Quote:
Originally Posted by madgravity34
I have no problems with VJs advices. But i learned: Trust is good, controll is better.
Printable View
Quote:
Originally Posted by madgravity34
I have no problems with VJs advices. But i learned: Trust is good, controll is better.
also wondering this.. even at 1.84 my comp reboots trying to run 3dm05..Quote:
Originally Posted by madgravity34
Thats not torture lol. That is the fast track to video card Valhalla by suicide lol.Quote:
Originally Posted by Jort
Viper
OVP isn't the issue but loaded Vcore folding down on Vcore current limiting is probably occuring.Quote:
Originally Posted by madgravity34
Viper
Quote:
Originally Posted by Slipknot
Both ATI and NV use 3DM03 MN and 3DM01 N for max GPU current draw checks. You can not use total PC power consumption to determine that as the CPU power compsumption varies widely test to test and skews the numbers.
BTW 10 passes through MN is not 25 passes lol.
Viper
There are several things that can cause a 7800's vcore PS to shutdown and trigger a reboot. OVP is determined in relative to the Vcore regulator VID pin programmed voltage config and is normally higher than 1.84 but tolerances can bring the down. The thing with OVP is it will happen the instant the Vcore exceeds the setpoint at anytime in any test.Quote:
Originally Posted by OgreX
More than likely you are getting a latched error on either total Vcore PS over current or Vcore PS mosfet phase (3 of them) overcurrent triggering the shutdown. That can happen at relatively low Vcore (less than 1.70) with some cards if the GPU draws higher than typical loaded current at a given Vcore and GPU load combined with tolerance stack up toward a lower than typical on one of the OC shutdown setpoints. The reason it shows up in 3DM05 Mother Nature is because the highest loaded average and instantaneous core current draws occur in it and/or 3DM01 Nature. If you watch carefully you will probably find the shutdown occurs at the same point in the test everytime indicating a Vcore PS shutdown on an instantaneous current excursion over the shutdown setpoint.
Viper
Fast way? I think not... I got the board like in October and benched then on 1.8v and 2.8v... 24\7 I use 1.55 and 2.3
I dont OC inside so there aint any overheating problems... just my watercooling on cpu wants to freeze all the time, this time getting dryice
Dang make up your mind lol. 1st you say 1.90 and 2.90, Then you say 1.80 and 2.80 and now you are down to 1.55 (which is low for 24/7) and 2.3.Quote:
Originally Posted by PlatoonSgtElias
If you were still running 1.90 and 2.9 like you said starting in October the odds are incredibly high you wouldn't still be running lol.
When you write suicide value voltages in a forum you need to note they were one time test values. There are a lot of users that cruise the forums looking for information on how hard they can push the envelope with their cards and get into trouble with good for the goose, good for the gander thinking.
BTW you pulled some damn nice numbers out of the 7800GTX if it is a 256mb version. Don't smoke it!!!
Viper
Are there alternative points to perform the VGPU mod?
I allready killed on of my 7800GTX :lol: (that did 635\1750) the damn ram module came of with the heatsink :D
But umm yeah i use 1.90 and 2.90 on TESTING only ;). what is good for 24\7?? I want 600\1600 for everyday use hehe :) I tought 1.65v and 2.5v for 24\7 was overkill? Or aint it? :o
Quote:
Originally Posted by ViperJohn
>You can not use total PC power consumption to determine that as the CPU power compsumption
Sure not, but in this test i saw a dependency between power consumption and the temperature of the GPU (and a larger drop down of the 12V in fillrate-test than in MN-test).
In Austria it was 4 a.m, so i was too tierd to run 25 times. I ll do this later, today.
But i saw 2 tendencies yesterday:
- The temp-graph (in Rivatuner) grows in fillrate-test faster than in MN-test.
- The temp-graph was stable at 46°C in fillrate-test, and unstable at 44-45°C in MN-test.
The fillrate-test was the first test i ran, so the air around the PC was warmed up for MN-test, and the GPU-temps were nevertheless lower.
I ll post later some graphs an values.
*Platoon*
On GOOD AIR 1.6v to 1.65v is about right. Dont know about mem....waits for VJ
*VJ*
Is there any way to prevent OVP or core shutdown from occuring? Does this have anything to do with the amount of amps that the system powersupply has? Or is this card-specific? My issue is that when using vcore above 1.75V the card will bench a single run and through the middle of the second the screen just turns off. The computer stays on, but it looks as if the card shut it self off. I figured it was peaks in voltage causing it to bump up past 1.85v, but I have never actually seen any jumps during benching on my radioshack multimeter.
Oh I have a copper big flower Prescott 3.2GHz cooler on it :) Ill but 1.65v on it then? But yeah.. what volts for memory? Mem heatsinks are one of the best i have ever seen... copper botom and aluminum fins
Tests are done, here the values:
Fillrate-Test (run 99 times):
http://www.web-upload.com/uploads/Fillrate.jpg
MN 03 Test (run 25 times):
http://www.web-upload.com/uploads/MN.jpg
There are only 3-4 peak values at 45°C. Most temperatures are 43-44°C. The temperature depends on the watereffect in MN. The more watereffects are drawn, the higher the power consumption and the higher the temepratures. In one test, the effect changes 5-6 times from low watereffect to much watereffect and back.
Yes but the solder point are even smaller and/or at the end of 401 size SMR'sQuote:
Originally Posted by Son1990
and that is damn small!
Viper
Oh damn that sucked! That was a way, way above average 256 card!!! You typically only see those core OC's from the 512 cards with their cherry cores.Quote:
Originally Posted by PlatoonSgtElias
1.625-1.650 loaded HP3D Vcore and 2.30 to 2.350 Vdd are good 24/7 numbers on WC. Remember you are only running those voltages when running in 3D mode on a 7800 card unless you are dumb enough to set the 2D Vcore to 1.40 in bios as well lol.
Viper
The issue is you do not know if the shutdown is being triggered by OVP (doubt it) or OCP (more likely). So far I have not found work arounds but I am not looking either. I do not run Vcore's high enough to trigger a latched Vcore PS shutdown in any case except for rare exception/odd ball cards with cores, that appear from the temps at the Vcore being run, to draw well in excess of typical current draw at the Vcore the PS is shutting down at.Quote:
Originally Posted by madgravity34
I have a 512 here now where the Vcore PS is shutting down in 3DM03 MN at the exact same place in the test every time with loaded Vcores over 1.70. The cards core is running 5-6C hotter than normal (true die) on a ViperVenom-II (245w TEC) compared to all other cards (and it's SLI mate). That is telling me the core is drawing well in excess of the typical modded current at the speed and Vcore being run. I take that as a warning sign to heed and not an reason to recalibrate the OCP shutdown setpoint with a core that has obvious current draw issues lol. Don't force it applies here.
Video card Vahalla is not a place I want a customers card to find!!! It is not worth a few extra Mhz of core OC from the 24/7 standpoint to change the setpoint protecting it. Besides the card is knocking on the door of 13K in 3DM05 now in an FX57 machine running 3.8Ghz.
Viper
Nice charts (I run them too), bad interpretation of the data in front of you lol.Quote:
Originally Posted by Slipknot
The fill rate test, like ATItools Scan for Artifacts, places a very steady state load on the GPU. What you are reasoning as unstable temps are the result of the variance in load on the GPU as the MN test runs through its scenes. You are only seeing the average temps in the graphs at the update frequency you have the graph set to update at.
Also the temp sensor response time to core heat caused by rapid changes in core current draw do not show up very well in any graph. There is always a lag between the temperature indicated versus the current being drawn at that moment in time and the true die heat being produced at that moment. You see a max temp of 45C in the 3DM05 graph but that was on a rising die temp. The core current draw at that point would have driven the die temp much higher (which it was instantaneously) had it been maintained steady state and/or the sensors response time was fast enough sense/to indicate it before the GPU load dropped, the current being drawn fell off a bit and the die temps dropped.
You can not exactly equate indicated average die temperature to rapidly changing core current draw and the localized heating it produces. The simple fact is 3DM03 MN and/or 3DM01 N produce the highest core peak current draws and that is why they are used to determine max core current draw and power consumption. The peak draws are not steady state but that are high enough and long enough to generally cause a 7800 core core to lockup (instantaneous localized heating) or to trigger a Vcore PS latched shutdown on OCP in those tests first when the card will run perfectly fine everywhere else.
Viper
n00bish question alert but when using autodetec in coolbits which core of an SLI is actaully being checked?my gtxs are votmodded, thanks to persivore ;) ,and I set the vgpu to 1.65. I notice whn i run coolbits auto the vgpu of card 1 is reading 1.63V or so but the vgpu of card 2 is only at 2d setting of 1.4V.
so in SLI when you overclock is it setting both cores at this same overclock or is it only the main gfx card that is overclocked?
I also had a mutlimeter in my hand. Exactly in that moments, when the water-effects were drawn, i read the lowest voltage. I do not have a good PSU (350W), but my advantage is, that i can read drop downs@12V. The highest dropdown i got in the fillrate-test (constantly 11.63V). In MN-test i got 11.65-11.73V. In that moments, when there were more watereffects shown as trees, gras... i got 11.66-11.67V. In other moments i got 11.69-11.73V.
You think, that my actualization-rate is too high. I will do this tests again. This time, i will check a log-file and set the rate to 10ms (a lower value is not possible). I hope thas low enough and i ll see what values i ll get.
Picture of the point?Quote:
Originally Posted by ViperJohn
Quote:
Originally Posted by Slipknot
Cant do 10ms measurement :(, 6MB logfile created, but the viewer is unable to open it^^, after 45min CPU-time canceld the viewer via taskmanager :(
anyone know how to do the vddq mod?
heard from hipro5 that it helps oc the ram
just be curious bout it
cheers
Vddq is probably being regulated by the same regulator as Vdd (the ISL6534). I would guess that FB1/pin 29 (or something connected to it) would be the point which needs modding.Quote:
Originally Posted by Victorshen
I haven't seen 7800-256 card yet where increasing the Vddq helped the memory OC and many where it hurt. There are always exceptions of course.Quote:
Originally Posted by Victorshen
Viper