PDA

View Full Version : WIN: 3D Mark Vantage Licenses!



LardArse
04-18-2008, 11:43 PM
http://img89.imageshack.us/img89/9792/vanuk7.jpg (http://www.futuremark.com/)

We all know about the impending doom... yes benching on Vista.. :rofl: While we await it's release, let's throw a small contest to give out 10 Advanced Edition Licenses!!

Contest is for every/anyone. One prize per person only. Well you bunch of geeks claim you know so much about Overclocking... well, share it with us!! Choose a topic on overclocking, any part of it (a particular voltage, particular form of cooling, etc etc) and write an INFORMATIVE article on it, and post it here! The top 10 informative posts shall win!

Contest open for 1 Month from now, or if 3D Mark Vantage is not released by then, then open until it's release.

Courtesy of the Quantum Force Team

massman
04-19-2008, 12:32 AM
Will those informative articles be used in any future articles/essays of Quantum Force or is it solely for this contest?

LardArse
04-19-2008, 01:02 AM
Will those informative articles be used in any future articles/essays of Quantum Force or is it solely for this contest?

solely for this contest massma, heheh i didnt think that far ahead, just wanted my free copy of Vantage :D

cadaveca
04-19-2008, 01:04 AM
I'm in!

Overclocking Basics, Top 5:


#1. Have a goal. Are you trying for a record or breaking hardware?:eek: Um...Yeah...nuff said.

#2. Prep your harddrive prior to clocking. Make an image...you might need it. Twice.:mad:
#3. Know your board. Left something on auto? It does what? Are you sure?:shakes:
#4. Multimeters and temperature probes are expensive, but so is dead hardware. ;)
#5. Screenshots, pictures, and video. Screenshots are good, but noone will dispute a video. Amazingly, screenshots can be made from video too...



:up:

massman
04-19-2008, 01:20 AM
solely for this contest massma, heheh i didnt think that far ahead, just wanted my free copy of Vantage :D

Great, I'll give it a shot!

alexio
04-19-2008, 01:26 AM
I'm in too :)

AndreYang
04-19-2008, 04:27 AM
i want it~~:D

manyga
04-19-2008, 04:30 AM
looks good!

rdrash
04-19-2008, 04:39 AM
I'll give this a go...for the fun of doing so...

BenchZowner
04-19-2008, 05:06 AM
Nice idea Sham.
I'd love to participate ( yeah, I do love writing tech articles & stuff ) but I'm not greedy to participate in a contest to win a license of something that I'll get for free anyways.

Gratz to Foxconn for making lots of steps into the right direction ;)

M.Beier
04-19-2008, 09:26 AM
Great :D
Im sooo in... :p:
When is deadline Peter..? Im just about to write two articles... one on calculation benchmarking, tweaks, effects of different settings....

And another on 3Dmarks... ;)

Hehe, btw, dont care about the price, just in it for the sport.. =)

EDIT:
SPI article in progress... Had 30 mins in the train, 2 pages down... Sneak preview:

SPI optimization.
The entire concept behind SPI, SuperPi, is calculating digits of Pi. This task uses the computers RAM and CPU.
When using the application SuperPi, which is a program, we want to make the two components used for calculating as available as possible, and as fast as possible.

In this article I intend to summon up the things one can do in order to get best possible results:
Overclocking
Timings
Windows optimization
“cleaness”
Ram availability
Cooling
Compromises
BIOS

Compromises:
Prior to the release of Core Duo & Core 2 Duo, people used either Intel Presscot 670, or the AMD 64 s939 for SuperPi. With A64 one had to coop with coldbug, especially in the 130nm processors this could be postponed by ease the IMC load.
Using double sided ram gains one best performance – as it enables 4-way interleave, which were supported by the IMC. The sticks for SuperPi were the low latency ram, typically BH5 – 2x 256MB which overclocked a tiny bit better – were single sided, and 2x 512MB, which allowed a lot better performance double sided. With a typical A64 setup, overclocked ofcause, one lost 0.12-0.18s by using single sided compared to double sided, however, this often allowed another 25 degree Celsius on the chip before the coldbug appeared.
So the choice was… 100~175mhz vs. more beneficial ram.

Cleanness:
When doing benchmarks, and hunting good results, one ALWAYS wants a perfectly clean install, as it gains higher performance.
When doing 2D benchmarks, one installs OS, and ONLY motherboard drivers, and the few applications needed for the benchmark; The benchmark program, MemSet, SetFSB, CPU-Z.
- The benchmark application is best off installed in C:\”application name”, directly in root.
Apart from this, one must put a big file somewhere on the partition, in order to do the Copy-Waza or CDT tweak which will be explained later in the article.

Overclocking:
Overclocking is a major topic, and already familiar to the potential readers of this article, so I’ll just summon up:
overclocking is clocking frequencies above their rated speed – This can be done within BIOS or OS.

BIOS:
In order to gain as clean install windows as possible, you have to disable all unnecessary components on the board, such as sound, LAN etc.
Also, its better to set the timings (explained in other chapter) to aggressive settings within the BIOS, as the “hidden” subtimings will be effected, unlike when changing in windows.
Also be aware that some BIOS is faster for 1M, not just the newest, usually for 32M, the newest is the fastest.

Cooling:
In order to gain the better frequencies, one needs to keep the equipment as cold as possible, or more accurately, finding the sweet spot, for instance, liquid helium is too cold for most electronics.
Cooling the CPU will allow higher frequency on the CPU itself, cooling the ram will typically allow higher frequencies, and/or tighter timings. Cooling the northbridge will result in it being capable of handling a larger amount of data stream – which can be the bottleneck for unleashing all the performance of the ram.

Ram availability
As mentioned previously in the article, calculation benchmarks are loading the CPU and the ram, thereby; keeping the ram “clear” is of great benefit to the benchmark.
This can be done in two ways, CDT or Copy-Waza.
The concept of the tweaks is moving data from one partition to another, as this uses the memory, and enables full performance for the calculation, this combined with the use of “MAXMEM”, which is set in msconfig, will benefit the result for large calculations.
Copy-Waza can be done by:
Transferring a file, ex. 800MB from:
C: -> D:
D: -> C:
C: -> D:
Personally I’ve used a image of a random game for this tweak, a windows image, file doesn’t matter, however, 800-1200MB is recommended.

As for MAXMEM; 600MB limit is still the recommended amount

Timings
Timings are very different from setup to setup, its very much a feeling of the system, as timings can easily be bottleneck, weakest link!
When optimizing ramtimings, one typically divides timings into two cathagories, ram-limited and IMC/NB limited, some timings, such as PL on core config - a timing that havent got much to do with the ram, same goes for 1T/2T.
Unlike the previous mentioned timings: we have the 4 main timings; CAS, ras to Cas, Ras precharge, tras - all limited by the ram itself.
Severe guides on how to configurate can be found on the web, yet, the best is to try yourself!
- The tricky part is, sometimes, lower isnt better, personally I prefer testing with the program "PiFast", as it always get influenced by change in ram timings, and thereby is uniq for testing if you've done something wise, or unwise, as BW reading programs isnt accurate enough.
When finding the minimum, you can figure what timings are effected by increasing voltage, some are, while others arent.

Windows optimization
Windows optimization can be done in various ways, each equally effecient, nowadays, some of the tweakers has released .reg files, that will clean everything for you in the regristry.
However; msconfig; MAXMEM, choosing amount of cores, killing applications.
Services under control panel / administrative tools.
reg (run -> regedit)

Typically first thing one does, if not having a .reg file to fix all:
1) Kill pause-monitor.
2) Select high performance (my computer, rightclick -> properties).
3) System cache.
4) Regedit -> Ram (HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contr ol\Session Manager\Memory Management).
5) Delete all uneeded applications, such as messenger.

Install tools in a folder on system drive, KEEP desktop CLEAN.
Install benchmark application in "C:\" root!

Affinity:
When using dualcore; select core#1 as only core for superpi, as its a single threaded application, and all your stuff will run on core#0.
When using quadcore; select core#2 as its the first core of second DIE, each DIE has shared cache, this makes the result a little better!
- Select realtime on the application afterwards and run :)

When running SuperPi, remember to CLEAN;
Run 2-3x 16K calculations, then run 1M or 32M, the one your up for!

Final words:
Higher FSB equals better score! Allows higher bandwidth!

WORD LIST:
AMD – Advanced Micro Devices
Intel – Integrated Electronics
IMC – Integrated Memory Controller
RAM – Random Access Memory
LAN – Local Area Network
PL - Performance Level
**Coldbug – A given temperature the component fails below.

---
Final update made CET(+1GMT); 23:45
Made it fast, however, yet complete, if wanted, I can redo and make it better at later time... Sorry :( Kinda forgot all about the deadline.

71 (Bryan)
04-19-2008, 01:11 PM
reserved =)

xlink
04-19-2008, 01:23 PM
HAHAHAHA

I was going to write something like this for a science class...

jas420221
04-19-2008, 10:49 PM
Pay $20....or write an article...Hmm, Im done with college, so $20.

EXCELLENT contest though, what a help for this site (right?)!!!!

M.Beier
04-20-2008, 01:16 AM
Pay $20....or write an article...Hmm, Im done with college, so $20.

EXCELLENT contest though, what a help for this site (right?)!!!!

hehe, I doubt any of us will do it for the price... ;)
More the fact that its a contest... Could be nice if FOXCON threw in:
1. License + GFX + MB
2. License + GFX
3. License + MB

s0lid
04-20-2008, 10:28 AM
Dice article reserved! :D
-Starts writing it here @ now, because i can edit it with my cell phone, school comps ect. Hope u understand :)

This is basicly for Cpu overclocking with dice.

Dice
What is Dice?
Dice also known as Dry Ice. Dry ice is basicly carbondioxide and it's sublimation point is at -79C. Because dice transform to gas every temperature above it's sublimation point, it's just perfect for overclocking since you don't have to empty the "pot" from liquids or any other melting results. Also Dice is guite cheap and with 10kg it's possible bench ~10-12 hours.

What we need?
5-20Kg of Dice
Pot(s)
Mostly any liquid including much of ethanol or it's freezing point is under -79c.
Insulation goods
Your rig
Silicon based thermal paste, Thermal pastes including silver or other metals doesn't work below 0c!
Gloves
Hairdryer
Of cource good pioneer spirit and couple sixpacks "install beer" if necessary :ROTF:

Pot
These metal "mugs" are basicly made from aluminium or copper. Their basic idea is contain the Dice and transform the heat from cpu or gpu to dice.
Copper is of course better material for this use since it's better thermal conductivity but it's much more expensive and more difficult to fabricate than aluminium.
There is basicly four types of pots:
Cpu, Gpu, Memory and North bridge

Picture from Cu CPU pot and Alu Gpu pots, pic by Sampsa:
http://plaza.fi/s/f/editor/images/X-20061231141756921.jpg

Insulation
Good insulation is always very crucial for long time benching. Because almoust every temperation below ambient produces condensation water.
Condensation water may kill your hardware what ever you are cooling with sub zero methodes, cpu, gpu, memory and especially motherboard.

Basic idea is to separate cold areas from air when condensation water is not problem.
However condensation water isn't an problem so long it stays frost or ice.

Good insulation goods:
Foamed plastic, Insulates well. Good product name: Armaflex
Household towels, "drinks" moisture
Diselectric grease, great stuff for insulating sockets or areas below foam on motherboards or graphic cards.

Preparations
Take piece of 8mm foam size of 9x9cm. Cut to it's center hole which is size of your cpu socket, now try it on and you can see than it doesn't fit there just like that. Now you need to cut holes for capasitors and inductors. If needed cut holes for mosfet coolers.

Now when your socket area insulation foam is done it's also good to make insulation from foam for your cpu pots mounting kits backplate, if any. Backplates insulatiopn comes between motherboard and backplate.

Insulating the pots, with cpu pot you need basicly just wrap foam around the pot and seal it with tape. But don't insulate the bottom and from inside.

I recommend to put some household towels between motherboard and foam because it drinks moisture.

Picture from well prepared motherboard, pic by OC_Windforce:
http://www.nowpug.com/bbs/data/pc_cpu/IMG_0002.jpg

Open your window because from 1Kg of dice sublimates 500litres of carbonoxide!

How to use

Now when you have insulated your motherboards socket area, mounted pot(s) and ready build your rig to your table it's time to bench with dice :)

Half fill your pot with dice grains and boot your system, you can handle dice a short time with bare hand but if you want use gloves or some small plastic shovel. Go to bios hw monitor and check than your cpu temp goes under ambient and 0C. Now it's time to add that ethanol to the pot(s), It helps temps going lower, only 1-2dl needed. If there generates some fog there is no need to worry it's only good thing. Now you are ready to bench with dice and have fun.

Example from Dice fog, my pic :)
http://allthemods.com/img/04/2789.jpg

Ending the dice benching

When you are ending the benching session make sure than your pots don't have anymore dice in it, your rig is off and you have removed the mounting kit. Take the hairdryer and start heating the pots and after short time try carefully twist the pot, if it moves try pull it off. Now take off all insulations from motherboard and check there is no ice or frost under them, otherwise take the hairdryer and dry them off but don't overheat your motherboard , GPU or the pot when it's still on the cpu!

If you have still some dice in your icebox or where u received your dice, just put them outdoors or your carage they will dissapear in couple hours if there is under kg of dice.

I Hope this was an informating article :)

Harshal
04-21-2008, 05:25 AM
Nice one, Sham :D
Honestly, I am too lazy to get into this :p:
Good luck to those who are in for it :up:

I34z1k
04-21-2008, 07:34 AM
Ooo I'll try :D

inCore
04-21-2008, 07:35 AM
Overclocking and the Effects of Electron Migration

http://i182.photobucket.com/albums/x157/gLipgLipgLip/intel-prescott-wafer-300x325.jpg

Previously, heat killed the CPU (Central Processing Unit) long before one were able to observe the phenomenon of electron migration. The only thing stopping overclockers in their pursuit for higher speeds was heat output and with good enough cooling, clocks of over 8 GHz could be seen with Pentium 4 processors. However, since the introduction of 45 nm CPUs, users are increasingly hearing of electron migration, a feared case of CPU death.

So, what is this electron migration we hear of?


To answers this question, we can begin to look at either the physics of voltage and overclocking, or several analogies which aid to explain the same concepts.

As stated by Ohm's Law, the voltage is equal to the current times the resistance, or:

V = I R

Let’s define the terms first:

V = Voltage

Voltage is more technically defines as a potential difference (p.d.), or the difference in electrical potential between two points. However, for the sake of simplicity, it can be said to be the pressure behind the electrons. A higher voltage means higher pressure, or more force (although it is not really a force, but imagine it as one) pushing the electrons around the circuit.

I = Current

Current is the flow of electricity. A higher flow means more electrons passing through the same point at the same time.

R = Resistance

Resistance is the opposition of a material against the current i.e. against the flow of electricity. This resistance is affected by many factors, such as the temperature of the material or simply the nature of the material itself. Conductors have a small resistance whilst insulators have an infinitely high resistance.

From the formula (V = I R) we can derive that when voltage is increased, the product of the current and the resitance increases too.

As we all know, to increase the CPU clocks after a certain FSB, the voltage requirements increase exponentially. This has partially to do with the non-Ohmic behaviour of the resistance involved – as voltage increases the resistance does not increase linearly. At a higher voltage, the resistance is less than before and there is therefore a higher current. This non-Ohmic relationship is often caused by the operating temperature and a higher temperature will cause a lower resistance.

The resistance can be isolated from the above formula by re-arranging the terms:

R = V / I

Extreme cooling methods, such as Dry Ice or Liquid Nitrogen (ln), stabilize the temperature and make the resistance Ohmic. This means that the V and I relationship is constant, and always equal to R. This enables a more predictable overclock.

Diagramatically, Ohmic and non-Ohmic behaviour looks like this:

http://i182.photobucket.com/albums/x157/gLipgLipgLip/Ohmic.jpg

http://i182.photobucket.com/albums/x157/gLipgLipgLip/NonOhmic.jpg

The two diagrams above show the relationship between I, the current, and V, the voltage. The gradient of the line at any point is equal to the resistance. The left diagram shows non-Ohmic behaviour and the right diagram shows Ohmic-behaviour.

Problematically, an electrical component, such as a wire, is only capable of a certain electron density (electron flow per unit area). The introduction of the new 45 nm CPUs have made these “wires” smaller than before and in order to keep the current density the same as for 65 nm CPUs, the current had to be decreased. In other words, as one can only control the voltage on a motherboard, the voltage had to be decreased to maintain the same current density.

The problems of electron migration are best explained using analogies:

1. A doorway is wide enough for a set number of people. This is the stock speed of the CPU, a number of people which will fit through the door fine, with room to spare. The voltage is a person pushing the people through the door from behind. At a low voltage, the person does not have to be strong, assuming that the people will fit through the door without trouble. However, as one increases the FSB, more people are trying to go through the door at once. They find it increasingly difficult and with a weak person pushing them from the back, they are unable to get through. This is the case of a too weak voltage when trying to reach a higher FSB. A higher voltage must be applied, i.e. a stronger person must be pushing them. This process can be repeated until higher FSBs and voltages. After a certain point, namely the maximum rated voltage of the CPU, the amount of people will start damaging the sides of the door. Adding an even bigger person pushing the people will enable the people to through, but the sides of the door will start to break. If an impossibly huge number of people is forced through, the door will collapse completely, and you are left with a fried CPU.

2. A similar idea can be applied to a river. Too much water flowing through will corrode the sides of the river until it breaks the river banks completely.

Naturally, these analogies do not take into consideration other factors such as the overclocking abilities of the motherboard, etc.


In conclusion, increasing the voltage will enable a higher overclock. It provides more total power and hence a higher FSB. The problems of electron migration are becoming more severe than previously, and as technology gets smaller as time progresses, one can expect an even lower tolerance of volts as we approach the even smaller nanometers.

I hope that this article has been informative, thanks for reading.

inCore

FrogBite
04-21-2008, 07:39 AM
Nice article, physics rules :D

Good luck with the competition.

LardArse
04-21-2008, 03:12 PM
I hope that this article has been informative, thanks for reading.

inCore

Very, that was a good read for sure!

Solid, waiting for the full Dice guide now :)

s0lid
04-21-2008, 09:03 PM
Very, that was a good read for sure!

Solid, waiting for the full Dice guide now :)

Done and it's s0lid with Zero not O :p

s0lid
04-23-2008, 10:29 PM
How it looks like than only me, M.Beier and inCore are going to get free licenses :D?

BenchZowner
04-23-2008, 10:58 PM
///joke/// starts writing various articles, pass them on to 10 different people to post them here, so s0lid won't get a license ///joke/// :p:

dinos22
04-23-2008, 11:07 PM
lol article writing while fairly easy and not so time consuming still aren't going to attract too many people participating as a 3DMARK challenge would ;)

good luck to guys that enter :up: :up:

mike
04-23-2008, 11:18 PM
How about the first price gets a professional license???

Just kiddin - awesome initial submissions so far!

LardArse
04-24-2008, 06:23 AM
:) comeon guys there's tonne of knowledge here, I'm sure the 10 copies will go out :yepp:

dinos22
04-24-2008, 06:32 AM
lol would recording a TV show work
i covered watercooling in the first episode :D

http://goodgames.i4memory.com/I4memory.com_on_ABC2.divx

here is a shorter and poorer quality video on youtube http://www.youtube.com/watch?v=ErjGoxY26sg

http://img81.imageshack.us/img81/3230/camerasetslidogik4.jpg

s0lid
04-24-2008, 06:37 AM
:) comeon guys there's tonne of knowledge here, I'm sure the 10 copies will go out :yepp:

Hehe if the contest were: "WIN: Foxconn Blackops" here would be 100 articles within first hour :ROTF:

loonym
04-24-2008, 07:02 AM
note: repost from a thread buried in WCG section.

Half assed work log, ready, set, go:

The base on the ThermalRight tower sinks are not close to flat and I wouldn't consider purchasing unless I was prepared to resurface. The IHS on recent intel cpus are flatter than I have seen on past models but they can still benefit from some wet/dry luv too :D Here's a recent work log for a TRUE:


I use 4 grits only. I don't care for mirror finish, only flatness. http://images6.theimagehosting.com/100_2426.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2426.JPG) So you see, 220,400,800 and 1000, along with the surface I work on.

I tape the paper down http://images6.theimagehosting.com/100_2428.ed7.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2428.ed7.JPG) this just simplifies things so much

the thing needs lots of 220 work http://images6.theimagehosting.com/100_2429.99e.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2429.99e.JPG)super convex center and concave outer, some say figure 8 but I like back and forth, rotating the cooler 45 degrees every 15-20 strokes, hold the cooler with only enough pressure to prevent chatter or tipping

finally ready to start with 800 grit http://images6.theimagehosting.com/100_2441.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2441.JPG) when the plating begins to disappear I make the reference lines so I can visually gauge my flatness as I progress, water on the paper becomes more important at this stage to eliminate chatter

almost ready for the 1000 http://images6.theimagehosting.com/100_2445.ecd.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2445.ecd.JPG) so slow, now I wish I did more 220 :D by this point I am sanding only in one direction, rotating the cooler 180 degrees every 20 or so and being careful to apply even pressure

wooo hoo finally the 1000 http://images6.theimagehosting.com/100_2449.f7f.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2449.f7f.JPG) I am pleased with this result

this is my 'polish' http://images6.theimagehosting.com/100_2450.50e.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2450.50e.JPG) like I mentioned, I don't care about mirror, just flat, clean and cool is ok

close enough for rock and roll http://images6.theimagehosting.com/100_2453.th.JPG (http://server6.theimagehosting.com/image.php?img=100_2453.JPG)

M.Beier
04-24-2008, 07:09 AM
Did a minor update... However, I think I have to re-write that copy-waza theory :p:

Apart from that, the timing recommendations will be fun :)
And XP walkthrough as well

TheKarmakazi
04-24-2008, 09:16 AM
rserved for multimeter article

ExodusC
04-24-2008, 02:07 PM
I guess I'll give it a shot... Post reserved.

Serra
04-24-2008, 07:33 PM
Ah drat, I'd been waiting for post #26 so I could have the top of the page :p

Spot reserved. Topic.... hmmm.... I think something on RAM timings versus frequency maybe.

s0lid
04-25-2008, 08:11 AM
Couldn't this contest be turned to win new QX9650 and Blackops mobo :/?
It seems than my DFI which i received today killed my QX and bunch of other hw :mad:

I've tested that QX and p4 631 with other mobo, no boot :(

Looks like i'm not going to play with dice this weekend...

E: phew everything else works but not QX, maybe DOA :/

[XC] leviathan18
04-25-2008, 09:24 AM
i will save my spot here

ghost101
04-25-2008, 01:21 PM
~reserved~

ahsalloum
05-02-2008, 04:36 PM
Reserved

s0lid
05-09-2008, 06:23 AM
when we'll get some licenses :D?

LardArse
05-09-2008, 06:45 AM
when we'll get some licenses :D?

05/18 i guess :)

Maarten-NForce
05-10-2008, 07:15 AM
Guess I'm gonna make an how-to damage your hardware, I'm kind a good in that :D
Nah, just kidding, voltmodding reserved :)

Beginner Voltmodding

Materials needed
-Soldering iron (18 Watt recommended)
-Some solder (Flux core solder)
-Variable resistor ( Multi-turns recommended, Brand: Helitrim)
http://www.dickbest.nl/webshop/images/uploads/balk2.jpg
-Some thin wire

Basics

Our example card:

http://overklokken.info/images/user/maarten/ISL65221.jpg (http://overklokken.info/images/user/maarten/ISL6522.jpg)
(Click for bigger)

For example, you've got yourself an Ati HD1337, and really wanna push him a bit more. The main thing is find the IC that is regulating the Vgpu. Known brands for example are InterSiL(ISL) and VolTerra (VT). Let's say there is an ISL6522CB IC on our high-end HD1337. The you gonna think logical. If it's really close to the GPU, then it's probably the Vgpu, if it's close to the Vdimm, then that's the IC for the Vdimm mod. Now you're gonna use your best friend, your mothe... uhr ahum: Google! Google on "ISL6522CB datasheet" and then you'll see a few datasheets. Click one, and open the (most of the time) PDF Document. I've came to this one:
http://pdf1.alldatasheet.com/datasheet-pdf/view/67657/INTERSIL/ISL6522CB.html
You'll see a datasheet, with the IC seen from the top. All the 14 pins are numbered. The pin near the "notch" is number one. One of the pins is called FB. This is the pin you are looking for. Measure the resistance between ground and this FB pin, multiple it by ~20 times, and you have your resistor. For example the resistance is 2KOhm, 2 * 20 = 40. Then you'll need to pick an 50 KOhm 20/15 turns variable resistor. How more turns the resistor has, how more accurate it is to change the Vgpu/Vdimm/Vyou-name-it. Now solder the Variable Resistor between the FB pin and the Ground, and you made your own Vgpu mod! The Vmem mod is actually the same, but then on a other IC.

So remember:
-Making a 0 Ohm connection between the FB pin and Ground is BAD! mmkay!
-If you have a Variable Resistor like in the image, don't use the pins on the outside, so use pin 2 and 3, or pin 1 and 2, but don't use pin 1 and 3, cuz that duzn't workz.

Measuring points

On the front off the card you'll see some coils. Turn the card on his back, and look at the coils. One of them is a ground, and the other one is a measuring point. Great, we found the measuring point of our HD1337!

OVP

Then the following two things; let's start with the OVP mod. OVP, aka Over Voltage Protection, everybody hates it, let's admit :D This, ánd the OCP (Over Current Protection) mod is a thing you (nearly) always can find in the datasheet. Let's take a look at our ISL6522CB Datasheet. When you go to page 5, you can see that the names near the pins are explained. As you can see, there isn't a OVP in the IC himself.

OCP

When you look further in the datasheet, you'll see "Overcurrent Protection" Right, this is what we're looking for. For the OCP mod we need to lower the resistance of the OCSET pin. You can see the following text:


OCSET

Connect a resistor (ROCSET) from this pin to the drain of the
upper MOSFET. ROCSET, an internal 200μA current source
(IOCS), and the upper MOSFET on-resistance (rDS(ON)) set
the converter overcurrent (OC) trip point according to the
following equation:
An overcurrent trip cycles the soft-start function.

So if you want to get rid of that damn OCP, lower the resistance of OCSET.

Deux
05-17-2008, 06:17 AM
Hope a PDF is ok. I can repost here if not.
http://deux.emrav.com/fsoc.pdf

JMKS
05-17-2008, 02:42 PM
The physics of overclocking

Have You ever thought about it, why such parameters as voltage and temperature are so crucial in overclocking? The obvious answer is - the CPU must have more electric power to operate at higher frequency. Yes, of course but what the proccessor "feels"? I'm going to answer it theoretically from the physics point of view, at least partially ;) (it is very interesting but to answer it all one must write hundreds of pages for sure :)).

The main term: propagation time delay (tp) - the time required for a signal to travel from one point to another. In case of that article it is a time between the signal on microprocessor changes from a value that means "0" to "1" (well, in fact it is more complicated, but it is not necessary to know all of this to overclock ;)).

http://img187.imageshack.us/img187/5755/25163209bx5.jpg

As we can see in that image, there are some unnecessary plateaus (where processor is just waiting for another cycle) - by overclocking we are shortening them :).

At given voltage and temperature we will reach a maximum frequency (fmax). The first conclusion - maximum frequency is inversely proportional to propagation time delay.
fmax ~ 1/tp
Pure physics as we can see :).

To increase fmax we must of course decrease tp.
How can we do it? This is the most interesting part.
Currently microprocessors are being produced in the CMOS technology, which means Complementary MOS (Metal Oxide Semiconductor).
There are 3 main parameters that affects tp in that technology: temperature (T, in kelwins), voltage (U, in volts) and fan-out.

http://img293.imageshack.us/img293/1492/86584046ye5.jpg

Fan-out - a measure of the ability of a logic gate output, implemented electronically, to drive a number of inputs of other logic gates of the same type. In our case it is a part we can not change (it is determined by manufacturer) so I will not elaborate more about this.

Voltage - as we can see tp is inversely proportional to square of U, so fmax is directly proportional to U^2. What does it mean for us?That by increasing voltage we can make our chip running faster. More important - that the fmax will rise like a square of voltage.
This is not that easy to increase frequency - as You can see in the image, we can not reduce tp to very small value (which will give us a huge frequency) - there is a horizontal asymptote.

Increasing the voltage is in fact not very difficult, so let's move further - to the "xtreme" section, by lowering the temperature :).
Many people thinks that it is only important to lower the temperature to avoid overheating. Wrong! As we can see in the image tp is directly proportional to T. The effect is simple: lower temperature => lower propagation time delay => higher frequency.

If we also remember one (approximated) formula to calculate power (it equals produced heat, so we may calculate the delta between heatsink temperature and processor temperature):
P="TDP"*(frequency/frequency_default)*(V/V_default)^2
we can approximate how CPU will overclock in various conditions from stock cooling to liquid nitrogen :).


That was, as I said before, the theory (and just a part of it), based on physics laws.
Practice is in fact often different ;). Real world is not ideal, that is for sure. And more important thing is, that those effects differs more on "xtreme" situations like very high voltage or very low tempearture. So I believe we can use those equations when approximating differences between stock cooler on standard voltage and some high-end air cooler with slightly higher voltage - in that case we should expect pretty good approximations.


But if You are curious to know:
http://img513.imageshack.us/img513/51/dsc00035jl5.jpg
if 2,2V on CPU will help...

http://img176.imageshack.us/img176/521/dsc00096te0.jpg
or if -196°C will help... (as we were formally "speaking in kelwins", so 77K actually ;))

well - You must test it yourself, good luck :D

I hope that I inclined You to some additional reflection between pushing voltages and freezing hardware :).

M.Beier
05-17-2008, 02:59 PM
I will finish mine later today ;)
Its 18-5, Im gonna fix it CET in time, hope it'll be accepted :)

s0lid
05-18-2008, 01:22 AM
How many minutes to go :D?

FrogBite
05-18-2008, 03:33 AM
8 hours.

loonym
05-18-2008, 09:02 PM
lol, I guess the prize wasn't a very popular incentive for most vs the effort. :p:

Movieman
05-18-2008, 09:07 PM
Wait? The article has to be on Overclocking?
What is that?
Dave the n00b..:p:

dinos22
05-18-2008, 09:07 PM
How many minutes to go :D?

:ROTF:

loonym
05-18-2008, 09:24 PM
Wait? The article has to be on Overclocking?
What is that?
Dave the n00b..:p:Just watch dinos22 movie and he'll show you :D

dinos22
05-18-2008, 09:33 PM
now now
let's play nice fellas :ROTF:

M.Beier
05-19-2008, 02:06 AM
Seems like fairly good odds getting a license :D

inCore
05-19-2008, 01:33 PM
Yeah, I'm actually going to be somewhat surprised if I don't get one. :p: Gotta love lazy XS members who consider typing more stressful than a days session with sub zero liquids.

Maarten-NForce
05-19-2008, 10:03 PM
So, like when it's going to be ended? The topic start says it ends the 19th of May?

M.Beier
05-21-2008, 02:04 PM
Any news?

mr_knowitall15
05-21-2008, 02:18 PM
Neat idea. I saw the thread and imagined a contest where highest 3DMark06 scores won. I thought, "oh awesome. A contest that rewards people who have enough money to buy all the best, no compromises hardware." As in: Oh you spend thousands and thousands on your rig, and quite a bit having exotic cooling made? Enjoy this $20 software. It made me lol.

THIS is a cool idea though. I like it.

LardArse
05-21-2008, 07:14 PM
hey thanks for participation!
i have pmed the winners the keys! Congrats! :up:

loonym
05-21-2008, 07:28 PM
If you would have given blackops as prizes you would have needed to hire extra judges :D

s0lid
05-21-2008, 09:24 PM
Daaamn this must be first time than i won something for a looong time :D

Maarten-NForce
05-21-2008, 10:53 PM
Daaamn this must be first time than i won something for a looong time :D

With ^^ :D

FrogBite
05-22-2008, 01:08 AM
So who won?

s0lid
05-22-2008, 01:21 AM
So who won?

Well i got my cd-key this morning :D

Solarfall
05-22-2008, 06:30 AM
Hehe if the contest were: "WIN: Foxconn Blackops" here would be 100 articles within first hour :ROTF:

:D:D:D im 100% sure that this would happen :yepp::yepp::yepp:... so far nice articles guys :up::up: keep em comming

LardArse
05-22-2008, 07:13 AM
3DMark Vantage Hotfix Released

Changes:

1. Off-line score viewing is now available in the Advanced Edition.

2. Score save / submit saved options are now available in the Advanced
Edition.

:up:

jmke
05-22-2008, 07:41 AM
that's the $20 version, right?