Jump to content

CustomBuild

Members
  • Content Count

    28
  • Joined

  • Last visited

About CustomBuild

  • Rank
    New Member
  1. Your temps are high. I run 2950MHz at 1.44 and my temps run ~41 at load (dual Prime). I run 3GHz @1.48 and load temps are 42-43. As it is my temps are a few degrees higher then they should be. I doubt your shut down problem at 3GHz is PSU related. You have a decient PSU with enough power. The ATX spec is +/-10% on the various rails so your's looks to be in spec. Odds are at 3GHz you did not have enough Vcore, HTT multi was too high, or RAM ratio or timings caused a failure. If you can get your temps in line I would try 3GHz again and just go through everything and make sure all the settings are correct.
  2. Past 250 you will need to lower your HTT multi. HTT can become unstable above 1000. At x4 your HTT=1000, anything above that will take it over 1000. Try lowering it to x3.
  3. Thanks for the ideas Politbureau. 1) It is possible my application is not good. When I changed nozzels I obviously had to remove the block. I noticed then that the spread was fairly even, slightly larger than a quarter, but was not transparent. It looked somewhat thick. While I am doing a "dab" in the middle, it is possible my dab is too big Although it could also be due to a concave IHS etc. I may reseat it and try using even a smaller dab. 2) My loop looks like that minus the res (I use a T-line). I was aware of the D5 heat dump so had the output going to the rad. 3) Right now, since I have fans on the outside of the rad pushing air through it, the fans are about 1/2 inch from the front grills. I figured this should ensure that most of the air pulled in is from outside, but the rad is not sealed to the case so it is possible (but not likely due to the amount of flow) some case air is getting pulled. 4) I had read that nozzel #4 tended to give better performance so that is why I switched to it. But again I did not see any improvement. 6) I also am using Zerex + de-ionized water (I did my homework). I targeted 95% water, 5% Zerex (1oz Zerex to 16oz water). 7) Now this could be an issue. I could have air trapped. I had a big problem with air bubbles/micro bubbles when I filled it. It usually takes hours to sort it way out. Lines look good now, but I could still have air in the rad. To the best of my knowledge water will absorb air over time (thats why pressure tanks need bladder/diaphram seperating the water from pressurized air) but it could be a function of pressure. I am hopeing that and air bubbles will get absorbed over time. What has been your experiance with air in the system? Yep, very similar setup, yet your temps were lower. I will probably pull the WB, try using less ceramique and see if that helps. Of course all this is based on the readings from the on chip sensor. According to some info I have seen from AMD the specs on them could easily account for the differece I see in my temps vs others. If my sensor just happens to read 2-3C higher than actual that would account for it.
  4. The rad is inside (front of the Stacker case), pulling room air in and through the rad. It does not appear to be airflow rated as I see no difference if I run the rad fans low, med, or high rpm (have them connected to a Sunbeam fan controler). Plus as Radodrill mentioned I doubled my rad size and did not see an impact. I am using a D5 with 7/16 Masterkleer. Tubing is short, clean, and no kinks. I am running the D5 on the highest speed setting (5) so it can't be too slow. With a 2-pass 240 rad it is hard to believe that the flow could be too much, but I suppose it could be possible, specially with a CPU only loop. I will try different pump settings and see if that helps. I can't swear that there is no air trapped somewhere in the rad or block. It is possible. It could take a number of days for any trapped air to be absorbed so I will have to wait and see on that. Right now I tend to agree with Radodrill in that most likely it is a IHS to WB issue. For my luck it is posible the IHS is not flat. I have not convinced myself it is worth sanding/lapping yet. My temps are not bad, just a little higher than I had wanted.
  5. While there are variations within steppings, you might want to double check that nothing else is holding you back. I have thw exact same stepping and get 2900Mhz at ~1.42V and can hit 3000MHz at 1.48V. Make sure you did not overclock your memory by mistake as your increasing FSB, and it is causing you to fail Prime.
  6. I actually push/pull, but I think that is probably overkill. My ambient (room temp) is 20C. It may be my delta's are normal. It is possible the posts I saw with lower numbers were naked CPUs, I don't remember.
  7. I am running a CPU only loop. I initially was using a GTS120 and a TDX with the stock (#1) nozzle. I was getting +7-8C detla (from 20C room temp) as measured by Core Temp and MBM. Under load at 3GHz I was getting ~+19C according to MBM and ~+21 according to Core Temp. I saw some posts that indicated people could get +6 at idle and ~+13 load. I wanted to push my temps a little lower for more OC attemps over 3GHz. I "upgraded" to a GTX240 and the #4 nozzle in the TDX. I was expecting a 3-5C change, but to my surprise I can't see any difference in temp delta. I idle where I idled before and under load (@3GHZ) I sit right at ~41C (+21). With the larger radiater, reduced pressure loss from the rad, and the smaller CPU block nozzel, I am not quite sure why I don't see any benefit. I also see no difference if I run the rad fans at 50cfm or 80cfm. That tells me it is not a heat capacity issue, but a heat transfer issue. I am afraid it has to be at the CPU itsself and my heat transfer between the die/IHS/and TDX is the limiting factor At this point I realy don't want to remove the IHS. Oh and I used Ceramique (dab in the middle). I don't have any other ideas at this point. Are my delta's pretty normal, or should I really be able to hit someting around +6 idle +13 load (1.48-1.5V range)?
  8. Yes, pretty much evryone is using water to get good temps at higher OC speeds. I just made the switch for exactly that reason. That plus the conversion was a fun project for the holidays
  9. Unfortunately you may have a bad processer then. It is not unusual to have a "weaker" core. In my case it is core 0. Core 0 will always fail first. So basically all my voltages are based on core 0. If I went by core 1 by vcore would always be lower for a given speed. But even so teh weaker core should still perform to a specific level. Your voltages are just too high I think for a good Opty. Or it could still be temps. Or you might have a bad connection between your die and the IHS (still a temp issue). Your heat sink might now be seated well, the thermal paste not applied correctly or well. Who knows.
  10. Made Prime Stable at 273MHz (DR546) @vdimm=2.6V. Again, I used Praz's timming as is: http://www.dfi-street.com/forum/showthread...1082#post601082 I posted my screen shot at: http://www.dfi-street.com/forum/showthread...4210#post684210 I have not tried to push it higher yet, so it may get better.
  11. I hit 3000MHz Also have some decient timmings on my Mushkind thanks to Praz's timmings. Now I need to download and run all the other 3dMarks to make the official OCC database post. VID = 1.35+110% ~1.48V
  12. I suspect your temp are holding you back. They are very, very high. Idealy you want core temps in the low 40's if not right at 40. I am surprised you see that much difference between SmartGardian and core temp with an Opty. I only see a few degrees diference at load (2-4C) between MBM5 and Core Temp. You have a good stepping so you should be able to get decient clocking out of it. Your volts also seem high for 2880MHz. I would have guessed that for your stepping your Vcore would be more in the 1.425V range. Make sure you followed the OC guide and worked your way up slowly. You can't really just jump to a clocking/voltage. Something just seems wrong with your setup.
  13. Sounds like you might not be running Prime95 correctly. If you don't run it correctly it will actually only use 1 core at 100%, uses page swaping, and everything crawls. The key is for each instance use custom (not the default blend). I select custom and change the min ftt to 1024 and memory to 700 (I have 2G). Basically you have to set the mem each instance uses to something less than 50% of your total memory. Prime95 seems to be the test of choice here.
  14. Do a searc in the OC data base for posts by Praz and he has what look to be some tight timmings for the Muskin XP4000s. I am trying them right now and so far initial mem86 tests ran fine at 256MHz and 261MHz using 2.6V. I have not tried Windows or Prime95 with them yet so don't know how stable they are on my sticks or not. Basically just started messing with them. EDIT: ok mem test looked good up to 270MHz at least (I stopped there for now). I am now running Praz's timming with dual Prime95. I am running 300FSB (3GHz), 180 divider (270MHz), 2.6V DRAM, 1.35+110% CPU.
  15. I used to think the same thing about the VID special (percents). Using a DMM I found that on my board: 104% was realy 103% 110% was really 109.6% 113% was really 112.7% If you build a little table in excel, you will see that the VID special actually does give you the ability to select voltages between 1.40 and 1.425 for example. While not necessary since if you could be stable at 1.416 (1.365+104%) you would also be stable at 1.425, I would rather run 1.416 if I could
×
×
  • Create New...