How high have you set you vmemm on 6800?

[Lesterp]

Did the vddr mod on a 6800nu and wondering how high others have set their vmemm voltage? set mine to 2.98v up from 2.7v default, i did the cap mod w/ a 3300/16v cap. wondering if 3v is too high.

[ZeroHero]

Did the vddr mod on a 6800nu and wondering how high others have set their vmemm voltage? set mine to 2.98v up from 2.7v default, i did the cap mod w/ a 3300/16v cap. wondering if 3v is too high.

have heard that around 2.8-2.85 wil be the best on a LE and a NU has the same memmory so I think it will be the same for the NU. More volt can give worser overclocks.

Did you got any higher overclocks with the vmem at 2.98?

[Lesterp]

yep - 25 more mhz.

[Lesterp]

anyone else do a vmemm mod on a 6800 or am i the only one?

[Del Gro Badeu]

i've broken my voltmeter so i can't measue vmem actually, will do it soon as possible ;)

[SikaRippa]

I have not modified the voltage myself, but I wonder the absolute maximum rating for that voltage is 3.6 V (stated in the hynix datasheet). I wonder that the clock speed can not be increased very much if you increase more than 2.98 V. What clock frequency you have now ? Do you know what memory chips you have 2.2 or 2.8 ns ?

There was a story in guru3d-page where somebody increased the frequency by tweaking the bios yo change the cache latency, but that can be dangerous (=card wont boot) if you put wrong values in respect of your memory-chips.

http://forums.guru3d.com/showthread.php?threadid=133245

[Lesterp]

i ran benchmarks briefly at 1020mhz artifact free w/ 3.3v but run the card at 2.98v at 983. i have hynix 2.8.

[***Deimos***]

I still can't get the solder to stick to the board. It is high in tin solder, and the silver one I have for electronics is lost somewhere. So the only mods I tried are the vmem and cap mod.
Used 10K resistor to go from 2.71V to 2.76V. With internal clock testing, went from ~790 to ~790.. no change, but with it disabled in rivatuner, dont see image degredation in rthdribl until 910-910 (before), and 936 (after). The extra voltage doesnt help much... I think my board is a very good candidate for the cap mod.

Added 4700uF 10V Fairchild (big mama) cap from TP1 (+) to TP4(GND). tried to stick the legs into holes but there are too big, so just got some clear tape and wound it around the card and got it to hold tight... result: card doesnt boot. I have a 3300uF 6.3V Rubicon which I have de-soldered from old Athlon MSI motherboard.... but have to make solder stick to get good connection.

[Nosfer@tu]

Sometimes it can be hard to geet the solder to stick on the boards, so use a Glasspen.

It is just like an ordenary pen but it has around 50-100 thin glas threads in it and you can remove the varnish so the solder will stick.

[***Deimos***]

Sometimes it can be hard to geet the solder to stick on the boards, so use a Glasspen.

It is just like an ordenary pen but it has around 50-100 thin glas threads in it and you can remove the varnish so the solder will stick.

I tried to scratch the surface of the TP pads with multimeter probe. It looked scratched, but wouldn't solder. But, its not a big problem because I can just stick resistor into TP13-TP14, and capacitor into TP9-TP10.

[***Deimos***]

I think effectiveness of cap mod might also diminish if you use long wires to connect cap... like the ones I solered to my cap... What kind of gains have you guys seen from it??

[***Deimos***]

Here it is for those who need it.
XFX 6800 with Hynix 2.8ns.. using NV5, and 1.5V BIOS.
Pot @ T13-T14 and cap 4700uF 16V @ TP9-TP10.
using rivatuner force perf, and disable internal clock test... only raising mem alone and watching for artifacts in rthdribl (blue spots).

kohms... V.... Mhz
Stock: 2.71: 910
5.33K - 2.79: 945
3.98K - 2.82: 950
2.96K - 2.84: 960
2.45K - 2.86: 960
2.45K - 2.86: 950 (with cap removed)
notice diminishing returns.

Possible limitations: NV5 doesnt have perfect contact with mem... gonna need to insert 6-10 tau shim. Plenty of AS5 used for GPU and mem. Additional 80mm fans on back of pcb where gpu, and where voltage regulartors in lower corner.

Conclusion.. this modding thing was really NOT worth it. Tried two caps, 2700uF and 4700uF.. at best 10Mhz gain. The resistor is much better bargain. If you have relatively low stock voltage, not too much trouble to stick a little fixed value 5K resistor. However, going past that, there are little gains (even with 2.95V would probably only get to 365).

Throughout most benchmarks, there is typically bigger benefit to GPU overclocking. DDR900 is sufficient for high scores.

[afireinside]

I ran up to 3.4vdd here but over 3.2 loved to randomly BS...

[ChronoReverse]

Did the mod a few weeks back. Bumped from 2.7 to 2.8V with a 4.7k resister.

I gained about 20MHz but that's it. Considering how little effect increasing memory from 700 to 900 gave in 3dmark05, this is almost negligible =(

[***Deimos***]

yes but thats only 3d05... didnt you test aq3, 3d01, q3 and ut2003 bench? (ie. anything else)