Confirmed Major Problems OC'ing Ivy Bridge - i7-3770K & i5-3570K

[jaquith]

I seen several Ivy Bridge articles by now, and I did notice some time ago odd and HIGH temperatures but the pattern is unmistakable and I'd say Confirmed - The IB has a major OC'ing problem. This is counter intuitive for the expected 'efficiency' of SB 32nm vs 22nm IB, I expected drops and higher obtainable OC's per both vCore and Temps?!

This is really bad, a 4.7GHz~4.8GHz @ 1.31v~1.36v vCore resulting in a near 100C on a Corsair H100 water loop! In comparison, my 6-core i7-3930K 4.8GHz @ 1.36v~1.40v is 70C on the hottest core using the same Corsair H100. That's at minimum 30C Hotter on the 4-Core/8-Thread Ivy Bridge vs 6-Core/12-Thread on Sandy Bridge Extreme.

The folks at EK, Koolance must be loving this and getting ready to count their money. Sure you can get a CrAzY high 7GHz+ on LN2 (Liquid Nitrogen), but I don't know too many if any Gamers playing BF3 running back and forth filling their LN2 pots??!!

ref - http://www.tweaktown.com/reviews/4663/asrock_z77_extreme6_intel_z77_with_ivy_bridge_motherboard_review/index11.html
ref - http://wccftech.com/intel-ivy-bridge-worse-overclocker-sandy-bridge/

 

[Uoppi]

So is 1.31V the minimum voltage required for a stable 4.7GHz~4.8GHz for the 3770k?

Can't remember which forum it was but just yesterday I saw someone had overclocked his 3570k to a stable ~4.8GHz with just a little over 1.2V (CPU-Z validated). No hyperthreading of course but I've understood even with 2600k/2700k you don't need much higher volts with hyperthreading on versus off?

 

[EzioAs]

Aren't those just engineering samples? The retail product shouldn't be something like that but who knows. Let's just wait for the true review around 2 weeks from now

 

[Uoppi]

Aren't those just engineering samples? The retail product shouldn't be something like that but who knows. Let's just wait for the true review around 2 weeks from now

The retail stepping is E1 which I believe was the one used in these latest benchmarks.

Don't know what the default voltage is but isn't it supposed to be something like 1.1V? So a jump to over 1.3V seems pretty high.

Edit: Here is the screenshot of the 3570k overclock (post #150): http://www.overclock.net/t/1242711/tweak-town-asrock-z77-extreme6-intel-z77-with-ivy-bridge-motherboard-review/140 .

 

[neon neophyte]

im not too surprised, even with sandybridge it was apparent that the cpu could heat up VERY quickly internally. even with a cold heatsink on it, the internal heat is still high.

seems this effect is only amplified for ivybridge. the days of heavy overclocking may be behind us.

 

[jaquith]

I simply hope Intel corrects the problem, Early Adopters be careful!

 

[Uoppi]

I'm still wondering why such high volts on those 3770k benches if the 3570k could be overclocked for a stable 4.8GHz @1.225V? Surely hyperthreading doesn't require that much more or does it? :O

 

[elbert]

Given the lower 22 nanometer 1.3 volts may be an over voltage. As the walls separating paths in the CPU get thinner so must the voltage passing though those paths get lower. The old 8086 worked on 5 volts which would fry these lower nm thin wall cores today.

Instead of trying to fix a symptom like the heat though revisions. Intel may have a harder task of finding something to lower voltage needed for higher clocks. This said the higher voltage for stable operation could be a factor of the motherboards needing better voltage regulation.

 

[jaquith]

There's also a relevant discussion more less starting here -> http://www.tomshardware.com/forum/page-303971_28_3500.html#t2505967 relevant statements I posted:

OC'ing 101 - you can set the vCore Fixed <or> Offset plus a variety of LLC & Current values depending on the BIOS. Therefore, unless the CPU is under full load you can 'show' e.g. 4.8GHz in CPU-z (C-States disabled) + Offset (close to stock vCore), or be 'real' about it and show the fixed vCore (C-States disabled or enabled) at a more realistic vCore....PLUS add another +0.01v~+0.04v depending on the LLC level/current.

4.7GHz @ 1.36v / 98C ; assumption = fixed - http://www.tweaktown.com/reviews/4663/asrock_z77_extreme6_intel_z77_with_ivy_bridge_motherboard_review/index4.html
4.8GHz @ 1.176v / 97C ; assumption = offset - http://www.tweaktown.com/articles/4621/intel_ivy_bridge_overclocking_with_the_core_i7_3770k_and_core_i5_3570k_cpus/index2.html

IB OC - here's my point and at least my deflated expectations of the IB CPU: I expected the IB vCore, regardless of how you played with your OC, to be considerably lower than the SB CPU. Frankly, just about all the Fixed vCores I've seen are (mid+) 1.3Xv for 4.6GHz~4.8GHz however my expectations were 1.1Xv or less with obtainable 'on air' OC's 5.5GHz~(lower)6.XGHz i.e. exceeding the current SB limits -- IB doesn't deliver (yet)! Instead, you are thermally limited on 'air' or 'water' to high 4GHz to very low 5GHz i.e. same as SB, but instead (unexpected) the IB produces much higher temps than the SB by minimum +15C to +30C for the same 4.5GHz+ OC which is a problem IMO Intel should fix.

'My' feeling is Intel rushed out the IB it's 22nm technology in attempt to combat AMD; Tick Tock.

Similarly, the SB-E had a plethora of similar screw-ups: CPU known errata C1/C2 VT-d, and Chipset X79 C0/C1 errata which I'm still trying to get a straight answer why the chipset required a revision, and lets not forget the most recent PCIe 3.0 compatibility from nVidia. Further, the SB-E upon release had known issues of high temps/higher vCore on non-ASUS MOBO's -- the best guess is 2%~5%+ with temp problems.

As I said, I noted high IB temps for some time but I needed to see more data and on different Z77's. I still recall the initial lack luster performance in the Gigabyte Z77's and my assumption was a BIOS issue; see - http://www.tomshardware.com/forum/329828-28-bridge-cpus-benchmarks-review-3770k-3570k My assumption was BIOS, and it seemed I was correct since other Z77's showed no such problems.

I've said it before and I'll say it again, now while AMD is more less no longer a threat to Intel, Intel needs to cool their jets and make damn sure both their Chipsets AND CPU's are spot on before delivering them to the public.

Bottom-line, I truly want the IB to be a killer CPU, and further I expect IF the Consumer IB vCore's are high that Intel re-release an errata update with the IB. Again, this spells-out being an Early Adopter as a risky proposition! We all remember the B2 errata fiasco, those with SB-E double whammy SB-E & X79 errata -- therefore since day one with SB->SB-E->IB(?) has been a bumpy road. Intel right now is like a double-bagger with a killer body...

 

[13thmonkey]

some Norwegian experience with retail chips. I'm going to do it anyway, I only want to hit 4.3-4.4Ghz so I'll be happy.

http://www.overclock.net/t/1242313/more-ivy-bridge-benchmarks-sandybridge-comparison-updated/160

 

[Uoppi]

Chipset X79 C0/C1 errata which I'm still trying to get a straight answer why the chipset required a revision

Curious to know if you've managed to find out anything of interest regarding the revision? It's supposed to be "available to customers" in a few weeks time.

With this apparent IB heat issue, it's looking more and more like X79 is the way to go for me. But I want the revised chip.

 

[jaquith]

Regarding X79, no concrete info, two guys that know more than I do don't know. Before I 'guessed' virtualization and IMO I probably am guessing right. Virtualization on the X79 is like using a 'screwdriver as a hammer' and is really meant for the C60x series and therefore not fully tested. Another SB-E issue is PCIe 3.0 validation for the nVidia GTX 600 series; for now nVidia is turning off PCIe 3.0 on the SB-E/X79; somewhere I posted a registry (hack) to re-enable PCIe 3.0 on the GTX 680's.

Lately with Intel it's one 'foul-up' after the other, it's almost funny.

 

[Uoppi]

Regarding X79, no concrete info, two guys that know more than I do don't know. Before I 'guessed' virtualization and IMO I probably am guessing right.

Not meaning to stray off topic too much but do you think efficiency improvements could be coming too (to any significant degree)? I've read about the IOH being surprisingly inefficient (?) and the IOH is what Intel is supposedly been fine-tuning.

Myself, I know next to nothing about the technical side of it but - like with Ivy - what I'd like to see is... less heat!

 

[jaquith]

The X79 is fine, IMO the vast (98%+) majority of folks are incorrectly install Windows 7 on the X79/SB-E. Here's my ATTO on the same Corsair CSSD-F180GBGT-BK 180GB SSD, but keep in mind I have this as my primary boot drive, Intel RSTE monitored, and SMART enabled (both slow the R/W).

X79 Jaquith (not my best ATTO either; the 2nd ATTO was done just now):
http://i1013.photobucket.com/albums/af254/Jaquith/ATTO_Sabertoth_X79.jpg
http://i1013.photobucket.com/albums/af254/Jaquith/ATTO_Corsair-GT-RSTE.jpg

P67 Legit Reviews (P67 is faster than Z68; the SSD below is a non-boot drive):
http://www.legitreviews.com/images/reviews/1804/atto-180.jpg

 

[neon neophyte]

one thing to be mentioned is, its possible this is all hokum in order to help clear out SB stock before IB arrives. send out some hot cpus and watch the SB just fly off the shelfs.

 

[13thmonkey]

there's some thoughts from overclock.net that these are 'oem' low bin chips that have been leaked, there's good ones and bad ones, good ones appear to be legitimate 'early release' as opposed to chinese early release.
I'm going IVB, don't want much of an OC, but do want PCI-E3.0 for my gpu upgrade which will be nvidia 7XX in late 2013 or whenever it is, long game its not so bad if not better than SB.

 

[4Ryan6]

Seeing as how any overclocking is a blessing given from the CPU manufacturer, meaning both camps could simply lock the adjustable parameters to operate in specification range windows and overclocking would simply be a thing of the past.

Unless hardware modifications are possible returning us to the Lazer Locked AMD days, don't really want to return to those days!

Meaning the overclocking draw is a sales factor enticement to overclockers only.

It's obvious now to Intel that they have shot themselves in the foot, meaning their desire to trump over AMDs Black Edition established CPUs by releasing the Intel K editions was obvious to grab that unlocked multiplier simplicity of overclocking edge AMD had.

Unfortunately for Intels greed they put in place their on limitations, when a budget or mid range CPU can be overclocked to competitive levels with their new Crown Jewel, there's a problem to be resolved on Intels part.

Of course the Crown Jewel can be overclocked but it's heat limited from all that's crammed under the hood, so I would venture to say the future released lower end Ks will suddenly become crippled and not perform as well overclocked as what's on the market right now.

The only true safety factor Intel presently has is very few overclockers have even discovered how to get all the Sandy Bridge Ks can deliver, and additionally that the overclocking community is a very small minority percentage of the entire computer using majority of the world as a whole.

Otherwise why buy the big dog that's being whipped by the little dog on steroids! :lol:

Note: This is my opinion and take on things, you may agree or disagree I don't care, you will not change my opinion, but it is your prerogative to try.

 

[jaquith]

To my understanding the initial IB ES were 77W but the shipping and latest IB ES are upped to 95W TDP. Story - http://www.nordichardware.com/news/69-cpu-chipset/45720-ivy-bridge-gets-95w-tdp-worse-overclocker-than-sandy-bridge.html Saturday April 14, 2012.

*note their statement "At 4.6 - 4.7 Ghz the temperature gets so high the processor throttles, meaning it clocks down to protect the processor from permanent damage. Comparing with Core i7-2600k that in many cases can do 4.8 - 5.0 GHz stable with a good after-market cooler, Ivy Bridge looks to be 200 - 300 Mhz worse than Sandy Bridge."

As far as IB vs SB vs SB-E, for Gaming it's 6 of one half a dozen of the other. Most folks have absolutely no need for a 6-core/12-thread CPU i.e. SB-E/LGA 2011. In 'my' case the more cores the better, my 'business' is Enterprise SQL. However, folks who do other forms of 'rendering' will feel the same, and the SB-E/LGA 2011 platform does offer 32-lanes of PCIe 3.0 to the GPU or at least in theory.

Some LGA 1155/Z77 + IB will offer a PLX PCIe switch to get, as I recall, 32-lanes of PCIe 2.0 with 4-WAY SLI support (x8/x8/x8/x8) with some burden of latency.

 

[colson79]

It's funny how after seeing one unofficial review with a IVY CPU they got in china from who knows where and it is confirmed that IVY bridge sucks. I will reserve judgement for the official release.

 

[jaquith]

No what's really fun is we are getting our info from China from an American based company. That's ROFL!

 

[dead_sincebirth]

I seen several Ivy Bridge articles by now, and I did notice some time ago odd and HIGH temperatures but the pattern is unmistakable and I'd say Confirmed - The IB has a major OC'ing problem. This is counter intuitive for the expected 'efficiency' of SB 32nm vs 22nm IB, I expected drops and higher obtainable OC's per both vCore and Temps?!

This is really bad, a 4.7GHz~4.8GHz @ 1.31v~1.36v vCore resulting in a near 100C on a Corsair H100 water loop! In comparison, my 6-core i7-3930K 4.8GHz @ 1.36v~1.40v is 70C on the hottest core using the same Corsair H100. That's at minimum 30C Hotter on the 4-Core/8-Thread Ivy Bridge vs 6-Core/12-Thread on Sandy Bridge Extreme.

The folks at EK, Koolance must be loving this and getting ready to count their money. Sure you can get a CrAzY high 7GHz+ on LN2 (Liquid Nitrogen), but I don't know too many if any Gamers playing BF3 running back and forth filling their LN2 pots??!!

http://i1013.photobucket.com/albums/af254/Jaquith/IB_OC_Temps.png
ref - http://www.tweaktown.com/reviews/4663/asrock_z77_extreme6_intel_z77_with_ivy_bridge_motherboard_review/index11.html
ref - http://wccftech.com/intel-ivy-bridge-worse-overclocker-sandy-bridge/

hi, i 100% agree with u "Intel" officaly anonced IB to be a low thermal and excellent overclocking processor , however from the very first day i was in doubt with there statement i shared a common physics principle in tweektown in order 2 make people understand tht why Ivybridges will must have higher thermal values and they will be never cooler than old sandybridge processors at anycost in case of normal clockrates , and ultimately worse in case of overclocking . how it happens let me tell u guys by a simple experiment and general laws of thermodynamics .

1). sandybridge : 32 nm , Ivybridge : 22 nm (note that point)
u see that there is a remarkable decrease in the chipsize of Ivybridge .

2).now as all of us know that a tea or any thing which is hotter gets cooled easily if the surface area of the tea is increased e.g by pouring it in to some plate or saucer .

now if we have 2 same vessals e.g a pot. lets boil 2 cups of water put one cup in to a steel coffee mug and other one in to a steel saucer and plate now put these respectively in both pots and cover the pots with 2 similar lids leave them for a minute or 2 after that touch both lids

u will notice that the lid of pot in which saucer was placed will be hotter than the one in which coffee mug was placed .


why ?

no just read the point number "2" it all happens due to increase in surface area the more surface area the more heat will evaporate which will ultimately warm the lid more quickly and with higher temperature than that of "liquid poured in a steel mug" which means that the liquid inside of a steel mug will be hotter but its lid will be cooler and the liquid in the plate will be cooler but its lid will be hotter...

now if we relate it with the Ivybridge processor . read the point no (1)


Intel will have to put the circuits closer in order to decrease the chipsize which will result in conjustion which will lead to less hottness escape and higher core tempereatures.

sorry to say Intel has not done his home work well on the desighning.

conclusion:
sandybridge is better than Ivybridge.

 

[13thmonkey]

hi, i 100% agree with u "Intel" officaly anonced IB to be a low thermal and excellent overclocking processor , however from the very first day i was in doubt with there statement i shared a common physics principle in tweektown in order 2 make people understand tht why Ivybridges will must have higher thermal values and they will be never cooler than old sandybridge processors at anycost in case of normal clockrates , and ultimately worse in case of overclocking . how it happens let me tell u guys by a simple experiment and general laws of thermodynamics .

1). sandybridge : 32 nm , Ivybridge : 22 nm (note that point)
u see that there is a remarkable decrease in the chipsize of Ivybridge .

2).now as all of us know that a tea or any thing which is hotter gets cooled easily if the surface area of the tea is increased e.g by pouring it in to some plate or saucer .

now if we have 2 same vessals e.g a pot. lets boil 2 cups of water put one cup in to a steel coffee mug and other one in to a steel saucer and plate now put these respectively in both pots and cover the pots with 2 similar lids leave them for a minute or 2 after that touch both lids

u will notice that the lid of pot in which saucer was placed will be hotter than the one in which coffee mug was placed .


why ?

no just read the point number "2" it all happens due to increase in surface area the more surface area the more heat will evaporate which will ultimately warm the lid more quickly and with higher temperature than that of "liquid poured in a steel mug" which means that the liquid inside of a steel mug will be hotter but its lid will be cooler and the liquid in the plate will be cooler but its lid will be hotter...

now if we relate it with the Ivybridge processor . read the point no (1)


Intel will have to put the circuits closer in order to decrease the chipsize which will result in conjustion which will lead to less hottness escape and higher core tempereatures.

sorry to say Intel has not done his home work well on the desighning.

conclusion:
sandybridge is better than Ivybridge.

Surely your conclusion should read 'I am smarter than all of intel'

Yes the watts per mm2 are higher than in sandy, about 10% higher. However if the heat spreader was perfectly connected then the heat spreader becomes a part of that surface area and the watts per mm2 go down.

The heat spreader is unfortunately very far from being perfectly connected, its a comparatively higher thermally resistant connection, this is the root of the problem.

However I have gotten my 3570K to 4.3Ghz, peaking at 70C in IBT 62C in game, with only 10mins work. This is about the same as a 4.5-4.6 Sandy. Given that all this was meant to be was a die shrink then there is nothing to complain about.

If they ever go back and resolve the heat spreader contact issue it will be much better. There's a report linked to on tweaktown indicating a 15C drop in temp at OC load by changing the tim between the die and the heat spreader, this indicates that the process is sound and that haswell will be cool if they use a good tim.

 

[4Ryan6]

This just makes My Water Cooling Solution much more appealing to those that have the room to accommodate it!

 

[13thmonkey]

This just makes My Water Cooling Solution much more appealing to those that have the room to accommodate it!

water can soak up an awful out of heat energy, and love the huge res. Haven't gotten into water yet, last time I considered it was in the days of the resorator (that big finned tower reservoir) for quietening an XP3200+ system.

If someone had a paid service for delidding and re-applying the internal tim with decent stuff, i'd go for it.

Link for the tim issue evidence http://www.tweaktown.com/news/24059/ivy_bridge_s_heat_problem_is_indeed_caused_by_intel_s_tim_choice/index.html

 

[4Ryan6]

:lol:

According to your own statements you don't clock past 4.0ghz, you could just mount a pin wheel and blow air inside your case occasionally to do that, can't you? :lol: