PDA

View Full Version : Overclocking Results - Acceptable?



squatchy
10-23-2012, 08:35 PM
Hello everyone,

So I'm pretty new to overclocking - but after doing some extensive research and such, I think I'm somewhat knowledgeable and won't do anything too stupid.

Anyway, I've managed to get the following:

4.5GHz overclock with VCore set to 1.225V (manual) and LLC set to ultra high (75%). This was tested to be stable with LinX and my peak temperatures were 86C for the hottest core.

I'm just wondering if this is an acceptable result? Do you think having it at this setting for 24/7 usage will be reasonable and won't damage my CPU over time? I know the temperatures are slightly high for everyday use, but I highly doubt I will ever get that high with any normal workload. Even AIDA64 Extreme doesn't drive temps as high as LinX.

My system specs:

CPU: Intel Core i7-3770K
CPU Cooler: Noctua NH-D14 CPU Heatsink
Motherboard: ASUS Maximus V Formula
RAM: G.Skill Ripjaws X 16GB (4x4GB) DDR3-1600 CL9 (9-9-9-24 CR2) @ 1.5V
GPU: Gigabyte GeForce GTX 680 OC
Power Supply: Corsair Professional HX850

Thanks in advance for your feedback!

Krindor
10-23-2012, 08:45 PM
During normal day work the cpu will never be pushed so hard so that overclock is succesful. If you're still uncertain run prime95 in blend there is probably no other program out there that pushes the processor to it's limit like pime 95. If the temps are still 86C then for daily use there shouldn't be any problems.
What are your temps in idle?

Zka17
10-23-2012, 08:45 PM
Which temps are you referring to and what application are you using to assess them?

Your most important temps are the temps of the individual cores... use an application which is able to show them... - then you could be on the safe side for 24/7 usage if your core temps are staying below 75C-80C during LinX...

The Noctua NH-D14 is a very potent cooler... - if you can supply enough airflow through it...

squatchy
10-23-2012, 08:55 PM
During normal day work the cpu will never be pushed so hard so that overclock is succesful. If you're still uncertain run prime95 in blend there is probably no other program out there that pushes the processor to it's limit like pime 95. If the temps are still 86C then for daily use there shouldn't be any problems.
What are your temps in idle?

Oh? Okay, I'll give it a shot. I've heard that LinX is more exhaustive, but I guess the only way to tell is to try both! I idle around 45-48C idle at a fixed voltage. I'm going to try using offset mode and get a stable overclock that way so that my cores can throttle down to 1.6GHz (and a lower voltage) when they aren't in use.

Thanks for your help!

squatchy
10-23-2012, 08:59 PM
Which temps are you referring to and what application are you using to assess them?

Your most important temps are the temps of the individual cores... use an application which is able to show them... - then you could be on the safe side for 24/7 usage if your core temps are staying below 75C-80C during LinX...

The Noctua NH-D14 is a very potent cooler... - if you can supply enough airflow through it...

I am referring to the peak core temperatures (I thought I mentioned it in my original post). I have validated this with both CPUID HWMonitor and AIDA64 Extreme's temperature monitor (individually, of course).

And yeah, the NH-D14 is working great. My idle temps are already 10-15C lower, and I think it's pretty impressive that I'm stable at this voltage given I'm at 4.5GHz.

When I'm up to it, I'll probably de-lid the CPU - I've heard this gives some phenomenal results.

Zka17
10-23-2012, 10:10 PM
Well, having 45-48C on the cores during idle at 4.5GHz, 1.225V, doesn't seems too good... it depends on the ambient temps, but at such low voltage, you should be at least in the mid 30s... again, depends on the ambient temps (I'm supposing around 24-26C)...

Regarding the de-lid of the CPU... well, there are some good experiences about that... and bad too... - I'm about to doing it, changing the internal TIM if you meant that... One thing is sure, we will loose the warranty... :)

squatchy
10-23-2012, 10:23 PM
Well, having 45-48C on the cores during idle at 4.5GHz, 1.225V, doesn't seems too good... it depends on the ambient temps, but at such low voltage, you should be at least in the mid 30s... again, depends on the ambient temps (I'm supposing around 24-26C)...

Regarding the de-lid of the CPU... well, there are some good experiences about that... and bad too... - I'm about to doing it, changing the internal TIM if you meant that... One thing is sure, we will loose the warranty... :)

I think it's pretty reasonable. I haven't gotten better than 38C for my peak temperature on the hottest core (idling) even at stock settings. So a 10C increase when my voltage is 0.225V higher and I'm running at 4.5GHz is pretty reasonable, in my opinion. My ambient temperature is about 22C. I suppose it's a little higher than what it could be because I have 4 HDDs in my system in addition to an SSD. All of them are 7200RPM drives too. If I removed them, I'd assume my temperatures would drop maybe 1-2C.

What is your peak core temperature under idle at 4.5GHz (what voltage are you running it at)? What is your ambient and cooling setup, also?

And with regards to the de-lidding, yes, it is a risky gamble. I'll probably wait, as my system is new and I don't want to unnecessarily risk it when there's no need to.

And no, the TIM is not the issue. It's the glue/adhesive they use to attach the heat plate to the die. It adds a lot of space inbetween the die and the heat plate and so, a lot of TIM has to be used to fill this gap. Removing the glue significantly decreases this distance, and so this is what yields the better temperatures (not the TIM itself).

Krindor
10-23-2012, 10:28 PM
And yeah, the NH-D14 is working great. My idle temps are already 10-15C lower, and I think it's pretty impressive that I'm stable at this voltage given I'm at 4.5GHz.

I'm quite jealous of you at the moment, the voltage is really nice, though the cpu seems to be a hot one. My 3770k is wc cooled and for a stable OC at 4.5 ghz I have to have 1.34v.

squatchy
10-23-2012, 10:39 PM
I'm quite jealous of you at the moment, the voltage is really nice, though the cpu seems to be a hot one. My 3770k is wc cooled and for a stable OC at 4.5 ghz I have to have 1.34v.

Yikes! What are your peak load temperatures for the hottest core? Something doesn't seem right with your chip - maybe it's one of the chips that just needs a higher voltage?

Yeah, I was a little surprised at how easily I achieved it. I had to tinker with the LLC for a little bit - so actual load voltage is around 1.233V. To be honest, I think this is the best that can be done with air-cooling if you don't want any degradation in performance over time.

I could probably push my overclock to 4.8 or even 5.0 GHz, but then I'm guessing even normal workload temperatures would start hitting 85C+ and that's not something I'm comfortable with.

Right now, I probably don't have the best air flow as my front intake is jammed with 4 HDDs running at 7200RPM. I think I'll add some side-fans and increase my intake, or maybe switch my top exhaust fan to an intake instead. (I have a Corsair 600T White case)

Krindor
10-23-2012, 10:55 PM
Peak loads are at a nice cool 65C though after 1 hour of Prime 95 it's 80C T_T But this is due to it being a watercooling build and I still need an adapter to install my last fan on my 3x120 rad. So soon I will be able to OC to 5 ghz with god knows how high voltage.....
But your cpu seems to one hot guy.... I recommend water cooling for your build :D Though it might be hard to fit in that case...

squatchy
10-24-2012, 12:30 AM
Peak loads are at a nice cool 65C though after 1 hour of Prime 95 it's 80C T_T But this is due to it being a watercooling build and I still need an adapter to install my last fan on my 3x120 rad. So soon I will be able to OC to 5 ghz with god knows how high voltage.....
But your cpu seems to one hot guy.... I recommend water cooling for your build :D Though it might be hard to fit in that case...

Why do you say its hot? I think that temperature is pretty standard given the voltage, no? I'll see what it is without all the HDDs installed - I have a feeling that this is having a larger impact than I thought.

If I'm able to get to around 4.5-4.6GHz with 85-86C as my hottest core - I think it's pretty good Anything more would mean I have a very good chip compared to the average for air-cooling.

I think my next build (4-5 years down the road) will have a water-cooled setup. For this one, I'm sticking with the good old breeze.

And, your peak isn't really 65C if it gets to 80C!! Anyway, this probably is due to the fact that all your fans aren't installed. Let me know what the temperatures are once everything is all set.

I'm curious to know what other NH-D14 users are getting as their idle/load peak core temperatures so that I can compare. Considering all I've heard about Ivy Bridge's terrible temperatures, I'm actually quite satisfied that I was able to make it this far with this voltage and temperature on air-cooling.

There are a few articles that show that anything beyond like 4.2 GHz starts to give diminishing returns in terms of MHz/power draw for Ivy Bridge anyway.

Krindor
10-24-2012, 12:39 AM
What I meant is that at your set voltage the temps are high. When I run my cpu with 1.25v I have around 40-50C at peak.
The reason for the temps to be at 80C is that the water gets heated up but remember this is after an hour and normally you don't run the cpu at 100% for an hour.
For a 4.5ghz with air cooling those temps are great due to the low vcore. But the temp/vcore is very high.

squatchy
10-24-2012, 12:48 AM
What I meant is that at your set voltage the temps are high. When I run my cpu with 1.25v I have around 40-50C at peak.
The reason for the temps to be at 80C is that the water gets heated up but remember this is after an hour and normally you don't run the cpu at 100% for an hour.
For a 4.5ghz with air cooling those temps are great due to the low vcore. But the temp/vcore is very high.

I suppose it may be a little warm - but it isn't a fair comparison to yours, because you have a water-cooling loop. It would only really make sense to compare with someone who is overclocking on air (and specifically, an NH-D14).

And with regards to how long you run your CPU - a good stability test usually takes about 1-2 hours, so whatever temperature you hit at your peak is what your maximum temperature is. You can't say that your peak is what you see within 10 minutes - that is also not a fair assessment.

EDIT: And also - this is quite important - what is the maximum frequency you can achieve with 1.225V VCore? It also isn't fair to compare temperatures if you are running at a significantly slower clock speed. It is obvious that you will have lower temperatures.

Zka17
10-24-2012, 11:49 AM
Well, I was looking on my notes to give you a comparison, but I didn't really run my 3770k on those frequencies... but I can promise you, when I will be ready to change the internal TIM in it, will do some pre- and post measurements (perhaps after the ongoing SuperPi challenge, as I'm using my mobo for that)...

Krindor
10-24-2012, 12:43 PM
I suppose it may be a little warm - but it isn't a fair comparison to yours, because you have a water-cooling loop. It would only really make sense to compare with someone who is overclocking on air (and specifically, an NH-D14).

And with regards to how long you run your CPU - a good stability test usually takes about 1-2 hours, so whatever temperature you hit at your peak is what your maximum temperature is. You can't say that your peak is what you see within 10 minutes - that is also not a fair assessment.

EDIT: And also - this is quite important - what is the maximum frequency you can achieve with 1.225V VCore? It also isn't fair to compare temperatures if you are running at a significantly slower clock speed. It is obvious that you will have lower temperatures.
Actually the frequency got close to nothing to do with the heat. The source of the heat is the amount of voltage needed and the higher voltage the hotter the cpu becomes. The heat should be the same no matter what the frequency is. Before I installed my water cooling system I used the stock cooler and I got around the same temps as you got with the same vcore ut with a freq of 4.0. So you're saying in this case that the Noctua cools the same as the stock intel cooler. Even if the freq where to increase heat he difference shouldn't be that much. By undervolting my vcore from 1.23-1.10 and overclocked my cpu to 4.0 I had shaved of 15C.

I just ran a test with both from cold startup. One with vcore 1.33 and freq 4.5, the other one still with vcore 1.33 but with freq of 3.9. There was no temperature difference at all. As I mentioned above vcore=heat while freq =/= heat.

Zka17
10-24-2012, 01:04 PM
Actually the frequency got close to nothing to do with the heat.

Actually, I do think that the frequency also contributes to the CPU's heat production... Increasing the frequency, the wattage will increase too - and that is translated in heat production...

Have you tried loading your system and register the max temps? The fact that you didn't observed temp differences in your test, it could mean that your cooling is over-provisioned at those settings...

I found this, it could help: http://www.extreme.outervision.com/tools.jsp

Krindor
10-24-2012, 01:31 PM
The difference in wattage between 4.0 and 4.5 at 1.33v was 12. I don't think that effects the heat that much. For a notible difference the multiplier would have to be at a 10 difference.

I used prime 95 to max the cpu. At idle both had cores at 34C and at max the cores where around 55C. When I change the vcore there's a clear difference of course but the multiplier isn't. I could redo the test with a higher vcore at around 1.45v and with a multiplier of 35 and 45 and see the difference.

squatchy
10-25-2012, 12:45 AM
Actually the frequency got close to nothing to do with the heat. The source of the heat is the amount of voltage needed and the higher voltage the hotter the cpu becomes. The heat should be the same no matter what the frequency is. Before I installed my water cooling system I used the stock cooler and I got around the same temps as you got with the same vcore ut with a freq of 4.0. So you're saying in this case that the Noctua cools the same as the stock intel cooler. Even if the freq where to increase heat he difference shouldn't be that much. By undervolting my vcore from 1.23-1.10 and overclocked my cpu to 4.0 I had shaved of 15C.

I just ran a test with both from cold startup. One with vcore 1.33 and freq 4.5, the other one still with vcore 1.33 but with freq of 3.9. There was no temperature difference at all. As I mentioned above vcore=heat while freq =/= heat.

Frequency definitely affects heat even if you are at a constant voltage. If you want proof, check out this article (http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge). Scroll down and see the results when the author fixes voltage at 1.25V and varies frequency. Obviously not as significant as changing the voltage, but there is a distinct, and noticeable difference when you vary frequencies as small as 300MHz.

Your results may be the same under idle conditions, but this again, isn't a fair comparison. My Noctua heatsink and the stock Intel heatsink don't show that much of a difference under idle conditions, but I see a distinct 10-15C difference under peak load temperatures at stock frequencies. Maybe it's an anomaly with my setup, but the Noctua cooler is definitely better.

Again, I can probably improve my temperatures if I removed my massive HDDs that sit right behind my front intake - but even then, the reasonable limit with air-cooling is still about 4.6GHz for 24/7 usage (i.e. keeping core temperatures below 75C).

Chino
10-25-2012, 02:31 AM
I used to run my i7 3770k without OC using it's stock cooler and after maybe a few hours of gaming, some of my cores reach as high as 70C. So you reaching 86C with an overclock of 4.5GHZ seems normal to me.

Just wondering, are you absolutely sure you can't run less voltage and still be rock stable @ 4.5GHZ? :p

squatchy
10-25-2012, 03:22 AM
I used to run my i7 3770k without OC using it's stock cooler and after maybe a few hours of gaming, some of my cores reach as high as 70C. So you reaching 86C with an overclock of 4.5GHZ seems normal to me.

Just wondering, are you absolutely sure you can't run less voltage and still be rock stable @ 4.5GHZ? :p

Yeah - I figure it's pretty normal for air-cooling. Comparing with a custom water-cooling loop won't give any meaningful information other than the fact that it's simply much better than air-cooling!

I just ran a few other tests and here are my results (using LinX for loading with maximum available memory):

3.9 GHz @ 0.984 Vcc measured (Offset mode with - 0.175V offset) - maximum peak core temperature: 59C
4.4 GHz @ 1.175 Vcc measured (Offset mode with - 0.050V offset) - maximum peak core temperature: 74C
4.6 GHz @ 1.295 Vcc measured (1.185V fixed Vcc in BIOS + 75% LLC) - maximum peak core temperature: 89C

With regards to your question about voltage @ 4.5GHz - I'm not sure. I was being lazy and since offset mode wasn't working anymore I guessed at 1.225V and set my LLC to 75% just to see if it was stable. I have a feeling that my case fans weren't running at full speed because in hindsight 86C is pretty high considering I'm able to get 89C @ 4.6GHz with a significantly higher core voltage.

So, you're probably right. If I tweaked a little more, I think I could get a stable core voltage as low as 1.22V. I doubt the average chip can do much better. If you can, then you must have a chip that came from a good batch!

As you can see from my results - 4.4GHz is sort of a sweet-spot in terms of voltages/temperatures. Trying to increase even one multiplier step above this requires a significant boost in core voltages (~50mV) and trying to increase 2 steps to 46 requires an astounding ~100mV and now core temperatures are just a little too high for every day usage in my opinion.

Theoretically, no normal workload will ever get temperatures as high, but still - I'd rather not take that chance when LinX gives out anything over 85C.

Once I buy a few more case fans, and maybe tinker around with my top exhaust fan (changing it to an intake), I'll re-test and post results. :D

Chino
10-25-2012, 03:42 AM
LOL. You can't get lazy when it comes to overclocking. Start over. :p Start at your default vcore and see how much MHZ you gain. A real golden chip can get around 4.4GHZ at stock voltage. My chip is normal at best. I can get to like 4.1GHZ at stock.

See how much you can get out of yours. And start there lol. If you're truly serious about overclocking your chip while protecting it. :p

squatchy
10-25-2012, 03:57 AM
LOL. You can't get lazy when it comes to overclocking. Start over. :p Start at your default vcore and see how much MHZ you gain. A real golden chip can get around 4.4GHZ at stock voltage. My chip is normal at best. I can get to like 4.1GHZ at stock.

See how much you can get out of yours. And start there lol. If you're truly serious about overclocking your chip while protecting it. :p

Haha, this is very true. But I wasn't doing anything crazy - voltages at 1.225 is a very reasonable value, and I was closely monitoring temperatures all the time. I think I hit it dead on, by the way - anything less than 1.225V gives me instability unless I significantly toy around with LLC or some other skew or PLL settings.

And if a real golden chip can hit 4.4GHz @ stock, mine is easily getting there. I think most chips will get this far anyway. You may be a little unlucky with yours. There's no way to determine what stock voltage actually is, because if you set everything to Auto and load your CPU at stock settings, it can pull a mind-boggling 1.25V which clearly isn't right. So, if we assume that stock voltages vary from 1.1-1.2V, then I'm getting very good results. :)

EDIT: My CPU can also throttle down to 0.725V @ 1.6GHz and that is something that I really like. Offset mode works great for anything less than 4.4GHz which is where I (and most other people) plan to stay for daily usage, and the power savings are a nice benefit :D

Chino
10-25-2012, 04:41 AM
The typical stock Vcore is around 1.05V. :p

squatchy
10-25-2012, 04:44 PM
The typical stock Vcore is around 1.05V. :p

I don't think this is true. I booted up my new PC and never saw stock voltages drop that low - even at idle. Maybe if the speeds dropped to 1.6GHz and it was idling it would go down that low. Try running your computer on factory settings under load and see what voltage you get.

Krindor
10-25-2012, 05:28 PM
I had my stock at 1.23 which ran pretty hot with stock cooler. The stock voltage is different for each cpu as well. To find out either clear cmos or reset everything to default in the exit menu.
If you have 1.05 as stock voltage then you have one hell of a golden cpu :D

squatchy
10-26-2012, 06:47 AM
I had my stock at 1.23 which ran pretty hot with stock cooler. The stock voltage is different for each cpu as well. To find out either clear cmos or reset everything to default in the exit menu.
If you have 1.05 as stock voltage then you have one hell of a golden cpu :D

I guess so. Haven't seen/heard anyone idling at this low of a voltage with stock settings, though.

Krindor
10-26-2012, 07:10 AM
Indeed, it is possible when undervolting at 3.9 but at stock voltage, Intel ain't that crazy

Amen_Ra
10-27-2012, 10:09 PM
Hello everyone,

So I'm pretty new to overclocking - but after doing some extensive research and such, I think I'm somewhat knowledgeable and won't do anything too stupid.

Anyway, I've managed to get the following:

4.5GHz overclock with VCore set to 1.225V (manual) and LLC set to ultra high (75%). This was tested to be stable with LinX and my peak temperatures were 86C for the hottest core.

I'm just wondering if this is an acceptable result? Do you think having it at this setting for 24/7 usage will be reasonable and won't damage my CPU over time? I know the temperatures are slightly high for everyday use, but I highly doubt I will ever get that high with any normal workload. Even AIDA64 Extreme doesn't drive temps as high as LinX.

My system specs:

CPU: Intel Core i7-3770K
CPU Cooler: Noctua NH-D14 CPU Heatsink
Motherboard: ASUS Maximus V Formula
RAM: G.Skill Ripjaws X 16GB (4x4GB) DDR3-1600 CL9 (9-9-9-24 CR2) @ 1.5V
GPU: Gigabyte GeForce GTX 680 OC
Power Supply: Corsair Professional HX850

Thanks in advance for your feedback!

Im guessing your on air cooling? I too am at 4.5Ghz but at 1.288 I have my LLC to extreme with water cooling my temps max out at around 55 to 60C.

Are you in a hot area?

I-Inc iF281D 28" LCD Monitor http://valid.canardpc.com/show_oc.php?id=2559164
Logitech Z-5500 5.1 THX Surround System http://speedtest.net/result/2246056731.png
Asus ROG Rampage IV EXTREME Motherboard
32GB Corsair Vengeance Quad Channel DDR3 1600Mhz
Seasonic X-1250W ATX Power Supply
Intel I7-3930K OCd to 4.5Ghz at 1.288Vcore stable.
Coolermaster CM Storm Full Tower Case
Thermaltake Water 2.0 Extreme CPU Liquid Cooling System
Sapphire Radeon HD 6950 2GB DDR5 Ocd to 1250Mhz
Soundblaster X-Fi Titanium Sound Card
Plextor SATA DVD-RW Drive Logitech G-700 Gaming Mouse

squatchy
10-28-2012, 04:51 AM
Im guessing your on air cooling? I too am at 4.5Ghz but at 1.288 I have my LLC to extreme with water cooling my temps max out at around 55 to 60C.

Are you in a hot area?

Yep - I have a Noctua NH-D14. I'll re-test at 4.5GHz though - because I get roughly the same load temperatures at 4.6GHz with your voltage for 4.5GHz (around 1.29V) as my 4.5GHz which leads me to believe I didn't have my case fans on full speed at the time.

Nope - my ambient temperatures are around 22C, but as I explained, I probably have some airflow issues as my front intake is blocked by 4 7200 RPM HDDs.

Also, you have a very good liquid cooling system - not really comparable to air cooling! For a NH-D14, I think my temperatures/voltages are pretty darn good.

samual
11-06-2012, 11:35 AM
I'm new at overclocking @ 4.6 ghz my temps were

core 0 - 82
core 1 - 90
core 2 - 90
core 3 - 85

This was after running prime for 4 hours. I'm using an H80 on a haf xm. I decided to go to full watering cooling. I had
a noctua c14 before and the temps were about the same. My opinion is if your going to overclock go to a full water cooling system.