Printable View
Big Thanks Too Catch22!! You really stepped up on this and its much appreciated.:toast:
I hear ya and I think I will start with lowering Vddq first. I will not mess with Vref at this point.
I will need to go get some 100k var's at Fry's after work. What would be the steps for Vddq? lower to 2.60v from 2.65v or go in smaller steps? Then increase o/c till I get artifacts and lower Vddq and repeat? How low can it go before problems?
As you lower it you'll get fat logs maybe and if it's too high sticks on wings. Best way i can describe it. I've ran vddq as low as 2.250 just to see what would happen. I just got done trying something for ddr. It didn't work as planned. But it was interesting and i got another idea i'll try. :)
Hey can someone try this for me. Disconnect the molex connector from the card. Then measure resistance of 5v to ground where that connector was. That back side of card would be easiest. I can't find what i believe is a fried cap and maybe i don't have one. I get 4.3 ohms btw That's kinda crazy if you ask me. Although that only equals about 21.5 watts i'd still like someone else to verify that resistance. thanks :)
Oh and it like like you'll need a higher value than 100k ohms. Maybe get a 250k ohm vario. My 100k ohm drops vddq a bit more than i'd like. Oh go in way smaller steps.
Macci!! whoa dude you got 10,008 marks in 3dmark03. Out standing my friend. I hope i helped a tiny bit. 10k i never would have believed possible with this generation of cards. Outstanding achievement there. :) :banana: :toast: :banana:
C22@Play, I can't find 1/8w resistors ( only 1/4 watt ) is that ok? and the VRs jump from 100k to 500k, will that work? I placed an order this morning , I'll call tomorrow and see if I can get some 500k Vrs added.
If MrCrusher can't ohm the 5v pin for you, then I'll pull my card in a day or two and take all the insulation off and do it! :)
btw, did you see the bridge repair Macci did.... :eek: :D
Catch22
Measured 5v at Molex to ground. starts off about 17meg then quickly goes to infinity. This is what i would expect a cap to do. If you can measure resistance then you melted the inside of one of your caps.
lol you should see my bridges. I'm missing a couple altogether. I don't use em. I think a better term would be "Buss fuse". It looks like macci's actually did it's job and fried itself instead of the card. It's thier sole purpose in life.
You don't need to pull the card to check the resistance for me. Leaving it installed is fine. 1/4 watt resistors are dandy.
Damnit dude that's not what i wanted to hear. I'd have to pull quite a few components to find the culprit. I've had caps die before and your estimation is dead on. When they fry they melt inside and you can sometimes get a dead short. Amazing i was able to run my ddr at 451. It also explains the problem people to include me were having when thier ddr was getting less and less of an o/c. Hmm actually thinking about it there aren't that many components i have to check. Maybe just 10 or so. But i will have to pull em. More than likely it's one of the big 6.3v caps. My psu rocks so i can probably get away with removing the 6.3v caps and not replacing em. Also i'm using half the wattage i was using before. Don't ask me to explain that one. It's a completely different mod. ;)
Edited // Maybe I'm not doing this correctly, but I get two different readings.Quote:
Originally posted by MrCrusher
Catch22
Measured 5v at Molex to ground. starts off about 17meg then quickly goes to infinity. This is what i would expect a cap to do. If you can measure resistance then you melted the inside of one of your caps.
Just re read your request again and I didn't measure what you wanted .... I'll do it again. Hopefully correctly this time! :)
Ok, here's what I did... turned the system off ... unplugged the molex connector to PSU. Read from the molex 5v solder pin at the backside of the card to case ground. I got 31+meg ohms and a slow steady drift lower towards zero. If that's not the right way let me know. I do it again! I just hate being a n00b. :p:Quote:
Originally posted by Hobocrow
Edited // Maybe I'm not doing this correctly, but I get two different readings.
Just re read your request again and I didn't measure what you wanted .... I'll do it again. Hopefully correctly this time! :)
Well it shouldn't drift to zero ohms. That would be a dead short. Let it drift until it stops somewhere and gimme that reading. I'd think anywhere from 200 ohms to 20k ohms would be about right.
The 200K Vario is in from pin 5 to pin 5 for Vddq. Voltage went from 2.65v to 2.63v and OC stayed at 405Mhz. I lower to 2.62v and OC drop to 395Mhz. At 2.60 the OC dropped to 388Mhz. This is going the wrong way.
BTW-I have been using ATITool to test with since I belived it is more objective then visual checking 3dmark2k3/nature. I leave ATITool running and as soon as I go too high or low with the volts the tool starts showing errors. You can probably clock about 10Mhz more before your eye starts to see the artifacts.
I then tried pin 5 to 200k Vario to ground for Vddq. Voltage went from 2.65v to 2.67v and OC went to 410Mhz. I raised to 2.68v and OC dropped to 399Mhz and at 2.69v OC was at 388Mhz. Again wrong way.
So the only success I had was raising Vddq to 2.67v for an extra 5Mhz.
I have not given up so I then tried raising Vref from 1.40v to 1.41v and OC again went down from 405Mhz to 399Mhz. At 1.42v OC went down to 395Mhz.
The only thing I did not try was lowering Vref because I am not sure were to run the other side of the 200K vario. For Vref lowering it's Pin 5 to 200K vario to were??
Thanks-Crusher
Ok, the system has been off for about a hour.. I unplugged the molex from the PSU and measured from the backside 5v solder pin. I set the DMM from auto range to 00.00m ohms and grounded to the case. The reading started @ 00.05meg and slowly, but, steadily climbed to 00.25meg, then really slowed and climbed to 00.28meg and stopped there. Hope that helps.. :)Quote:
Originally posted by CATCH22ATPLAY
Well it shouldn't drift to zero ohms. That would be a dead short. Let it drift until it stops somewhere and gimme that reading. I'd think anywhere from 200 ohms to 20k ohms would be about right.
OK i guess i really do need to find that fried cap then.
OK you were supposed to move vddq and Vref both. Looking at your results it looks like your vddq and Vref like this math
Vref=(vddq/2)+0.065
so if you set vddq at 2.60 set Vref at 1.365
If you want raise em both as well. They are extremely picky
OMG i've been running that tool for the last 20 minutes and it's still trying to find max ddr o/c with no artifacts. WTF how long does this test take. It's down to 388.13 now. Will it ever end or is it the end?
So using that utility with my voltages at default i get 459/388. I'm set up for game mode and using 6xAA and everything max like AF etc etc. So at least that utility is good for setting an o/c for playing games. Although i hardly ever o/c for games i may try it. Back to Nature 3dmark2k1 for testing benching o/c though. :)
C22@Play, I just turned the system off again and with the DMM set the same and same measure points. This time I got the reading starting @ 01.14meg ohms and then a steady rise to 01.73meg ohms! I have no idea wtf that means...hope it helps!
Catch
Were do I get the Vref voltage. I can't solder to the pad in front of the IC??
Thanks
MrCrusher, I just re-read the math post from C22 and then your best setting for oc'ing your card. Looks like your best was: 2.67v(vddq) / 2 =1.335v (vref) +.065 for a 1.40vVref!
Wasn't that where your Vref was/ is? Maybe something else needs tweaking next?
My stock voltage is Vddq=2.66v & Vref=1.41v so 2.66v(vddq) / 2 =1.33v (vref) +.08 =1.40v Vref. 0.08v is also what i am measuring at pin 5 on the Vref IC. I also understand now (da!) that ishould be changing these voltages simultaneously.
I decided to use 500k vario's to get a finer adjustment.
I tried going down in volts. Set Vddq=2.64v and Vref at 1.40v and OC went from 405Mhz to 398Mhz. Then set Vddq=2.62v and Vref at 1.39v and OC went from 405Mhz to 395Mhz. This is the Wrong way.
I tried going up in volts. Set Vddq=2.67v and Vref at 1.41v and OC went from 405Mhz to 410Mhz. Set Vddq=2.68v and Vref at 1.42v and OC went from 405Mhz to 420Mhz. Ok, now we are getting somewere. Set Vddq=2.71v and Vref at 1.44v and looked up at screen just in time to see the opening scene from the Matrix. You know the one with all the 1's and 0's and x's going up and down in columns.
I shut the computer down rapidly and put evrything back to stock volts. Reboot and got the bios screen with no problems and windows started booting. As soon as the log in box came up the screen did the matrix thing again. This time when I reboot the Bios screen was repeating itself across the screen with a shadow effect. One of the words I could read was checksum error! I have the bios backed up on disk so I flash the bios hoping this would fix it and rebooted the computer. Get the bios screen with no problems and went into bios to change the CPU back to stock also. Reboot computer and no problems on start up screen. But as soon as windows got to Log-in the matrix was back.
Am I screwed? When I fried the memory on my 9700pro over-volting (I know, should have learned the first time!) I never saw the bios screen again it was just completly scrample from the second it was turned on. I measured all the voltages and all are correct at stock volts. I can go into the motherboard bios screen and sit there without any problems but as soon as windows start all hell breaks loose.
This morning I put my 9600xt in to see if OS was screwed and had no problems running some slowass benches with the 9600.
What next?? or is this the excuse for R420 and I limp along on the 9600 for next 4 weeks???
MrCrusher it looks like you're gonna be limping along for a month or 2. But the card may still be alive. Check Vref and see what it's at. It may be freaking out now. If it's not at 1.405 or so, you know like 1.265 or 1.650 you are screwed. My card finally went and died on me too. I feel for you dude. It's either that bad cap or one of the hexfets went. At least that's my best guess. I'd have to use a R9700pro as a backup. So i went and bought a Hercules R9800XT and just now stuck it in. Completely stock and gpu v core and vdd both are well below average i can run at 465/400. Case temp about 28C atm. hmm lemme check volts real quick live.
Vgpu=1.788
Vref=1.430
VDDQ=2.667
VDD=2.690
So gpu Vcore is a bit under stock of 1.80 and vdd is below stock of 2.80. But vddq is a bit high stock is 2.65 and Vref is high, stock is 1.403 or so. I like it btw. In fact i'll leave vddq alone and raise Vref a bit more. Then raise vdd as needed. I'll wait on Vcore until i got the time to solder and replace the stock cooling etc. Seems like a decent card so far. I'd say slightly better than my previous one.
Oh the card that isn't working well now i'd had the vddq/Vref at up to 3.20/1.65 and didn't have problems running it that high. Except o/c wasn't very good. in fact i've accidentally had Vref at up to 2.75 before without killing it. Screen did the maxtrix of course. But it was still readable and all was well when i went back to normal volts. Crap speakers going wack on me. Gotta look into it. L8z
MrCrusher, sorry about the card ... I hope you get it going again!:(
C22@Play, geezzz, same for you, but, your new one sounds really promising.
My 500k Vrs should be in next week, then it'll be my turn. In the mean time, I lost my HDD ... @ startup all it does is thump, thump and whin! I got stuff on there that's not backed up, of course!
I lost so much :banana::banana::banana::banana: when my hdd's fried :( hehe
Maybe i should have said try theese mods when your card's ddr isn't O/Cing like it used to. As in my o/c dropped from 438+ to 405 max. Then i modded it and ran my benches at 451 or so.
Maybe heat is what really messes theese cards ddr up. I tried vdd mod and i was able to o/c from 400 to 410 but score was worse. Hardly any artifacting though. Tried raising Vref and that raised score. What's wierd is vdd over 2.725 is where scores gets worse. I noticed the hexfets for vdd, vddq, and Vref are all different and put out more amps. No wonder the ddr chips seem so much hotter too. I can fix it so they put out less amps. But i'm leaving this card with stock cooling for a while in case it fries the ddr. I may do the V gpu mod though. Kinda curious what this card can do before i do my 6 or so V core mods.
I got some extra hexfets. So i'll take the old bad one off and try to solder a good one on. I just gotta find one with same pinouts and since it's Vref one with more amps. Personally i think that's the only one that needed more amps. Might as well try it.
Now for tonight it's back to Dungeon Siege Legends of Aranna :)
:banana::banana::banana::banana:.... :eek: I was raised on a farm... that's enough sex from a life time! :p: I killed this HDD benching it @ 86+ Mhz... jezz IBM just can't make a good product! ;)Quote:
Originally posted by CATCH22ATPLAY
I lost so much :banana::banana::banana::banana: when my hdd's fried :( hehe
I'll check back in when I get my Vrs!
OK i was taking a break from gaming and decided to figure out what exactly is fried. Looks like pin 2 on the FET which is G-1 and which goes to DH-1 on the isl6522 chip isn't reading correctly when i ohm it out. Basically it's a dead short between pin 7 and 8 and 2. 7 and 8 is where 5v is connected btw. So pin 2 is getting 5v as well. Sounds like a complete melt down inside the FET. I hope it isn't the isl6522 controller chip. I doudt it is. I'm reading the pdf files and some notes on what extra parts i got lying around or on other dead cards. Well the closest match i can find is the FET from an R9700pro's gpu Vcore. It's a bit more amps. But that's what i want anyways. The pin out does match too. Here's the difference between the fried FET and the one i want to install. The first one is the dead one btw.
STmicroelectronics STS8DNF3LL
RDSon=17meg ohm I=8.0 amps@25C and 5.0 amps@100C
Fairchild? FDS6898A
RDSon=14meg ohm I=9.4 amps@25C and 9.4 amps@100C? (should be lower)
So they look to be a close enough match and about what i wanted to increase current. I hope it works. :)
hmm.. i got a strange problem with my sapphire 9800 pro 256 with the 9800 xt layout, core and memchips. after flashing the xt bios it will run 412/365 but at the very end of '03 it starts artifacting. so that's when i decided to get a reading on my vgpu voltages. well what can i say in idle/desktop mode it's pretty normal, reading 1.778v which is close enough to the 1.78v that's default with an xt-core at 412/365. but once i fire up any 3d-app like 3dmark 01 or 03 or aquamark or any kinda game, voltages drop. under heavy load they went as low as 1.711v which is bad and probably wouldn't help getting rid of the artifacts.
seeing that i decided to go at the xt-vgpu mod described in here. hooked everything to the card, and adjusted the pots to 20k so i'd get 1.800v. on startup it would show me 1.78v (woot, 0.02v more than before) so into windows we go. benching a little and the volts drop again, but not quite as much as before.. only to ~1.75x volts. hm... might raise it a little more. ok up to 1.850v... on booting it says 1.780v again - wtf? right after i enter my pwd on the login screen the voltage goes up to 1.83x. wooot.. but well, it drops back down to 1.780v soon enough running 3dapps. well the drop while loaded is still there, but it seems to stabilize a bit.. ok. death or glory it is. 1.90v captain. said and done... what do you know? same thing as before.. 1.78v booting, 1.88x v logging in. 1.78 in 3dmark, heavily loaded 1.76x v. yea crapper.
i didn't do any other mods on my card yet and i hardly think i will do so since this card seems to be a firstgrade pos.
catch22atplay, you don't happen to have any magic up your sleeve that might cure this biatch? i've posted some pics of my card and have a whole thread about these type of downlabeled but fully functional xt-cards (except mine, maybe) here.
my card together with my ghetto vmod is featured on page 7 here.
any insights on what to do are highly appreciated. but if you think it's hopeless, i'll rma it on monday and wait for the r420. ;)
C22@Play.... I didn't understand alot of that.... see what I mean! ;)Quote:
Originally posted by CATCH22ATPLAY
OK i was taking a break from gaming and decided to figure out what exactly is fried. Looks like pin 2 on the FET which is G-1 and which goes to DH-1 on the isl6522 chip isn't reading correctly when i ohm it out. Basically it's a dead short between pin 7 and 8 and 2. 7 and 8 is where 5v is connected btw. So pin 2 is getting 5v as well. Sounds like a complete melt down inside the FET. I hope it isn't the isl6522 controller chip. I doudt it is. I'm reading the pdf files and some notes on what extra parts i got lying around or on other dead cards. Well the closest match i can find is the FET from an R9700pro's gpu Vcore. It's a bit more amps. But that's what i want anyways. The pin out does match too. Here's the difference between the fried FET and the one i want to install. The first one is the dead one btw.
STmicroelectronics STS8DNF3LL
RDSon=17meg ohm I=8.0 amps@25C and 5.0 amps@100C
Fairchild? FDS6898A
RDSon=14meg ohm I=9.4 amps@25C and 9.4 amps@100C? (should be lower)
So they look to be a close enough match and about what i wanted to increase current. I hope it works. :)
Anyways, I've got a dead 9800p if you want I'll send it to ya and you can rob it for everything you need ... if you'll send the remains back to me afterwards! :)