Looking for input/thoughts on my first attempt at a stable OC

B

Betrayer00GT

Limp Gawd
Joined
Feb 25, 2007
Messages
350
Hey, all.

This is my first OCed system. I made my first attempt back when the 300a slot Celerons were going to 450 on a regular basis, but mine wouldn’t do it for me. So, here’s my next try. I am looking for assistance in getting to 3.2 on the stock Intel HSF (rest of watercooling still with UPS).

This is what I am running:

E6600
ASUS P5N32-E SLI Plus (the 650i/570 SLI frankenstein model)
Kingston HyperX PC2-8500 512MB x2 (should have a couple 1GB OCZ SLI PC2-8500 modules later today, as well)
Stock Intel HSF cooling (the VGA waterblock is on its way, no watercooling loop again until it shows up next week)
XFX 8800GTS 640MB

I have an image that shows where I am at, but it is almost 500KB in size. I apologize in advance for the size of this image, but I run a 2560X1024 desktop and have nothing to resize images with on this XP install. I will only be running this XP install for less than 24 hours. Just until I can pick up my copy of Vista Home Premium later this morning, when the place that has it now is open.

This image has the Nvidia monitor, CPU-Z, Prime95 for each core, SpeedFan 4.32, and the task manager running with more information than I will be documenting, below:

sofarsohappy.jpg

For those who do not like pretty pictures when words would do (there are more listed voltages in the image, though):
FSB: 333Mhz
Core: 3Ghz
DDR2: PC2-8500 (1066Mhz) at 5-5-5-15-2 @ 2.2v (haven’t tried anything else, yet)
Graphics card OCed to 574Mhz core and 866Mhz (1733 effective) memory
1.3625v CPU (not necessarily as low as possible - Need to find where Prime95 dies to test next step down in vCore for same level of stability)
Each core @ 65C (SpeedFan 4.32) - motherboard reported CPU temp is at 69C

Prime95 has been running on both cores now for an additional hour and a half, or more, since this image. I believe I am at >2.5 hours with it, at the moment.

You know, I would like to see a stable 3.2Ghz, but I have not seen a combination at that speed (so far) which doesn't error out on one of the cores during Prime95 in "0 minutes". I can POST all day long at 3.2Ghz/1138Mhz (PC2-9100) @ a 1.3875 vCore (same timings as above), though. I haven't given up/tried everything, but it may not be doable on this stock Intel HSF.

I probably won't get a chance to try much more on this XP install, as I may have moved on to Vista Home Premium before I get the chance. In all honesty, if this is the wall I hit on stock air cooling, I would be satisfied. Not thrilled, but certainly satisfied. The point of going the OC route was to beat the best (stock) Core2Duo available with an e6600. Big banner: “Mission Accomplished”

So, any obvious pointers anyone can provide on getting to 3.2Ghz on this stock Intel HSF with this system? I would appreciate some things to try in getting that to work. Or, have I really just hit a wall regarding a stable OC point, with this combination of parts and cooling? I would also be interested in guidance on current voltages. I have tried lowering/raising the HT bus and FSB, but have not been able to try all the zillionis of combos for everything yet. Just not enough time. I've only been at this for a few hours. I try to leave things at Auto, if I don't have a decent understanding of where to put them.

So, anything?

[edited to correct vCore number]
 
Betrayer00GT said:
Hey, all.

This is my first OCed system. I made my first attempt back when the 300a slot Celerons were going to 450 on a regular basis, but mine wouldn’t do it for me. So, here’s my next try. I am looking for assistance in getting to 3.2 on the stock Intel HSF (rest of watercooling still with UPS).

This is what I am running:

E6600
ASUS P5N32-E SLI Plus (the 650i/570 SLI frankenstein model)
Kingston HyperX PC2-8500 512MB x2 (should have a couple 1GB OCZ SLI PC2-8500 modules later today, as well)
Stock Intel HSF cooling (the VGA waterblock is on its way, no watercooling loop again until it shows up next week)
XFX 8800GTS 640MB

I have an image that shows where I am at, but it is almost 500KB in size. I apologize in advance for the size of this image, but I run a 2560X1024 desktop and have nothing to resize images with on this XP install. I will only be running this XP install for less than 24 hours. Just until I can pick up my copy of Vista Home Premium later this morning, when the place that has it now is open.

This image has the Nvidia monitor, CPU-Z, Prime95 for each core, SpeedFan 4.32, and the task manager running with more information than I will be documenting, below:

sofarsohappy.jpg

For those who do not like pretty pictures when words would do (there are more listed voltages in the image, though):
FSB: 333Mhz
Core: 3Ghz
DDR2: PC2-8500 (1066Mhz) at 5-5-5-15-2 @ 2.2v (haven’t tried anything else, yet)
Graphics card OCed to 574Mhz core and 866Mhz (1733 effective) memory
3.625v CPU (not necessarily as low as possible - Need to find where Prime95 dies to test next step down in vCore for same level of stability)
Each core @ 65C (SpeedFan 4.32) - motherboard reported CPU temp is at 69C

Prime95 has been running on both cores now for an additional hour and a half, or more, since this image. I believe I am at >2.5 hours with it, at the moment.

You know, I would like to see a stable 3.2Ghz, but I have not seen a combination at that speed (so far) which doesn't error out on one of the cores during Prime95 in "0 minutes". I can POST all day long at 3.2Ghz/1138Mhz (PC2-9100) @ a 1.3875 vCore (same timings as above), though. I haven't given up/tried everything, but it may not be doable on this stock Intel HSF.

I probably won't get a chance to try much more on this XP install, as I may have moved on to Vista Home Premium before I get the chance. In all honesty, if this is the wall I hit on stock air cooling, I would be satisfied. Not thrilled, but certainly satisfied. The point of going the OC route was to beat the best (stock) Core2Duo available with an e6600. Big banner: “Mission Accomplished”

So, any obvious pointers anyone can provide on getting to 3.2Ghz on this stock Intel HSF with this system? I would appreciate some things to try in getting that to work. Or, have I really just hit a wall regarding a stable OC point, with this combination of parts and cooling? I would also be interested in guidance on current voltages. I have tried lowering/raising the HT bus and FSB, but have not been able to try all the zillionis of combos for everything yet. Just not enough time. I've only been at this for a few hours. I try to leave things at Auto, if I don't have a decent understanding of where to put them.

So, anything?
Click to expand...



It looks like you are probably stuck with one of the latest E6600's - L640+ batch from the new plant - they have lower tolerances than the previous batches, and tend to require more voltage, and produce more heat... and as a result, don't OC as high.

The intel HSF just isn't going to cut it, your cores are too high, you need to back it off and replace the cooler. If you are loading @ 65c at 3.0GHz you are definately lowering the life of that CPU and possibly it's tolerance when you do cool it correctly, these are NOT the 90nm technology you are probably familar with, 65c is not okay - some will say no more than 60, and absolutely never more than 70. 65 is not worth it definately. It does seem slightly high for a stock HSF - but again it could be due to your batch - I would quickly lap the bottom of your HSF and use some AS5 or x23 grease and spread it with a razor blade and try again.
 
cerebrex said:
The intel HSF just isn't going to cut it, your cores are too high, you need to back it off and replace the cooler. If you are loading @ 65c at 3.0GHz you are definately lowering the life of that CPU and possibly it's tolerance when you do cool it correctly, these are NOT the 90nm technology you are probably familar with, 65c is not okay - some will say no more than 60, and absolutely never more than 70.
Click to expand...

Their mobile chips do run up to 100c without destroying themselves...

But, your point is even more valid, considering the OP is using speedfan to measure the temperatures. Op: Try Coretemp or intelTAT. Speedfan isn't always that accurate (unless it's the newer beta).
 
Arcygenical said:
Their mobile chips do run up to 100c without destroying themselves...

But, your point is even more valid, considering the OP is using speedfan to measure the temperatures. Op: Try Coretemp or intelTAT. Speedfan isn't always that accurate (unless it's the newer beta).
Click to expand...

mobile chips are manufactured to completely different tolerances (much more rugged).
 
I have received my copy of Vista Home Premium, and my 2x OCZ SLI-Certified 1GB PC2-8500 modules (not the platinum, the black ones). They have both been installed since yesterday, and everything continues to be stable.

Due to the heat issues, I have discontinued the search for a faster stable OC, until the 8800 waterblock shows up and I can add that to my loop then get the loop installed after its leak test.
 
Ok,

I have had the OCZ RAM in with the Kingston (both seem to be Micron D9 chips, from what I can find around the Internet) for several days, now. Vista is installed and I like it. And, as of 36 hours ago, the WC loop is in the computer. I have a DangerDen Maze4 CPU waterblock, and a Swiftech MCW60 waterblock on the 8800GTS.

Though the 8800 was easily good for 575Mhz (heat went beyond 73C under load, but passed clock test in Ntune) under WinXP, now that it is in Vista (same, stock air cooling) it does not usually even pass the clock test for its stock setting. Ntune finds an optimal frequency for the core of 487Mhz, but won't set it below stock. Very odd. Especially since this exact same behavior persists now that the WC loop is installed and the idle temps went from ~60C to <40C. That is a big difference. Still, the "optimum" frequency for the card is still detected as the exact same 487Mhz. Leads me to the obvious conclusion:

Nvidia's Vista drivers are the only constant, here. They apparently suck pretty bad.

Pretty obvious, but still sucky.

Anyway, now that the WC loop is in, I believe I have found the highest, stable overclock on the CPU that I am comfortable with. Higher requires too much voltage/generates too much heat with my loop for my taste.

I am now running my = or > week 640 e6600 @ 1.475v and 3.3Ghz. It has yet to complete a >4hr run on Prime95 at that voltage because I was lowering HT, NB, and SB voltages at the same time and ran into a BSoD problem when I got down to 1.4v on the first two. I anticipate that has been resolved by moving back up to 1.5v on those two. We will see in a couple hours one way or the other on that. I have also had the RAM operating at 1066Mhz in all cases. That speed (with 5-5-5-15-2T and 22TRC latency settings) seems to be a piece of cake for the two OCZ 1GB modules and the two Kingston 512MB modules I have installed. Very cool. :cool:

I must say I am a bit disappointed about my overall results, but I now understand that the <630 week C2D chips were much better overclockers than the more recent production runs, like mine. I am actually fine with that. My goal with overclocking was to achieve a higher CPU clock using a much cheaper e6600 vs the stock clock of the extreme edition (2.93Ghz). I have succeeded at that. I am even quite stable @ 3.3Ghz. Check this out for stable voltages and temps (the temps are better @ 3.3Ghz w/ the WC loop than @ 3.0Ghz w/ the Intel HSF) for my >5h 5m dual Prime95 small FFT run:

happinessis.jpg

And the full shot showing the Prime95 instances, after I stopped them (hi-res, large file size / 2560X1024, ~650KB):

stability.jpg

I hope to get a new one of the stability screenshots @ 1.475v. 1.46875v died out on one of the cores at 51 minutes of Prime95, small FFTs. Oh, well. Once I can demonstrate stability at 1.475 vCore (seems likely), I will move on to reducing memory latency at my chosen PC2-8500 memory speed. Wish me luck with that. :D

Well, I am happy and that is all that really matters in this endeavor. :D

Any thoughts/impressions/tips are welcome.
 
to reduce image sizes just email them to yourself w/ Outlook
Windows will ask to reduce the size,. click yes and send
 
1.475 vCore turned out to only be stable to almost 3 hours of small FFTs. I found that out once I eliminated the BSoD issue.

So, the stability.jpg is where I am at now. I now run my BlackIce Extreme II radiator's fans at a slightly lower RPM, since the most heat I can possibly generate is with Prime95, no application I use (game or whatever) will generate that much heat. My push/pull 4x 120mm fans are running between 1700 and 1800 RPMs. Full blast is 2000-2100 RPMs. I get good heat dissipation with the lower range, and pretty minimal noise. Works for me.

Additionally, I was unable to get stablility with lower timings on the RAM, even running at 2.35v (measures as 2.4v in the BIOS, which is the most my OCZ are still covered under warranty with). I have run Prime95 in "blend" for three and a half hours with no issues using 5-5-5-15-2T @ 1066Mhz with 2.2v. That is good enough for me. Lower would have been nice, but is not required in my book.

So, the only "benchmarking" I have done yet is to run 3DMark06 with the new overclocking settings and now on Vista instead of XP. My 2.0 and 3.0 rendering scores dropped between 20 and 50 points a piece, but that is not bad. That is with out changing the Core or Memory clock on my 8800GTS, I am cool with that much of a difference under Vista. The real payoff was a CPU score that went from 2092 @ 2.4Ghz/DDR2-800 to a score of 2749 @ 3.3Ghz/DDR2-1066. This is the first concrete evidence that I actually accomplished something here. :D

Very nice.

I appreciate all the reference information that was available here. Also, Merc was very helpful in PM with getting my Swiftech MCW60 installed and working with an MC8800 surface-mount component heatsink kit installed to cover the rest of the contact points from the stock cooler. Thanks go out to him for his suggestions, they worked wonderfully in my case!

banana.gif
 
cerebrex said:
mobile chips are manufactured to completely different tolerances (much more rugged).
Click to expand...

Not really as true as you'd expect. The tolerances are suprisingly similar. It's the same materials, under the same processes, with the same transistor creation method. The difference between the two chips, however, are extremely exagerated. Just look at the TDP of the X6800 versus the E6300, or the TDP of the Quad cores... Identical chips (or 2 of them in one package) very different heat tolerances.

OP: Glad you got it stable. Is it possible for you to run at 4-4-4-12 @ DDR2-800mhz? You'll see more benefit from tighter timings than faster speed.
 
Arcygenical said:
OP: Glad you got it stable. Is it possible for you to run at 4-4-4-12 @ DDR2-800mhz? You'll see more benefit from tighter timings than faster speed.
Click to expand...

Yes, I ran it at 4-4-4-12-1T for a while at 3.375Ghz (highest CPU Clock I ever saw Windows with).

I thought this would be better as well, but I love seeing that 1066Mhz during the post. Kinda retarded, but that's where I'm at.

I could get over that I am sure, if someone has suggestions on additional benchmarking software where I could run with loose PC2-8500 timings and tight PC2-6400 timings and acutally see the benefit. Any suggestions, anyone?
 
Hmm...

I switched my setup from

1466FSB/DDR2-1066 @ 5-5-5-15-2T, 22TRC 2.2v (all else set to Auto)

to

1466FSB/DDR2-800 @ 4-4-4-12-1T, 17TRC 2.2v (all else set to Auto)

and in SuperPi @ 1M I went from 15.706sec (best) to 15.80X (best). That is not the right direction. Either way however, not much of a movement at all.

I have a feeling that this is because when you compare 4 clocks @ 800Mhz and 5 clocks @ 1066Mhz you find that the number of nanoseconds with 5 clocks @ 1066Mhz is actually smaller. Not by much, you would need 3.75 clocks @ 800Mhz to be equivalent, which is both impossible and very close.

Now, 1T vs 2T is an improvement, and a rather large one, but so is 1066Mhz vs 800Mhz, so I guess it turns out to be pretty much a wash.

Any other benchmarks anyone can think of for this? Now, I am curious.
 
You must log in or register to reply here.
Back
Top