I just finished building my new socket 939 system. the specs are as followed: Athlon 64 3500+ proc Gigabyte GA-K8NSNXP-939 BFG Geforce 6800 GT x2 512 sticks of Corsiar XMS 3200 DDR-Ram 36.7 Gig Raptor drive Those are the important parts. The system is being cooled with water, and this is my first WC system. To cool the system I'm using (for refrence): DD TDX 1/2" Barb CPU block DD 68NV 1/2" Barb GFX block DD 12v In-Line pump Black Ice Dual Micro Radiator No res, only a fill and bleed kit As for fans in the system, I have 2 fans on the front on the radiator sucking air in. On the other side of the radiator I have 2 more fans pulling air in (fans:radiator:fans:front of the case). On the back I have 2 more fans pushing air out. I have a side blow hole in take also right over the CPU block. And the PSU fan. Now my problems, I have three of them. I'm making 1 thread because, well its on the same system. Problem #1: The bios shows the temps of the CPU being 12C.... 12C, almost always. Some time as low as 9C. Thats at idle. Now, I did some searching and it seems like the average temp of a 939 3500+ is about 50C-80C on air. Since this is my first water cooling set up, I'm not aware of what kind of temp drops I should be getting. I know it would be cooler than air of course... but 12C seems low. I tryed running MBM but it doesn't support my mother board. Only the socket 754 Gigabyte boards. Could this temp possible be right? If its not, can anyone reccomend another program to try and use to check the temp? Or do you think I should flash the bios? I block feels only slightly warm, and the radiator feels warm. But comforable warm. Not so warm that I have to move my hand or anything like that. Problem #2: I'm getting random blue screen errors. The error that comes up and says "Windows has encountered a problem and closed itself to prevent further damage to your system", its something like that. Then theres some stuff about if this is the first time seeing it, and running virus scans, etc. Then at the bottom it gives the 0x000000XX (There might be an extra 0 in there), and the next line says it dumped the physical memory. I'm pretty sure this is a RAM problem. It doesn't happen at any set time, or during any set thing. I got it once playing the EQ2 beta, and onces opening up IE. The Gigabyte board is set up so that its Slot1:Slot2:Slot3:Slot4. Slot1 and 2 are a set up dual channel and slot 3 and 4 are another set, so they don't alternate. Theres also kind of close. The ram heatspreaders touch. They don't push on each other, but they do touch. Right now I have the ram in slot 1 and 2. If I only put 1 stick in slot 1, either stick, the problem seems to stop. It I put 1 stick in channel 1 and the other in channel 3, it seems to stop but the ram is now not in dual channel. So it seems to me that either 1, the ram is getting to hot, or 2 dual channel is messing up. Any other ideas? Problem 3: Overclocking. I tryed to overclock last night, just to see what I could do. Even though I'm un sure of the temps. But heres what I found... it plain sucks. Heres a pic of my bios at the max settings I could post at: I can get the clock speed up to 209, and then after that I have to push up either the CPU voltage (shown above) or the HT voltage (+.1). I can't go over a 214 clock speed though. It wont boot at 215, and upon start up the mobo automaticly kicks the speed down to 200 and any voltage changes back to auto or normal. Adding extra voltage to either the cpu, ht, or ram wont work. Or combos of those three. It will just kick back down. Have I hit my max? Or is there anything else I could try? Also with overclocking, after I was done I tried to boot back into windows. I got the error "Windows is either missing the following file or it is corrupt." the file is that dang WINDOWS\system32\config\system file. I never shut the computer down during start up or shut down. So i'm not sure if my over clocking attempts messed it up, or if something else did. But I had to reinstall windows (no big deal really). I'm planning on trying to put it back at 11.5x214 tonight before I install xp pro (I'm running home right now, but I just got pro). And see if it happens again. I thought I read some where that someone else overclocked there 3500 and got SATA problems, so I'm pondering it was that. But any ideas on what to do about overclocking as well? Thanks for any help on these problems!
you need to latest bios to fix the temp problem and there are known issues with corsair ram in this board. you need to run your ram in slots 3and4 plus take a read of this http://www.houseofhelp.com/forums/showthread.php?t=28664&page=1&pp=15&highlight=setting+bios its the official Corsair forum as for your ocing, id try get ya system stable at stock b4 you go messing with that.
I had the same problem witht he physical memoryt dumps and i finally solved it like 2 days ago. My motherboard was detecting hmy ram timings wrong so it was making the system unstable. what you wanna do is go into your bios and look for dram timings and change them to 2.5-3-3-6(leave the rest on auto) now my system runs great. as for the mobo heat problem. you just need the updated bios because the sensoris reading it wrong
Problem 1, dont trust bios temps full stop. For overclocking, ive heard quite a few problems with that board. Use google to see what you can dig up. (When i was looking for something else, i came accross a forum full of problems for it) Also, the standard multi for the 3500+ is 11x. Not 11.5. The multipliers are only unlocked downwards only, so you can use anything under and including 11x. As for the SATA problems, it depends on the board. I know for mine, SATA ports 1&2 are not locked, which stuffs up your o/c if your using them. It also corrupts data on HDD's, which sounds like it already has in your case. I also heard that this was a problem with the NForce chipset itself, so may well be common to your system too. Try the other ports/controller. Use clockgen or a64tweeker to make sure the locks are working. (Start to up the HTT and you should be able to see if they change) Try setting the agp at 67 or 68. My board doesnt lock if left on 66. Cant really think of anything else. There should be forums out there with lots of info for you to try specific to your board. The CPU has nowhere near reached its limit, so i wouldnt worry on that score. I got mine to boot the other day at 270x10 but not for long. It gets crazy hot, which shouldnt be a problem for you. Is it a 90nm chip you have? If so, then even better. These are frequently hitting 3ghz (On Vapo) at around 1.5v.
Thank you guys for pointing me into the right direction. I did loads of reading last night on other fourms and such and got most of the problems solved. Well, I flashed the bios to the newest bios (F4) and that solved the temp error problem. And I moved the ram to slots 3 and 4 and lowered there times. After installing XP Pro, all of my games, programs and documents I have yet to get a blue screen of death. So that seems good. I figured out that pushing the HTT over a 1000 can cause the SATA corruption. I before I reformated and installed XP Pro, I pushed the multi up to 5x220, saved the bios, the computer restarted and it set it back down to 5x200. And XP was corrupted again. So that solved that issue. After I installed Pro I started to overclock. No other programs installed. These were my results... all of these were stable enough to get to the desktop, go to the start menu, load Windows Media player, and play the sample song. There all with the HTT at 4x. 11x245 @ 1.700v on the CPU, +.2v to the ram, and +.3v to the HTT (thats the max voltage for the cpu, the ram, and the HTT.) 10.5x245 @ 1.600v, no ram or HTT voltage mods 10x244@ 1.700v, no ram or HTT voltage mods I was dissipointed with the multi at 10x. I really thought I'd be able to get over 10x244. Thats only 2440 MHz... but no matter what I did it wouldn't get to the desktop with it over 244. Not lowering the HTT multi to 3x, not maxing out the voltage at 1.7v. Not even adding voltage to the ram or HTT. I mean, 10.5x245 is faster than it with a lower CPU voltage and no other voltage mods. Which puzzles me. I decided to go with the multi at 11x. And right now i'm running at 11x235 with the cpu at 1.600v, the ram and HTT at +.1v. I decided to run it at this so that I could back the voltages down a bit. I think that I might be able to drop the ram and HTT voltage down to stock to be honest. But I just wanted to make sure that I was stable for installing all my games and such. I'll probably tweak it a little more later and run some benchmarks. Once again thanks for all the help guys! EDIT: Also, its a 120nm 3500+, not a 90nm
You mean 130nm. You really have to do some proper testing BEFORE you set yourself a final speed. Windows Media Player isnt exactly stressing your processor, so get yourself Prime95 and run it for at least 10 hours. If you get errors, you've got an unstable system, and a fairly useless overclock. High overclocks are great for benchmarks and so on, but for real use you need stability.
alastor, Thanks for your concern. I realise that playing 1 song in WMP isn't the end all to be all stablity test. But I had no internet connection, and I didn't have any benchmarking tools to run. No sandra, no aquamark, no cpu-z, no prime95. (I lost my utilities cd I burned all that stuff onto =\). I mearly meant that those were the speeds that I was able to actually use windows with. Any higher and I can't even get into windows. So those are my breaking points it seems. I do have other things to try, like lowering the ram clock speed to see if that helps, but for now those are my limits. Like I said, I backed the speed down to hopefully be at a stable speed. And I've backed even further down now to 11x230. Now I'm starting to run Prime95, play some games, and get some benchmarks. Those speeds were just the start to know how high to go. Now i'm trying to start lower, and see how high I can go stable. I just wasn't going to run Prime95 for 24 hours at each speed I tried you know? Thanks for your concern and words though. And yes 120 was a typo.
Hard luck with the CD Still, nearly 2.6ghz is a pretty impressive overclock and a good target to aim for I think. Good luck!
Aye. All the forums I read before I bought the proc said they were running at 2.5 Ghz stable, so that was what I wanted to atleast try to get to. 2.6 is definitly a nice spot to stay stable at if I can, and I don't think I'll go over that.
Personally, i wouldnt use Prime95. I dont trust it. I have read lots of problems with different hardware making P95 fail even at stock speeds. For me, if i can loop sandra or 3dmark for a while, or long gaming sessions dont cause any problems, its stable. It does seem odd that you can get a higher HTT with 11x rather than 10x. It suggests to me, that your memory might be holding you back. The best way to overclock these things, is to do one thing at a time, and find all your stable maximums. You can also tell what the weak link is. Use dividers on the RAM and use a low CPU multi, to see how high your board will go. Should be at least up around 300 at least, most boards have up 450 selectable in the bios. Then, again using a RAM divider, aim for the highest HTT/multi you can stablely to see what your CPU is made of. Personally, id see how high you can go with a 10x multi. If you can reach well past 244, you know your ram isnt up to the job. Then its time to start shopping for RAM that can do your HTT 1:1 for some blindingly good bandwidth. You can use the dividers without killing your performance though, its not like it was on nforce2. You have already seen that your CPU can do nearly 2700 @ 11x245. So if you got RAM that could handle 270(DDR540) and ran at 10x270, it would be pretty damn awesome. Edit - You can also use a low CPU multi and use memtest86 to see what the maximum your ram can handle. Oh, and dont worry about the HTT multi. Dropping it even to 3x doesnt make the slightest difference in performance.
Sounds like a plan Dan The only thing is, you said: "You have already seen that your CPU can do nearly 2700 @ 11x245. So if you got RAM that could handle 270(DDR540) and ran at 10x270, it would be pretty damn awesome." The problem with his mobo, as it seems, is that this board is really picky with RAM. I got that impression from all the other forums and topic specific threads I've gone through. It seems that blue screens of death are common, because of RAM, when using low latency high preformence ram. People that are using RAM with like 2-2-2-5 timings are having to relax those settings to be stable. My RAM right now is running at 3-3-3-8. It should be running at 2-3-3-6. Heres a link for you . If I could get it to work... (If it is the ram, I'll look through the "Safe List" for ram thats know to be stable with it, and see if theres some DDR 540)... then the only other problem I have is the voltage. To get it at 2700 GHz, everything is maxed... I'll be a little worried about frying the CPU out in a month =\ After 2 hours of Doom 3 the CPU temp was 42C. But the worry comes that the extra voltage is just going to drasticly kill my chip really early. But I'll see what happens! I'll start looking into it again tomorrow.
Dont hold me to it, but i think 1.7v isnt really that bad. Have you got an option for 1.65v? Most reasons ive heard for these chips dying early is memory voltage funnily enough. Anything more than 2.9v and your on dangerous ground. I know people use OCZ DDR boosters with there EB and stuff, but i wouldnt. I remember reading something from AMD a while back, saying the onchip memory controller on a64's are quite sensitive to memory voltage for some reason, and the majority of cpu deaths are really the memory controllers going pop. If you feel uneasy about it though, then do whats comfortable. At the end of the day, the CPU performance is hardly a limiting factor at the moment anyway, the only reason im overclocking mine is because its good fun. As you say, picking RAM for your board is gonna be a pain in the arse. Same with mine, they are really picky with what you put in them. Saying that though, MSI at least seem to be working on it. With each bios release the compatability gets a little bit better. This might be the same for your board, id wait around and do a lot of reading and see what newer bios releases are like. Samsung TCCD seem to be overclocking really nicely at the moment, in particular OCZ Plat Rev2 3200. The Hynix DTD-5 based 3700gold rev3 also works quite well. You will have to read up on what ppl have bought and had success with on your board though, it could be completely different. Try not to blindly trust online reviews either, always try and find someone that actually has the RAM in their system. (Ive seen lots of people get stung after reading a certain review on memory, and not being able to overclock for toffee)
The max cpu voltage is 1.7v, so I can't go over that. There is a 1.65v option, but at 11x245 it wouldn't boot at 1.650v. It goes 1.x00, 1.x25, 1.x50, theres just no 1.x75. I'll google that fact you gave me about the memory controller being what goes bad on the chip first, see what other people see is safe as well. Thanks for the info! The DIMM voltage wont go higher than +.2, so I think that what ever I max out at will be pretty safe.
