Is 1.52v really safe for Sandy Bridge?

[SickBeast]

It's like leaving a 100w bulb lit.

It's true and it probably doesn't cost much, but when you consider the degradation to the CPU, 5ghz doesn't seem worth it to me.

To me that's something you should do if you need that extra performance for something. In my case I'm a gamer and I'm completely GPU limited until my next upgrade.

 

[nyker96]

Did you smell something funny in the air?

 

[SickBeast]

No. Why would you even ask that?

 

[Joseph F]

I used the Cooler Master paste that came with my old 212+.

The way I install paste is to put a small dab in the middle of the CPU that's about the size of a grain of rice. Then I saran wrap my index finger and spread out a thin layer over the core.

On stock volts at 4.4ghz I'm only hitting 55C.

It's a cheap heatsink as well. I paid $45 for it.

Just so you know, Cooler Master paste sucks. I would recommend some HeGrease. It will probably get you an extra ~5C of cooling, if applied properly.

 

[chimaxi83]

I've hit 5.3 GHz highest speed on both my cpu's, my 2500K at 1.56V and 2600K at 1.55V. For stable 24/7 operation I've been relatively lucky with 4.8GHz at about 1.37V.

I custom water cool and 1.55V load temps hit nearly 85C

 

[SickBeast]

Just so you know, Cooler Master paste sucks. I would recommend some HeGrease. It will probably get you an extra ~5C of cooling, if applied properly.

Interesting. I'll have to get some. Thanks.

 

[Idontcare]

I've hit 5.3 GHz highest speed on both my cpu's, my 2500K at 1.56V and 2600K at 1.55V. For stable 24/7 operation I've been relatively lucky with 4.8GHz at about 1.37V.

I custom water cool and 1.55V load temps hit nearly 85C

Yikes, what are the water temps like at that point?

 

[SickBeast]

I've hit 5.3 GHz highest speed on both my cpu's, my 2500K at 1.56V and 2600K at 1.55V. For stable 24/7 operation I've been relatively lucky with 4.8GHz at about 1.37V.

I custom water cool and 1.55V load temps hit nearly 85C

Wow, those are some fast CPUs. You're brave giving them that much voltage.

 

[ehume]

The summary page of the most careful, most thorough series of reviews is here. Phobia HeGrease was indeed the top finisher of those tested so far, but the Prolimatech PK1 was only a bit behind, if that's easier for you to get.

Vapor thinks HeGrease may be the same paste as Gelid GC Extreme (well-reviewed here, but that review is going on 3 years old now). GC Extreme can be found with free shipping here, when it's in stock.

I've used both the GC Extreme and the PK1, and don't see much difference. They're both very easy to work with.

 

[Joseph F]

Interesting. I'll have to get some. Thanks.

Anyway, here's a bar graph showing the performance of HeGrease compared to AS5 and Indigo Xtreme:

Remember that Indigo Xtreme is an unconventional liquid metal-based TIM, and costs $19-$20 per application for LGA1155.

 

[Kenmitch]

It's true and it probably doesn't cost much, but when you consider the degradation to the CPU, 5ghz doesn't seem worth it to me.

To me that's something you should do if you need that extra performance for something. In my case I'm a gamer and I'm completely GPU limited until my next upgrade.

Most of the time my rig sits at idle but ready to pounce if needed

I"m not a gamer anymore but I do have a couple of benchmarks that I run once in awhile just to see cpu/gpu system stability. As far as I can tell the load on the cpu is tiny. The temps are insanely low and the vcore is lower than if I run Intel Burn Test for example.

Using AVP D3D11 benchmark for example as it was the first shortcut I found. Most likely not a heavy hitter as it runs like butter!

Here's a screenshot of voltages,temp,power draw, etc. 4.8ghz overclock.

In this shot I did some Intel Burn Test runs. Compare the max values to the above AVP run. 4.8ghz overclock

It looks like it's a huge win for gaming vs IBT as far as temps, voltages, watts, etc.

To me it looks like the maximum overclock as far as power draw, wattage used, etc. is only drawn when it's trully needed. Supply and demand type of thing.

Wouldn't a person be more succeptable to possible chip degredation during extreme prolonged stability testing more than everyday use? I'd think that depending on what your rig is used for the more leanient it would be to vcore to achieve maximum load stability.

 

[Idontcare]

That's impressive they are zeroing in on indigo extreme, I've used indigo and it really does deliver temperature improvements (4C decrease versus using NT-H1 for me). But it's a pain to clean up when you reseat your HSF.

 

[Joseph F]

That's impressive they are zeroing in on indigo extreme, I've used indigo and it really does deliver temperature improvements (4C decrease versus using NT-H1 for me). But it's a pain to clean up when you reseat your HSF.

What amazes me most about the stuff, is that it's $10 for a 3.5G tube, whereas IX is $20 for a single use. D:

 

[TidusZ]

I don't even run my 65 nm C2Q at 1.4, let alone 1.5. Your asking for problems imo.

 

[Joseph F]

I don't even run my 65 nm C2Q at 1.4, let alone 1.5. Your asking for problems imo.

Voltage doesn't always decrease linearly (or at all) with a process node shrink. I don't know why, but I've found it to be that way.

For example, the 65nm Phenom X4 9600 had a stock voltage around 1.2-1.25V, while the 45nm Phenom II X4 940 had a stock voltage around 1.3-1.4V, if I'm not mistaken.

 

[Grooveriding]

It's high. My chip does 5ghz at a shade under 1.5v and is stable for all but linx. I can run prime etc, but linx crashes unless I go up to about 1.53 or so.

I believe those last few 100mhz that take the big voltage increases are just not worth it. It's satisfying as an enthusiast to see a bigger number, but it's way more than you need for anything but a serious encoding rig. After playing with my setup I found my chip will run 4.7 with only 1.32V which is pretty darn amazing for a SB-E. For every 100Mhz on top of that it takes me about .05 to get there.

I think the best bet is just find the sweet spot for voltage to speed and stick to it. I just run 4.7 @ 1.32 now and my temps are drastically better than they were @ 5.0 1.49. Went from high 50 to mid 60s loads to low to mid 50s. I'm actually amazed at your temps on air, because I get similar temps on water @ 5ghz with almost the same voltage you are using.

 

[Idontcare]

What amazes me most about the stuff, is that it's $10 for a 3.5G tube, whereas IX is $20 for a single use. D:

IX is $10/use. Its 2 applications in the $20 package.

I look at it like this - we are already making a series of costs/benefits decisions in leading up to buying our system and a non-stock cooling.

When we contemplate purchasing a NH-D14 vs an H100 for example, we are making a costs/benefit decision on the basis of added cost to the overall expense of our system (a $700 rig becomes a $770 rig if you add a D14 into the mix)...and we justify the 10% cost adder on the basis of the benefits of the lowered temperatures (aesthetic value) and higher performance (higher OC because of lower temps) and lower overall operating cost (lower power consumption because of lower temps and lower Vcc).

So in this situation, the costs/benefits decision for sticking with $10 goop versus $20 IX comes down to the performance benefits. If the $10 goop is NT-H1 then you could spend $10 more and have your system perform up to 4°C cooler.

I've already bought the $700 system because I wanted performance X. I then bought a $70 3rd party HSF because I wanted performance X+Y.

And now I'm down to arguing whether or not I want to spend $10 more versus $20 more for something that can take my temperatures down another 4°C...do I want performance X+Y+Z and does the difference between $770 versus $780 make performance X+Y+Z not worth it while performance X+Y is worth it.

In my case my system was more like $1000, and I paid $100 for cooling, why would I bother dickering over $10 goop versus $20 IX at that point? Give me the 4°C cooler stuff, for sure.

(But I won't use IX going forward of that other goop can be easily had through Newegg because then the costs/benefits would say the benefits aren't there anymore )

 

[Idontcare]

Voltage doesn't always decrease linearly (or at all) with a process node shrink. I don't know why, but I've found it to be that way.

Voltage changes with node changes are actually not tied to the node shrink itself.

What determines the necessary voltage for the xtors to switch is something that is worked on in parallel to the rest of the stuff being done to physically shrink the lateral dimensions of the transistor.

In other words, what the voltage at 45nm be lower than that of 65nm is not because the xtors were 45nm, but because there was a team of engineers who were working on the xtor parametrics (gate oxide thickness, channel engineering, stress engineering, etc) at the same time but all process techniques that could have been implemented in 65nm to lower the voltage of 65nm just as it did for 45nm.

That is why voltage decreases with node shrinks do not seem to make a nice trend, its because the cause-and-effect you are thinking is involved is actually non-existent.

 

[pelov]

Voltage changes with node changes are actually not tied to the node shrink itself.

What determines the necessary voltage for the xtors to switch is something that is worked on in parallel to the rest of the stuff being done to physically shrink the lateral dimensions of the transistor.

In other words, what the voltage at 45nm be lower than that of 65nm is not because the xtors were 45nm, but because there was a team of engineers who were working on the xtor parametrics (gate oxide thickness, channel engineering, stress engineering, etc) at the same time but all process techniques that could have been implemented in 65nm to lower the voltage of 65nm just as it did for 45nm.

That is why voltage decreases with node shrinks do not seem to make a nice trend, its because the cause-and-effect you are thinking is involved is actually non-existent.

But, correct me if I'm wrong, isn't it a necessity? Off the top of my head, the die size reduction (in what would essentially be a straight shrink) would benefit (or even require) a decrease in vcore, correct?

 

[Blades]

I'd really like to put nylon hoses on my H80 (stainless steel hoses would be silly) and submerge the radiator in ice water.. Wonder what would happen then..

 

[videogames101]

I'd really like to put nylon hoses on my H80 (stainless steel hoses would be silly) and submerge the radiator in ice water.. Wonder what would happen then..

condensation

 

[BallaTheFeared]

Idontcare...

I'm about to order some thermal paste, I've been getting Shin-Etsu X23 for awhile because I also like to use it on my gpu's. It's non conductive which is what got me away from AS5. I use it on my cpu, gpu, vram, and vrms

I'm not really concerned about my i5-2500k's temps, even at 5.2GHz they're laughable, however I can always use lower gpu temps.. My i5-2500k is strange in so far as it doesn't care what temps it's running at they don't provide any benefit outside of peace of mind.

My 470's on the other hand benefit from lower temps, would the plastic application thingie thing thing of IX be large enough to work on a large 40nm gpu IHS?

Price difference would basically be $27 IX + Shin vs $14 Shin x2

 

[SickBeast]

Balla what voltage do you need to run yours at 5.2ghz? Do you run it at that speed 24/7?

 

[Idontcare]

Idontcare...

I'm about to order some thermal paste, I've been getting Shin-Etsu X23 for awhile because I also like to use it on my gpu's. It's non conductive which is what got me away from AS5. I use it on my cpu, gpu, vram, and vrms

I'm not really concerned about my i5-2500k's temps, even at 5.2GHz they're laughable, however I can always use lower gpu temps.. My i5-2500k is strange in so far as it doesn't care what temps it's running at they don't provide any benefit outside of peace of mind.

My 470's on the other hand benefit from lower temps, would the plastic application thingie thing thing of IX be large enough to work on a large 40nm gpu IHS?

Price difference would basically be $27 IX + Shin vs $14 Shin x2

Hey Balla, good question!

I thought about doing that for my GTX460 (since I had two application packs and only used one for my 2600k), but the application process and the curing process really looked to be super finicky and I decided I wasn't going to do it and risk destroying my 460.

Instead, I went the "pop the IHS off" route for my 460, and then used NT-H1 on the bare die. Worked great (10°C temperature improvement).

With this experience, I will always be popping the IHS off my video cards until such time that the IHS are welded on like Intel does theirs.

 

[geokilla]

How can Intel even tell that you were running your processor outside of normal specifications? Like the chip failed due to overclocking?