^ yes pencil easily kills gddr.. mightier than the sword indeed
try 3.3k 1/4w resistor @ MVDDC and 3.3k 1/4w resistor @ MVDDQ and see how much you get out of that
^ yes pencil easily kills gddr.. mightier than the sword indeed
try 3.3k 1/4w resistor @ MVDDC and 3.3k 1/4w resistor @ MVDDQ and see how much you get out of that
Last edited by NapalmV5; 01-23-2009 at 09:45 AM.
Does anyone have screen shots of 4850's volt or pencil modded in crossfire?
CPU:Q6600 G0 @ 3.825
Motherboard:Asus P5E X38
Memory:2x2GB OCZ Reapers DDR2 1066
Graphics Card:Asus 4850
Hard Drive:2xSegate 500gb 32MB Cache raid0
Power Supply:Xion 800W
Case:3DAurora
CPU cooling: D-tek Fuzion V2 (Quad insert removed)
GPU cooling: mcw60
Monitor:24" LG
1/2w is just able to handle more current then 1/4w, it isn't worse at all...
990FXA-UD3 | FX8350@4.7Ghz | Asus HD7870 | 2x 4GB Crucial Ballistix Tracer 2050Mhz 8-8-8-22 | AX850W |SSD Vertex3 Max IOPS 120GB | Auzentech Forte + TAPCO-S5
EK Supreme Full-Gold | XSPC RX240 + EX120 | MCP35x | 3x Koolance Blue Led @PWM | Tygon Black 1/2 | Bitspower Compression | @ FM CM690 I
The current flowing through the feedback line is most likely in the µA range. That means that the 1/4W resistors will be just fine. To reach this limit of 250mW of power dissipation on the resistor itself, you'd need a very high voltage.
P= 250mW
I= 20µA
=> U = P/I = 12500 V = 12.5 kV
Just as an example.
Now let's say we have a feedback voltage of 1.2V, it comes down to this power dissipation:
P = U*I= 1.2V * 20µA = 24µW
This might not be 100% accurate, but just to give you a clue. The actual current flowing through the feedback line could be explicitly stated in the corresponding data sheets.
And I didn't even take into account that the current is divided into two parts when connecting a resistor in parallel.
Quote from one of our professors:
"Reality is hiding in the imaginary part."
Interesting software vmod here: http://www.ocxtreme.org/forumenus/showthread.php?t=4427
-
"Language cuts the grooves in which our thoughts must move" | Frank Herbert, The Santaroga Barrier
2600K | GTX 580 SLI | Asus MIV Gene-Z | 16GB @ 1600 | Silverstone Strider 1200W Gold | Crucial C300 64 | Crucial M4 64 | Intel X25-M 160 G2 | OCZ Vertex 60 | Hitachi 2TB | WD 320
i5 2500K @ 4.9GHz+ 8GB G-Skill RipJaws DDR3-2000 @1600Mhz CAS 6 Asus P8P67 Pro CrossFire 6970's @ 950/1450
Xeon X5677 @ 4.5Ghz 6GB G-Skill RipJaws DDR3-2000 @1600Mhz CAS 7 Gigabyte EX58-UD5 4870x2
i7-880 @ 4.2Ghz+ (still playing) 4GB G-Skill RipJaws DDR3-2000 @2300Mhz CAS 9 Asus Maximus III Formula MSI Hawk 5770
I've had my 4870 at 1125 mem (4500) often for several days/nights at a time (usually idle at very long times) and I haven't had a problem yet. But unlike some people around here, I know better than to say that memory dying from speed is just a 'myth'...since I know first hand what it's like for Pentium 4's and Core 2's to degrade, even from just a slight voltage jump (while others will tell you that i'm lying or I should get out of the overclocking community and use a console instead (yes, people have been that harsh before).
I don't know -why- stuff degrades like this. I did see someone lose 200 mhz (!) on a very mildly overvoltaged E8x00 with just stress testing (prime/Intel burn test) at 1.32v....and it was well cooled...
And then others say their chips were highly voltaged with zero degradation...
Seems to me to be down to each individual card/cpu/memory sample I guess (except for those northwood P4's...a LARGE # of people had those things simply die or degrade).
Sorry I couldn't answer more satisfactory...
My card was doing 850/1200 on stock volts no problem, and its taking 1.400 Vcore Load to get it to run at 875! Im testing stability now at 860/1200 in ATI tool, ill see how much i can get out of it. But it looks like more volts is doing very little on my card...
The modding arena!
Im not sure why I uploaded this
Assorted board with lots of 10 turn pots!
All done.
I soldered the other side of the pot onto a random pad for a cap that lead to ground as you can see. Nice and secure by itself, a dab of glue makes it solid.
Its running at 880/1200 right now, but its taking 1.48 Vcore to get it to be stable at that. Is that kinda high volts for the stock cooler?
EDIT: If I lower the ram down to 1100, the crysis FPS dosent go down, and it only takes 1.4 Vcore to be stable.... Hrm...
Last edited by ZOMGVTEK; 02-01-2009 at 07:13 PM.
Hello DuncanOc!
I have the same video card Gigabyte hd4850/zalman
Please tell me if you identify what rezistor must pencil and were is the point
for measure final GPU voltage.
We can't make software vmod with riva tuner because Ncp5392 controller
is not in database.
Gamer HUD help you to OC or is only marketing gadget.
Man, here's the IC:http://www.xtremesystems.org/forums/...postcount=1390 but I dont have a DMM right now to identify which resistor to pencil.Maybe ceemic can help you
Sorry, I haven't been reading this thread a while. You have to find a place, where You can pencil FB to GND. It can be a SMD resistor or ceramic-cap.
Vmods - what can possibly go wrong? ©
Dudes, do hardmodes, easier and more reliable
Vmods - what can possibly go wrong? ©
I like hardmode. If you know were must solder a trimmer res. I be glad!
Guys, I presume you'd be interested in this bit of info, so just a pointer to it..
It seems that it's possible to force 2D clocks for GDDR5 memory in BIOS after all, and AVOID the flickering issue.
http://forums.techpowerup.com/showth...19#post1191319
So far only one user reported it so it needs further confirmation, but he says that keeping 2D clocks for memory at 450MHz while 3D clocks are at 900MHz - there is no flickering, and temperatures/consumption in 2D is lowered.
I'm thinking that it's maybe 2:1 ratio so for those overclocking it could be that it's not 450 2D/1000 3D but 500MHz 2D/1000MHz 3D (I'm only talking about memory clocks here).
I'm just sending out the word for now, as I'm selling my card in 2 days and don't want to bother 8or risk) with flashing just before buyer arrives, but I'll probably test these once I get new card later next week (I hope).
Cya!
Intel Core 2 Duo E4400 @ 320x10 (3200MHz) w/ Scythe Ninja rev.B + 120mm fan | Gigabyte GA-P35-DS3R (Intel P35 + ICH9R) | 4x 1GB Kingmax MARS DDR2 800 CL5 | Sapphire ATi Radeon HD4890 | multiple random HDDs | LG GH20NS15 SuperMulti, SATA | LG W2361V, 23“ LCD, 1920x1080 | Aplus CS-188AF case w/ 250mm side fan | onboard 7.1, with Logitech X-540 5.1 speakers | Chieftec 450W /w 120mm fan | Win 7 Enterprise 64bit
The pic has allready posted: http://i40.tinypic.com/2ryq43m.jpg
You can use one of the two points which are connected to pin17. Measure default resistance between FB and GND. Then multiply the number with 20 and that is the VR You can use.
Vmods - what can possibly go wrong? ©
Got a DMM on the way, Haven't looked over this thread cos it'll take hours, but i have a Sapphire 4850 with Dual Slot cooler, non reference board.
Are there any pencil mod's for this card?
Hi all this is my first post on this great Forum!!!
I have a Sapphire 4850 Toxic
I have started by testing the stock core voltages and I have then done the Bios mod.
Now a few things are puzzling me.
The core voltage after the bois mod hits 1.27 Volts and the card will clock to 795 MHz
I thought the most the Bios would give was 1.21V at the card.
But the strange thing is the memory. The card has Samsung 0.8 ns chips which are clocked out of the box at 1100 MHz but the data sheet from Samsung quote they are 1200Mhz chips with a max voltage of 2.05 +/- 0.05
My memory voltage is 2.13v at stock and the chips won’t even do 1105 MHz without crashing?
Is there a way of getting the memory voltage down or do you think raising it up a bit more will help the chips clock higher.
Thanks.
Bookmarks