9800XT vmods

OK screw it i'll break the NDA on the reset issue. You better listen up. R1596 sets current limit. Well it's one of the items that set it. Default value is 40k ohm. Just pencil that sucker in and set it at 20k ohms. Hell i run at 8k ohm myself. When you fry your card don't say i did it. hehe BTW that R1596 goes from I limit pin # 13 to ground. So if you want solder a vario from resistor side closest to pin and ground. Us a 50k ohm vario i say. That should fix reset issues for 10-30 extra mhz and should also ally you to set Vcore a bit higher. If you still have problems i have 2 more current v mods i use. It's still a work in progress. So no offial details as of yet. Let's just say i ended up replacing 2 resistors because i needed to increase resistance and not lower resistance. That let me set my Vcore to 2.30 and max o/c and no resets ever. Personally i think everyone has been doing the gpu v core mod wrong all this time or if not wrong let's say very imcomplete. I'm about to try another method. I'm always modding something. :)

Maybe some of you guys don't know what Catch22 is saying..so I made some quick pics of it.
Just like catch says use a 50kohm...and check with your multi where you read 40kohm on the r1596 res (You can do that with your pc off) and solder a wire to it.

But pics talk more than thousend words
http://www.xs4all.nl/~jvlk/vlimit9800XT1.JPG http://www.xs4all.nl/~jvlk/vlimit9800XT.JPG

Thanx run for pointing this thread out to me...glad to see some updated info here.

Hey catch, I pm'd you too about this already. Will this mod on res r1596 help with the volts going all over the place at anything over 2v?? IT seems that when I set my trimmer to anything above 2v, the card goes berzerk under load and volts go all over the place(up to 2.1) and cause lockups.

So just use 50k pot between ground and left side r1596. 50k pot needs to be set at 20k? Is this the best setting? How about other values? Lower res on pot means higher vcore limit?What does setting at 20k translate to in terms of the vcore limit on the chip.

Thanx in advance for any info catch....your knowledge on this stuff is impressive.

BTW--I measured the ohm's on the left sied res1596 and I get 41.1k

Ah i kinda figured a team puss member would have a pic . JVLK has it helped you much. When i first told Runmc how to do it and told him to tell the rest of team puss i haven't had much feedback wether it helps a lot or just a little. The first person i showed this mod to was macci months ago then dj then team puss. Been keeping a lid on it. But i think it's about the right time to share. FYI the mod helps me a lot. Especially since i don't use a phase change setup on my card. I'm currently trying to get my Vcore to run happily at 2.30 It runs at that voltage but it doesn't score better or help o/c any. I decided to use the vid select pins. I set vid pins for 1.80 which gives me 2.290 Just did that about 20 minutes ago. I was hoping that might help. But it made zero difference. I only have my socket 370 heatsink on the card atm and it's running just fine. Do not do this if your card likes to reset. I had to do other current limit mods to run the Vcore that high. I'm about to do 1 more mod for current limit. Then another mod if this one don't help. Then another one after that. er then another after that and so on. hehehehe Hell i've tried to kill this old card by shorting out stuff from psu 5v to whatever on the card and it just won't die. I sorta get the feeling i'm gonna kill this gpu core yet. BTW anyone ever try raising SSTL2 voltage. I raise it on all my cards. I always thought it helped. Makes for one ugly looking bench though. I run it way too high. Hmm i think that must be it. Doesn't help o/c so don't ask about the mod. I just like modding sheit. Usually i do that mod when i feel the voltage is way too low. The r9800xt's are set correctly btw and about time ATI

Quote:

---

Originally posted by k|ngp|n
Thanx run for pointing this thread out to me...glad to see some updated info here.

Hey catch, I pm'd you too about this already. Will this mod on res r1596 help with the volts going all over the place at anything over 2v?? IT seems that when I set my trimmer to anything above 2v, the card goes berzerk under load and volts go all over the place(up to 2.1) and cause lockups.

So just use 50k pot between ground and left side r1596. 50k pot needs to be set at 20k? Is this the best setting? How about other values? Lower res on pot means higher vcore limit?What does setting at 20k translate to in terms of the vcore limit on the chip.

Thanx in advance for any info catch....your knowledge on this stuff is impressive.

BTW--I measured the ohm's on the left sied res1596 and I get 41.1k

---

ok you said lockups. You don't mean the card resets? You mean it locks as in freeze frame thingie? If it locks you're probably screwed. lower gpu temp is all i can say. Voltage going up is normal during a bench. Voltage should actually go down a bit for Vdroop while under load and go up at idle. But ATI didn't quite follw the controller guidelines. My card runs the same as yours except mine doesn't lock until i start getting a screen full of white dots. That means i need more current. If i and everyone else cooled off the hexfets that would get you more current. Mine get burning heat even with heatsinks. Put your finger on em and see how hot they get. Perhaps your cooling is fine for core but not good enough for hexfets. ALso each card is different. I bought this card when the r9800xt's first came out via preorder. My volts mods helped but time actually helped the most. ATI vid cards for me take about 6 months to burn in. I was running 2/4 7 with all my v mods on and gpu v core set to a stable amount. I think it was 1.975 when i first bought the card. Then i ran it at 2.025 for a couple months. Then 2.225 for a couple months. Usually just with my heatsink. But gpu temp still stays fairly low compared to a lot of cards i've owned. Geez i'm rambling again. hehe

Hey catch..........Glad to see you responded here too :)

I definitely can tell when it hard locks. Usually, a little more voltage on the trimmer takes care of that. My main problem is that with higher volts, the card looses video signal and black screens( reset switch does nothing, I have to power down everything than power back up to get it to work again...then all is well) Is this what you mean by the card reseting?

This card has run 594/445 and completed 3dmark with barely any noticable artifacts. It needs right around 2.02-2.03 to do it. The problem is that when I set trimmer to 2.03 in windows right before the benchmark, as soon as I open up 3dmark, the volts jump to 2.1 or so and stay there. This card will only run 3dmark with the precise voltage needed for whatever clocks it's set too.
Even at lower clocks on the core like 570-590, it would black screen if the volts were set too high(usually over 2.03). If I set the volts to say 2.1 before bench, it will run for like half of car-lo then black screen....but there are 0 artifacts.
There is NOTHING on the card that doesn't feel cool to the touch....even while 3d01 is running with 1.99v core/2.9ddr.
My cooling is a chilly1 r507 mach1. I took the card off and teset the unit for the tenth time to make sure it wasn't the cooling and I get a temp of -55c on the evaporator(baker)

My PSU is pc power and cooling 510. My rails unloaded at the molex are 5.22,12.27, 3.37. I checked the 5v during a bench yesterday, and it was at 5.04-5.06 for whole bench.3.3v@load reads like 3.34 and 12v loaded reads around 12.04.

I totally understand that every card is different, as all of the cards I tested in the last few weeks have all had different limits.
I have tested 5 modded XT's now with this cooling and all the rest had way more artifacts when at any clocks over 573 on core.
This one doesn't, thats why I am thinking it might go farther if can get it to take more volts without shutting off on me.

When you say hexfets, do you mean the ones that are circled with yellow or the ones with green in my pic? Im assuming yellow ;)
http://pic8.picturetrail.com/VOL256/...6/46184867.jpg
All of these are mostly cool to the touch during benches unless I go for over hour at xtreme settings they maybe slightly warm but thats it.

Ok your gpu core is resetting due to current limit. Doing my r1596 mod will help take care of that. Remember raising voltage increases current.
P/E=I
E/R=I
quare root of P/R=I

where E=volts, R=ohms, P=watts, and I=current. ohms is the actual resistance of the gpu v core.

Lemme know how much the mod helps. We need plenty of feedback. If you still have problems i got a couple more things/ideas that i may show you.

oh btw looking at your pic the upper right square inductor you circled damn near gives me 3rd degree burns on my card. Mine are NOT Falco's like in that pic. Perhaps mine are rated a bit differently than yours. Looking at other areas of your card a lot of things are different. Mine is an ATI. Who makes yours. Also can you link me to a bigger pic of that card. Both sides if possible. I'm very curious what they did to it. ta dude.

Quote:

---

Originally posted by k|ngp|n
When you say hexfets, do you mean the ones that are circled with yellow or the ones with green in my pic? Im assuming yellow ;)
http://pic8.picturetrail.com/VOL256/...6/46184867.jpg
All of these are mostly cool to the touch during benches unless I go for over hour at xtreme settings they maybe slightly warm but thats it.

---

Green and yellow circles are inductors. Hexfets are the 8 legged chips. There are 8 of them located to the left of your green circles. gpu v core uses current sharing. It has 2 independant channels. 4 hexfets are for each channel. My problem atm is channel 2 goes to minimum mode and channel 1 carries all the load. If i can get channel 2 to not go into minimum mode perhaps i can get above 600 core with 0C water. 582 is sucking it. But it's the highest core o/c i've seen with 0C water so i shouldn't complain.

Catch- Thanks for breaking your silence.

50K Vario on board. Where do we start at? 40K on the Vario?
When adjusting the Vario what are we reading res on the r1596 or a voltage?

Maybe you can think about breaking the silence on the vref mod?

Excellent Mod!

I set the Vario at 40k and measured the resistance at r1596 at 21.2K ohms.

Booted up, set Vgpu at 2.0v and ran 3dmark2003.

No reset!!!!!!!!!!!!!!!!!!!!

Ran all the way and volts stayed stable with only .02v fluctuation

This got me from 490 to 500Mhz without dots.

Next set Vgpu at 2.21v and no resets and 520Mhz no dots.

At this point temps got up to 50c during 3dmark run.

What's the safe max temp and Vgpu before frying 9800XT????

Is there any reason to adjust Vario if 9800 is no longer reseting?

Thanx for the info catch.......I to did this mod on res 1596 and early indications looks good :) I have only run a few benches so far, going to see what is the max voltage it will take now without shutting off on me. I did it a little bit different than MR. Crusher though, I put a 50k ohm vario set to 20k between ground and left side of r1596 where I measured 41.1k before I attached the vario. Card is working fine......better than before.

582 core on water is very impressive catch. What a difference better cooling makes on card limits.....
I can go 580 core on this card with only 1.84v on r507...you guys are pushing some mad voltage.:eek:

I

Kingpin....What temps do you get at the back of your card with that 507a prom on it (Stock volts and clocks)

w00t!!
Finally broke 600 core on this card :)
I can't say 100% for sure, but I think your mod has helped me catch..
After mod, I tried 580/405 @ 1.86v to see what happen and volts only changed maybe .01-.02 during whole bench. Bumped a lil and lil more and just got 600 on core 3d stable almost 0 artifacts(which I still can't frickin believe) at 2.03v. Last run was the real test.. highest my volts got for entire test was 2.04! That was a first in a week on this particular card so I am pleased.

Hey jvlk...hmm I don't bother with a temp sensor on this setup. I boot raddy prommy first for like good minute or two, then boot up mach2 when the ram chips get nice and cold.
The only temps I can give you are that the baker block gets to -55c unloaded-- and just like my chilly1 mach2, the temps just don't really rise that much under heavy load.

I have had a few XT's on this thing and the fastest/most volts was at 625/445 and 1.99 corev/ 2.86 ddrv and even at those clocks and juice, the card and all components on pcb have always felt very cool to the touch. 2.16v doesn't make the card really any warmer.

Quote:

---

Originally posted by CATCH22ATPLAY
Yes you did it correct in that the right side of R1592 and or R1597 is to be connected to ground. Doesn't matter which resistor you use. Both the right sides of them resistors are tied together on the right side. Soldering from the right side of R1592 would actually be the more correct way because the left side of that resistor is connected to ground. So for a resistor divider network you picked the perfect.

---

Thanks for sharing the info on the R1596 resistor mod...That's probly going to help alot of ppl, myself included. :)

I still have questions concerning the Vgpu mod.... I did the BMG mod with the Vr connected to the (x) side of each resistor: (x)R1597 and R1592(x)

I think Opp did this: (x)R1592(x)
Charlie did: R1592(x) --> Grd
Creative: ? pics don't load
JVLK did: same as Opp
Runmc did: BMG mod
K/ngp/n: I don't know
Mrcrusher: I don't know

So, ..... which one is correct? :confused:

I did mine just like this....cant see where his wires are going to in this pic but each wire goes to a leg of a 10kvr on my card.

http://home.comcast.net/~rbmccall200...8/site1501.jpg

Hobocrow they are all correct.

MrCrusher leave it alone if it isn't resetting any more. If it resets when you try a higher voltage then adjust vario.

My gpu temp at 582 core was 29C at idle according to rivatuner and 49C idle according to ATI. FYI

Quote:

---

Originally posted by CATCH22ATPLAY
Hobocrow they are all correct.

MrCrusher leave it alone if it isn't resetting any more. If it resets when you try a higher voltage then adjust vario.

My gpu temp at 582 core was 29C at idle according to rivatuner and 49C idle according to ATI. FYI

---

Thanks C22@play.... You've been helping ppl before I met you ..that was several years ago! Although, you really difficult to understand, you are the most wise! :) Thank you ...I'll try the R1596 pencil.... ;) Regards..HC :toast:

Hehe thanks dude. My typing skillz and my math are sometimes very confusing. :) Well i hope it gets your score up to #4 again or better. ;)

Catch22atPlay - Thanks for all your help. Much appreciated. The resetting was really bugging me. Now I've got the volts, I got to cool the gpu more. Either cool my water or break down and get a Mach II.

catch...

Unfortunately, I already did up the card and I have it back in my setup now. I have a bunch of dead stripped down ones here I can take pics of though. They are all different.

Out of the five I have here, no two have the same components on the pcb. In fact, the chips look different on two of them. There is a different ATI print stamp on the chip on these. These two also turned out to be the worst overclocking cards...real bad.

I can take some good high res pics of all them if you like, just let me know bud.

what does it meen when the screen gets full of random colored blocks??
I was cooling it with a Vapo and no hardwired mods at all and woke up yesterday morning to a screen full of green and blue blocks that covered the screen, even in the BIOS page. I put the factory HS and fan back on and the same thing, so I tryed it in another machine and no diffrence.
RMA`d it yesterday and expect a replacment in a week or so....but did it happen because of something I did when mounting the Vapo.....did I OC it too far?.....or do they just go bum from timer to time??



P.S. I`m sure glad it did it before I did the volt mods!!

Rukee your name fits very well.
It's called condensation. Take it out and get a room heater and blow warm air on it for a couple hours. Sticking it in the oven helps also. Watch the temps though. After it's dried use alchohol and cotton swabs to clean the whole card. Also the agp slot may have condensation as well. You'll need to make something to stick in the slot. I use a piece of cardboard from a box and alchohol. I've had cards that wouldn't even turn on come back to life doing that btw. Don't RMA a card that isn't dead yet. Yours was probably just fine and just needed a toweling after it's bath. :)

btw, the current limit is there for a reason ;)

certain components dont like to run at 2.28V/657 :D

http://www.akiba-pc.com/broken.jpg

Macci was that the diode? D30 to be precise?