Author: Aaron Gordon
Date: 18:10:51 07/03/03
Go up one level in this thread
On July 02, 2003 at 20:40:43, Vincent Diepeveen wrote: >On July 02, 2003 at 17:51:47, Aaron Gordon wrote: > >Just because you can work with your hands at a microchip doesn't mean you know >anything from testing a chip to be not losing bits and bytes. > >It's even intel that has problems with it. They had to clock back itanium2s from >1Ghz to 800Mhz. > >Now you would of course clock that thing back to 1Ghz, because it doesn't blow >itself up nor crashes within 24 hours :) > >But i am sure you will never understand the above :) You still don't understand I clocked the chip *BACK* enough for it to be perfectly stable. Perhaps you should re-read the previous post I made. >>On July 02, 2003 at 11:24:12, Vincent Diepeveen wrote: >> >>>>Since the cores are all technically 2.4GHz, running one of these cores (from the >>>>hand-picked XP 1700+ chips) at 2.0-2.4GHz isn't overclocking at all. This is >>>>similar to taking a P4-3.06 remarked to 2.4GHz, then modifying it back into a >>>>3.06.. is that overclocking? No. >>> >>>It is overclocking by definition. A product you bought to be running X you run >>>at y > x. >>> >>>It even more emphasis to the average user to not overclock himself. >>>Overclocking is a dangerous business. >>> >>>A big problem of overclocking is that the reliability becomes less of for >>>example the FPU. Now you might care shit for this, because you only 'test' it >>>for a few seconds, but the reason those cores are clocked for example at 2.2Ghz >>>where you find out that they 'work' for you at 2.4ghz is because some parts at >>>the cpu are not handling 2.4Ghz very well. >> >>Only test it for a few seconds? Read my webpage AND previous post. I test for 24 >>hours on single CPU chips and 48 hours on dual capable chips. >> >>>A good example is the itanium2 cpu's 0.18 which were clocks 900 and 1 Ghz. Intel >>>has either all ordered them back to factory or clocked them back to 800Mhz, >>>because they found out that in some calculations users lost bits. >>> >>>It is exactly this that will happen at your chips. >> >>Just because you can program and babble on about matrix garbage doesn't mean you >>know a single thing about overclocking. The chips *ARE* stable. Not a single one >>has had a problem after I've shipped it to a customer. The methods and equipment >>I use ensure full stability under and situation. As I mentioned previously I >>have a 'cushion' of MHz. >> >>I clock the chip back a bit after ensuring stability. When overclocking the chip >>will go from completely stable to extremely unstable over a small temperature >>range when you're on the 'edge' of stability. I let it get very warm (55C+) then >>find the chips 'edge'. I then back the MHz up just a little >>and test until its completely stable. After this I knock the chips MHz back a >>significant ammount, ensuring complete stability. >> >>Intel increases voltages (from 1.50v to 1.525 & 1.55v) to help increase yields. >>This helps the chips run higher, and if you wanted to get down to the >>engineering level you could consider that a mild form of overclocking. Taking >>lower yield chips, boosting the voltage and selling them faster. Intel has been >>keeping the chips right on the edge at times (not all of the time of course).. >>this is why they had to recall the P4-3GHz C, Itaniums and P3-1.13GHz. They know >>what the cores are doing (guesstimate) then sell the chips slightly lower. >>Sometimes they guess wrong & end up with an unstable chip. If they tested the >>chips before hand like I do, they wouldn't have this problem. Of course there is >>absolutely no time for AMD/Intel to run extended stability tests on every cpu >>they produce. It would cost too much money and take a massive ammount of time. >> >>I go the extra mile by testing the chips more than Intel/AMD would ever hope to >>test their chips, then I do as I stated in the previous paragraphs. >> >>You don't know the minute effects of voltage changes on a chip in reguards to >>overclocking & heat, I do. If you need me to show you some proof mathematically, >>I can do that as well. >> >>When trying to find calculations on a cpu you never, ever calculate by the chips >>"rated" speed, the core is almost always much better. Take the latest >>hand-picked 1700+ (1.46GHz) chips. Lets say after 24-48 hours of testing we find >>the core can do 2440MHz at 1.75v, completely stable via Prime95/BurnK7 for hours >>and hours at 60C (via Standard heatsink and fan running low rpms, to raise cpu >>temp to 60C intentionally). >> >>Todays chips will do exactly 100% more at their absolute maximum if you drop the >>temperature 160C. Thus a 1.47ghz rated chip with a 2.5GHz core (40C maximum >>stable temp for example) can run 5.0GHz at -120C. Using this constant, 0.625 (or >>160/100), you can figure out the following. >> >>Lets take our example cpu from before, (max 2440/1.75v @ 60C), we can do this: >> >>We know the normal user won't run 60C, with average cooling (0.16 c/w heatsink) >>and figure in an ambient case temp of 30C, and the cpu at 2440/1.75v being >>86.509 watts. You can figure out the chips wattage by doing, >> >>Overclocked Watts = Default Watts * (Overclocked Mhz \ Default Mhz) * >>(Overclocked Vcore \ Default Vcore)² >> >>I used the numbers from an Athlon XP 2700+ (2167MHz, 68.3 watts, 1.65v) and >>scaled it to 2440MHz, 1.75v. So, the wattage is accurate. >> >>Now, back to cooling. You can figure out the CPU temperature by doing this: >> >>cpu temp(celsius) = (wattage * c/w) + ambient (celsius) >>43.84144 = (86.509 * 0.16) + 30 >> >>So, 43.84144C cpu temp with a 0.16c/w heatsink/fan (this is 0.16 degrees celsius >>increase per watt). Using the constant of 160 per doubling, we can now figure >>out that max cpu speed (with the same stability as 2440/1.75v/60C) at 43.84144C. >>We do: >> >>2686.41804MHz = 2440 * ((((60-43.84144)*0.625) / 100) + 1) >> >>So, at about 43.85C it's capable of running 2686MHz, and that is the SAME >>stability as 2440/1.75v/60C. >> >>Now, upping the voltage to 1.85v also helps stablity. Lets first calculate the >>slight MHz drop from the temp increase at 1.85v. If 2440/1.75v == 86.509 watts, >>2440/1.85v == 96.678 watts (still cooler than a 3GHz P4!). This will result in a >>cpu temperature of 45.46848C and a new max stable speed of 2661.60568MHz. >> >>Now, to figure out the speed increase from upping the voltage from 1.75v to >>1.85v, we do a simple linear equation: >> >>2813.70~ = 2661.60568 * (1.85 / 1.75) >> >>End result, our "example" CPU is capable of doing ~2.8GHz questionably stable >>and is clocked down to 2.4GHz. The CPU will be completely stable beyond any >>doubts. >> >>This is what I do, the chips run perfect. >> >>Please take your asinine babble elsewhere, Vincent. >> >>>So for those people who sometimes run crucial software, they cannot use your >>>chips at all. >>> >>>That i personally am busy with computerchess and that everything is an >>>approximation there and all is integers and not floating point, makes me simply >>>an exception. >>> >>>Despite that i didn't do effort to clock my 2.1Ghz chips to 2.2Ghz. >>> >>>Best regards, >>>Vincent
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.