9800XT Vdd mod

I figured I'd pass on a little info on the 9800XT Vdd mod. The 9800XT uses the same ISL6522 controller as the 9800pro. Resistor values have changed however. The FB input is pin 5 and ground is pin 7. If the resistance from pin 5 to ground is R2 and the resistance from the Vdd output to pin 5 is R3, we get Vdd=.8x(1+R3/R2). R3 has a code of 13B, which is 1330 ohms. R2 has a code of 73A, which is 562 ohms. This gives a default Vdd of 2.69V. I measured 2.68V. The ram is spec'd at 2.80V+/-.15. Notice that the default voltage is actually below the 2.80V nominal voltage spec for the ram. Connecting a 10k pot from pin 5 to ground gives us:

Rpot/Vdd
9.8k/2.80V
5.0K/2.90V
4.0k/2.96V
3.5k/3.00V
3.0k/3.05V
2.5k/3.12V

Vdd is generated from the standard motherboard 3.3V supply, so I suspect it would be difficult to fry the ram (Vdd max should be less than the 3.3V supply), but I'd still be careful.

In the picture the yellow/black pair is for Vdd monitoring and the red/black pair is where I attach the pot. I've got enough air moving in my case that the backside plate does an adequate job of cooling the ram, but I'll probably eventually switch to regular heatsinks. BTW, the 3 pin connector actually fits though the hole in the plate, so the soldering was done and then the plate reattached. I'll most likely just replace the 562 ohm resistor on the board with a 487 ohm (for 2.98V) and then remove the wires.

Here's a picture of the front:

What kind of clocks are you getting before and after the vmod?

your insane. 'nuff said.

niiiicceeeeeeeee.....
looks fun!
RESULTS Please!

The 9800XT's ram is much better than my 9800pro's, but the core isn't much better (nothing a bit more voltage wouldn't improve I suspect). I just ran through 3DMark2001 at 486/412 with no artifacts. That's with Vdd=3.00V. At 486/419 or 493/412 I'm starting to see minor artifacts, but of course the test runs fine. If I run the core above 479 with 3DMark03 I start getting small white dots in Nature. I have a spot in Unreal2 that I use for "Sparkle" testing. With 4xAA and 8xAF-quality it'll show sparkle when nothing else I've found will. With those test conditions I start seeing minor sparkle at 378 memory. 384 shows quite a bit of sparkle in this test, but it looks clean everywhere else. That's why for gaming I just run 472/378 (normally 4xAA and 8xAF-Quality). The GPU is cooled with room temperature water.

Looks like they bumped up the memory subsystem! :slobber:

Just to be clear - we are talking Vddr here, yes? Or is VDD something different?

As I expected...420 on the mem with a slight voltage increase.
I bet this mem will go 440+
Jason, this should be the card for YOU, with the mem oc luck you have... ;)
C

Quote:

---

Originally posted by xDUCK
Looks like they bumped up the memory subsystem! :slobber:

Just to be clear - we are talking Vddr here, yes? Or is VDD something different?

---

Yes, Vdd is the ram core voltage and Vddq is the io voltage. Vddq should never be greater than Vdd, according to the ram spec.

BTW Bmg,
Thanks for the clear detail and instructions on the new hardware. It's refreshing to see as some of the guys here hide their new hardware and don't give any tips on it until they've exploited it themselves...
C

Can we get a screenshot of this location?

Quote:

---

Originally posted by STEvil
Can we get a screenshot of this location?

---

The worst spot I've found is right at the beginning of the repair mission in Unreal2 (I forget what it's called, I'll check it tomorrow). Just turn to the left and look at the ground after Isaak stops talking. It's unclear to me exactly what makes this spot so bad. You'll need to push your video ram hard, turn on high AA and AF, everything maxed. 1024x768. You probably need a reasonably fast cpu to push the video card. I've seen this with the 9700pro, 9800pro and now 9800XT. One of the old Q3 maps used to be my standard sparkle test, but this is much worse.

Let me know what you find.

bmg:

Did you get Half Life2 together with your card?

Quote:

---

Originally posted by Manoj
bmg:

Did you get Half Life2 together with your card?

---

ROFL. Half-Life won't be released until February 2004. I highly doubt he got HL2 with his XT.:rolleyes: :p: ;)

half-life...

http://www.overclockers.co.uk/acatal...agx_2d019_2das

it is on preorder,with halflife2

sorry,you just get a coupon for it,so you can get it free when it will be released

Quote:

---

Originally posted by GJoz
sorry,you just get a coupon for it,so you can get it free when it will be released

---

Then I guess the only main advantage the RAdeon 9800 XT will have is in the benchmarks.

http://www.3dchips-fr.com/News/actua...200310092.html

96pro and xt sure got a difference.

Quote:

---

Originally posted by GJoz
sorry,you just get a coupon for it,so you can get it free when it will be released

---

Just a coupon. The box says it's good for the full single and multiplayer game. I can't remember the details, but basically the game will be shipped to you (or you can download using steam) and the coupon has the activation key. http://www.ati.com/gitg/promotions/h...e2/coupon.html

Nice info there bmg :)

Whats the Vddq voltagelevel on this card?
ATI lowered it from like 2.8V to 2.65V when they went from R9700pro to 9800pro so is it even lower on this card? :D

Have you tried mem at 3.3V or higher yet to see if incresaeing it past 3.12v gives any extra boost?

3.0=410
3.1=420
3.2=430
3.3=440 i guess

whats the max vddr you guys would run 24/7?

Quote:

---

Originally posted by saaya

whats the max vddr you guys would run 24/7?

---

That's exactly what I was wondering. I'm sure 3.0V is fine since the 24/7 spec covers up to 2.95V. I'm guessing 3.1V might be ok 24/7, since it's only 5% above 2.95V. What's everyone's opinion?

well, i don't know but at 4.1V it won't run for longer than approx 24 hours benching. I have first hand experience:D

24 hours at 4.1v... wonder how long it would go for 24/7 @ 3.6v? heh..

Quote:

---

Originally posted by STEvil
24 hours at 4.1v... wonder how long it would go for 24/7 @ 3.6v? heh..

---

I'm still a coward and running at 3.0V. Actually,3.15V is the max I can get. I'd probably need to boost the 3.3V level of my system power supply to get more.

well, sorry i should've said 24 hours of total time benching let's say 3x8 hours or maybe little longer. at 3.9V it was doing fine for a couple of days cooled by Leadtek's copper block from FX5900 Ultra cut up and glued with AS Alumina and a 120mm delta screamer blowing across the board. GPU cooled by Prometeia r404a.

well, i think 3.7V is pretty safe if you slap some decent (fabricated) ram sinks. If you have any dead leadtek ultra board, i strongly recommend that you use it on your raddy's memory. keep in mind that's my opinion and i'm not saying it safe on all raddy's.
btw, it was club3d with stock v-mem at 2.93V.

So if vidmem will stand 3.7 24/7, why wont system ram?

Quote:

---

Originally posted by STEvil
So if vidmem will stand 3.7 24/7, why wont system ram?

---

who said it won't?:D seriously, i don't know maybe because the system ram doesn't need 3.4v+?

Doesn't it run off the 3.3v rail? If so how would you get it higher than 3.3v? Wouldn't you have to tweak your PSU to give more than the usual voltage?

Heres my unmodded results:

440/398 seem to be stable, error free clocks with normal default cooling, case closed... only tested the clocks really fast, but atleast these work. Played ~20 minutes Max Payne 2 @ 1600x1200 4xAA 16QAF without _ANY_ issues. Running 3Dmark @ 475-480/~435 caused massive errors, but it didnt crash.

:toast:

The core oc is bit disappointing, but mem scores are what i expected. Gonna keep it unmodded for now.. maybe add Vmods and watercooling later when i get the watercooling system installed on my CPU.

PTK

So in other words Radeon 9800 XT provides better memory overclocking than Radeon 9800 Pro.

I know with the Radeon 9800 Pro unmodded, I managed 480mhz core and 378mhz mem.

I am going to order the Asus Radeon 9800 XT tomorrow and I may have it by the end of the week.

My Radeon 9800 Pro will be sold in a couple of days time.

I have found the Comanche 4 demo to be a good test for 'sparklies' caused by o/c'ing of the core......the bit right at the beginning of the test on the hill will sparkle if the core is pushed too far......and its quick to keep going over until you can get it right.

Thanks

Quote:

---

Originally posted by pc ice
ok did i read this thread right and is it ok to give the mem on my 9800 pro that i care nothing about 3.7 volt's becuase i already gave it 3.6v but i turned it down quickly. didnt bench it or try to o/c the mem at that voltage im only giving ithe mem 3.21v right now and i get only 345 mhz but on the core im giving it 2.1v @-25 temp's and i get 520 mhz... if this sound's like a crap card some one say some thing so i can return it. thnx

---

Well you may have done like me.i had a card that was running 562core all day long.
i turned the voltage from 2.04v to 2.11 trying to catch the big boys on the orb.
and then i had a lockup..not thinking nothing strange(since i get them alot when pushing the card)the card only runs 519 now.somehow i killed something inside the card with the extra voltage.other than that it works great.but i lost alot on the top.thought it was the motherboard,finally tried it on a board where i used to keep it at 562 and finnally figured out,i must have over voltaged it....thats the only thing i can think of now that i did to it.
so i wouldnt go over 2.06max.

Great Job BMG

the black wire next to the yellow one is that the point you use to ground your multimeter?

Quote:

---

Originally posted by pc ice
i know im going to rma this card so i went ahead and gave it 2.24v and it took it well i turned it up while benching in 3d mark 01 at 520 and it cleared it up.and ran fine i know not to expect a long life of this card giving voltages like that but i buy them with all intention's of testing and returning as long as there are no burnt black spot's or any thing noticable. when im done with them. and i went ahead and gave the mem 3.7v and it helped alittle but onle got me to 380.

---

You guys in the US are lucky.

You can take risks like that and still get to replace the card for a new one. In Hong Kong there are no replacements due to overclocking.

If the card is damaged it is a total loss of $500. Money is burnt.

you cant always RMA it. Sometimes it smells too bad.

Hey BTW, have any of you guys managed to RMA a card that was soldered on?

Quote:

---

Originally posted by pc ice
all the time i cant even count how many time's..and do a good jod on clean up and you will be good to go..

---

Lucky guys.

When I was in the US studying for my Bachelor's Degree, I used to get surprised how people used to buy something and keep the receipt and then after a few days return it. It was too easy.

I think these things can only happen that easily in USA. Nice!

bmg,

do you see that around 486 is the limit of this core? reason i ask is because my core will do 470 with stock voltages and stock cooling. i was hoping for atleast around 500 core with volt mod. my memory does 400 without a hiccup. i would volt mod the board, but if for only 16mhz, i dont think its worth it.

TIA, daos

Quote:

---

Originally posted by daos
bmg,

do you see that around 486 is the limit of this core? reason i ask is because my core will do 470 with stock voltages and stock cooling. i was hoping for atleast around 500 core with volt mod. my memory does 400 without a hiccup. i would volt mod the board, but if for only 16mhz, i dont think its worth it.

TIA, daos

---

I'm not necessarily a good person to answer this, as I'm more of a gamer than a bencher. That said, with room temp water cooling my 9800XT is artifact free at 493 core and 1.82V (1.77-1.78V is stock). If you're artifact free (3DMark03 nature water is a good check) at 470 with stock cooling and voltage that seems pretty good to me, as I couldn't get above 445 under those conditions. You might want to ask the question on the overclocking forum where it'll be seen by more of the hard core overclockers.

Quote:

---

Originally posted by pc ice
all the time i cant even count how many time's..and do a good jod on clean up and you will be good to go..

---

Ha...ha....ha... I like this man!!! ha... ha... just take off the mod and put them in stock condition and rma it.

Once I had epoxy a big HSF on top of my GF4 Ti4200's core and it won't come off for good!! I figure it out how to remove the HSF. I carefully burn the HSF with gas stove...... after 15 minutes, it finally come off but the core came along with the HSF.... damnn I said. I think the epoxy glue really do the job well done.

So while it still hot, I carefully twist the core from HSF.... yep it finally come off. And I glued again the core to VGA PCB and return it to the shop. The shop took care of me very well, so I opt to paid the difference with FX5900 Ultra and straight went home for some moddin again.

Quote:

---

So while it still hot, I carefully twist the core from HSF.... yep it finally come off. And I glued again the core to VGA PCB and return it to the shop. The shop took care of me very well, so I opt to paid the difference with FX5900 Ultra and straight went home for some moddin again

---

The things some get away with....:rolleyes: :D

Will SMC Grabbers work for this mod???? :confused:

TIA,

Scott

i have my vmem done with SMD grabbers but it just gave me more ... artifacts! on stock voltage i can do 490/400, with 3.3V in mems i can't even get 390! i got to try it again, i probably gave too much voltage!

SO many people got less memory overclock after the volt mod but got improvement on the core.

Maybe it is best not to do the memory mod for people that are thinking about it

Its just my opinion

The memory is the easy part too :)

i have gave it 3.1v aand i am not getting past 415mhzmem

core is easeliy doing over 500mhz....

with the XT most people actually get less overclock after the ddr mod. :(

Can anybody tell me, is there VRef mod on 9800XT? Just wondered because of voltage controller the same as for vdd is near to the voltage controller of vdd itself. If you understand what i mean. :)

I lower v ref and vddq on my card. I raise VDD to 3.075 without ramsinks. Seems theese cards are not all the same. What works for other people didn't work for me. So i ended up doing what i normally do. Which is mod every damn voltage i can find a way to mod. The ddr on my card would only do 405 with just vdd mod. Experiment a bit and find what works for you.

CATCH,
Do you have pics/info on the Vref and Vddq mods???????
I been looking for a while and have not found any info.
I am stuck at 400 no matter what the Vdd (2.8-3.3volts)

Quote:

---

Originally posted by 2PiZzA
Can anybody tell me, is there VRef mod on 9800XT? Just wondered because of voltage controller the same as for vdd is near to the voltage controller of vdd itself. If you understand what i mean. :)

---

Bump, because this is a very important question, imo... I think others have done this mod....please share! :)

Just pencil in a random resistor ther bud. ;)

Random Resistor?????? Which resistor and where??

Crap now my mem is screwed again. From 451 down to 432. Hell even default 365 was messed up for a bit. I'll have to redo mods and see what actually does work. Gonna have to check all my capacitors and see if any are burned out. Vmods didn't do this. I screwed up the ddr months ago O/Cing. I was amazed when i managed to get the stuff to 451 after messing up the ddr. Now back to the drawing board. Hmm perhaps i should take better notes.

Hi everybody ,

I just got my 9800 pro 128mo from sapphire with r360 , hynix 2.8 , xt pcb and heatsink of the pro 256mo , my gpu voltage is 1.55v and keep 407mhz with atitool , my mem voltage is 2.48v and keep 371mhz with atitool , i found a pencil trick for gpu voltmod and I wonder if it's exist one (pencil trick , don't be an expert in soldering) for the mem voltmod, any help will be appreciated, thanks ,tomati

Hi,

I made some test ; I put 1.75v (pencil trick) on the gpu and got 420mhz with atitool , when I started the mem (staying the gpu @ 420mhz ) I got only 353mhz (ram voltage is now @ 2.45v , don't know where I lost the 0.03v) but when I put the gpu @ 378mhz , I hit 378 for the mem ( best then first time with 2.48v ??? , so strange) , so I wonder if it's a problem with the heatsink (stock one) , because when I try the gpu @ 420mhz , the ram was hot (I can stay my finger on) and when I try the ram @ 378 mhzit was less hot but the gpu refuse to increase and stay @ 378mhz.


then 420mhz(1.75v)/353mhz(2.45v)

378mhz(1.75v)/378mhz(2.45v)

so I'm still looking for a pencil trick for the ram , because I can't put mds grabber on it (no place) and I don't want to solder for the rma in case ; I wonder if the R2 (73A) is the good one to pencil because is the one to decrease , I will try and give you my feel back , see you later , tomati

Hi,

So it's the good one resistor "73A" , I got 2.65v now but didn't seems to help , need better aircooling for my gpu and ramsink , maybe it will help , little disapoint because i got only 403mhz for the gpu @ stock voltage (1.56v) and 345mhz for thez ram @ 2.65v , don't understand !!! , yesterday with 2.45v I got 353mhz when the gpu was @ 420mhz with 1.75v , need more test, see you , tomati


edit: just retry @ stock voltage 407mhz@1.56v and 365mhz@2.44v , then I suspect the aircooling to be responsible of the bad performance when I increase the voltage gpu & ram

hi,

I do the pencil trick , put 1.66v on the gpu , din't touch the ram , put an artic vga silencer and poor ramheatsink, flash the card with the xt sapphire bios and now I hit the 440/380 , then it was well an overheating issue , so I guess that I can go over because I'm only at 60°c max in burn under atitool .