Help with my first OC...stable but some questions

[andyo]

By the way, the instructions for AS5 say there is a break-in period. I'm gonna wait it out a few days and then check on my AS5 if it's well spread.

Thanks again.

 

[bjp999]

Andyo -

Sounds like you've made some good progress. AFFINITY is important. You want each of your separte Prime95s to be configured with affinity to a separate core. That way, each prime is working on one core. You also want to run large FFTs on each one.

When both cores are going full bore, that's when you take your temps with TAT. I would not go much higher than 60C. 65C is pushing it, and 70C is too hot.

1.6V is also higher than I would recommend 24x7. I would drop back to something that lets you run in the 1.55v - 1.5625v, with temps maxing out around 60C. There are no guarantees with OCing, but at least you'll have lots of company at that level. I don't know of anyone running 1.6V @ 70C 24x7!

BTW, how was did the lapping go? 5C reduction is on the low side, but maybe your HSF was pretty flat already. I'd try reapplying the AS5 again like you said, and it might go down more.

Keep up the good work!

 

[andyo]

Well, right now I'm not running 1.6v and 70° anymore. I'm doing 1.5750 and 65° max, and I am gonna wait a bit for the purported break-in period of AS5.

The lapping actually was harder than I supposed. I had everything, but the 400 grit took like 2 hours. My heatsink was a bit convex, but most of all it had a big scratch right through the middle that I noticed only after the lapping revealed it. From 800 up to 200 it went quick enough.

I'll check on my AS5 in a few days, but I'm thinking of getting a better cooler too, even if it goes down another 5°C I don't think it's enough. I would like to go as low a vCore as possible, I'll probably try liquid, that'll give me some experience on that front too.

 

[bjp999]

Originally posted by: andyo
Well, right now I'm not running 1.6v and 70° anymore. I'm doing 1.5750 and 65° max, and I am gonna wait a bit for the purported break-in period of AS5.

Better. Just be careful not to let it get too hot.

The lapping actually was harder than I supposed. I had everything, but the 400 grit took like 2 hours. My heatsink was a bit convex, but most of all it had a big scratch right through the middle that I noticed only after the lapping revealed it. From 800 up to 200 it went quick enough.

Hope you mean up to 2000 not 200! Hopefully you were nice and shiny at the end. It gets faster and you do a better job after you've done a few.

I'll check on my AS5 in a few days, but I'm thinking of getting a better cooler too, even if it goes down another 5°C I don't think it's enough. I would like to go as low a vCore as possible, I'll probably try liquid, that'll give me some experience on that front too.

The AS5 "cures" in about 200 hours (less if you let it go through heatup and cooldown cycles more frequently). I noticed about 3-4 degrees drop. Remember that every time you remove the HSF, you have to thoroughly clean the AS5 off with alcohol and start over iwth the AS5 (otherwise you get air bubbles).

A new HSF is not a bad idea - yours is not bad but not in the top tier.

I've considered water cooling but decided it was not for me. I estimated the cost of what I wanted at about $350+ for good setup and CPU / GPU cooling. And the max benefit I might get is 400MHz, more likely 1/2 of that or less. Weighing that benefit against horror stories like one I read about a guy that developed a leak and came home to a puddle around his computer and a fried CPU (it overheated after the leak), , I decided to stick with air. If you go with water, there is an article that explains how to make you computer shut itself down if the temps start to rise too high.

 

[Ibn]

Try a different HSF with AS 5. I OC my 6600 on the same board just recently and this is what it's been looking like.

http://ct.pbase.com/g6/35/597035/2/70304825.uN15iMpP.jpg

Using a Zalman 9500 on there right now and at full load it's 58C.

 

[Link]

Is an E6600 doing 3.3ghz with 1.30V(bios) good overclocking? It ran the Ortho for about 6 hours without an issue.

 

[andyo]

Well, yeah hehe, I did mean 2000, not 200. I'm considering, if not water, at least the tuniq tower, it seems from reviews that it has a significant advantage to the big typhoon, zalman 9500 and other similar ones. It actually kinda smoked them on the xbitlabs review. What gets me is that I had actually checked out that cooler before, but just quickly dismissed it for the big typhoon, which is very good, but only as good as all the second-bests, it seems. Also, I went for the ease of installation. The Big Typhoon VX is just as easy to install as the stock cooler. I wish they came out with a tuniq tower with straight on LGA 775 compatibility too.

Still waiting for the break-in period, so far nothing seems to change. I just realized 200 hours is more than 8 days, I'll just wait. I'm only turning my pc on while using it, and then hibernating so it will record programmed HDTV.

Another, kind of related thing, my mobo does not seem to want to wake up after a S3 sleep when overclocked (to 3.6 GHz). When not overclocked, it does work. I haven't tried with in-between overclocks, though. Hibernation works better for me, and I don't have trouble with that.

 

[andyo]

Originally posted by: Ibn
Try a different HSF with AS 5. I OC my 6600 on the same board just recently and this is what it's been looking like.

http://ct.pbase.com/g6/35/597035/2/70304825.uN15iMpP.jpg

Using a Zalman 9500 on there right now and at full load it's 58C.

What's your vCore set at? Are you using TAT to monitor the temps? I assume you are also using the p5W Deluxe

Thanks.

 

[bjp999]

Temperature reductions after applying AS5 are a little subtle. I noticed a 1-2C reduction over the course of the first day, but didn't notice much more over the next week. I just went and looked after several weeks and did notice that temps were down another 1-2C. Don't bank on that making a huge difference.

 

[bjp999]

Originally posted by: Link
Is an E6600 doing 3.3ghz with 1.30V(bios) good overclocking? It ran the Ortho for about 6 hours without an issue.

Sounds like a good CPU. Mine needs +.1v to be stable at 3.3GHz.

A good overclock a compromise between speed and heat. By pushing volts and FSB, you get a faster CPU, but heat becomes a problem. Combatting the heat is what OCing is all about. Your OC appers great for keeping a cool system (you don't post your temps so not sure) and should be good for a longterm solution with lower risk of component failure compared to someone that is pushing like andyo and I.

BJP

 

[Ibn]

Originally posted by: andyo

Originally posted by: Ibn
Try a different HSF with AS 5. I OC my 6600 on the same board just recently and this is what it's been looking like.

http://ct.pbase.com/g6/35/597035/2/70304825.uN15iMpP.jpg

Using a Zalman 9500 on there right now and at full load it's 58C.

What's your vCore set at? Are you using TAT to monitor the temps? I assume you are also using the p5W Deluxe

Thanks.

vCore is set at 1.4V. I haven't been using TAT to monitor the temps; I've been using CoreTemp instead for it. You're correct that I'm also using the P5W DH Deluxe MB.

 

[andyo]

Originally posted by: Ibn

vCore is set at 1.4V. I haven't been using TAT to monitor the temps; I've been using CoreTemp instead for it. You're correct that I'm also using the P5W DH Deluxe MB.

Man, that's unbelievable. The Zalman 9500 and big typhoon shouldn't be too far off. But anyway, I already ordered the Zalman 9700, I'll see what happens.

Also, to bjp999 (or anyone else) I just replaced my 500 W thermaltake W0093 for a 600 W Coolermaster i-green (supposedly much more efficient). The temp seems to have dropped 1-2°C right after. Is it possible that the PSU did it (maybe slightly better power management)? The PSU seems to be running much cooler and effortless than the previous one, but I don't think its temp should affect the CPU's temp, since all the air is coming out from the Big Typhoon and some of it is sucked by the PSU, which is right above it. Don't think any significant amount of heat, if any, could transfer from PSU to CPU.

 

[Canai]

Efficiency is pretty much power loss to heat, and aside from better quality components, a better fan / airflow through the PSU, and sleeved cables, I'm not suer what else could have caused the temp drop. PSU heatsinks are often close to, if not touching, sides of the enclosure, heating up the sides of the box.

I've noticed that in 3dmark and everest's benches, my c2d e6600 @ 2.66 running on 42% FSB OC (380x7) performed as well as it did when it was clocked at 3.4, and at a lot less vcore and temps (21ºidle/35ºload die and 30º/45ºcore with mobo at 23º/32º) on stock vcore.

I've got a Scythe Infinity with a beefy 120mm strapped to the side pulling air through it and an 80mm intake on the case blowing air into the other side.

 

[bjp999]

Andyo -

I think the 1-2C may be due to a slightly better fans in the PSU. Or taking the thing apart you may have arranged the cables a bit differently so that air is ciculating better.

Your perception that the PSU is running "effortless than the last one" is not something you could know by observing it. You'd need multimeters and all sorts of expensive test equipment to tell. Maybe the fan runs quieter ... that's about all.

 

[andyo]

Well, it's probably only perceived, but that its temp is much lower should be at least a vague indicator that it's making less of an effort. More efficient=less effort for the same work, isn't it? Also, my previous PSU was making somewhat of a low, slow-pulsating humming noise. Since I quieted down my case so much, it was really annoying. There's a fridge in the room, when it went quiet, I could hear the humming of the PSU in my PC. Now it's all so much better.

Also, I think the fan is not actually quieter, I might just switch it with a quieter one (the one I previously put on the old PSU). That way I can also monitor its RPM, and by it, I can tell a little bit about how much "effort" it's putting out.

By the way, the zalman 9700 is arriving tomorrow. It seems the zalmans already arrive very nicely polished. Do you think I should try to lap it too?

Thanks for everything so far.

 

[bjp999]

Lapping is up to you. You might want to install it first if only to get a temperature reading to compare it to after lapping. Of course if the temps are very good, you might skip the lapping all together.

An important point, you will want to clean the AS5 off your CPU as well as you can before installing the new HSF. I have some swabs (link) that I like to use (after using qtips to get off the bulk) that seems to deep clean very well. You might search for other techniques that others have found effective. Qtip clean is probably ok, but I can clean until an isopropyl dipped qtip stays pure white after hard rubbing, and still wind up with a black (isopropyl dipped) chamois swab after just a couple swipes.

I'm keeping my fingers crossed for you with the Zalman!

BTW, I ordered an HR-03 GPU cooler that will be delivered tomorrow. I don't like running my 1900xtx as hot as it runs currently. Hope it cools the GPU without warming my CPU!

 

[andyo]

Well, I swapped the heatsink. What an annoyance to have to get the mobo out. Anyway, the temps show almost the same, maybe 1°C lower. But this is the 9700 un-lapped and the AS5 freshly applied. I assume the temps should go down a couple of °C in the days to come. But that's not enough, so I think I'm gonna check on the AS5 and then maybe lap my CPU and HSF. Actually, the HSF was very smooth, and by looking at reflections of far away objects you can see if it is deformed in any way. It was as plain as a mirror. In any case, I'll try and check both the HSF and CPU.

One thing that I noticed when taking out the Big Typhoon, is that the AS5 was spread more towards the edges, suggesting that either the HSF or CPU were convex-shaped. Since I lapped the HSF, it should be the CPU. Do you have any suggestions on how to lap the CPU, and what precautions should I take as opposed to lapping the HSF?

Thanks.

 

[bjp999]

Not sure there is a great way to tell if a HSF is flat. You can lay it on its side on a piece of graph paper and see if the reflected lines are perfectly straight. I can tell you that my HSF was shiny and passed this test with flying colors, but it took quite a lot of lapping at 220grit to get it all uniform.

There are 2 schools of thought on lapping the CPU. One group thinks that you should take all precautions to keep it dry. The other group says its going to get wet no matter what and not to worry about it. I subscribe the the 2nd school. I used the small black thing that the CPU came in to hold it and keep the back pretty clean, and lapped at 400 grit until all the nickel material was gone, and moved through the different grits up to 1500 grit. I used a little dishwashing detergent when lapping to keep it from sticking. When using the 1500 grit paper (the finest I used), I used lighter and lighter pressure until I got the reflective finish. Lap in one direction, and keep flipping 180 degrees. You don't want to flip 90 degrees! (I did this at first and couldn't figure out why I wasn't gettting good results). I didn't use toothpaste or brasso or any such stuff after the lapping - I wanted the bare metal polished smoothly with the sandpaper. All it all it was very easy compared to lapping the large HSF, although the technique was the same.

When I was done I ran the whole CPU under the kitchen sink (I didn't want any copper dust to be anywhere on the CPU). I then dried it gently with a paper towel. I dried my HSF with a hair dryer, and while I was at it I used it on the CPU as well. (I was VERY careful not to let the CPU get hot.) I also cleaned the contact area of HSF and CPU with isopropyl alcohol.

Use a very small amount of AS5 - size of 1 to 1.5 grains of rice. There are no gaps to fill, and you don't want a layer of AS5 separating the HSF from CPU, just enough to fill in the micro crevices, but with lots of metal to metal contact! When mounting the HSF, give it a slight twist back and forth a couple times under moderate pressure to grind the AS5 into the crevices and cause it to spread. Then mount without lifting. Make sure, whatever the mounting technique that the Zalman uses, that it you get it mounted correctly!

Remember that lapping the CPU voids the warranty. Unless you are concave, you may only get 1-2C.

Originally posted by: andyo
One thing that I noticed when taking out the Big Typhoon, is that the AS5 was spread more towards the edges, suggesting that either the HSF or CPU were convex-shaped.

This is surprising. I've never heard of anyone reporting a convex IHS. Although convex is probably better than concave, lapping probably will help either way.

BTW, how is your CASE cooling. If you are hot it could be because it is hot inside your case. Do you have a good, cool case with good airflow?

 

[andyo]

If you're still around...

I've given up on 3.6 GHz. Maybe later on I'll get to a better cooling solution. The Zalman 9700 seems to be pretty good though. At idle there's only about 1-2°C difference between the (lapped) Big Typhoon, but during workload it seems to cool much better, though it could be the cool winter here in L.A. The weather dropped temp just about when I switched coolers, so I don't know for sure.

In any case, I couldn't get 3.6GHz Orthos-stable unless I got to 1.6V, which I want to avoid. The temps weren't too high, actually in the low 60's for workload, but still I don't know if too high a Vcore could be damaging in the long term.

What I'll do now is to set the highest Vcore I want to deal with (1.55V) and then just up little by little from 390 FSB until I reach the limit. I guess from 3.510 GHz to 3.6 GHz is not that much of a difference, but from 1.55V to 1.6V it is. So there. If you're still around, do you have any comments on this?

Thanks again, and happy work-free days (If you have such luck).

 

[bjp999]

Hey Andyo!

Good to hear back from you!

It sounds like you have made a very smart choice, and that you have an excellent overclock. The last 100MHz is nothing but a little pride. You are smart to stop at the point that Vcore and temps get out of hand.

I bought a new case (Antec 900) and did a review of how I modded it. Thought you might be interested. It dropped my temps 5-6C! Here is the link.

I am taking off the rest of the year also. Happy Holidays!

 

[andyo]

Hi again, finally, 26-hour Prime95 stable at 391 FSB (3519 MHz).

I had to give into 1.5625V, though. At 1.5V, I wasn't stable even at 389 FSB, and I didn't want to have to go below 3.5 GHz. I almost had it, Prime95 ran for like more than 8 hours in small FFT test, but broke down on large FFT tests after between 1 and 2 hours. One time it lasted just a bit more than 2 hours. On Orthos, I couldn't even get the "blend" test to run more than 10 minutes under this config.

So now my specs are:

1.5625V vCore, 391MHz FSB x 9
memory: 2 GB Corsair XMS2 TWIN2X2048-6400 at 2.0V, 4-4-4-12 (tweaked from 1.9V 5-5-5-12 stock), running at 1:1
MCH, ICH and FSB termination are in auto, I couldn't find any differences if I tweaked them.

Is the Orthos blend test OK to test both CPU and memory? Right now I have had it running for about 2 hours, but I had to stop it to restart the computer for another reason. Prime95's blend test doesn't work for dual core, and I think the Orthos test should be the same, but optimized for dual.

Another thing, it's probably nothing, but when I start my computer, if I go anything below FSB 400 and 1:1, in the POST section I get "PC 5700" about the memory, but if I go 400 and 1:1, then I get "PC 6400", which is my actual spec. If anything, it should show technically "PC 6256", but I guess this figure doesn't matter and the POST can only show some numbers.

 

[bjp999]

Originally posted by: andyo
Hi again, finally, 26-hour Prime95 stable at 391 FSB (3519 MHz).

I had to give into 1.5625V, though. At 1.5V, I wasn't stable even at 389 FSB, and I didn't want to have to go below 3.5 GHz. I almost had it, Prime95 ran for like more than 8 hours in small FFT test, but broke down on large FFT tests after between 1 and 2 hours. One time it lasted just a bit more than 2 hours. On Orthos, I couldn't even get the "blend" test to run more than 10 minutes under this config.

So now my specs are:

1.5625V vCore, 391MHz FSB x 9
memory: 2 GB Corsair XMS2 TWIN2X2048-6400 at 2.0V, 4-4-4-12 (tweaked from 1.9V 5-5-5-12 stock), running at 1:1
MCH, ICH and FSB termination are in auto, I couldn't find any differences if I tweaked them.

Is the Orthos blend test OK to test both CPU and memory? Right now I have had it running for about 2 hours, but I had to stop it to restart the computer for another reason. Prime95's blend test doesn't work for dual core, and I think the Orthos test should be the same, but optimized for dual.

Another thing, it's probably nothing, but when I start my computer, if I go anything below FSB 400 and 1:1, in the POST section I get "PC 5700" about the memory, but if I go 400 and 1:1, then I get "PC 6400", which is my actual spec. If anything, it should show technically "PC 6256", but I guess this figure doesn't matter and the POST can only show some numbers.

1. Orthos just runs 2 copies of Prime95, one on each core. That is what you should be running for processor stability testing. Some people have two copies of Prime95 and can run it just like Orthos does. But running A SINGLE prime95 is not a good test. It will run on one of your Cores only, and will not generate the heat nor the stress that Orthos will generate.

2. I have read conflicting reports of whether you should run small or large FFT, but I run large.

3. The blend test does not stress the CPU as much, but does use some memory so is perhaps a good quick test of memory. memtest86 is the right way to test memory.

4. Don't incease voltage or OC your memory until you get your processor stable. It just creates one more variable. Even after you are stable, try to not raise the memory voltage much if at all. I was able to drop timing at stock voltage. I've seen many reports of people having problems on Conroes because they DO overvolt the memory. (Some MBs are not smart enough to figure out stock voltage or your memory - in this case set to the manufacturors spec).

5. I found that raising the FSB voltage to 1.4 helped a little in achieving stability. If you are close (Orthos running for an hour), then you might find kicking this in makes you stable. Maybe not.

6. The BIOS reports your memory at PC5700 / PC6400 based on a range of FSBs. Don't worry about what it says.

7. Did you check out my link from the prior post?

 

[andyo]

Hi,

I did run one prime95 for each core. I have two copies on two different folders. I also ran Orthos, and dual-prime95 seemed to get errors quicker than orthos, in both large and small FFTs. From all the tests I ran, it seems large FFTs makes the test crash more quickly, but it could be a memory thing. What I meant that prime 95 won't work for dual cores is only for the blend test. Even when you run one for each core, with the blend test only one CPU seems to be stressing, while the other doesn't do anything. CPU usage is around 50% or just a bit more in Task Manager, and only one core shows 100%. I think maybe it's because one core is using up more than half of my available RAM, and the other core can't use it. Orthos seems to anticipate this.

By the way, the blend test I said I would run on Orthos crashed, with this same configuration, but this time it lasted about 2 hours 12 minutes, much more than the 7-9 minutes it lasted with the same config, except 1.5Vcore, so I don't think it's a memory thing. Also, when I ran all the tests, I did have the memory at stock speed, voltage, and timings. Only for the final, 26-hour dual prime95 (large FFTs) test I had the memory tweaked, and it passed. I'll try memtest86, and I'll try the blend test again tweaking FSB termination to 1.4, too.

I did check your post, but I forgot to mention that I also just got a new case, I got a Cooler Master, which is pretty solid. I went for it for silence. Now I'm having a bit cooler temps and less noise, mainly because this one has a vent just very close to the Zalman 9700's heatsink.

Thanks for everything, I'll report back when I get past the 24-hour Orthos blend test for good.

 

[gerwen]

Thanks for this thread guys. I've learned a lot through your trials and tribulations.

I've my first system upgrade in 5 years waiting under the tree for me, and thanks to this post, and many others here, i'm hoping to have a fun time overclocking. (I haven't overclocked since my Celeron 300 long ago.)

 

[BoboKatt]

Aye I just read through this whole thread. My e6600 on my eVGA 680i also requires close to 1.6 to hit 3600Mhz. Thus I too dropped it to around 1.56 and now working out the kinks to keep it at around 3500MHZ.

The interesting thing is that if I drop it down to say 3300MHz the amount of Vcore needed drops just do disproportionably that it just makes me think the extra 200 MHz is not worth the heat and possible damage.

I am using the Tunic Tower myself and AC5. I don't think I did a very good job of spreading the goo on the CPU.. but anyhow now at around 1.46v at 3400mhz after a few days it idles at 37c and normal work environment brings it to roughly 41C and under Orthos it hovers around 49-51. This is just fine for me. Like you upping to 3.5 or 3.6, although possible really brought the temps up at almost 60c on Orthos. Coming from an AMD X2 4800+(running 2700Mhz) with a Big Typhoon that NEVER EVER went over 40 even on hours of Orthos, it just scares me.

My case is probably not helping as I have an original Sonata. There is no actual fan in the front sucking in air, just the one inside drawing air away from the HD and one in the read under the PS. Anyhow might consider that Antec 900 or the Cooler Master.