PDA

View Full Version : Max 24/7 Vmem for 8800GTS? - PROGRESS! POST #7



Timbosan
12-07-2007, 05:33 AM
Im currently at 1.95v vmem on my GTS (G80). How much voltage can the memory ic's and power circuitry of 8800GTS handle 24/7 aircooled?
The highest factory vmem ive seen on a GTS is 2.08v, so i'd think that is safe longterm.
What have you people run &/or consider safe longterm?

ic3m4n2005
12-07-2007, 06:30 AM
Same question here, but i'm on full cover waterblock, so what are you gyus giving your GTS on mem and core ?

largon
12-07-2007, 07:28 AM
I reckon there's no definitive answer.
I'd go no farther than 2.1V. And in the end, Samsung GDDR3 OCs these days pretty well even at stock volts. Used to bench @ 2.3V. Overvolting gained 3DMark06 stability from 1107MHz upto 1161MHz.

N1ck R1mer
12-07-2007, 10:12 AM
Im currently at 1.95v vmem on my GTS (G80). How much voltage can the memory ic's and power circuitry of 8800GTS handle 24/7 aircooled?

In my opinion, you should install more effective cooler first. With the stock cooler, video memory heats up much by GPU, so after changing the stock cooler to the Thermalright HR-03 Plus (http://www.thermalright.com/a_page/main_support_installation_hr03-plus.htm), I've noticed, that mem now takes higher stable freq. even with the standard 1,87 V.

eclypse
12-07-2007, 12:27 PM
heh i dont know what they were thinking in that pic on step 12! lol.. Lay the heatsink right across the end of the card where the caps are. heh i wonder if that card still works after that?

Nice lookin HS though.

Timbosan
12-07-2007, 09:22 PM
In my opinion, you should install more effective cooler first. With the stock cooler, video memory heats up much by GPU, so after changing the stock cooler to the Thermalright HR-03 Plus, I've noticed, that mem now takes higher stable freq. even with the standard 1,87 V.

Thanks for the responses. I was thinking i would need better cooling, i do intend to either mod stock cooler or get the HR-03 plus. However i have read when GPU is watercooled, the ram ic's only get to about 30*C when benching, measured by an IR thermometer. Wouldnt having the GPU heatsink in contact with the ram ic's actually heat them up rather than cool them since it gets so much hotter?

Ive read GTX's (superclocked evga) have 2.2v vmem, Gotvoltage's GTS has 2.08v stock, hence my question what can they take, as they seem silimar but overall GTX's have slightly higher vmem.

Then there is the effect some people experience with decreasing ram clocks as they up the voltage. This may be heat related, and most likely is due to proximity to GPU & the fact the IC's & basically forced to operate at the GPU temp due to heatsink design, or may be power related, as as the voltage is pushed up, the current ripple would also increase as the PWM is nearing its limit of efficiency. It may be that cards capable of 2.1v+ have better components or simply provide cleaner power to the ram ic's.

Timbosan
12-08-2007, 01:39 AM
PROGRESS!!

I Increased Vmem to 2.00v (idle & load), and increased Vgpu to 1.41v/1.36v (idle/load). The core can now do 684 stable, with the shader at 1620 and memory at 2106.

Previously it wasnt stable at these speeds with <2.00v on the memory. When the gpu hit 85 - 87*C it would lock up, both in games and benchmarks.

I think what was happening is as i had to raise Vgpu to 1.41v, it was having the effect of heating the ram up more - i think by increasing the Vmem 2.00v i have stabilised the mem at these higher temps.

Some people have the problem that their ram is not responsive to voltage increases, while others have good gains but find it is sensitive to heat. If you do have the ram that responds to voltage increases, as in my experience, its a matter of finding the Vmem required to stabilise the ram at whatever temp your cooling & gpu clock & voltage makes it run at.

Point in question , during a game this arvo, the gpu hit 95*C !!! There were no slowdowns or artifacts, i had no idea it was that hot till i alt+tab 'ed out and saw the riva monitor ( see screenie 1).

I also gained a hundred points in 3dmark03 and 06 from the 9 MHz core increase (675 to 684). I didnt alter the shader yet. I hope that once i properly cool this space heater masquerading as a 8800GTS, i may get even more out of it.

gOtVoltage
12-08-2007, 02:18 AM
Yeah thats definatly a cooling issue....Looks good so far...:yepp: I had a bad Night i mean real bad.....like in almost 3-4K beacuase of PG&E surge...Serveral other people in my nieghbor hood had same problems i did last night.....So bunch of us are filing claims for repairs to our electrical relatrd items!

Timbosan
12-08-2007, 02:40 AM
yeah i definately need better cooling! 95*C load is unacceptable, and is holding back my clocks something fierce.

you must be devastated mate (i posted in your thread) damn them make them pay for EVERYTHING you lost.

gOtVoltage
12-08-2007, 03:03 AM
yeah i definately need better cooling! 95*C load is unacceptable, and is holding back my clocks something fierce.

you must be devastated mate (i posted in your thread) damn them make them pay for EVERYTHING you lost.

Yeah I think im gonna tell them that when they replace the two affected pc's ,I WANT NEW mobo// chip// mem and vid cards WITH power supplies..ALSO two water pumps..... I have reciets so they should...sucks its good stuff too or i should say was..

Timbosan
12-08-2007, 03:07 AM
oy yeah of course, you are not to know what is damaged, and cannot risk testing individual components in known working hardware, incase it damages the working hardware! full system replacement is a given in this situation. they cant ask you to pull it apart and risk your stuff that works testing it, they wouldnt even know you could do that. to most people pc's are just a box you buy & use (and never open :))

gOtVoltage
12-08-2007, 03:24 AM
Ima give reciets and then bargain shop...i know i can save some towards the phenom setup....but that wont be for at least 2months untill any of us know on those things any ways so im in no rush,:D