A reworked OC Very Long

Evening all,

Having an overclocked rig is just a great thing.  But as I have learned more over the last few weeks, it really made me wonder if the OC done by DS could be improved upon.  So let me start from the very beginning.

I ordered my system with Vista Ultimate, a q6600, with DS overclock.  Stage 2 cooling, XMS 800 memory, and an 8800GTS512.  All stuffed into an Ultra case.

When my box arrived, and upon boot up the first time, I was amazed at having a 3.6 OC.  So I did what just about everyone else does and downloaded all the windows updates, nvidia hotfixes but I did not change any of the drivers that the machine came with.  I loaded up COD4 and was amazed at how fast and smooth the game played.  I played for hours (not a very good game player yet, but am getting better)  The system locked up hard, requiring a hard reboot.

So I got the apps of CoreTemp, CPU-Z, and Prime 95 and started doing some testing and monitoring.  The first go around with Prime told me that under full load the temps were 70*.  NOT GOOD!!  On top of that now I was starting to have problems at boot and at shutdown.  I had taken CoreTemp and had created a task schedule for it to start at boot.  (This program will not run from start-up folder at boot).  When the system would lock up, a hard boot was usually required.  At boot, sometimes I would not get a desktop, but would still be able to alt-tab out to task manager to quit tasks.  Even being able to do that was no help, as it would now hang at shutdown, requiring another hard boot.  When the next boot came up, I would get a desktop that looks normal, but the task scheduler would not get CoreTemp started, and I would have no Internet access.  Click restart, and would again hang at shutdown, hardboot,  select start windows normally would finally get a system that behaves like it should.

These were the 3.6 OC setting in bios

CPU Core  1.55
CPU FSB  Auto
Memory  2.0
nForce SPP  1.45
nForce MCP  1.6
HT nForce SPP  <-->  MCP  1.45

The CPU was cooking and nothing I could do would lower the temps.  I was frustrated, as most with a new machine that is not working correctly would be so I called Josh at DS.  He had me try a couple of things that had no effect on anything finally I just had him help me over the phone to reset my OC to a more reasonable level.  He had me change the Multiplyer from 9 to 8 and change the CPU core voltage to 1.5 everything else stayed the same.  So now I have this new OC of 3.2 pretty stable all around but still with random hangs at boot and shutdown.  Also have temp in the 65+ column under stress from Prime95.  Still not good.

I talked with one of the Forum members, and he suggested a few things that I could try.  Namely Running Auslogics Disc Defrag, and CCleaner.  I was only 1% fragmented but let it clean it up anyway.  The CCleaner on the other hand found multiple problems both in the file system, and the registry.  If I remember correctly almost 1,000 instances.  I let it fix everything.  I believe this cured alot of the boot hangs, and shutdown hangs.

This morning I had the house to myself, and decided to see if I could get my voltages down so my temps would come down with it.  I started lowering the CPU Voltage down by 2 increments in the bios.  Each requiring a reboot, check the voltage windows sees thru CPU-Z and start Prime 95 using the small option.

Voltages tried:

Bios        CPU-Z voltage        Temp under full load

1.4875        1.432                67*
1.475        1.416                67*
1.4625        1.4                67*
1.45        1.392                65*
1.4375        1.384                65*
1.425        1.368                65*

As you can see I was getting nowhere fast.  I had dropped the voltage .064 volts but had only dropped the temps by 2*.  I was also getting concerned about how low one can take the voltage.  I did a google on voltages and it took me to an overclocking site where I saw that a few of the guys were running voltages of around 1.25  Then next jump went down quite a bit.
                                                                                        Voltage under Load of Prime95
1.3000        1.256                Failed at start of Prime95
1.2875        Did not load to desktop
1.30625        1.256                Failed with immediate reboot
1.3125        Coretemp failed to load
1.29375        1.248                Failed at start of Prime95
1.31875        1.272                Prime95 ran 1 minute then Failed
1.325        1.272                Prime95 ran 2 minutes then Failed            1.224
1.33125        1.280                Prime95 ran 3 minutes then Failed            1.232
1.3375        1.288                Prime95 running for 1.5 hours no problems        1.240

With Prime95 still running as I write this, my temps under full load are 58* and have been steady as a rock for over an hour and a half.  So I guess the whole story behind this is if you are seeing higher than expected temps, you may be able to lower voltages to reduce the heat that is being produced by your processor.  I dont know if even Crysis can stress the processor like Prime95 can.  I would sure like to hear any input anyone has on this subject.




Edited by Roger - 23 Feb 2008 at 9:38pm

the q6600 has a x9 multiplier so he's at 3.6ghz.

if you can run it for 8+ hours without errors than it will be ready to handle anything you throw at. nothing stresses your cpu more than a stress test.

1.35v is the default for core 2s so you have an excellent chip right there.

Roger, I'm looking over the Prime tests running in that screen shot. You should try and isolate the CPU with a run of small FFT's in the torture test selections. This will not put much stress at all on the system RAM, but will really push your CPU. With the small FFT's running you should also gain better insight into your max temperatures. In my testing of various overclocks, I have also had much better luck running the RAM synchronized than not.

Once you have run small FFT's for about 20 minutes, stop the test, and run a blend for 2 hours or more to put a more balanced stress on the system. In this way, you will know if the failures are coming from your CPU settings, or from your RAM settings. There may be better ways, but this has worked for me.


Good the see the random hangups have lessened in severity.

I tend to push my RAM speed a bit faster than the 1:1 to help compensate for latency. The timings you list are pretty stock. The full timings are probably something like 4-4-4-12 2t 3-24-5-8 7.8us running at 2 to 2.1V.

RAM timings are tricky. It's hard to get the best result from RAM, and it requires just as much patience as overclocking the CPU.

The way I do my timings, and it has worked for me so far: A-B-C-D 1T up to 770 MHz 2T past 800MHz E-F-G-H 7.8us

A+B+C=D
E+G=H
F=2*D

The actual relationships are more complex, but if you stick to that you shouldn't violate any of the requirements for timing relationships between the values. I use SiSoft Sandra's RAM bandwidth and latency tests to see if I am doing better or worse with a timing set, then see if it's stable on a Prime 95 blended FFT test with a known good CPU overclock.

Morning all,

Well,  got up this morning and came out to see how Prime was doing.  At 5.5 hours core 4 had an error.  Bummer.  I am going to go in and bump up one more notch and give it a rerun.

I also ran 3DMark06 for a score of 13461 3DMarks.  Is this any good?  I am not going to do anything with pushing the card until I get Riva tuner profiles set up.

Mythius,  where are you at now, and where are you trying to get to?  Sounds like you are getting close.  For me, it was just a few clicks up on VCore to get into an OS.  Lotta trial and error here.

Might as well look at my ram timings while I am in there and see if I can bump them any.  Thanks all for the help.  Greatly appreciated.

Roger

Correct me if I am wrong, but what Roger is going through (time & effort)to find stability with his CPU is what DS is supposed to do when they build your rig and we ask for the CPU to be overclocked with stability? Don't we pay for that...maybe I missed something in the early posts... 

P-Dog,

Yanno, looking back at how my box arrived, I was not dis-satisfied.  I was really rather surprised at the temps that I had when under a Prime95 load tho.  I gladly paid the 25 bux for the overclock.  That only amounted to 1% of the total cost of the system.  What It did give me tho is a good base, with typically known settings.  Since heat, and later instability became an issue I decided to go from 3.6 to 3.2. 

I have lowered my voltages in Bios by 10%.  Going from 1.5 to 1.35, under load the vcore was around 1.4 and now is 1.248.  Big, Big diff in temps now.

When I got this rig, I really abused it the first weekend.  COD4 for about 6 hours, and Crysis for an entire day.  One helluva stress test any way you look at it.  For a user that maybe only plays for a couple of hours at a stretch, and not really taxes the machine, it probably would have been a great machine!  I also think that part of my instability issues were around one of the MS updates that either did not install correctly, or did not uninstall correctly.  Tyler turned me onto a prog CCleaner that seemed to resolve the instability problems.  But on the other hand, heat may have played a part in it also.  I just wont be able to know until I start trying again for the 3.6 on air.

I know I am not the only one that has recently recieved a 3.6 oc.  I was really expecting 3.0-3.15,  3.6 was a very pleasant surprise.  I also know, that DS, does not have the manhours to do what I have been doing the last 2.5 days.  If they had to devote the time you could have probably tacked another 500 bux onto the cost of the rig, and more days in stage 2 and 3.  I am confident that if and when I run into that proverbial brick wall, that DS will be there for me to help me pick up the pieces and get me back on my feet.

On an extreme oc such as the 3.6 there could possibly be an upgraded northbridge and cpu cooler, I know there are other options out there.  But when it hits the test bench, and its obvious that the box can run that fast, all it takes is a phone call from DS to tell you that the Arctic 7 may not get the job done well enough.  I would have popped for an extra 30-50 bux in a heartbeat!  Liquid cooling on the other hand is not for me.  To much maintenance and to many things can go wrong.  Plus the cost involved.

Sorry for going on like this, just wanted you all to know where I stand.

Roger

I think the main thing with liquid cooling is the price tag. I'm sure everyone would *love* to drop $900 on a dual loop cooling system to run their SLi rigs to the wall, but this is not within the realm of reality for many. Even $300 for a CPU only cooling solution would cause the price of a system to go from affordable to out of reach for some buyers.

Considering the addition of a couple of high CFM fans and a higher performance CPU HSF can be done for under $100, I don't see where liquid cooling is required. I derive a sense of accomplishment from a successful attempt to adjust or tune my machine to ever higher levels of performance. I simply disagree with the notion that something cannot be improved upon, and could not envision spending $250+ on an enthusiast motherboard, designed from the ground up to support end user tweaking and customization and then leaving it be. I would view that as "missing something". Others are fine with that, and that's great. Systems like the ones DSO builds can be many things to many users.