My first i720 OC @ 4.2ghz with Corsair H50 - small writeup and review

H

hobag

n00b
Joined
Nov 2, 2006
Messages
39
[H]i guys,

So here's my first OC with the i720 platform. Haven't done any overclocking since the Opteron days and only recently retired that machine for this setup. I know I'm a little late to the i720 game, but thought I'd atleast contribute to the archives.

Ok, so my system specs are in the sig so I'll go straight into some of the details:

Prep:
The CPU and H50 base was lapped. The base was lapped first with 400 grit, 800, then 2000. Since the base was already pretty flat, I didn't spend a whole lot of time on the base. Not quite a mirror shine, but signifcantly smoother and shinier than stock.

Then I lapped the CPU. Not very flat to begin with, the heat spreader was concave and I noticed that when used in stock form, only the edges of the cpu were making direct contact w/ the base. This was mitigated inefficiently by using more TIM in the middle vs the edges. Process was more or less the same, except I started w/ 220, then 400, 800 and 2000. Almost mirror finish, but decided it was flat enough and stopped. Nickle plating was completely removed (obviously).

The TIM I used was AS5, applied using the line method across the middle and then I improvised by putting a small dab of AS5 on each of the 4 corners of the heatspreader. Logic is that when the line is spread by the compression force, the TIM might not reach the corners. I've used the "even-layer" method before, and found that it was less efficient and the contact is less uniform.

The H50 was seated with the least amount of "twisting" as possible to keep the spread uniform. The base was clamped down pretty tight - all 4 screws were tightened in a cross pattern 1/2 turn at a time until each of the 4 screws could not be turned with a reasonable amount of force.

Cable managment is key, and I did the best I could. With a midtower, space is tight and airflow is not as easy to achieve. As of right now, the case has no stray cables in the middle and air flows pretty freely.

The Cooler:
The H50 is setup in an exhaust method (blowing hot air OUT of the case), in the back of the case underneath the PSU. The radiator is sandwiched (push/pull config) by a 120mm SilenX case fan, and the stock 120mm Corsair fan the H50 came with. 2 front 120mm case fans are on intake duty (another SilenX + a CM case fan). The only other exhaust is from the 120mm PSU fan.

The OC:
The method I used is pretty straight forward; Adjust settings in BIOS, boot into Win7 x64, load up HW monitor, CPU-Z, task monitor (to ensure that CPU and RAM are at 100% load during testing), and then start Prime95 torture test with custom settings.

Using the above routine:

Control: All stock settings @ 2.66ghz and ran prime95 error free for 10hrs to ensure everything is solid. 27-29c idle / 58-60c load

3.2ghz, 20x160 @ 1.2v, speedstep off, HT on, Ram profile set to XMP (8-8-8-24, 1.5v): Same routine, stable after 10hrs of prime. 30-32c / 60-63 load

3.6ghz, 20x180 @ 1.2v, ss off, HT on, XMP: prime stable for 10hrs, 33-35c / 62-65c

4.0ghz, 20x200 @ 1.225, ss off, HT on, XMP: prime stable 10hrs, 35-36c / 73-75c

4.0ghz, 20x200 @ 1.225, ss off, HT OFF, XMP: prime stable 10hrs, 30-33c / 59-61c

4.2ghz, 20x210 @ 1.25, ss off, HT off, XMP: COD MW2 stable but not prime stable, 34-35c / 65-67c

4.2ghz, 20x210 @ 1.275, ss off, HT off, XMP: running prime now. Will know by end of day.

I have a feeling that it won't pass prime for 10hrs, and that it will need a boost to 1.3v. What has really shocked me is the difference that HT makes; over 15c on load difference on some occasions! That's incredible, seeing as how HT isn't useful in most games and daily usage, it's almost a no brainer to keep it off for best OC results.

Now, after doing some reading, it seems most people are already at 1.3v + when they're at 4ghz. I just want to verify if I'm doing something wrong or cheating somewhere, because I seem to be doing just fine with 1.225 at 4.0ghz. Or is it possible that I just lucked out with an extremely stable chip?

I am extremely pleased with the H50 and what it has done so far. Temps are well in the acceptable range and my case has tons of breathing space due to the small form factor. For those of you who have been lead to believe that the H50 isn't up for 4ghz+ OCing shouldn't be so quick to write off this little guy and drop big $ on huge WC system or giant HSF combo. I'm aware that you can get better temps by going either route, but 4.2ghz at 70ish degrees on load is pretty darn acceptable to me, especially for the price of an H50. Extremely good value and very easy to install.

However, those are temps with HT off. I gather that those who want to keep HT on might need a different cooler if they want to keep their temps below 75c at load @ 4.2ghz, but I don't think most gamers will care about HT. But just for reference, I will turn HT on and do a few more tests just to see what the temps would be with the H50 stressed to the max.

I have pics of the lapping process, the rig setup and screen shots of the OC's. These I will post as I pull them off my digicam and when I'm not writing at work =P

You comments, tips and questions are welcome =)

FYI - I tried 4.4ghz very briefly last night and got it to boot into win7 @ 1.3v and surfed the net a little before I got a BSOD. I have a feeling 4.4 will be the tough barrier, and I will need to fiddle around with quite a few more settings in BIOS before I get that stable. Anyone have some settings that has worked for them and wouldn't mind sharing?
 
you have a nice chip but those temps aren't the best considering you have pretty low voltage and have HT off. trying running that chip at 4.2ghz with HT on and at 1.3vcore to achieve prime95 small fft's stability.

you will get temps in the high 70's low 80's. This was the case for me :( so i switched to a megahalems and haven't hit 70c under the same conditions.

edit: the H50 is a good cooler and as you saw at 4ghz with lower voltage and ht off its very capable and its a great choice. but once you start cranking up the voltage and clock speed its struggles. I mean it only has a 120mm rad so how much heat dissipation can you expect. It does very well for its size and ambition but it cannot compete with to the top tier heat sinks above 4ghz+ (with HT on) in my experience.

Good write up.
 
Last edited:
Yup, totally see where you're coming from. I'm still in the early stages of testing thoroughly, but I'll keep updating this thread as I go along.

What are your BIOS settings for 4.2?
 
my settings for 4.2ghz are pretty straight forward

21 x 200
1.30vcore
1.84vtt
1.34 dram/qpi

speedstep and all that other cpu stuff disabled (like hardware pretcher etc).
i leave c1e on so the computer can go into s3 sleep though.

thats it.

hopefully you can get away with lower volts.
 
Is that your voltage reading under load or at idle?
 
Last edited:
the H50 is def. a very capable cooler. i run the pump at 100% and cant hear it. two scythe SSF21G (i think... something like that) the 1800 rpm with a shroud on each side is a very good cooling solution. i am still doing full blown water cooling in the near future though. just cuz it sounds fun. =P
 
hobag said:
[H]i guys,

So here's my first OC with the i720 platform. Haven't done any overclocking since the Opteron days and only recently retired that machine for this setup. I know I'm a little late to the i720 game, but thought I'd atleast contribute to the archives.

Ok, so my system specs are in the sig so I'll go straight into some of the details:

Prep:
The CPU and H50 base was lapped. The base was lapped first with 400 grit, 800, then 2000. Since the base was already pretty flat, I didn't spend a whole lot of time on the base. Not quite a mirror shine, but signifcantly smoother and shinier than stock.

Then I lapped the CPU. Not very flat to begin with, the heat spreader was concave and I noticed that when used in stock form, only the edges of the cpu were making direct contact w/ the base. This was mitigated inefficiently by using more TIM in the middle vs the edges. Process was more or less the same, except I started w/ 220, then 400, 800 and 2000. Almost mirror finish, but decided it was flat enough and stopped. Nickle plating was completely removed (obviously).

The TIM I used was AS5, applied using the line method across the middle and then I improvised by putting a small dab of AS5 on each of the 4 corners of the heatspreader. Logic is that when the line is spread by the compression force, the TIM might not reach the corners. I've used the "even-layer" method before, and found that it was less efficient and the contact is less uniform.

The H50 was seated with the least amount of "twisting" as possible to keep the spread uniform. The base was clamped down pretty tight - all 4 screws were tightened in a cross pattern 1/2 turn at a time until each of the 4 screws could not be turned with a reasonable amount of force.

Cable managment is key, and I did the best I could. With a midtower, space is tight and airflow is not as easy to achieve. As of right now, the case has no stray cables in the middle and air flows pretty freely.

The Cooler:
The H50 is setup in an exhaust method (blowing hot air OUT of the case), in the back of the case underneath the PSU. The radiator is sandwiched (push/pull config) by a 120mm SilenX case fan, and the stock 120mm Corsair fan the H50 came with. 2 front 120mm case fans are on intake duty (another SilenX + a CM case fan). The only other exhaust is from the 120mm PSU fan.

The OC:
The method I used is pretty straight forward; Adjust settings in BIOS, boot into Win7 x64, load up HW monitor, CPU-Z, task monitor (to ensure that CPU and RAM are at 100% load during testing), and then start Prime95 torture test with custom settings.

Using the above routine:

Control: All stock settings @ 2.66ghz and ran prime95 error free for 10hrs to ensure everything is solid. 27-29c idle / 58-60c load

3.2ghz, 20x160 @ 1.2v, speedstep off, HT on, Ram profile set to XMP (8-8-8-24, 1.5v): Same routine, stable after 10hrs of prime. 30-32c / 60-63 load

3.6ghz, 20x180 @ 1.2v, ss off, HT on, XMP: prime stable for 10hrs, 33-35c / 62-65c

4.0ghz, 20x200 @ 1.225, ss off, HT on, XMP: prime stable 10hrs, 35-36c / 73-75c

4.0ghz, 20x200 @ 1.225, ss off, HT OFF, XMP: prime stable 10hrs, 30-33c / 59-61c

4.2ghz, 20x210 @ 1.25, ss off, HT off, XMP: COD MW2 stable but not prime stable, 34-35c / 65-67c

4.2ghz, 20x210 @ 1.275, ss off, HT off, XMP: running prime now. Will know by end of day.

I have a feeling that it won't pass prime for 10hrs, and that it will need a boost to 1.3v. What has really shocked me is the difference that HT makes; over 15c on load difference on some occasions! That's incredible, seeing as how HT isn't useful in most games and daily usage, it's almost a no brainer to keep it off for best OC results.

Now, after doing some reading, it seems most people are already at 1.3v + when they're at 4ghz. I just want to verify if I'm doing something wrong or cheating somewhere, because I seem to be doing just fine with 1.225 at 4.0ghz. Or is it possible that I just lucked out with an extremely stable chip?

I am extremely pleased with the H50 and what it has done so far. Temps are well in the acceptable range and my case has tons of breathing space due to the small form factor. For those of you who have been lead to believe that the H50 isn't up for 4ghz+ OCing shouldn't be so quick to write off this little guy and drop big $ on huge WC system or giant HSF combo. I'm aware that you can get better temps by going either route, but 4.2ghz at 70ish degrees on load is pretty darn acceptable to me, especially for the price of an H50. Extremely good value and very easy to install.

However, those are temps with HT off. I gather that those who want to keep HT on might need a different cooler if they want to keep their temps below 75c at load @ 4.2ghz, but I don't think most gamers will care about HT. But just for reference, I will turn HT on and do a few more tests just to see what the temps would be with the H50 stressed to the max.

I have pics of the lapping process, the rig setup and screen shots of the OC's. These I will post as I pull them off my digicam and when I'm not writing at work =P

You comments, tips and questions are welcome =)

FYI - I tried 4.4ghz very briefly last night and got it to boot into win7 @ 1.3v and surfed the net a little before I got a BSOD. I have a feeling 4.4 will be the tough barrier, and I will need to fiddle around with quite a few more settings in BIOS before I get that stable. Anyone have some settings that has worked for them and wouldn't mind sharing?
Click to expand...

Please show Pics of 4.2ghz, 20x210 @ 1.2 CPU-Z and RealTemp.
 
Last edited:
hobag said:
FYI - I tried 4.4ghz very briefly last night and got it to boot into win7 @ 1.3v and surfed the net a little before I got a BSOD. I have a feeling 4.4 will be the tough barrier, and I will need to fiddle around with quite a few more settings in BIOS before I get that stable. Anyone have some settings that has worked for them and wouldn't mind sharing?
Click to expand...

I need ~1.49V to hit 4.4GHz with HT enabled and 1.475V to hit 4.5GHz with HT off on my i7 930. Temps get up there even under my 3x120 radiator, 75-80C. Don't be afraid if you know what you're doing.
 
I got my I7950 at 4.2 with 1.375 volts on the CPU. I got both HT and turbo on. It pass 25 runs of lynx and left prime running for about 5 hours. Thats good enough for me. Wish I could get my temps better but Im not on water. They idle around 43 - 46 , in long gaming sessions they hit 55 , and in Lynx it hits 92 C. Prime for some reason only makes it get up to 88-90.
 
any issue with running low 90*C?

The CPU throttles at 100* C --- so you are close to throttle temp....I hit 100*c a few times with my chip in testing and have ordered a H70 to put an end to that. I just got my chip last week, it overclocks to 4.2 at about 1.35 and 4.0 at about 1.2875. Neither of which my N520 Hyper Cooler Master cooler can keep pace with - unless I turn off HyperThreading (15*C cooler without hyperthreading)
 
You must log in or register to reply here.
Back
Top