RussianHAXOR
Folding Points:
Posts: 11791
Current Blue Ribbons: 3
Joined: 11/1/2005
From: San Diego
Status: offline
|
I recently read a few over-clocking guides in here and I disagree with some of the practices within them. Instead of de-railing those threads I thought I post my own thoughts on the subject. I wrote this for the 680i motherboard though it could apply to any.
Basics.
The Intel Core 2 line up runs on a 266Mhz quad-pumped FSB (QDR1066).
QDR refers to the quad-pumped FSB, but is often refered to as simply the FSB. So, 266 x 4 =1066. This motherboard (680i) officially supports up to a QDR of 1333, 333 x 4 =1333. Anything beyond this is icing on the cake.
e6300 1.86GHz 266MHz x7 (variable down)
e6400 2.13GHz 266MHz x8 (variable down)
e6600 2.40GHz 266MHz x9 (variable down)
e6700 2.66GHz 266MHz x10(variable down)
x6800 2.93GHz 266MHz x11(variable up and down)
Edit: By manipulating multipliers I'm betting the sweet spots for the chips I have tested seems to go like this.
e6300 2.3ghz 1333QDR L/S 667Mhz Ram===my granny can do this.
e6300 2.8ghz 1600QFR L/S 800Mhz Ram===Safe zone.
e6300 3.15ghz 1800QDR L/S 900Mhz Ram===pushing your luck.
e6400 2.66ghz 1333QDR L/S 667Mhz Ram===my granny can do this.
e6400 3.2ghz 1600QDR L/S 800Mhz Ram===Safe zone.
e6400 3.59ghz 1800QDR L/S 900Mhz Ram===pushing your luck.
e6600 3.0ghz 1333QDR L/S 667Mhz Ram===my granny can do this.
e6600 3.38ghz 1500QDR L/S 750Mhz Ram===Safe zone.
e6600 3.6ghz 1600QDR L/S 800Mhz Ram===pushing your luck.
Hope this helps someone. All of the above tested configs were achieved using the link/sync 1:1 FSB ratio...your results may vary.
Hope this helps some of you.
I’m going to assume you are a bit familiar with your BIOS and changing your FSB…etc. If your not, the manual explains most of this quite well and I've included a step by step further down. Rtfm.
#1 Mutipliers. I see no reason with this motherboard for anyone to be lowering their multiplier; you paid for that CPU’s performance. If you want, I’ll swap you chips, mine is a 7x and I wish I had coughed up the extra $40 for an 8x everyday. The official FSB support on this board is 1333. The closer you stay to that the better off you will be. That being said, you could make an argument for dropping the multiplier on the e6700 or e6800and maybe the e6600, but only under some extreme over-clocking circumstances. At the end of the day I feel it is better to over-clock the CPU, not the motherboard. The northbridge is probably running hotter than your CPU on this motherboard. Every Mhz in FSB you squeeze out just furthers its stress. Remember this is an Intel chip, the memory controller is where? Inside that blistering hot little northbridge that's where, and that is why I suggest leaving your multiplier alone. Faster CPU, slower FSB and Memory Mhz if possible. Two different paths to one goal, performance and stability.
#2 Old school. Turn off all those special features. Don’t let your motherboard help you…it’s not that bright. Turn off CPU Thermal Control, C1E Enhanced Halt State, Intel Speedstep, all Spread Spectrums, SLI-Ready Memory, Linkboost, and Smartfan. Later, after you get a decent stable over-clock, feel free to enable these and test your results, but for now turn them off. The two biggies here that I think are killing people are SLI-Ready Memory and Linkboost. These features automatically over-clock Ram and PCI-e respectively. Are you going to let the people who brought you the fabulous Ntune automatically adjust a single setting on your motherboard? I didn’t think so.
#3 Linked and Synced baby. In every single over-clocking guide I have seen the first thing they say to do is unlink the FSB and Ram. I strongly disagree. This was a common over-clocking practice during the P4 days of the past, when the FSB was 800Mhz and anything faster than DDR 400 was tough to come by. By today’s standards that would be 1066 and DDR2 533, are any of you running DDR2 533? I didn’t think so. Even I managed to scrounge up some DDR2 800.
So linked and synced your FSB to Ram ratio will be like this…
QDR 1066 =DDR2 533
QDR 1333 =DDR2 667
QDR 1500 =DDR2 750
QDR 1600 =DDR2 800
QDR 1800 =DDR2 900
QDR 2133 =DDR2 1066
This provides what I consider a true 1:1 FSB/Ram ratio, see images below. I know this is going to upset a lot of people that bought uber fast Ram. Hear me out (or write your own guide, lol). You should be able to run your Ram at whatever speeds you want, but as many have noticed this (speeds above 1000Mhz (this may be better now that p25 is out, I'm testing)) seems to be an issue with the 680i at this time. For the sake of stability I have found this to be a better option, without sacrificing much, if any, performance. Proof is further down.
#4 There’s more than one way to skin a cat. People who did by the uber fast Ram rejoice, I wrote this section for you. As many of you know with Ram the faster you go (in Mhz) the higher (worse) your timings have to be. Alternatively, the lower you go (in Mhz) the lower (better) your timing can be. So you can still get that $400 out of your Ram. Is this a work around to an issue prone motherboard, yes it is. I apologize. If you decide to follow the above steps (linked/synced) many of you will be running your 1066Mhz capable Ram at a lowly 800Mhz or so. No problem, dump those timings. I have tested two such scenarios and got the same benchmark scores either way.
I ran my DDR2 800 at 5-5-5-15 2t, and I ran again at DDR2 533 at 3-3-3-9 1t, the results were nearly identical. I used 3d Mark ’06. If you run a memory bandwidth test, then the DDR2 800 will win, but I don’t care…I’m after overall performance at the end of the day, and I don’t care how I get there.
In another such comparison I had “bigdawginva” (much thanks) try it with his DDR2 1000 Ram. With an over-clock to 3.37Ghz on an X6800 with DDR 750 at 3-3-3-9 linked/synced he scored 15233 in 3d Mark ’06.
The same system, same over-clock, unlinked DDR 1000 at 4-4-4-12 he scored a 15208. He actually gained a few points trying it linked and synced, but the scores are well within the margin of error. My whole point is that they didn’t drop when running Ram slower than advertised if you adjust your timings accordingly. Also note what Everest sees as a 1:1 FSB...interesting.
Pictures courtesy of “bigdawginva” get that guy a BR!
Only advice I have is be careful messing with the 1t/2t setting. 800Mhz and up use 2t, otherwise you can probably get away with 1t. I also never mess with the bottom half of the Ram settings in the BIOS; I would leave those on auto. Results may vary depending on how fantastic or ******tastic your Ram is.
#5 Automatic voltages. They are a good guide, but not the law. Not a lot to say here, but if the auto voltage is anything its overkill. I over-clocked the other night to 2.8Ghz (e6300(400x7)QDR1600) with all the voltages left on auto. The system gave my CPU a healthy 1.46v. I started rebooting and testing for stability to see at how low a voltage I could be stable at. I ended up being stable at 2.8Ghz with 1.2v. This shaved almost 10 degrees Celsius off my CPU temps. Now don’t get crazy, not everyone will boot at 1.2v, for sure you 4meg cache fellas won’t. I’m merely attempting to show you that the auto voltage settings can be flawed and are not custom tailored to your system. I found similar results dropping other voltages as well. I also found that increasing some voltages beyond what they automatically go to helped, as it the case with the FSB voltage. Every motherboard out there will be a bit different, if you’re running a bit hot, try some voltage changes. You never know. Intel's official stance on high temps seems to be 65c, while I think this is probably a bit conservative, I wouldn't dare let my system hit those temps. Ntune seems to report the temps a couple degrees cooler than reality. If you are constantly in the ~60c range I suggest you clock down or improve your cooling solution. (Remeber #2, Smartfan can be finicky with some fans, some fans will stop spinning at the 50% setting, while some continue at 0%)
#6 Nturd. This is easy, never use it for anything. Except maybe monitoring, maybe. It used to be okay for the VGA over-clocks. I had been using it with my 2 old 6800s, but ever since I upgraded to 2 7900gs…BSOD all day (and yes with the same driver). This tells me it has issues. Over-clock your motherboard in the BIOS, period. Use Coolbits for the VGAs until Ntune gets re-vamped.
EDIT: A new version of Ntune has been released while I was writing this, 01/25/07, I’ll edit if things change….yea right.
EDIT2: I fired up the above version of Ntune, wow. My initial response is...still sucks, but I'm going to keep testing. I tried the GPU over-clocking utility on my SLIed 7900gs, keep in mind they are 500/690 at default speeds, Ntune feels that my "optimal" speeds should be 950/820. Back to the drawing board ladies. Good thing eVGA has a lifetime warranty. 
EDIT3: It's so bad I'm not even going to bold this statement.
#7 Be conservative. As always over-clock with extreme caution always keeping an eye on your temps. I’m willing to bet everyone can easily hit a QDR1333, with a little tooling I’m sure most could post up a QDR1600, if your CPU is up to it. Always use/turn on your northbridge fan to at least 75%...if it’s too loud for you at that speed, quit over-clocking or replace it. I doubt that I am alone in thinking the northbridge, while over-clockable, is running too hot and the cause of many of the problems plaguing this motherboard. Otherwise good luck, not everything I have written here will help everybody, but it has helped some, and might help you. Perhaps the bugs will get ironed out of the 680i and you will be able to run your Ram unlinked at 1066, heck some of you can, but until then try this if you’d like. I’ve been rock solid since day one.
UPDATE: I originally posted this in the CPU thread, but it seemed relevant here also.
#8 Why can't my CPU do that? Understanding the computer industry, it’s manufacturing practices, and a bit about economics. CPUs are manufactured on wafers of silicon. These wafers are a set size, so only so many CPUs can be manufactured from one wafer. What does this mean for you? It means that, given the same architecture, it doesn’t cost Intel or AMD any more or any less to create CPUs of different speeds. For instance, an e6300 running at 1.86 GHz cost Intel the exact same amount of money to produce as the e6400 running at 2.13 GHz. Why the speed difference you ask? It’s just the way the chips, pardon the pun, fall. Believe it or not CPU manufacturing is not a perfect science. On any given wafer’s worth of CPUs, about 10% don’t even run! The rest are put through stability tests of different sorts and run to their maximum clock levels. Once each chip has been pushed to the failing point, the speed is backed off to a more conservative level, and viola a CPU is born and stamped with a given clock speed (AKA binning). Why are some of these clock speeds so conservative? Mainly because Intel, for example, has no idea where that chip is going. It could end up in my rig with an Arctic Cooler 7 Pro and four case fans…or it could end up in an insufficiently cooled Dell with a flaky proprietary power supply. Either way Intel’s name is on the line as far as this chip functioning as advertised, so they generally err on the side of extreme caution. I digress, in theory, on the same wafer of silicon; you might get some chips that run at 1.86 GHz, all the way up to 3.43 GHz. Neat huh? However, that is not the end of the story. Both chip manufacturers are ultimately in the process of making money. So when a new architecture is introduced, a lot of the times the first ones released to the public are set to very conservative clock speeds, that way in a few months they can introduce a “new” faster CPU, and charge a premium for it once again. In reality, a lot of the time, this “new” chip isn’t new at all. It’s just another chip that for what ever reason ran faster than is wafer-mates. Another interesting thing happens in the CPU industry. It involves large OEMs (like Dell) and large retail stores (like New Egg or Comp USA). If they wake up one day, call Intel, and ask for 100,000 CPUs of a given moniker. For argument’s sake let’s say the e6300. What happens if Intel only has 60,000 e6300s lying around? You might have guessed it, 40,000 e6400s just magically became e6300s. This is one reason why the lower number chips, as a whole, appear to be better over-clockers. As far as the price difference goes, the higher clocked chips naturally sell for a higher price for two reasons. Demand, and mathematically speaking, there are going to be fewer of them, so supply. In the above circumstance, market share, and sheer volume of sales will usually persuade both AMD and Intel to occasionally re-badge a chip. Sometimes it’s just marketing, for instance if AMD hadn’t been making the faster chips as of late, I’m willing to bet Intel would have charged nearly double for everything in its current Core 2 Duo line up. I’m sure that Intel dropped the price just to give AMD fits and regain some of the enthusiast market share it had lost. Remember, competition is your friend. There are of course a lot of other little things both companies do to make and save money. Special edition chips with the multipliers unlocked (think extreme edition Intels), enabling or disabling some of the cache, and renaming server level chips and selling them as desktop parts (think original socket 740 AMD64s) are just a few of these practices.
I hope what I have stated here today hasn’t overwhelmed anybody. I also hope this helps to answer the age old question of “why does your CPU over-clock better than mine, they are both the same?” Try to remember also, everything stated above applies in part to all of your other computer components. Ever wonder how eVGA and BFG can produce the same Video Card as another company, but with higher clock speeds? Basically, the same thing…they paid a premium for NVIDIA’s “best” chips. Motherboards and some of their parts also fall into the above discussion. There are a lot of other mitigating factors that will ultimately decide the performance of a given CPU, VPU, or whatever; but the above is the key logic behind the entire over-clocking mindset. When I over-clock my poor e6300, in my mind, I’m not running it out of spec…I’m running it right where it should have been, LOL.
Due to popular demand (mostly from noobs , we have all been there) I have now made a BASIC step by step BIOS over-clocking "slideshow." Not everything is included in what you are about to see (ie: memory timings)...I have given you just enough information to be dangerous. This is an example, not everyone will be able to run like this. This is just an example! You still need to keep reading, and be reasonable...meaning your not going to hit 4Ghz with an e6400 on air.
When you see this hit delete, if you can't figure this out...I'm afraid I cannot help you.
Navigate to the Advanced Chipset Features.
We will start with System Clocks. Hit enter.
Here you will find your multiplier settings, an bunch of useless (advanced) stuff, and the Spread Spectrum settings. Disable all spread spectrum settings, leave the rest alone. Hit escape.
Navigate down to CPU configuration. Hit Enter.
Disable what I have disabled for now, you can mess around with this on your own time. Hit Escape.
Navigate down to Nvidia GPU... and Linkboost, disable them both.
Navigate to FSB and Memory Config, now it gets fun!
Set the FSB Memory Clock Mode to Linked. Then select the FSB Memory Ratio. Hit Enter.
In this sub-menu select Sync. Hit Enter.
Take note of the default speeds. With my method of over-clocking you will not have to adjust your memory speeds. If what I have just said confuses you, go back and re-read my guide.
In this EXAMPLE, I'm going to set my QDR to 1600. This is a 399FSB. Notice how this will set your memory automatically to 800Mhz. NOT ALL CPU's WILL RUN AT THIS QDR, I HAVE AN e6300...IT HAS A LOW MULTIPLIER, THEREFORE IT IS SLOWER...IF YOU HAVE ANY THING HIGHER THAN AN e6400 YOU NEED TO TAKE BABY STEPS (like 1333). DO NOT JUMP DIRECTLY TO A QDR OF 1600. IF YOU DO YOU ARE AN IDIOT, AN I WILL NOT TAKE RESPONSIBILITY FOR WHAT COULD HAPPEN TO YOUR SYSTEM. 'Nuff said, not trying to be mean...
Hit Escape.
Let's make sure our fans are up to speed before we invoke this over-clock. Find the System Monitor. Hit Enter.
Hit Enter again to adjust the SmartFan settings.
Make your's look like this. You can do your own tweaking later.
Hit Escape twice, press F10, select Yes, hit Enter. Unless you went crazy with the QDR speed you should boot into Windows. If it doesn't boot into Windows, or boots into Windows differently than normal, there is a good chance you are not stable. You will need to lower your over-clock, or change some other settings. That, however, is too complex to explain right now. You will have to learn as I and all others did...by trial and error.
The lost art of over-clocking is not buying the most expensive/best stuff and hoping it goes fast...it's buying the cheapest stuff and making it go fast.
Depending on response I might make another step by step for memory timings, but after a while it will come to you. If you have any questions regarding my "theories", feel free to post, I'll respond ASAP. Thanks all.
< Message edited by EVGATech_JacobF
-- 6/15/2007 4:02:59 PM >
|
|
Products
Printable Version
Overclocking with the 680i (pwnzor's thread revived) - 
New Messages
No New Messages
Hot Topic w/ New Messages
Hot Topic w/o New Messages
Locked w/ New Messages
Locked w/o New Messages
Post New Thread