View Full Version : 9800pro vmod worth it?
illogical06
04-13-05, 09:40 AM
Well, thats basically the question. I want to know if i can break 500mhz core if i vmod that baby. I gotta cpu hsf on my gpu so i can handle some higher temps. I already get pretty darn good clocks outta this card but i'm wondering if i vmod would be a worthy risk...as my soldering skills aren't grand but i'll figure a way to make it fool proof....(i hope)
3dMark scores are in the sig, the latter of which artifacted slightly because my RAM was a lil higher then it should have been. I was @ 470/410 clocks, 252 fsb, rest in sig.
Edit: Sapphire card with 2.8ns Samsung Ram i believe.
illogical06
04-13-05, 01:23 PM
bump
nicspits
04-13-05, 04:11 PM
well what temps are you getting? you should be able to do it and break 500, if your core can handle it as that is a great OC already. And don't worry about soldering, buy some SMD Grabbers (google it) and you'll only have to solder to ground, which is super duper easy with a 30w iron(you won't be able to get it hot enough and make it stick with 15w if you're using one of the larger points. could just buy more grabbers).
Edit for crimedog: multimeters are wonderful things, look for the OCfaq guide and follow his recommended voltages. I believe it was 1.85 With good air cooling, might have been 1.8 but yeah, get a thermal probe, multimeter, the grabbers, and you'll be set to break 500.. One of the best parts, you can always remove the grabbers and use them for future volt modding projects, hehehe!
crimedog
04-13-05, 05:08 PM
if you don't break the card it's worth it, yes
tomslick
04-13-05, 05:19 PM
well what temps are you getting? you should be able to do it and break 500, if your core can handle it as that is a great OC already. And don't worry about soldering, buy some SMD Grabbers (google it) and you'll only have to solder to ground, which is super duper easy with a 30w iron(you won't be able to get it hot enough and make it stick with 15w if you're using one of the larger points. could just buy more grabbers).
I agree. I got some smd grabbers for the vdim on my motherboard at frys outpost.
illogical06
04-13-05, 10:13 PM
Well slight problem fellas, those clocks in my sig ARE 3dmark stable but apparently not CSS stable. But that's beside the point
Those pretty high 3dmark scores (for 9800 pro) you see in my sig were doable by the means of slight artifacting. I didnt watch most of the 3dmark01 bench b/c i didnt see any artifacts at first, but i watched the 03 and their were slight artifacts in it. I said oh well, lets see what i get and thats where that score came from (i'd be listed on second page of pro if it wasnt flashed :))
3dmark 01 clock : 479/420 3dmark 03 clock: 470/410
Now my card seems to be unstable even @ default xt speeds of 412/365....i may have damaged it i guess, but it wasnt even artifacting that bad...really it wasnt.
To make sure its not my cpu or ram oc, i went to default settings and also default xt speeds. I then left on the CSS menu and went to work, when i came back it was froze.
What do you guys think?
nicspits
04-13-05, 10:34 PM
possibly a little toasted
illogical06
04-13-05, 10:40 PM
I ran ati tool for 10 minutes scanning for artifacts @ 435/385 and none showed up. Max temperatures reached 58C core and 39C environment. These cards get ALOT HOTTER then that. As stated before, the artifacts weren't that bad, thats why i continued....you still think it might be toasted?
Btw, this card has been running @ 355-385 minimum for basically its entire lifespan if that helps at all. I THINK i have 2.8ns Samsung ram, its the most common found in the sapphire 9800 pro's. I believe Hynix is also found, but no cheap stuff.
nicspits
04-13-05, 10:45 PM
very strange. with my 9800np, I've gotten kicked out of games due to errors, even when ATI tool and 3dmark(all versions) tested fine. The solution was to lower my FSB but you say you've reset things to default settings and it still does that.
Perhaps try reinstalling your drivers or installing the latest once.. or even go Omegas and see what happens. Quite the odd situation. If none of that works, try underclocking and see what results you have.
As speed increases, cooler temperatures are needed to maintain stability so while 59c may not be that hot, at those speeds AND temps, it might not run stable. Are you still on the stock cooler with generic thermal paste?
illogical06
04-13-05, 10:46 PM
Not even close, i have a modded cpu hsf and bga ramsinks with a 120mm blowing on the card from the top plus a xp-120 w/ panaflo blowing on the backside. AS5 too.
nicspits
04-16-05, 12:37 PM
still, you'd need watercooling to really get that high and maintain stability, peltier even as you will see my results in my sig (though it's a np). Oh, and my max temps @ 1.875 volts through my card still gives me temps <30C always.
illogical06
04-16-05, 01:54 PM
Wow man, i just checked out that np, very impressive clocks you got. Well i havent had a crash in awhile, i'm at pretty much default clocks for everything (except pro is @stock xt clocks). Is it possible to overwork a card and then just let it 'chill' for a bit to regain its composure?
felinusz
04-16-05, 03:12 PM
About voltmodding your card, it is definitely worth doing.
~1.85V is a conservative overvolt to start with on your 9800 Pro's core, with high-end custom aircooling like you have. I strongly suggest that you try scaling the overvolt higher than that, in order to see whether your card sees any stable overclock gains from it. Every CPU/GPU is different, many see stable overclock gains from seemingly absurd overvolts. :)
Stability is your best indication of a safe overvolt/temperature/overclock - your hardware will tell you when it is getting too hot by losing stability, and artifacting. If its stable, its safe (in my experience and opinion). This rule of thumb does involve keeping track of how your specific hardware reacts to large overvolts and higher temperatures - every CPU and GPU is different. If you don't pay attention, stuff can fry.
It is hard work, and it pays off.
As for overclock gains after voltmodding, I would gander that you will probably see at least another ~20 MHz out of the core with a generous overvolt. 500 MHz on the core is definitely attainable with an above average card and very good aircooling.
That Samsung memory (if it is Samsung, then it is 8E die Samsung) should do 400+ MHz for you with around ~2.85V of VDDR (the rough average overvolt 'hot spot' for this specific type of memory). With my old 9800 Pro with Samsung 8E-GC2A die RAM, I found that any overvolt above 2.85V actually degraded stability and my overclock. This behaviour is quite common with this memory. But play around with it, especially if you aren't sure whether you have Samsung or Hynix.
You are probably unwilling to flash your card back to a 9800 Pro BIOS, but doing so might help your memory overclock quite signifigantly, if your card does have Samsung 8E die memory on it. If you want to give it a shot, check out this thread HERE (http://www.ocforums.com/showthread.php?t=295383)
All that said, if your card is still acting up, I would not voltmod it untill you get it running smoothly again.
One thing to note, is that the on-die temperature sensors on R360 cores are notoriosuly innaccurate. Your stability and overclock degradation problems may be due to overheating because of poor contact between your custom heatsink and the card's core. I would remove your heatsink from the card, and make sure that there is a good TIM die imprint on the heatsink, which tells you that the core is making good contact with the heatsink.
Wow man, i just checked out that np, very impressive clocks you got. Well i havent had a crash in awhile, i'm at pretty much default clocks for everything (except pro is @stock xt clocks). Is it possible to overwork a card and then just let it 'chill' for a bit to regain its composure?
Sounds to me like your processor/FSB overclock was causing the crashing. Slight GPU/GDDR instability manifests itself in the form of artifacting; crashing is uncommon unless your card is very unstable. Overclock your card, keep your processor/FSB at stock, and see if it will run stably for you.
For some tips on thorough system stability testing, you might find it worth your time to check out the light blue link in my signature, "The Definitive Guide to Stability Testing" :)
Good luck with your card! :)
Like Felinusz said, volt modding is definitely worth it....... my Pro-> XT was able to run a 430 core with a modded Arctic Silencer, and after the vmod I was able to run 480 stable. In your case, I think that you would get much higher than I did since your card gets almost to my overclock on stock voltage.
ViperJohn
04-18-05, 02:42 AM
still, you'd need watercooling to really get that high and maintain stability, peltier even as you will see my results in my sig (though it's a np). Oh, and my max temps @ 1.875 volts through my card still gives me temps <30C always.
Not hardly lol.
Viper
illogical06
04-18-05, 10:29 AM
I shall vmod my card, btw, i dont think 450 is stable for gaming...only benches.
MrCooper
04-18-05, 02:28 PM
I shall vmod my card, btw, i dont think 450 is stable for gaming...only benches.
Definetely vmod your 9800 Pro 1.85Vgpu/3Vddr....Now that should make things a little more stable now. :cool:
felinusz says it all i volt modded a 9800 pro and was very satisfied with the results. also you got those fans blowing over the card that will keep your resistor cool and your voltages stable, which is very important. so... go for it. to nicspits, i did the solder just fine with a 15w soldering iron. a little trick i used was to coat the wire with solder, then touch it to the solder point, then touch thesoldering iron to the wire. the solder liquifies and drips down the wire onto the solder point. it also helps to add a little extra solder to the point. good luck. oh yes, and it do think you can hit 500.
illogical06
04-19-05, 11:48 PM
Sweet guys, imma vmod my card this weekend...or actually probably thursday if im not busy. I'm going to use smd grabbers for this one but nice tip on the soldering, i wish i tried that for my 9600xt.
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.