The document also states OVP is fixed to 1.7V.
And since other people have had their cards upto 1.45v it makes me wonder what's really wrong if a card trips @ 1.1_v...
trez0r,
Have you measured vGPU voltage after the card has tripped?
If it hits 0._v then it's some protection mode that causes it. If not...
You were not supposed to see this.
Did "other" people solder the same place (mod)? I haven't found any place on the internet where someone shares his vmod on 96gt. (And i did search a lot). Can someone share his experience here?Originally Posted by largon
I think i measured the voltage after the screen went blank and it was something 1.0x V if i recall correctly..
Last edited by trez0r; 02-27-2008 at 06:23 AM.
If vGPU doesn't drop to near 0v then the tripping is not due to OVP.
You were not supposed to see this.
You can use my nickname, I won't make me mad at all
People, I found voltmod myself, it's not online yet since it is not a reference card, we still have to explore the possibility's of this early test sample. It won't be of much use to you either way, VRM's are totally different.
With ADP3208, voltage is directly read in the VRM and then compared to 10% and -10% higher/lower voltages then VID, aint that causing problems here?
I just made an experiment.
In idle i lowered frequencies to around 600/1400/1600. Then i launched atitool artifact scanner and started scanning. While scanning i upped the voltage to 1.18v - no blank-out. I started raising core freq all the way up to 840. No blank out - no artifacts. Next i raised memory to 1160 - all well. Then i started upping shader - i think i got to 1600mhz when card shut itself off. After the black screen i measured the vgpu it was 1.045V. So to conclude this is very unlikely OVP. Looks more like OCP what do you think?
what was using type of wr resistor?
for example 200 k
I hear with g92 you could add a bit through the bios, just like the old 6800/7800 trick. Anyone try nbitor yet?
EVGA 780I P02bios
E8400@3.6
BFG 9600gt SLI 710/1000
2x2g GSkill ddr2
Sceptre 20.1 naga
Antec SP500/TT Sli psu
op1e: look at my post #5
kaptan54: not sure if i did understand your question..
I asked to you how many kohm pots?
but I learned them
for gpu 100k
for mem 220 k
is that right?
the last question
show me voltage measure point
pls give me your bench score with vmod
Maybe we can find something from below....
And this Sunday i will try it....
(The vid code table is captured from adp3207 pdf file.)
Last edited by fatbird2046; 02-29-2008 at 11:20 AM.
Hihi, exactly what I was thinking about, yes it should work out just great if the located spots are correct.Maybe we can find something from below....
And this Sunday i will try it....
(The vid code table is captured from adp3207 pdf file.)
Also regarding this news, my GT was running closer to 840MHz at 1,45V then instead of 780MHz. PCIe 110MHz
@fatbird I am waiting for your test
Hi,
I haven't contributed to the forum earlier but I thought I'd share my experiences now.
I have a 9600gt and did the VR vGPU mod first and had the same problems described earlier.
Yesterday I tried the vid-mod and it worked like a charm! Thanks so much for the table and photos! I first tried 1.3V and while it worked great, it was a bit too much for stock cooling and 3dmark06 failed at the end of the second game test.
So I went for 1.2V and now it's stable. Temps still reach 75C top so I guess I have to change the cooling. But the bottom line is that with the vid-mod, the ocp doesn't seem to kick in and everything works.
I'll run some tests today with a vmem mod.
Thanks again!
The problem with the traditional voltmod via FB pin is not OCP related but rather OVP.
Ok, so how come the OVP doesn't kick in as soon as I crank up the voltage? It runs fine for a while but crashes when the heat builds up.
With the vid-mod, I don't seem to have this problem. This far the best results are 820core / 1080mem, 3dmark06 stable. Yielded 12963 points. I'll try to push the mem little bit further still before changing the cooler. 13k with stock cooling is my goal.
There is a 10% upper- and underlimit made in the VRM. Whenever the external GPU core voltage exceeds those limits (compared to VID config), then the VRM will generate a fault and prolly shutdown your card. I've messed with this kind of mod before, would make most sense if this is what caused the problem. Then again, it's your experience, I haven't tested voltmods with this board design so if you say its OCP what can I do about it to proof you might be wrong
plese put ss bench scores
13k reached finally. Not with stock cooling though as I didn't want to stress the core with too much heat anymore. The core clock was stable with the stock cooling but I added a vmem mod to get 1107 mem clock.
800 / 1.2V core, 1107 / 2.1V mem:
I'll try to up the vgpu to 1.3V next. I already tried 1.4V but it was too hot for me. I want 24/7 air cooling. Oh, the tests are run in a closed rig.
<edit>
1.3V didn't give any higher clocks so I reverted back to 1.2V. I'm happy with my 800MHz overclock, maybe I'll try to up the memory a bit. With 1.2V vgpu the temperature stays below 60C in 3dmark06 (800MHz core) and I guess the power consumption isn't bad either but I'll have to measure that later.
</edit>
Last edited by Xargo; 03-01-2008 at 05:07 PM.
Nearly 50% perfect....
1.2v 790mhz ok!
1.4v is not stable! System keepS reseting over 770mhz.I think current limit should be modded too(100% NOT because of hot i think).Give me a couple of days and i will find it.
Last edited by fatbird2046; 03-02-2008 at 01:04 AM.
Thanks! I just meant 1.4V was too hot for my cooling. The chip ran at 77C in 3dmark06. 1.3V was around 63C but didn't give any performance boost so I reverted back to 1.2V. I'm looking forward to your current limit mod, maybe it will give better results with 1.3V.
Nice find!
So you just short the correct configuration and get the voltage indicated in table? Cool! Will try this out a bit later..
In the case of 9600GT w/ ADP3208A VID pins are relatively unpractical to exploit for purposes of increasing vGPU as the complexity of the mod goes far beyond that of the variable resistor -method:
Stock vGPU is 1.1500V = 0011100, this means VID pins 2, 3 and 4 are closed. One can't possibly increase vGPU by just shorting other VID pins (0, 1, 5, 6) because all VIDs coding vGPU higher than stock require breaking atleast one of VIDs 2, 3 and 4.
And, with VID pins one can't tune the voltage like with a VR.
Last edited by largon; 03-02-2008 at 03:16 AM.
You were not supposed to see this.
DIP switches on board is possible the fastest way to do so.
True.
But complexity - thus chance of error/failure - would also increase by a large factor.
VID code -method is just much more inconvenient than the VR-method:
- vGPU higher than stock always requires removing atleast one bridge
- 10-20x more complex to solder than VR if the user needs multiple options higher than stock vGPUs
- offers no on-the-fly tuning
You were not supposed to see this.
I thought this IC did allow on-the-fly VID change, that's at least what the datasheet tells me.
Posting Permissions
Reply With Quote
Bookmarks