This is unreal - never have I been so dumbfounded

felinusz · Oct 18, 2004

I finished doing a VDD, and VDIMM voltage modification to my NF7-S a few (six?) hours ago, as well as an L12 mod for my mobile Barton.

My NF7-S board, which couldn't clear 211 MHz on the FSB stably, regardless of voltage or BIOS, even with my RAM which is specced for 217 MHz, is now merrily running memtest86 at 250 MHz on the FSB.

The VDIMM is at 3.1V, the VDD is at 1.82V.

250 MHz failed test 2 of memtest with a CAS 2.5 (I'm using a Dual channel gigabyte of OCZ PC3500 EB RAM), but is on #4 now with a CAS 3, and loosened timings - looks promising, and I'll raise the voltage if I need to for 250 MHz :drool:

.

I have a few questions for the memory guys.

How does this make sense? The only explanation I can think of is the increased chipset voltage (stock limit is 1.7V).

How high can I "safely" raise the voltage with this RAM (I see this annoying question a lot, but would still appreciate a third-person opinion specific to my RAM)?

And lastly, what are some optimal timings to aim for at 250 MHz, again with this OCZ PC 3500 EB RAM?

I'll have some screenshots in the morning (after schoooooooooool

), I've been up too late playing with my new voltage/FSB wall, and need to sleep right now

.

Thanks a ton for any advice!!

flapperhead · Oct 18, 2004

felinusz said:
I finished doing a VDD, and VDIMM voltage modification to my NF7-S a few (six?) hours ago, as well as an L12 mod for my mobile Barton.

My NF7-S board, which couldn't clear 211 MHz on the FSB stably, regardless of voltage or BIOS, even with my RAM which is specced for 217 MHz, is now merrily running memtest86 at 250 MHz on the FSB.

The VDIMM is at 3.1V, the VDD is at 1.82V.

250 MHz failed test 2 of memtest with a CAS 2.5 (I'm using a Dual channel gigabyte of OCZ PC3500 EB RAM), but is on #4 now with a CAS 3, and loosened timings - looks promising, and I'll raise the voltage if I need to for 250 MHz .

I have a few questions for the memory guys.

How does this make sense? The only explanation I can think of is the increased chipset voltage (stock limit is 1.7V).

How high can I "safely" raise the voltage with this RAM (I see this annoying question a lot, but would still appreciate a third-person opinion specific to my RAM)?

And lastly, what are some optimal timings to aim for at 250 MHz, again with this OCZ PC 3500 EB RAM?

I'll have some screenshots in the morning (after schoooooooooool ), I've been up too late playing with my new voltage/FSB wall, and need to sleep right now .

Thanks a ton for any advice!!

eb ram responds to voltage nicely. also extra voltage to the northbridge helps with fsb stability. 2 things, the eb gets hot so make sure u have a fan on the ram. secondly most stock voltage to the nb is around 1.4 to 1.6 max.
so 1.82 is quite hi but acceptable, however i would definitely put an aftermarket nb cooler. the swiftech 159 is great but expensive, the thermalright nb1-c (copper) is real nice and good price (17.99)

flapperhead · Oct 18, 2004

flapperhead said:
eb ram responds to voltage nicely. also extra voltage to the northbridge helps with fsb stability. 2 things, the eb gets hot so make sure u have a fan on the ram. secondly most stock voltage to the nb is around 1.4 to 1.6 max.
so 1.82 is quite hi but acceptable, however i would definitely put an aftermarket nb cooler. the swiftech 159 is great but expensive, the thermalright nb1-c (copper) is real nice and good price (17.99)

as for voltage, use the minimum amount to get ur best oc (sweet spot). for example if 3.0 gets u 250 and 3.2 gets u 253, i would stay with the 3.0v
i am not an amd guy so i m guessing to try and run the latencies as low as possible. the intel chipsets dont need extremely low latencies to perform well.. goodluck

felinusz · Oct 18, 2004

Thanks a lot for the info flapperhead :beer:

.

I should update that 250 MHz *Froze* memtest86 on test #6, and locked my BIOS into hig. Looks like I need more voltage.

My RAM is still cool to the touch strangely, although it's getting picky about timings for stability. 240 MHz had only 8 errors in memtest on test #5 after 6 hours, with a 3V VDIMM - unstable, but very close. 233 MHz has passed memtest, looped, for 3 hours now, at 2.9V. All this with "stock" C2.5-2-3-8 timings.

I'm ordering a NB-1C right now, my Northbridge is getting hot :-/

.

I'll pick up a few cheap 80mm fans while I'm at it, for RAM cooling.

I'm going to get myself a stable FSB within 3.1V, and then post some screens. I'll aim for a stable 250 MHz (this BIOS' max FSB

) after I get those fans, and a NB-1C installed.

Do any Socket A AMD users with this RAM have any advice for me on timings I should be trying out?

cetoole · Oct 18, 2004

Have you done vtt mod to go with vdimm, might help.

felinusz · Oct 18, 2004

240 MHz @ 3.1V VDIMM, C3-3-4-10 has passed memtest86 for 6 hours, and is now priming at 240 X 9, just to be safe.

233 X 11 = 2563 MHz failed Prime95, even though 233 MHz passed memtest86 with flying colors at 2.9V, and C2.5 timings. At 233 X 10 = 2330 MHz, Prime95 was fine.... :-/

. Looks like a power issue, seeing as how I'm running my chip at very low speeds, and ridiculous Vcore voltages to rule out processor instabilty inside Windows. I really hate how memtest86 will pass, yet Prime95 will fail in Windows at otherwise obtainable processor clockspeeds with a high FSB - my chip will do 2700 MHz (200 X 13.5) at 1.9V.

My VDIMM fluctuation is insane, low of 2.93V, high of 3.14V, when set to 3.1V via my volt mod.

I haven't done a VTT mod to my board, and am unsure whether or not my board could take it. My board is a PCB Revision 0.51 (I think..... not sure, as I couldn't find that sticker in the corner - I have white PCI slots, and black RAM slots).

Could you link me up to a solder + VR method of doing the VTT Mod? I really don't want to cut traces to do the VTT tracking mod.

I'll probably be raising my VDIMM to around 3.3V later on, and a VTT mod to ~1.65V will almost certainly be neccessary to hold a stable vltage that high.

larva · Oct 18, 2004

felinusz said:
I have a few questions for the memory guys.

How does this make sense? The only explanation I can think of is the increased chipset voltage (stock limit is 1.7V).

Why in heaven's name would one do mb mods if they didn't help?

felinusz · Oct 18, 2004

Why in heaven's name would one do mb mods if they didn't help?

Well, I'm more sort of shocked at how much they helped. 211 to 240 stable is a very big difference for 0.1V on the chipset, and 0.2V on the VDIMM.

Shuzzy · Oct 18, 2004

felinusz said:
Well, I'm more sort of shocked at how much they helped. 211 to 240 stable is a very big difference for 0.1V on the chipset, and 0.2V on the VDIMM.

your right it is. i love reading threads like this knowing i have the same memory in my setup thats gonna get a voltage shock in about a week. :santa:

cetoole · Oct 19, 2004

Here you go from tictac's site, the 4 known ways of doing the vtt mod for NF7: http://s6.invisionfree.com/underclock/index.php?showtopic=100

pwnt by pat · Oct 20, 2004

The pcb revision is on the very top of the board on the back of it by the i/o. If you do tictac's method then you run the risk of frying your motherboard - moreso than the trace cut mod.

vdimm=vi/0 = risky on .51, guarenteed working on .52
pot vtt = guaurenteed on both

This is unreal - never have I been so dumbfounded

felinusz

Senior Overclocking Magus

flapperhead

Senior Member

flapperhead

Senior Member

felinusz

Senior Overclocking Magus

cetoole

Member

felinusz

Senior Overclocking Magus

larva

Inactive Moderator

felinusz

Senior Overclocking Magus

Shuzzy

Member

cetoole

Member

pwnt by pat

Member

Similar threads