8800GT voltage mods

[babalouj]

Well I put the 77X resistors on there, did the correct OCP mod, then adjusted the vmem to 1.86V and vgpu to 1.4V. I gained 30mhz core, 0 mhz shader, and 50mhz memory bringing me to 880/2100/2100 @ 57C loaded. Not too shabby for a free mod.

[jason4207]

Has anyone had trouble overclocking their memory to previous stable levels after the gpu volt mod? My memory could do 1000 before, now I'm having trouble with it @ 975. I think maybe my core hates 800mhz even if I pump the voltage beyong 1.3v. It seems stable if the memory is left at 950 though. I have Hynix chips (HY5RS123235B) which are rated to do 1000. Stock memory voltage is 2.00v. Any thoughts?

Also, what are the symptoms of OVP kicking in?

Yes. I was stable at 1026 before modding, and after doing the vmem buck I was stable at 1056. After I cranked up the vcore I had to dial the memory back down to 1044.

Well I put the 77X resistors on there, did the correct OCP mod, then adjusted the vmem to 1.86V and vgpu to 1.4V. I gained 30mhz core, 0 mhz shader, and 50mhz memory bringing me to 880/2100/2100 @ 57C loaded. Not too shabby for a free mod.

I wasn't sure if the 77X would work right, but like you said it's only 2%...it can't mess it up that bad. Glad you got it working!

[nvrmind]

Do you think it's just the Quimonda chips that are benefiting from the vmem buck mod or Samsung and Hynix as well? I guess I could try pencil modding the memory to see if it responds to higher voltage and if not I could try the vmem buck mod. Kind of annoys me I'm losing like 50 mhz on the memory.

[natedog420]

ok well i dont have the resistors i took off for the ocp mod i ended up losing them. i have an old motherboard. what do i need to measure to find new ones. can i just solder in one of the new ones since the stock ones are gone? i run 1.45 volts usually so i need the 1.2k like you say.

[nvrmind]

Its odd, I've been up to 1.45v on my MSI with no indication of the OVP activating. Doesn't seem to matter as my card won't clock any higher after 1.29v. What voltages are considered unsafe on these cards?

[Mysterfix]

Hey, maybe you guys can help out a bit. I have a freakishly good 8800gt. It runs STOCK at 800/1900/1010. So I decided to Vmod it, maybe i can make it even better. I attatched the resistor as per the instructions ont he first page of this forum, put 1.3v through it (vf900 cooling) and can push it up to 900/2100/1000.

This is all good, but I get MASSIVE artifacts after a while. So I put my overclock back a bit, and still artefacts. I then run my old overclock, and STILL get artifacts - when before I would get none. THing is, I get them after a while, which leads me to believe that a part of my card is overheating or something (other than core/memory).

It did this before my vmod, so I put heatsynks on the moffsets, and pretty much every other big black part on my card and slapped another fan on those and it worked. Obviously, it doesn't do anything now.

Anybody got any ideas?

I'ts probably your ram getting too warm and causing the artifacts, which is why it doesn't do it right away.

[jason4207]

Do you think it's just the Quimonda chips that are benefiting from the vmem buck mod or Samsung and Hynix as well? I guess I could try pencil modding the memory to see if it responds to higher voltage and if not I could try the vmem buck mod. Kind of annoys me I'm losing like 50 mhz on the memory.

I have no idea about the Samsung or Hynix. I've been waiting for someone to post something. Your card is behaving like mine, though, so my 1st guess would be that vmem buck would help you. No way to know unless you try!

ok well i dont have the resistors i took off for the ocp mod i ended up losing them. i have an old motherboard. what do i need to measure to find new ones. can i just solder in one of the new ones since the stock ones are gone? i run 1.45 volts usually so i need the 1.2k like you say.

They are 78X. If you see SMR's on your MoBo w/ the same label they might work. If not maybe you can find them online. Only way to measure resistors on the Mobo is to remove them first.

You don't need 1.2k for 1.45v. 3.8-4.2k would work fine for that. You shouldn't have any issues until ~1.625v idle w/ 3.8-4.2k.

Its odd, I've been up to 1.45v on my MSI with no indication of the [s]OVP[/s] OCP activating. Doesn't seem to matter as my card won't clock any higher after 1.29v. What voltages are considered unsafe on these cards?

Have you tried running AquaMark3? That seems to be a good stress test for OCP.

[nvrmind]

I haven't tried AquaMark3. OVP doesn't seem to matter in my case since my card won't respond to higher than 1.29v. I was hoping to get the shader to around 2000, but 1825 seems to be it's limit. My card is one of the lower clocking cards I've seen on here. The voltage mod got me 90mhz on the core and 50 on the shader (800/1825). Not too bad, but I was hoping for more. Now if I can just figure out this memory issue.

I'll try the pencil mod first with these Hynix chips. If increasing mem voltage doesn't help, than I guess I'll go ahead with the vmem buck mod. I measured the voltage around 1.9v+.

[scottc19]

I haven't tried AquaMark3. OVP doesn't seem to matter in my case since my card won't respond to higher than 1.29v. I was hoping to get the shader to around 2000, but 1825 seems to be it's limit. My card is one of the lower clocking cards I've seen on here. The voltage mod got me 90mhz on the core and 50 on the shader (800/1825). Not too bad, but I was hoping for more. Now if I can just figure out this memory issue.

I'll try the pencil mod first with these Hynix chips. If increasing mem voltage doesn't help, than I guess I'll go ahead with the vmem buck mod. I measured the voltage around 1.9v+.

This seems really odd, how do you know that more voltage doesn't help,ie what kind of crashing are you getting. Does your card still read voltage after it crashes?

You don't need 1.2k for 1.45v. 3.8-4.2k would work fine for that. You shouldn't have any issues until ~1.625v idle w/ 3.8-4.2k.

So if you use 1.2k, what kind of headway do you think you will get then like over 2v? I used the 1.2k on mine for around 2 weeks without problems, but finally decided to rma since ram never ran stock. I am going to do this mod again on my set of 8800gt now and I guess if they can be pushed to 1.625, then it seems there isn't any point to go all the way down to 1.2k. Seems like 3.8-4.2 will do that trick. I am looking for 1.55max, so you think 4.2k will do or will I need the 3.8(ie is 1.625 with 3.8 or 4.2?)

[nvrmind]

I've gone up to 1.45v. 3D06 locks up above 800/1825. The strange thing is though I've been able to run 800/2000 without artifacts on ATI tool @ 1.29v for an hour and I can play CS:S without lock ups. Other times CS:S will lock up as well. My multimeter will continue to read the correct once 3DMark crashes, which is why I don't think it's OVP. I've had very inconsistent results so far. Temps are just fine, load is 40C running ATI tool.


**EDIT**

I was just able to pass 3DM06 @ 800/2000/950 @ 1.29v. Very odd. The only thing I can think that made a difference was changing a bios setting. My Biostar TA770 A2+ has a GPU setting in the bios that puts a wattage limit on the PCI-E slot. The default setting is 75W. I changed it to 135W. Perhaps that setting has made the difference, as a volt modded 8800GT is definitely eating up more than 75W. I'll keep testing and report back.

[EightyDuce]

nvrmind run AquaMark and see if it triggers OCP, thats about the easiest way of doing it.

[aussie-revhead]

Can you guys tell me how much difference there is between a vmodded GT and GTS at same clocks ? Which of these will score the highest in 3D06 ? I just want to know if its worth spending the extra $100+ per card .

Any help would be great.



Edit - looking at some of the clocks in this thread the GT clocks much better than my GTS , is there a brand/model of GT to look for ? One with better overclockability ? Better ram maybe ?

Cheers .

[EightyDuce]

GTS has more SPs than the GT so it should be faster clock for clock. Also, GTS will OC higher than the GT.

As far as brands go, they're all reference cards, for the most part, so should OC about the same. It's all about luck of the draw.

[jason4207]

So if you use 1.2k, what kind of headway do you think you will get then like over 2v? I used the 1.2k on mine for around 2 weeks without problems, but finally decided to rma since ram never ran stock. I am going to do this mod again on my set of 8800gt now and I guess if they can be pushed to 1.625, then it seems there isn't any point to go all the way down to 1.2k. Seems like 3.8-4.2 will do that trick. I am looking for 1.55max, so you think 4.2k will do or will I need the 3.8(ie is 1.625 with 3.8 or 4.2?)

I don't know the equation, but VJ told me 3.8-4.2k would give you 25% headroom. I know ~1.3v idle is where OCP kicks in, so 1.3 x 125% = 1.625v.

Largon said 1.2k gives you 50% headroom, so 1.3 x 150% = 1.95v.

I used 3.92k, but haven't gone any where near 1.625v yet, so I can't tell you for sure. I think you'll be fine for 1.55v w/ 3.92k or 3.8k if you want to be sure.

Someone correct me if my interpretation is flawed.

Can you guys tell me how much difference there is between a vmodded GT and GTS at same clocks ? Which of these will score the highest in 3D06 ? I just want to know if its worth spending the extra $100+ per card .

Any help would be great.



Edit - looking at some of the clocks in this thread the GT clocks much better than my GTS , is there a brand/model of GT to look for ? One with better overclockability ? Better ram maybe ?

Cheers .

It varies from card to card. My GT did a little better on the core, but the GTS did a lot better on the shaders. The GTS has 3 phase power, so my gut tells me it would handle the increased voltage better. Also, the GTS does much better on memory clocks. At the same vGPU, and w/ tweaked memory my GTS scores 1200 points higher in 3DMark06 than my GT. Plus you don't really have to do the OCP mod on the GTS (unless your going extreme cooling).

Ideally you want a GTS w/ Samsung memory. I have no idea how you'd be able to identify one, though. Perhaps a more recent date on the box would give you a better chance at getting the Samsung chips?

[EightyDuce]

Ideally you want a GTS w/ Samsung memory. I have no idea how you'd be able to identify one, though. Perhaps a more recent date on the box would give you a better chance at getting the Samsung chips?

Is it just the new GTS that have a chance to come with samsung chips or the GT aswell?

[largon]

I don't know the equation, but VJ told me 3.8-4.2k would give you 25% headroom. I know ~1.3v idle is where OCP kicks in, so 1.3 x 125% = 1.625v.

That can't be the case. OCP can't be tripped at idle (theoretically it can but not in reality). Otherwise the card would trip at stock freqs and volts as soon as it enters 3D load.
-> OCP means that the phase controller monitors the amount of current that runs through the phases, once it detects the preset OCP-value has been exceeded the controller forces a shutdown on the mosfets. This means that in practice OCP can only activate under load.

Level of current that flows through the phases depends on three things thus OCP trip-point can be achieved only by increasing the following variables - unless the trip-point marker is altered ofcourse:

1. level of load
power consumption = C_device_constant * voltage^2 * freq

Variable "C_device_constant" varies according to the activity of the chip (= idle/load)

2. frequency
More clocks = more current draw

3. voltage
This is because the current draw of a device increases slightly as voltage is increased. In reality this has only a marginal effect.

Largon said 1.2k gives you 50% headroom, so 1.3 x 150% = 1.95v.

We can't be sure since Primarion doesn't share the datasheet for PX3544, but I bet the equation isn't straightforward as this, and as stated previously, voltage doesn't affect OCP much.

[jason4207]

LOL...I didn't mean 1.3v idle triggers OCP, but that if you have ~1.3v idle then when loaded and under the right conditions OCP will kick in. I didn't give load voltages, b/c they tend to vary based on the program in my experience.

But, the rest of your write up puts things better in perspective. There is no absolute max voltage obtainable based on the R value of the SMR's used in the OCP mod. You also have to consider the higher frequencies you'll be running as well.

Can you estimate what values of R should be used for OCP override at different max voltages based on average clock speeds obtained at those voltages and assuming the worst case load?

Example:
.
.
.
3.92k-->good to 1.55v as long as clocks are below 950
3.8k-->good to 1.6v as long as clocks are below 975
.
.
.
1.2k-->good to 2v as long as clocks are below 1200
.
.
.

[largon]

jason4207,
Sorry, I have no idea...


Damned Primarion, why do you sit on the datasheets?

[jason4207]

Oh well. I guess one could just use a 3.92k SMR to be safer to start w/, and then if OCP becomes an issue at desired higher voltages and frequencies then one could just solder another 3.92k SMR on top of the other one effectively cutting the resistance to 1.96k.

[EightyDuce]

Oh well. I guess one could just use a 3.92k SMR to be safer to start w/, and then if OCP becomes an issue at desired higher voltages and frequencies then one could just solder another 3.92k SMR on top of the other one effectively cutting the resistance to 1.96k.

SMR sandwich? Also from what I understand, the GTs do not really respond to voltages higher than 1.5v, so there isn't much of a point to go crazy on OCP.
Unless I misread something, which is entirely possible.

[adpr_02]

I'ts probably your ram getting too warm and causing the artifacts, which is why it doesn't do it right away.

Nop, downclocked it to below stock, and still got the artifacts. I then tried lowering the voltage to 1.2v and still got the artifacts.... Raising the voltage to 1.32 made me get a black screen very late in the game.... does the OVP mod actually work?

[EightyDuce]

Nop, downclocked it to below stock, and still got the artifacts. I then tried lowering the voltage to 1.2v and still got the artifacts.... Raising the voltage to 1.32 made me get a black screen very late in the game.... does the OVP mod actually work?

Could be a physical issue with your card. And yes, the OCP (not OVP) actually works.

[Mysterfix]

Nop, downclocked it to below stock, and still got the artifacts. I then tried lowering the voltage to 1.2v and still got the artifacts.... Raising the voltage to 1.32 made me get a black screen very late in the game.... does the OVP mod actually work?

Are you using the stock cooling? I would take off the heatsink and check for good contact. I've had 2 older cards that acted like yours both had issues with the HSF not getting good contact.

Yes, the mod works but it's for OCP (Over current protection) which kicks in waay before OVP.

A really easy way to check if it may be a heat issue is to turn off your computer for a few minutes to let the card cool down. Then fire it back up and start a bench loop. Time it to see how long it takes to artifact. After you've timed it from a cold boot try it with the card already warmed up. If it artifacts quicker when warm you know you have a heat issue.

[Demo]

Anyone know where the vGPU and vMEM read points for the Palit 1Gb 8800GT ?

[largon]

Demo,
Yes. Need something to mark them on, like say, a pic of the card for example.