Evga Voltage Tuner

[fadetoblack]

i set the clocks on both my stock gtx 280s (water cooled) to 713/1512/1296 and bumped the voltage up to 1.35 using rivatuner. after running ati tool for some time, the temps on either core don't exceed 52c, however the vrm phase 1-5 temps are a different story. on gpu0 the max vrm phase temp i saw was 83c, but on gpu1 max was 50c. i'm not sure why the phase temps have such a big delta.

i'm hesitant to try anything over 1.35v cause thats apparently the highest the evga voltage tool will let you go, but im tempted since the gpu core temps are still reasonable. do you guys think i should try bumping it up a little or leave it at this?

[LiquidReactor]

Try 1.3V or lower and bump up your core clocks. Even if one of your cards caps at 1512 shader you should still be able to reach 756 core on both of them.

[soundood]

Dear KELL5 and the Co, if you think that my reasons are completely illogical, please stop attacking me and prove it by actions. You have easy way to prove your words by writing your own tool doing exactly the same things you're demanding me to do. Wrting your own Windows MMIO driver is a question of 1-2 weeks. Writing your own HAL accessing ATI/NV I2C buses based on open source Linux ATI/NV drivers is a question of 1-2 weeks too. Having that in hands, you can use the information provided in this thread to implement VT1165 support layer. This is a question of 1 hour. Attaching the slider to it is a question of 5 minutes. Do it, if you can do something besides trolling. Easy money, you can even sell it to some vendor.

so true, if you want something done, then do it yourself.

for what its worth, i cant wait till vendors implement this feature across the board, maybe under a Alt + F12 secret option for 'expert' mode.

and never had the chance to say it, but nice work on rivatuner Unwinder

[T_Flight]

i set the clocks on both my stock gtx 280s (water cooled) to 713/1512/1296 and bumped the voltage up to 1.35 using rivatuner. after running ati tool for some time, the temps on either core don't exceed 52c, however the vrm phase 1-5 temps are a different story. on gpu0 the max vrm phase temp i saw was 83c, but on gpu1 max was 50c. i'm not sure why the phase temps have such a big delta.

i'm hesitant to try anything over 1.35v cause thats apparently the highest the evga voltage tool will let you go, but im tempted since the gpu core temps are still reasonable. do you guys think i should try bumping it up a little or leave it at this?

How is your card watercooled. What block or block sink combo are you using? If you have a unisink do you have a fan blowing over it? That card is getting great core temp by the way!

[STaRGaZeR]

Dear KELL5 and the Co, if you think that my reasons are completely illogical, please stop attacking me and prove it by actions. You have easy way to prove your words by writing your own tool doing exactly the same things you're demanding me to do. Wrting your own Windows MMIO driver is a question of 1-2 weeks. Writing your own HAL accessing ATI/NV I2C buses based on open source Linux ATI/NV drivers is a question of 1-2 weeks too. Having that in hands, you can use the information provided in this thread to implement VT1165 support layer. This is a question of 1 hour. Attaching the slider to it is a question of 5 minutes. Do it, if you can do something besides trolling. Easy money, you can even sell it to some vendor.

So you call it trolling. You call constructive criticism trolling. And then you play the ofended programmer game against people that you know they have 0 idea about programming. That's fine. I've already said, and this is pure logic, that RV is your software, and you can do whatever you want. Seach my posts and see if you can find anything that forces you in some way or something like that. Your negative is perfectly tolerated, allowed, blabla, but that doesn't change the fact that is illogical. I don't know if this will offend you, but you're acting as the typical open source programmer that receives a suggestion, and when you say no and then people tell you "please do it because of this, that..." you say "do it yourself troll".

Now to the point, as I've said in previous posts, RV is already a dangerous tool that has a specific section for all the important stuff with a big warning directed for noobs. Because of this when you say "I won't put it there because people could fry their cards" is 100% illogical. If noobs fry the card by going there and then they complaint to you, just ignore them. They chose to ignore your warning, it's their responsability.

[fornowagain]

Where is the logic in triggering OVP in this situation? As far as I understand VT1165 principles, the sense of OVP is preventing output voltage from going beyond Vnominal + OVP (hardwired and defined with configuration resistors). Vnominal is output nominal voltage, i.e. the voltage VRM is expecting to see on output, this is the value you are seeing at RT's hardware monitoring graphs and this is the value you're programming into VRM. So for example if VRM is programmed to give 1.4V on output and resistors define OVP of 200mV, OVP condition will be triggered when real output voltage goes beyond 1.6V. This makes sense for hardware voltage mods, when you change the voltage without altering programmed VRM's nominal voltage. In this case you change voltage by altering nominal voltage yourself. And VT1165 supports much higher voltages than 1.37.
It sounds more like and OCP for me. Maximum current for each VRM phase is also hardwired with configuration resistors.
I've already posted it in other thread: you can "ask" VT1165 to report OCP/OVP conditions by reading its register 13. Bits 0 and 1 are OCP/OVP states. Take a note that these bits are read-only status bits, you cannot change them.

I'll have to look at little more closely. The load current is much higher in game than the resting load current at 3d, so I'm not sure its OCP. What I know for sure is that the vcore measured from the back of the card never exceeds 1.35v volts no matter what value I use. This is a trace from 'in game' loads compared to resting loads at different voltages. It doesn't pull any more current in 3d 1.52v than 3d 1.35v. And were it moves to 1.52v in resting mode the current is well under even the 1.26v stock load current.



Is this the right OVP register etc? Not sure, returns the same 01 regardless of voltage selected.

Anyway 1.06v (2d register) to 1.55v gives ~1.35v. Upto 1.35v its very close to the DMM values in 0.0125v steps.

[ChinStrap]

This thread is a great example of why parents don’t take their little kids to the grocery store.

Kid sees *NEW FACY CEREAL* they HAVE to have – Mom and Dad say no, it might be bad for you.

“BUT MOOOOOOOOOOM!!!!!!….”

Get over it guys, no means no. at the end of the day your opinion doesn’t matter in the fact at all, no matter how you try to sell it.

If the programmer doesn’t want to add a slider, that is 100% his prerogative – believe it or not, whatever you say isn’t going to help that fact. Might be hard to swallow, I know this IS the internet – but it’s the truth.

[fadetoblack]

How is your card watercooled. What block or block sink combo are you using? If you have a unisink do you have a fan blowing over it? That card is getting great core temp by the way!

basically it's setup like shown in my sig (DDC3.2 w/ XSPC Res Top -> PA120.3 (3x YL D12SM) -> D-Tek FuZion GFX 2 -> D-Tek FuZion GFX 2). i don't have any fans blowing over the unisinks. i have a fan intake at the bottom of the case but i dont think that makes too much of a difference.

Try 1.3V or lower and bump up your core clocks. Even if one of your cards caps at 1512 shader you should still be able to reach 756 core on both of them.

lower the voltage? i get artifacting otherwise, and in fact im not even ure 713 is stable at 1.35, i think i would need a lot more voltage to get up to 756. i may just have bad overclockers.

[Bo_Fox]

This thread is a great example of why parents don’t take their little kids to the grocery store.

Kid sees *NEW FACY CEREAL* they HAVE to have – Mom and Dad say no, it might be bad for you.

“BUT MOOOOOOOOOOM!!!!!!….”

Get over it guys, no means no. at the end of the day your opinion doesn’t matter in the fact at all, no matter how you try to sell it.

If the programmer doesn’t want to add a slider, that is 100% his prerogative – believe it or not, whatever you say isn’t going to help that fact. Might be hard to swallow, I know this IS the internet – but it’s the truth.

Drop it, all right.. you trying to be my Daddy or what? It's just fun beating on a dead horse that died for some illogical reason, that's all. Unwinder said that it was beating on his own dead horse, and some of us enjoy it for kicks and giggles, for a little while, at least.

Unwinder never responded to my last post asking him if we need permission to add a mod to RT, adding voltage sliders. Like with D3DOverrider and StatisticsServer, they're both "add-on's".. it would be nice if somebody could jump in and add this voltage slider add-on. I do not have the programming experience to do that, but if somebody could teach me real quick how to do it, I'd love to as long as it takes less than a week or so. Anybody? Maybe I should ask Ray Adams, since he already added voltage sliders to ATI Tray Tools, but it only takes 4850 cards to 1.35v, which is most likely due to the OVP kicking in at 1.35v. When I had an X1900XTX and HD3870, I preferred ATiTrayTools over RT anyways..

[STaRGaZeR]

Maybe I should ask Ray Adams, since he already added voltage sliders to ATI Tray Tools

What sliders?

[W1zzard]

the comments in this thread are really sad and will help turning possible new freeware coders away from this scene

[Diverge]

the comments in this thread are really sad and will help turning possible new freeware coders away from this scene

Exactly!

It's one thing if a coder is asking for feedback or requests for additional functions. It's also one thing to ask... once. But wrong to keep persisting, and not taking no for an answer; regardless of the reason the coder gives.

Most people who create freeware probably do it cause it is fun for them, and serves a general purpose to themselves and others. When it becomes a job, or when people complain, then I'm sure that weighs heavily on if they 'want' to continue building on what they already created.

[BababooeyHTJ]

the comments in this thread are really sad and will help turning possible new freeware coders away from this scene

I agree, how are you going to treat the man who wrote Riva Tuner that way? Its a great app. I honestly expect better at XS.

BTW, I also love Ati Tool. I use it all of the time.

[BenchZowner]

Can you guys drop the silly Voltage sliders debate ?
If you can't, pick up your cellphone, call your friend and go out for a drink.

[LiquidReactor]

basically it's setup like shown in my sig (DDC3.2 w/ XSPC Res Top -> PA120.3 (3x YL D12SM) -> D-Tek FuZion GFX 2 -> D-Tek FuZion GFX 2). i don't have any fans blowing over the unisinks. i have a fan intake at the bottom of the case but i dont think that makes too much of a difference.



lower the voltage? i get artifacting otherwise, and in fact im not even ure 713 is stable at 1.35, i think i would need a lot more voltage to get up to 756. i may just have bad overclockers.

Well, I guess you have to play around a lot more with your system to figure out stable clocks for both cards. My GTX 280 can do 730 core, 1512 shaders at stock 1.19 vcore and at 1.24 vcore I can max out the core clock to 756.

Can you guys drop the silly Voltage sliders debate ?
If you can't, pick up your cellphone, call your friend and go out for a drink.

There shouldn't be voltage sliders incorporated into RT for obvious reasons. Way to many people will try to overvolt their gpu to 1.6V and then try to crank up the core speeds to 1000 and beyond in the process possibly damaging their gfx. Plus doing it the way that was described in this thread isn't hard at all, if someone has a problem following a few simple steps they shouldn't even be messing around with overclocking at all.

[CryptiK]

I'm hoping support for the GTX285's does not take to long to develop. Something I found odd was (if the readout can be assumed to be correct) is the VRM current draw when the vgpu was increased. Under normal circumstances the current should increase with increased voltage, however it did not seem to do so beyond 1.35v. Fornowagain seems to say that if you set over 1.35v, rivatuner reports the voltage increase, but when measuring at Vgpu measure point the voltage never exceeds 1.35v. Is this correct?

You can see current draw increase from ~33A at 1.26v to ~41A at 1.35v, however going above that seems to cause a reduction in current to between 37 - 40A.

[Diverge]

Since there are no datasheets available for the volterra VRMs, I was curious about what other registers might be; So I read them all at various conditions to try to figure out what each one is.

Here's what I found for 280gtx. Not really much... I was trying to see if there was a OVP/OCP register.. but didn't see any other registers change but the ones I marked in red. I only went as high as 1.44V w/o a load (3d app started, but not run - furmark).

My reasons for trying? I was just curious

[fgw]

ok folks,

just to put some fuel into the fire ...

[Diverge]

ok folks,

just to put some fuel into the fire ...

Pretty cool. I thought of making my own GUI in visual C#. But I was lazy and once I find my voltage/settings, it's easy enough to just run a batch file.

But good job! I'm some people will be bothering you for it

[Anth Seebel]

Funny how a person can come in here and tell a programmer that he doesnt approve of how they handle the software that they wrote.. um.. hello ?

Thanks unwinder for your contributions, many of us here appreciate it..

[Bo_Fox]

ok folks,

just to put some fuel into the fire ...

You rock, man!!!

To others still playing with fire here, I'll start taking the lead by NOT playing with fire anymore. Hopefully you guys will follow and stop playing with fire (that is, either defending or criticizing Unwinder after he has insulted human intelligence by making awfully poor excuses) here in this thread. Started NOW.

[mikeyakame]

What has playing with fire got to do with anything? He just wrote a simple c++ app to do the calculations. Same thing as doing it with a script or by hand, just a little less cumbersome.

[W1zzard]

I'm hoping support for the GTX285's does not take to long to develop. Something I found odd was (if the readout can be assumed to be correct) is the VRM current draw when the vgpu was increased. Under normal circumstances the current should increase with increased voltage, however it did not seem to do so beyond 1.35v. Fornowagain seems to say that if you set over 1.35v, rivatuner reports the voltage increase, but when measuring at Vgpu measure point the voltage never exceeds 1.35v. Is this correct?

You can see current draw increase from ~33A at 1.26v to ~41A at 1.35v, however going above that seems to cause a reduction in current to between 37 - 40A.

gtx 285 seems like a tough nut to crack for voltage control because the voltage controller has no I2C support. only VID pins which are probably controlled via GPU GPIO

[slapmehard]

this is my "new volt tool" in action
its 100 u.s.d to download (only note pad riva is free)
here at work...i'm still working on the sliders part but it copies and pastes well

[CryptiK]

gtx 285 seems like a tough nut to crack for voltage control because the voltage controller has no I2C support. only VID pins which are probably controlled via GPU GPIO

I have no problems hard modding it, it's probably the easiest mod I have ever done in fact, and I'd also like to add some extra caps to the card, so if it doesn't happen I won't be too concerned. It just would be nice to be able to test voltages above 1.30v without triggering OCP.