Question for experienced overclockers? 3570k

C

chronicfx

Limp Gawd
Joined
Mar 29, 2012
Messages
286
I have a 3570k I have been trying to push a little further and a little further every-now and again. Obviously the dream is the 5GHz chip. Probably not happening without some de-lidding and water but I do think I have the chip to do it.... eventually.

My question is one that I know that noone knows the exact answer too and is asked probably a little bit too much. So I am looking for opinions and experiences etc.. With pushing the limits of your chips 24/7. I am primarily a gamer on a single screen at 2560x1440 and I plan to go SLI GTX680 by xmas. I only play for 2-3 hours per night and the computer is idle or asleep for the rest of the day. Having 2 young children really limits the daytime video games. My current plan is to skip haswell, the one after that, and then upgrade, much like my last chip was a Q9550 and I skipped the nahelem? and sandy bridge.

Currently my chip is at 4.8GHz with a D14 air cooler. My voltage is 1.320v under 8k ffts (temps 90 degrees C) and during gaming sits at 1.352v - 1.360v (LLC 3 for asrock board). I have been thinking about running 4.9GHz as a 24/7 the only thing is my 8kfft voltage moves to 1.352v (95 degrees C again LLC=3) and my gaming voltage 1.392v. I am really debating whether to move to 4.9 for 24/7 and what this voltage will do for the longevity of my chip.

I have read the overclocking guides for ivy and the majority say on air 1.45v is the max recommended but you will never get there because of heat. Is this still an accepted value? My desktop is functional at 5GHz also starting at 1.4v and I bet I could get it stable at under 1.45v (I would need to de-lid and buy WC kit for this or maybe the extra 10 degrees de-lidding will afford me is all I need...).

So now that you have background..

Experienced overclockers I ask you.. "Do you feel" that this chip will be alright for three years subjected to these voltages? and what will I see if my chip is not alright? Will the multi and voltage just need to be updated at some point?

Also I am aware that I have a lower end mainboard (I had a beer budget with champaign taste.. SO I had to cut somewhere) and Sin shits all over it over at OC so I am prepared for it to decay at some point unless he is wrong.

Also these voltages are overnight prime 27.7 stable except for 5GHz so I shouldn't need to tweak them (fingers crossed) I didn't prime 5GHz because cinebench was 89 degrees (8.01 score) and I didn't want it to throttle.
 
I'm surprised your temps are not hotter for your voltage. What speed and latency of ram?

Delidding and direct die contact will lower temps for maximum performance and life of IB
is a must to me.

Your voltage seems high for 24/7 right now. I run 4.7GHz stable with 1.2v for safe operation with ram 2400mhz Cas 10.


I plan on delidding soon.
 
Last edited:
My 2600k has been pushed close to the limit for the last year and a half at nearly 100% nearly the entire time folding. The only "break" it gets is when I log on to play a few games at night and that's about it.

That being said, a little over a year ago everyone said you shouldn't push the SB's much past 1.4v, and I haven't. At 4.8Ghz my 2600k runs about 1.392v, which is actually more voltage needed when I started overclocking it.

In July and August, I actually backed the chip down to 4.6Ghz, because I noticed the temps were getting up to 82-85C in the middle of the day, (with central air running in house all day). Other than that, this chip has had few breaks. I think the heat hurts your chip a lot more then the voltage, (but then again when you push more volts, you get more heat).

I will say this though, for gaming, You'd probably need, at a minimum of a SLI/Crossfire 7970/GTX680 rigs to tell any difference whatsoever, between 4.2Ghz and 4.8Ghz with my 2600k. When the SB/IB chips are running that fast, the video cards quickly become the bottleneck and probably stay that way until you have well over $1.2k invested in graphics cards. The other night I watched my usage monitor and it actually hovered at 14% while playing Smite online with a couple friends, that's with an overclocked Radeon 6970.

My first chip I overclocked was a P2-200, overclocked to 233Mhz. I have overclocked every chip I've owned since then and have never killed one. But then again I never tried to push my chips past the limit. To me the limit for SB chips would probably be 90C and 1.4v+. Yes you can run them that hard, but you probably won't be running it too long that way. Check overclockers.net or a few other sites, there have been plenty of people that killed SB and IB chips doing this.

The only person that will ever know the difference between 4.8Ghz and 5Ghz is you. You probably couldn't tell the difference if you set the computers next to each other and played on both.
 
My ram is CL9 1600MHz at 1.500v. My 4.7 and my 4.8 settings actually do not have a large diff in vcore. This is because I need to enable PLL overvoltage for 4.8 and over but it is not needed for 4.7. My PLL is set to a firm 1.65v in bios I haven't tested lowering it more I just set it there one day and it worked and its low so I left it alone. I don't have hours on end to overclock, just about 2 hours a night to change voltages, multis and "mini-stress", then stressing overnight and I have to cut out gaming to do it. I actually have not tested 4.7 with PLL overvoltage enabled yet but I bet it needs less volts. Whats your personal limit for vcore and temp Sonda5?
 
chronicfx said:
Whats your personal limit for vcore and temp Sonda5?
Click to expand...

My vcore limit is based on temps.

I stress test with LinX with all memory in windows diagnostic mode. I do the standard all memory 20 pass test.

At 4.7GHZ with 1.2v my temps on my 2nd core are low 90s and rest in low to mid 80s. This is as high as temps as I want to ever go with LinX on stable IB system. In real life everyday operations my IB cpu will never see this temperature since no other application is designed to heat up cpu like LinX.

With these settings seems perfectly stable. I would like to be no higher than low 70C for this linX test but IB is extremely hot. I blame the poor IHS, from uneven IHS surface to poor TIM. Intel IHS on IB is a mess. I plan on delidding soon improve temps as much as possible.

I think once I delid and go direct contact I will be able to run 5GHZ with similar temps to what I get now at 4.7ghz.
 
I agree with you on that. But regardless you are going to need alot of voltage to make that jump from 4.7 to 5. Its not gonna happen with anything under 1.375v at the min. if you have a great chip. Is that a voltage you would be comfortable running at 24/7 if IBT hit a max of 90? I have been using LINX too but I am starting to get swayed the other way, I am finding that whats stable in linx seems to need alot more voltage to get through a night of prime without WHEA errors. Over the last couple of months I am slowly switching back to prime.
 
chronicfx said:
Experienced overclockers I ask you.. "Do you feel" that this chip will be alright for three years subjected to these voltages? and what will I see if my chip is not alright? Will the multi and voltage just need to be updated at some point?
Click to expand...

I think it will depend on the chip, really. I've had some where the OC "degraded" after a year the chip could no longer reach waht it used to. I really had this issue with my old A64 and A64X2 systems. Eventually, I just had to lower the multiplier/FSB until I reached stable operation again. My A64 6000+ actually got bad enough that I had to down clock it all the way around - FSB, multi and RAM.

I'm sure most people don't share the same sentiment, but I don't plan on my chips lasting 3 years. If they do, awesome, go me. If they don't, meh, buy another. I tend to abuse my chips a bit, so when one starts showing wear, like my A64s did, I look to upgrading. With the pricing on the current crop of i5/7s, it's kind of no big deal.

Oh, and I can totally tell the difference between 5.1 and 4.8ghz. 4.8 actually feels faster.
 
chronicfx said:
I agree with you on that. But regardless you are going to need alot of voltage to make that jump from 4.7 to 5. Its not gonna happen with anything under 1.375v at the min. if you have a great chip. Is that a voltage you would be comfortable running at 24/7 if IBT hit a max of 90?
Click to expand...



Will depend on temps.
 
baldrik said:
I think it will depend on the chip, really. I've had some where the OC "degraded" after a year the chip could no longer reach waht it used to. I really had this issue with my old A64 and A64X2 systems. Eventually, I just had to lower the multiplier/FSB until I reached stable operation again. My A64 6000+ actually got bad enough that I had to down clock it all the way around - FSB, multi and RAM.

I'm sure most people don't share the same sentiment, but I don't plan on my chips lasting 3 years. If they do, awesome, go me. If they don't, meh, buy another. I tend to abuse my chips a bit, so when one starts showing wear, like my A64s did, I look to upgrading. With the pricing on the current crop of i5/7s, it's kind of no big deal.

Oh, and I can totally tell the difference between 5.1 and 4.8ghz. 4.8 actually feels faster.
Click to expand...

If 4.8ghz feels faster then 5.1ghz, then you're likely unstable at 5.1.
 
Even if your temps and voltage's are within a safe boundary, at that voltage and %OC it is still going to degrade the life expectancy of your cpu.
 
The short answer is, we don't know yet.
The only conclusive "safe" place for vcore is at stock.
The people that point at how high the VID chart goes as the safe limit are out to lunch, that's just how high the voltage controller can go natively. It has nothing to do with how much the CPU can take.

Personally I'm keeping mine under 1.3v, over that it gets awfully hot awfully fast while implies more leakage than at lower voltages. In theory this could mean more electromigration is happening. It also might not, of course.
 
Jorona said:
If 4.8ghz feels faster then 5.1ghz, then you're likely unstable at 5.1.
Click to expand...

That's my thinking too, I just havent had a chance to fiddle with it anymore. It primed for 18 hrs and did IBT at the maximum with no errors on either.

Edit: Now that I think about it, IBT showed 130-133 Gflops @ 4.8 and 110-18 Gflops @ 5.1. Something is not right at 5.1 for sure.
 
Last edited:
baldrik said:
Edit: Now that I think about it, IBT showed 130-133 Gflops @ 4.8 and 110-18 Gflops @ 5.1. Something is not right at 5.1 for sure.
Click to expand...

It's unstable somewhere, either your RAM is holding you back or you need to add a little bit of extra v to that cpu or the nb, have you turned off those power throttling settings in the BIOS such as speedstep, c1e, turbo power limiter, hardware adjacent cache and eist?
 
Odds are it's throttling at 5.1. Generally instability manifests itself with an error, not a drop in flops.
Getting too hot and throttling however will drop the flops.
 
chronicfx said:
I have been using LINX too but I am starting to get swayed the other way, I am finding that whats stable in linx seems to need alot more voltage to get through a night of prime without WHEA errors. Over the last couple of months I am slowly switching back to prime.
Click to expand...

If you are getting higher temperatures in Prime95 while under 100% vs LinX, then you need to update the binary packs for LinX that are released every so often..Here is a link for them directly from Intel..

I personally use IBT, which is the same as LinX, just with a cleaner interface...I run my cpu with "Custom Settings", where I specify it to use 15.5GB (of 16GB) of ram, and schedule at least 50 passes..I let it run overnight, and if there is any instability, trust me, you will find it quickly this way..Every system I have ever used is tested this way, and once those 50 passes are completed with 98% ram usage as well, I deem it to be 100% stable..

Now, to answer some of the questions you are wondering about...I run an Intel 3770K de-lidded, with an Enzotech Sapphire CPU block (beats or equals the top 3 modern blocks with flow rate and cooling)..I just rebuilt my loop and went with using IC Diamond TIM directly on the die..Which ASRock MB are you using? Can you please post the temperatures of your Vregs and Southbridge?

I am using an ASROCK Z77 Extreme MB that I honestly feel is one of the best Z77 boards out there when it comes to features vs price..My board cost me $89 out the door @ MicroCenter..I choose this board over two higher priced Asus and two Gigabyte boards...They did not offer anything that justified the minimum price price jump of $65 for the next cheapest board..It has been rock solid in everyday use, and I expect it to do the same when overclocked..Anandtech highly praised this board, as it was able to attain the exact same max overclock speed as the other more costly boards, and did so with ZERO ripple with a 3770K @ 4.8Ghz..That is very important when pushing for the bleeding edge, as I plan to do.

It is worth nothing that while the Vreg heatsinks were nice and evenly attached to every IC, the Southbridge heatsink seemed to wobble if you pressed down on a corner..I decided to remove it, and found it had hardly any TIM...I cleaned the stock paste off and applied a nice thin layer of Arctic Cooling MX-2..At the stock speed of 3.5Ghz (all power saving options off) my MB temperature is reported @ 28.5C..
I am going to dive into finding my max overclock starting tomorrow, so I will of course report back with my results if you would like..

I think it will depend on the chip, really. I've had some where the OC "degraded" after a year the chip could no longer reach waht it used to
Click to expand...

I have overclocked every CPU from my Pentium 100, through every generation of Intel and AMD cpus since, and NEVER had an overclock "degrade"..When I see people post this, I assume that the o/c was never truly stable in the first place, or the MB/PSU is not delivering enough clean, ripple free voltage to the CPU..A lot of older MB's used inferior Chinese caps, especially before all solid caps became the norm on any decent mid to high range MB..

*Snip*The people that point at how high the VID chart goes as the safe limit are out to lunch, that's just how high the voltage controller can go natively. It has nothing to do with how much the CPU can take.

Personally I'm keeping mine under 1.3v, over that it gets awfully hot awfully fast while implies more leakage than at lower voltages. In theory this could mean more electromigration is happening. It also might not, of course.*snip*
Click to expand...

There are several things wrong here, and I don't blame you for thinking these things are truths rather then opinions, as many people on various forums spout..

The VID table is actually quite a good guide to determining a starting voltage for a given o/c speed, as it is programmed by Intel to increase in a certain fashion..The cpu does not just randomly create it's own voltage, it follows the table that Intel specifies..This has been accepted pretty much as close to a fact as possible based on Intel whitepapers...

Just saying that your cpu gets hot "awful fast" without telling us what kinda of cooling you are using on the cpu, the amount of airflow your case provides, and what speed you are running the memory controller at makes it kind of hard to interpret what values you are referring too..people tend to run higher VTT values with IB in order to hit DDR2300+ speeds, since IB does make a use for these higher clocks where SB was happy with DDR1333/DDR1600..

Electromigration is often painted as the big bad villain lurking just around the corner, waiting to pounce on your system and destroy it quickly..In modern IC/transistor designs, "EM" potential is estimated by High Temperature Operating Life (HTOL), and the design of modern parts takes this into account..In addition, various metals are much better/worse @ resisting the effects of "EM", such as copper, which is very resistive..

I have an AS in EET, and my cousin is a P.E. with a Master's in EE as well..We have discussed this topic numerous times, along with my favorite professor who instructs many of of the upper level Engineering courses I am taking to finish my BS in EE...

If one looks @ the Wiki for "EM", the quote "When operated within the manufacturer's specified temperature and voltage range, a properly designed IC device is more likely to fail from other (environmental) causes, such as cumulative damage from gamma-ray bombardment." pretty much sums up what they both had to say on the topic..

Many people forget that modern CPUs are designed to work @ extremely high temperatures (ie 103C for IB) before they throttle since most OEM boxes have piss poor airflow..Taking these factors into consideration, I HIGHLY doubt that even with a VCC of ~1.4V-1.45V, one should worry about killing their CPU within a 5-7+ year window..Hell, even if they last exactly 5 years, @ that point it will be at least 2, maybe 3 generations old..

*Snip*I agree with you on that. But regardless you are going to need alot of voltage to make that jump from 4.7 to 5. Its not gonna happen with anything under 1.375v at the min. if you have a great chip. Is that a voltage you would be comfortable running at 24/7 if IBT hit a max of 90?*snip*
Click to expand...

Finally, I personally will be completely comfortable running my cpu @ 5Ghz with up to 1.45V, even if the temps do reach 87-90C under IBT..I highly doubt they will get anywhere near that high considering the amount of radiator space and de-lidding, but I would still run it that way 24/7 even if it did..
 
Thanks for the vote of confidence cc. I will try for 4.9 without a delid but it may be hot. Although only in linx. It would probably just hit 68 ish gaming. Though i may be upper 90's during ibt. Is 2.54 a better version? I had been using 2.53
How is your 480 treating u? I still have my two 280's.
 
chronicfx said:
Thanks for the vote of confidence cc. I will try for 4.9 without a delid but it may be hot. Although only in linx. It would probably just hit 68 ish gaming. Though i may be upper 90's during ibt. Is 2.54 a better version? I had been using 2.53
How is your 480 treating u? I still have my two 280's.
Click to expand...

I honestly do not think you are going to get anywhere near 4.9ghz with the stock TIM in place, unless you are running a full water (not H100 etc) setup..Nearly every review out there peaked @ 4.6-4.8Ghz, and they were hitting temps of 90+C, so the chips were throttling..

My 480 is doing just fine actually..it is faster then an o/c'd 570, thanks to the additional Vram, and is pretty close to a 580 in most games..I had a 6970, but I made the mistake of buying one that didn't use a reference PCB, thus no GPU block, so I sold it to a friend...I really don't see any difference speed wise..
 
chronicfx said:
Thanks for the vote of confidence cc. I will try for 4.9 without a delid but it may be hot. Although only in linx. It would probably just hit 68 ish gaming. Though i may be upper 90's during ibt. Is 2.54 a better version? I had been using 2.53
How is your 480 treating u? I still have my two 280's.
Click to expand...



Just a heads up.

My 3570k is a hot one and after I delidded and went direct contact on die without IHS my average core temp dropped 12.5C under load.

I was happy with the drop in temperature but it still runs hot. I have been doing alot of research on i5-3570k for over clocking and for the most part it looks like i5-3570k run much hotter than i7-3770k.


Right now my loop is hurting because I just sold a 360mm radiator and am running a gpu and cpu with a single slim HWlabs stealth 120mm radiator.

delidtestrig120mmradiator.jpg



I have an Alphacool Cool NexXxus UT 60 420mm radiator on order at FrozenCPU. As soon as they are in stock it is being sent to me.

For now my little 120mm radiator is doing ok.
The only reason I am running it like this is because I sold the 360mm radiator that was in the same loop last week.

Did this test a few hours ago.

20 run LinX pass in Windows Diagnostic Mode
4.8 GHZ
TIM IC Diamond 24


stable4p8.jpg


http://valid.canardpc.com/show_oc.php?id=2532673

http://hardforum.com/showthread.php?t=1664940&highlight=Intel+Burn+Test
 
Last edited:
That's what I am talkin about! Good overclock and nice voltage for that clock! I would run that clock and not worry about it.
 
chronicfx said:
That's what I am talkin about! Good overclock and nice voltage for that clock! I would run that clock and not worry about it.
Click to expand...



This may be my sweet spot. Hopefully my radaitor and loop mods I have planned will lower temps. Im also thinking about trying some Liquid Ultra. Just not certain is much better than IC diamond 24.
 
From what I have read they are not very different. The liquid pro may be slightly better by 1-2 degrees. People do say that IC diamond may "scratch" the surface of the die. I don't want to get you worried though, see what you can dig up on that.
 
There is a thread on over clock.net called ivb 3770 Tim testing (pic heavy) and on page 5 schmuckley mentions the possibility of scratching and killing your die using ic diamond. I would just link it but I am just learning how to use this iPad and have no clue how to copy and paste links yet.

If you do decide to remove the ic diamond I think it is important not to wipe it off but to kind of soak it off using a solvent like acetone I read. But definitely do your own research on it unless you are a seasoned ic veteran. I have never used the stuff myself but hear it has great performance at high temps.
 
chronicfx said:
From what I have read they are not very different. The liquid pro may be slightly better by 1-2 degrees. People do say that IC diamond may "scratch" the surface of the die. I don't want to get you worried though, see what you can dig up on that.
Click to expand...



I've used it on bare die without any major scratches does have some very minor minor scratches but the dies feels smooth. Really no bad scratches and the gpu works great. Come to think of it I did probably cause the scratches when I cleaned it off with a paper towel.

I'm going to switch. 1-2 degrees isn't much but if it will keep the die perfectly new condition and cools I'm going to switch. Only reason I went with IC Diamond is because I wanted the best and it's not very pricey. I've had good results from IC Diamond.

When I clean off the IC Diamond I will soak it down with a few drops of Goo off and dab it with a damp micro fiber rag. I will not smear if off and possibly scratch it. IC Diamond is like thick gum .

IC Diamond should re do their formula and make it into a super smooth synthetic liquid formula. Something that goes on like liquid and creates a nice bond like liquid Ultra.
 
Last edited:
One more question. Is that a purposeful bus speed bump to 100.5 or is that just a fluctuation?
 
chronicfx said:
One more question. Is that a purposeful bus speed bump to 100.5 or is that just a fluctuation?
Click to expand...

That is what my MSI Mpower Z77 runs at default. It is fluctuation. It is set to 10000 in the BIOS settings.
 


My 3570K easily overclocks to 4.6 GHz, but above that you're operating beyond the design limits of the hardware and any performance increase will be trivial.
 
reb00tin said:


My 3570K easily overclocks to 4.6 GHz, but above that you're operating beyond the design limits of the hardware and any performance increase will be trivial.
Click to expand...

Define easily? Just curious about what design limits you are referring to? I find the post to be very vague, perhaps I am just not understanding as easily as you do.
 
I've rethought the IC Diamond scratching problem and I think I did have a serious problem with a bad batch of IC Diamond.

This happened a few months back and I didn't now what happened at the time and I did an RMA for the block with the manufacturer and the manufacturer wasn't sure what caused it either.

Now I think the problem was caused by a bad batch of IC Diamond that had a chunk of hard diamond dust that cut into the block and left a mark on the die of my HD7950.

Photos from the problem that I had that I think was created by some chunky diamond crystals were in my IC Diamond.



Pit in block
ICDiamondPitproblem.jpg


Mark on die with some micro scratches.
markondie.jpg


This is the only case that I know of anyone having a problem like this with a any block or IC Diamond. I was very careful when I applied the IC Diamond and surfaces were completely clean. The block was brand new. Only had the IC Diamond on it for a few weeks before I took it off to do some loop adjustments and I found the problem.
 
Last edited:
damn! you probably could have mounted that chunk and given it to your girl :) i would have been worried to use it again. nice of them to replace the block if it was the tims fault.
 
chronicfx said:
Define easily? Just curious about what design limits you are referring to? I find the post to be very vague, perhaps I am just not understanding as easily as you do.
Click to expand...

Easily, I put +0.04V offset in BIOS, everything else is on Auto for a stable 4.6 GHz. That's 1.2 GHz above stock. Motherboards, for example, are designed to operate within a certain range, the manufacturer's specs will tell you what these are.
 
chronicfx said:
damn! you probably could have mounted that chunk and given it to your girl :) i would have been worried to use it again. nice of them to replace the block if it was the tims fault.
Click to expand...



At the time there was no solid explanation for what happened and the manufacturer was nice about it.

Thinking about it now I think the TIM was the problem though looking at the photo of the block it looks like there may be different piece of metal soldered onto that block and then a coast of copper over it. Mabey the TIM pushed through the coat of copper?
 
You must log in or register to reply here.
Back
Top