to make memtest #5 error free.

On these I started out 1/2, get a one byte error at offset xxxx0 or xxxxx8 so it is one byte of one stick (8 bytes wide sticks).

Lowering the Vtt only made it worse, unlike the CH5. But raising it above...say 1.6V instead of 1.55V got fewer errors. Im still tweaking to get 0. Just thought this was a weird one, to mention.

Good thing I have the pot mod, to dial in adjustments :D

to make the vtt = vref safe... just cut the vtt leg on dual channel n mosfet and solder it with 120 ohm resistor to vbt sense on winbond ic... so that the vref voltage wont mess up with dual channel n mosfet....

for MBFaqs VDimm mod and Cut-trace Vbt how high should the 3.3v rail be to make 3.2v-3.3v VDimm stable?
________
buy silver surfer

Quote:

---

Originally posted by Rabbi_NZ
for MBFaqs VDimm mod and Cut-trace Vbt how high should the 3.3v rail be to make 3.2v-3.3v VDimm stable?

---

Around 3.55-3.65v or so.

Quote:

---

Originally posted by Hell-Fire
Around 3.55-3.65v or so.

---

Are there any components that will heat up significantly more due to higher 3.3v rail?
________
volcano classic review

Nope, everything stays happy. I'd go 3.6V. Im kinda bummer, got one of the Superflower 550W supplies.... no 3.3V pot. Ill have to do it the hard way.

Are you guy's al running with L12 mod?

Because my max fsb on low multi is much higher then with high multi (13-14) even if the max cpu clock isn't reached bij 150 mhz...

Can someone post a pic of the vref : vbt/vtt mod?

Is it necessary to do the DP vdimm mod or will the motherboardfaq one do?


I was also wondering if i connect vtt with vref the vref mosfet would burn?

Otherwise I'll just cut the pcb and place a pot between it:D

i just got confirmation from fellow member at Polish forum and a member here - Tonic - that vref does not track after doing vdimm mod "MBfaq's way" hence cut-trace vtt-mod is needed.
in other words, using 4th method of vtt-mod (vref = vtt) seems to be useless unless one does vdimm mod "Dirty Punk's" or "MM's" way.

I'm wondering, anybdy have done the "Dirty Punks" mod?
Wanted to ask you guys, if i do the Dirty Punks mod, do i've to remove the MBfaq's Vdimm mod rght?.
I've also done the Vtt with "Cut Trace" method do i've to remove that as well and do the Vref=Vtt mod?

And with the "Dirty Punk's" method will i be able to change the Vdimm voltage from BIOS ??? For example, if i change from 2.7 to 2.9 from BIOS will it change the Vdimm voltages? or it will be constant with the 3.3v rail volts?

Cheers,
Byron

Nope, Vdimm will always be what your 3.3v is you don't really need to remove the mbfaq mod, but it will be useless anyway......, I thinks its the best you remove the vref=vtt mod, but i'm not sure about it..

It is really best NOT to tie Vref and Vtt together. The load on Vtt is much higher, driving all those resistor packs you see along the DIMM sockets....the Vref regulator only drives a tiny load of the Vref signal to the 3 DIMMs pin1.

Stick to MBfaq Vdimm mod and cut-trace/pot Vtt mod. Then up the 3.3V rail on your PS. Nothing bad will happen with this method.

One thing is for sure.... my NForce boards DO NOT like BH5 ram at all. CH5 is soooo happy though. Abit has done such a horrible job in the bios, Im thinking these 512M sticks are suffering from serious BAD SETTINGS-itis. I just dont know how you guys ever got such great results. 2 1Gig sets from zippy, brandnew (was dated/labeled from back in the "GoogleGear" days, so on the shelf ALONG time).... Im sooooo confused.

CONFIRMED... 2 more sets BH5.... D328DW-45 datecode 0304, 2 Gigs of it... 3Gigs total now, all sux...wtf.

Anyone have a suggestion as to a AMD motherboard that just plain WORKS with BH5, something cheap I can test in? Just assume for a minute that the bios code is anti-BH5 for the moment, what VIA/SIS/etc board is known good for high FSB with BH5. I dont have a P4, esp a 2.4C, so it aint an option.

So far NVidia wont even return a phone call or email inquiring HOW to even inquire about getting technical information. Im gonna have to get mean. To think years ago I had Motorla falling over itself to give me NDA and latest 68040 family and PowerPC info, wtf is going on these days. All I want is to design with thier product..... they dont want customer, it would mean a higher tax bracket? :D

I have no issues with my HyperX 3200 (BH5) in my NF7 V2 boards,

But I had a hell of a time with BH5 in my A7N8X V2...it loved CH5 for some reason.

You could always send a set of you 2x512 for me to run in one of my boards to see what happens. If you decide to do so, I will send you my ram to mess around with while I test yours.

I am assuming you have tried 2x256 BH5 to see if there is a difference versus the gig config.

The prob with VIA/SIS, as I am sure you know, is that there is not a pci/agp lock except at certain intervals. So 133/166/200 should work fine, but those in between fsb's cause out of spec pci/agp. PCI being the worst obviously.

i think i said that many times before... OCing 1GB of BH-5 on NF7-S is a b.... !
i have 2 kits of Mushkin's LvL II...
2x256 max prime95 stable @ 242FSB (244 benchable), 2x512 max prim95 stable @ 217FSB (220 benchable).
one stick of 256MB doesn't get me higher that 244 benchable and one of 512MB is good for 232FSB.
also 512MB in DIMM3 plus 2x256MB in DIMM 1 and 2 (total 1GB) is good up to 230FSB.

so conclusion is that there's something in BIOS that makes BH-5 incapable of high OCs with 2x512MB.
i know that there's some people on my native forum that run their 2x512 Hynix sticks at 240ish FSB with no probs...

Quote:

---

Originally posted by uwackme
Stick to MBfaq Vdimm mod and cut-trace/pot Vtt mod. Then up the 3.3V rail on your PS. Nothing bad will happen with this method.

---

That's what i've done but my ATI VGA seems not to agree with high 3.3v rail volts (3.64v).
System crashes for no reason, "no signal" error appears just before the crash (hence i think its the VGA...) when i run memtest86 with the 3.3vrail on its just fine tho.

Cheers,
Byron

uwackme the best you can do is try your 2 x 256 BH-5 on a lanparty, imo.
BTW, just got 1 MHz (from 243 MHz to 244 MHz) with Vdd@2,2V
243 - 6 hours memTest - 1,9V Vdd - 3,2 Vdimm
244 - 2 hours memTest - 2,2V Vdd - 3,2 Vdimm but still not 100% stable
BIOS - tictac D22a1
Greets

that one may indicate that OverVoltageProtection kicked in on your PSU which causes it to shutdown...
most of the time u can tell that by Power and HDD leds being lit but system was shut down (no fans spinning).

I may take you up on that Hell, I'll let you know.

I dont actually have any 2x 256M BH5, my goal in bothering at all was to get higher than my 1Gig of CH5 was able to with the full gig in. So I have no way of knowing if this is BH5 period or a doublesided stick issue. I tend to think it is a DOUBLESIDED bios settings issue, since the sticks DONT SCALE at all. Going to 11,3,3,2.5 does NOT yield any change in results.

I would think, if the sticks were marginal, that looser timings would yeild a little change for the better in FSB, but it has 0 effect.

I also only have 1 stick of CH5 512M, Ill borrow a couple Corsairs CH5 512M sticks tommorrow for testing the doublesided issue a little, though 2x256+512 is supposed to be the same as 2x 512.

On a side note: I have a friends NF7 in for testing. It too dont like my BH5, but also, with absolutely IDENTICAL mods/heatsinks/ram and CPU, it does not get anywhere near the max FSB that my two NF7's get. Where I can memtst 11,2,2,2 @ 250, he cant memtest at 220 11,2,2,2....needs 11,3,3,2.5 to error=0.

Guess he just has a lemon of a SPP chip. When I say mods, I mean every mosfet/sb sinked, Swifty on NB, alpha on CPU, same exact vmod's same voltages, etc. Pegged at 3.3Vdimm his just dont get there.

On a bright side, his burnt in CH5 2x 256 does the same 11,2,2,2 @ 250Mhz in my two NF7's.

Hell, you think it's a case of some SPPs liking BH5 and some not? As well as there must be memory controller settings being used in the bios code that are just not compatible with this stuff. BH5 is a bit exotic for the mainstream customers of the 1 MILLION per MONTH they make of these boards.

uwackme inspit of buying more memory why don´t u buy a DFI lanparty?
And try your BH-5 on it...
GL

Quote:

---

Originally posted by uwackme
I may take you up on that Hell, I'll let you know.

I dont actually have any 2x 256M BH5, my goal in bothering at all was to get higher than my 1Gig of CH5 was able to with the full gig in. So I have no way of knowing if this is BH5 period or a doublesided stick issue. I tend to think it is a DOUBLESIDED bios settings issue, since the sticks DONT SCALE at all. Going to 11,3,3,2.5 does NOT yield any change in results.

I would think, if the sticks were marginal, that looser timings would yeild a little change for the better in FSB, but it has 0 effect.

I also only have 1 stick of CH5 512M, Ill borrow a couple Corsairs CH5 512M sticks tommorrow for testing the doublesided issue a little, though 2x256+512 is supposed to be the same as 2x 512.

On a side note: I have a friends NF7 in for testing. It too dont like my BH5, but also, with absolutely IDENTICAL mods/heatsinks/ram and CPU, it does not get anywhere near the max FSB that my two NF7's get. Where I can memtst 11,2,2,2 @ 250, he cant memtest at 220 11,2,2,2....needs 11,3,3,2.5 to error=0.

Guess he just has a lemon of a SPP chip. When I say mods, I mean every mosfet/sb sinked, Swifty on NB, alpha on CPU, same exact vmod's same voltages, etc. Pegged at 3.3Vdimm his just dont get there.

On a bright side, his burnt in CH5 2x 256 does the same 11,2,2,2 @ 250Mhz in my two NF7's.

Hell, you think it's a case of some SPPs liking BH5 and some not? As well as there must be memory controller settings being used in the bios code that are just not compatible with this stuff. BH5 is a bit exotic for the mainstream customers of the 1 MILLION per MONTH they make of these boards.

---

I will be here if you decide to either send me your ram for testing or me send you mine.

I have had CH5 and BH5 run good in this board so far.

I have a gig set of the KHX3500 (CH5) that runs 235 just fine...but never really tried to push much past that.

I am with you on the "double sided" issue with the ram. In my experience, even 256 sticks that had ICs on both sides did not scale as well on my NF7 as the singles did. :confused:

To your last paragraph:

I certainly believe that the bios is a major issue with the NF2 boards, and more importantly the NF7 for our uses. Older BH5 would not work for chit in my A7N8X or NF7 boars...but newer sticks would. Whereas the CH5 modules would work in either revision of either board regardless of its manu date.

Makes me wonder if all those extra memory settings in the DFI bios is the thing that is helping those owners achieve such high fsb without the need for mods. There are many accounts of DFI boards hitting well beyond what a NF7, or any other NF2 platform for that matter, right out of the box.

As far as the SPP goes, there has been widespread speculation for quite a while about certain steppings being pretty damned touchy about what ram you use and the timings. I mean, your friend is running CH5 at a reported 2-2-2-11...whereas RAS-CAS 2 for CH5 ddr was always folklore for the NF2 chipsets. Later people were saying it was a bug that the bios was reporting 2 for this setting, but in actuallity it was 3 instead. I could never get any of my 4 sticks of CH5 to even think about running at RAS-CAS 2 in either my A7N8X or NF7 (both boards being newest rev).

One of my NF2 boards loves BH5 and the other loves CH5. Both SPPs are C1 revisions...so figure that one out if you feel like getting a headache. I still fall back on the old saying that anything mass produced, whether they are all "alike" or not, will produce several that are simply not up to par with the standard.

Both are the Ultra 400, but both respond totally different to BH5 memory.

I got ahold of 2x 512M BH6 sticks, that are burned-in for a year, they act exactly like the D328DW-45's..... same memtest errors, one byte screwing up, at the same FSB/timings.

Looks like NForce2 boards....except DFI, sort of... have nasty limitations with 2x 512M sticks installed. Double sided is not a happy camper.

One test I think I need to figure out..... a mod to force a stick to ACT like it is a single sided 256M stick... maybe disable one bank's chipselect signal (cut an etch) or short to Vdimm. Point being, 1/2 of the doublesided chips should then act just like a 256M stick. if so then that would prove (as the full 32 chips would be installed, 1/2 disabled, but loading the bus) that there is settingsin the bios that need strrrreeeetching out for doublesided sticks.

But at least given I have $650 in ram sitting here, Im not holding a bag of lemons.

if the KNOWN good BH6's had acted alot better than the KHX3000's, then I'd be worried.

Quote:

---

Originally posted by uwackme
I got ahold of 2x 512M BH6 sticks, that are burned-in for a year, they act exactly like the D328DW-45's..... same memtest errors, one byte screwing up, at the same FSB/timings.

Looks like NForce2 boards....except DFI, sort of... have nasty limitations with 2x 512M sticks installed. Double sided is not a happy camper.

One test I think I need to figure out..... a mod to force a stick to ACT like it is a single sided 256M stick... maybe disable one bank's chipselect signal (cut an etch) or short to Vdimm. Point being, 1/2 of the doublesided chips should then act just like a 256M stick. if so then that would prove (as the full 32 chips would be installed, 1/2 disabled, but loading the bus) that there is settingsin the bios that need strrrreeeetching out for doublesided sticks.

But at least given I have $650 in ram sitting here, Im not holding a bag of lemons.

if the KNOWN good BH6's had acted alot better than the KHX3000's, then I'd be worried.

---

There is certainly some interesting issues with NF2 and the gig configs to say the least.

Like I said before, my BH5 2x512 config wouldnt run for chit...but 2x512 CH5 is another story.

If only we could get a few more memory timings available in the bios.....I am fairly convinced that this is what helps the DFI line.

You are surely sitting on a cache of nice ram man..lol.

Hate to see you short out half a side of one of those nice babies.