Asrock xtreme 4 2500k 47x sustainable (Hopefully) OC

[Lex_Sinistre]

Thanx to 4Ryan6,Leaps From Shadows, and Bear claw. Sorry for all of the idiot posts I made with stupid questions to everyone else. I finally got something that looks like it works, Here are the specs I am using:

Motherboard Asrock extreme 4....................................Bios 1.80

Settings entered in UEFI:

TWEAK TAB

Load Optimised CPU OC Setting.................................[Default] \OFF

CPU Ratio Setting....................................................[Manual]

Max Ratio...............................................................[ 47 ]

Internal PLL Overvault.............................................[Disabled]

Intel Speed Step.....................................................[Disabled]

Turbo Boost Power Limit..........................................[Manual]

Short Duration Power Limit (w).................................[ 200 ]

Long Duration Power Limit (w)..................................[ 200 ]

Long Duration Maintained (Seconds).........................[ 56 ]

Additional Turbo Voltage..........................................[+0.004]

Curent Core Limit...................................................[ 200 ]

BCLK...................................................................[ 100 ]

Spread Spectrum..................................................[Disabled]

DRAM CONFIG
[9,9,9,25, (tras)1N] Set manually to specs;

VOLTAGE CONTROL

Power Save Mode.................................................[Disabled]

CPU Core Voltage.................................................[Offset Mode]

Offset Voltage......................................................[- 0.005]

CPU LLC..............................................................[Level 5]

Dram Voltage......................................................[ 1.50 ]

PCH Voltage.......(1.059)......................................[Auto]

CPU PLL Volts......................................................[1.791]

Vtt&VccSA.........................................................[Auto]


ADVANCED TAB-> CPU CONFIG:

Active Cores.....................................................[ALL]

Hardware Prefetch............................................[Enabled]

ADj Cache line Prefetch......................................[Enabled]

Enhanced Halt..................................................[Disabled]

C3&C6 Support................................................[Enabled]

CPU Thermal Throttle.......................................[ Enabled ]

IVT.................................................................[Disabled]

NEMP.............................................................[Disabled]


I am reasonably hope full with this one, I ran a good 2 hours of Aida 64 extreme CPU Stress, Cache Stress, Mem, Stress. The first thing I like about it was that it sat at 47x all the way through and did not hunt around as it has previously (So that is a learning for me and an apology to all of you that had to waste time reading my stupid question as to whether Aida and Prime and OCCT had some default in them that made ur clock speed hunt like a ferret at a drainpipe convention)

Now I acknowledge that a short test like this on Aida is just the first question answered as to the systems real "Stability" I have more to do yet I know.

The good things I liked from the test were:

Max Core Temp 75; this was a oncer as well usually ran between 68 and 72 (Still have a cut price Aircooler; Can't wait till my Antec H20 920 Arrives in the Country.. bit worried it will be

noisy after reading more reviews)

Max Vcore under Load 1.336; Lowest Vcore under Load 1.320; as a general observation it spent a lot of its time at 1.328v


Max Vcore at idle 1.018; Lowest Vcore at idle 0.998v

Here is a link to the CPU - Z validation Site if you are interested.

http://valid.canardpc.com/show_oc.php?id=1902261

Good Luck to all. Thanks for the help it is an amazing site you guys have here. Any Comments on improvement appreciated.

Kindest Regards
Lex

 

[4Ryan6]

That cooler you have coming is definitely going to help your temps! It would be interesting when you get it, if you record your temps just before changing it out and let us know the temp difference after installing it.

 

[Lex_Sinistre]

Thank you for your encouragement 4Ryan6. You too Leaps from Shadows ( LFS).

I have an update; remember I said I have to test it with a heavier programme Well what follows is what happened and it is really quite interesting, I think.

First Setup (SET) as in my previous post above.

TEST: OCCT 3.0.0 / Custom/ Large Data Set/Duration 1 Hour.

RESULT: 1 minute into test Donald Duck Quacking at me in fits of laughter and Fail Error Displayed

My Conclusion: As the error occured at the virtual outset of the test it looked like insufficient voltage when serious computing began so it means my turbo did not keep up at initiation; probably needs more backgroud support to overcome vdroop. Need to correct CPU LLC setting.

Second SET RAISE CPU LLC from Level 5 to level 4.

TEST: OCCT 3.0.0 / Custom/ Large Data Set/Duration 1 Hour.

READINGS:

Vcore: Min 1.280v; Avge 1.288; Max 1.336

MaxTempsC: #0Core 79, #1Core 84, #2Core 84, #3Core 82. (This is well above what LFS advised me way back when).

RESULT: 31 minutes into test Donald Duck Quacking at me in fits of laughter and "ERROR in CORE #2" Displayed.

Conclusion: Well it is running hot, but I will up the CPU LLC again.

Third SET RAISE CPU LLC From Level 4 to Level 3

TEST: OCCT 3.0.0 / Custom/ Large Data Set/Duration 1 Hour.

4 minutes into Test I noticed Core Temps of 81,87, 88, and 86 with a vCore of 1.368v

Test Stopped manually.

Conclusion: TOO much voltage getting in here and the Temps looked like they are heading for the Tj so this is no good. need to drop LLC and Make the volts up somewhere else.

Fourth SET: Lower CPU LLC from Level 3 to Level 4, Raise; Additional turbo Voltage from +0.004 to + 0.008v

TEST: OCCT 3.0.0 / Custom/ Large Data Set/Duration 1 Hour.

Vcore: Min 1.272v; Avge Range 1.312v-1.344v, Max 1.352v

MaxTempsC: #0Core 62, #1Core 65, #2Core 63, #3Core 65. (I couldn't believe it myself, but it honestly knocked the heat back that much.

Result: Only 17 minutes remaining and the Duck took off on me again Message "Error Found Core #0"

Conclusion: Those temps look really tasty, previous CPU LLC increase was a bust for heat so a little more on Additional Turbo Voltage Limit should do it.

Fifth SET: Raise Additional Turbo Volt Limit from 0.008 to + 0.012.

TEST: OCCT 3.0.0 / Custom/ Large Data Set/Duration 1 Hour.

Vcore: Min 1.328v; Avge Range 1.336v-1.344v, Max 1.360v

MaxTempsC: #0Core 79, #1Core 83, #2Core 84, #3Core 82. (Back into the Too high zone inrelation to Tj and LFS advice).

RESULT: "TEST COMPLETED WITHOUT ERRORS" So B****r the Duck!!!

FINAL OBSERVATIONS:

In the succesfull test I was able to watch right through and noted that at 17 minutes to go OCCT started testing for a figure like 13407.00% of Computer, and then at successive intervals for 13406.00%, and 10020.00% etc. I note that it was at 17 minutes my set up [Fourth SET], with the attractive Coretemp figures, threw out an error code. I am Tempted to revert to that setting and run a test in Prime. OCCT was testing at an incomprehensible figure at that stage (To myself as well as my Com).

Alternatively I have a figure which has passed OCCT that I should be able to run 47x when my cooling solution arrives.

A further alternative is to search for another magical Combo to get the clock speed without the heat. I think my Fourth Set up achieved this by just staying ahead of the point where CPU LLC actually kicked in so I was surfing a sweet spot until the 13407% stuff from OCCT.

Any one who has any advice on this situation please give me some feedback.

Reards
Lex

Footnote: I need to remember that when you start believing in magic you are going to go wrong 99% of the time - I fell for the temptation knocked the ATBLimiter back to 0.008 and ran prime..no problems but those magical heat figures not there any more.. I was getting 79, 82, 82, 79 this time.. so I guess I misread My earlier Temps.. must have been looking at an average column. So no cheating available to me, got to think of another plan or be resigned to waiting for the Water Solution to have a 24/7 47x Rig.

@4Ryan6... looks like I hit the wall this time... the thermal efficiency wall.lol.
Lex

 

[4Ryan6]

^ Is all this with the new cooler?

I use prime95

 

[Lex_Sinistre]

^ Is all this with the new cooler?

I use prime95

No Still on Air until 23 July.

I will use prime when I get the water.. because at those temps I do not want prime running when I am asleep....

Currently I am running it at 45x 24/7 with Max temps of 74,75,74,73, and Avge temps of 68, 69, 68, 68.5 (Tested during a 4 hour no error run on Prime/ Large Data set/ rounding number check. Vcore max 1.302, Vcore Avge 1.288

I actually found an interesting thing when setting this speed up. (interesting to me because I am a complete noob and still learning)

The ONE CONSTANT is that these chips all idle at 1600mHZ. So do all the normal settings (Disable and enable as above), Set PCU LLC to level 4,Dial up your desired Multiplier (45x in this case), set ADD Turbo volt limit to 0.004,.... then lower offset volt value till it BSODS. WAIT till com clears and returns to UEFI. then take offset up a level until it boots into windows, then open up Internet explorer, navigate to Toms Hardware.. or a site where it is Reading material, (Law Reports) use scroll button to see if page hangs.. if it does then you will need to go up one more offset step.

When it is booting into windows and Browser is not hanging then you have the lowest base ( Lowest voltage to achieve stability at idle) to start from for your OC. On my machine I do not need to add more than the minimum .004 Additional TurboVolt Limit on a -0.060 offset, to get the 45x on air as just described

I am recording all the data as you requested and will submit my POST ANtec Kuhler Install figures here in a comparison chart.

Regards
Lex

 

[4Ryan6]

^ Interesting Discoveries!

 

[Kralnor]

Yes, thanks for sharing your discoveries.

I have nearly the same CPU/mobo, just a Z68 Extreme 4 instead, but the UEFI settings are identical to those that you described.

I wasn't aware that LLC could have a noticeable impact on temperatures (4C difference between LLC level 3 and LLC level 4) - that's good to keep in mind.
I just jacked LLC right up to level 1 straight away because auto LLC didn't seem stable (defaults to level 5). However, the core temps also went up considerably.

Please keep us posted.

 

[Lex_Sinistre]

Yes, thanks for sharing your discoveries.

I have nearly the same CPU/mobo, just a Z68 Extreme 4 instead, but the UEFI settings are identical to those that you described.

I wasn't aware that LLC could have a noticeable impact on temperatures (4C difference between LLC level 3 and LLC level 4) - that's good to keep in mind.
I just jacked LLC right up to level 1 straight away because auto LLC didn't seem stable (defaults to level 5). However, the core temps also went up considerably.

Please keep us posted.

Thank you for your kind reply. In respect of the CPU LLC settings I notice that what happens is that the LLC controller sets in various baseline default voltages so the higher you go (Level1 being the highest) the higher the supporting default voltage.

When you get to the stage that you are seeking the "smallest value for the maximum voltage" ("SMV") to get past prime 95, then having your CPU LLC set too high affects your "Average" voltage by running its support too close to the SMV, or even above it!!!. I have found consistently that a lower CPU LLC setting and a higher Additional Turbo Limit let my machine operate further below the SMV during the course of a run in Prime, only calling for the SMV on the hard parts. This results in a lower average Vcore voltage draw, and a corresponding reduction on Temps.

Regards
Lex

 

[Lex_Sinistre]

I am confirmed in my view of LLC manipulation by the following:

Voltage Regulation

Our voltage regulation testing will focus on the various voltages and the differences encountered between what is selected in the BIOS and what is measured by a digital multi-meter (DMM). Thanks to the ProbeIt feature we didn't have to go poking & prodding everywhere, since all the voltage read points are located in one convenient spot.

These measurements were taken at stock system speeds and with C1E, AMD Turbo Core, Spread Spectrum, disabled in the BIOS. Just to clarify, the Vcore (LLC) section is the Vcore results with Load-Line Calibration enabled, and Vcpu-nb (LLC) is the Vcpu-nb with Load-Line Calibration enabled as well. Here are our findings CHART AT SOURCE WEBSITE - SEE BELOW (Lex)




To be honest, we were quite surprised to see how well the ITE IT8721F controller performed in reading the various voltages on the motherboard. Even taking the margin of error of our multimeter into account, we can safely conclude that the software voltage monitoring tool of the ASUS Crosshair IV Formula is spot-on with our digital multimeter readings.

Diving further into these results, we can see that our CPU experiences significant amount of Vdroop when LLC is disabled. Generally speaking, Vdroop is a term that is used to describe a drop in CPU voltage at times of high CPU load. While some overclockers view this as a negative aspect which will lead to instability of the processor, others see it as a safety threshold to ensure that the CPU voltage never exceeds the voltage set in the BIOS. Nonetheless, ASUS has included Load-Line Calibration (LLC) as a function on the Crosshair IV Formula, which should technically help resolve this Vdroop “problem”.

Unfortunately, when looking at the voltages recorded when LLC is enabled we can see that the Crosshair IV Formula tends to overshoot the voltage set in BIOS. Technically, LLC should help stabilize CPU voltages to ensure that it remains constant at times as experienced in our review of the ASUS Maximus III Extreme. From this data, it looks like the LLC function simply increases CPU Voltage at times of load to maintain stability – a practice that is currently used by AMD and their Turbo Core technology.

This creates one big messy problem since unlike AMD’s Turbo Core technology which balances voltage and frequency to ensure that the CPU remains within its TDP specifications, ASUS’ LLC function will undoubtedly cause the CPU to create more heat through the voltage increase. Without adequate cooling, the increased temperature of the CPU will now pose as a source of instability and counter the “benefit” of the voltage increase. To summarize, the voltage overshoot caused by LLC increases core temperatures, which in turn could lead to increased instability.

Source:http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/33811-asus-crosshair-iv-formula-am3-890fx-motherboard-review-17.html

 

[4Ryan6]

One note is all motherboards handle LLC differently and you have an ASRock, plus LLC is a factor directly regarding the cooling solution, if your cooling solution can handle the heat then LLC is less a concern on the operating end of things.

I went through many cooling solutions for the 2500K trying to reach my goals until I finally took this route which has shockingly allowed pretty much anyway I desire to run the 2500K.

 

[Lex_Sinistre]

One note is all motherboards handle LLC differently and you have an ASRock, plus LLC is a factor directly regarding the cooling solution, if your cooling solution can handle the heat then LLC is less a concern on the operating end of things.

I went through many cooling solutions for the 2500K trying to reach my goals until I finally took this route which has shockingly allowed pretty much anyway I desire to run the 2500K.

Ok I promised to do it so Here goes my findings on the efficacy of the Antec Kuhler 920 compared with an "Aywun" aftermarket Air cooler.

The most I felt comfortable running on the air cooled system was a multiplier of 46. In both cases ran a 4 hour Test in Prime/Large Data Set/ Rounded Number Check. The Temps and Vcore figures are recorded as follows:

Aywun Air:

Temps C max 83, 88, 88, 85
Temps C Avge 77.2, 81.3, 80.6, 79.1
Vcore max 1.336 Vcore Avge 1.315

Antec Kuhler 920:

Temps C max 62, 65, 62, 62
Temps C Avge 54, 58.2, 55.6, 55.7
VcoreMax 1.352, Vcore Avge 1.304

This represents a big difference to me; between 20 -25degrees C cooler than before. It gives me confidence to tackle the 47x and perhaps beyond.

One thing I read a lot of about the Antec Kuhler when researching as to whether or not to buy it was the "Noise Factor". Well here is my experience. I agree that if you go to the extreme setting and wind these babies up they do Howl. I think that is the basis of the concern. I think it is only fair to note that at 46x my water temps during a 4 hour Prime run never got above 35 Degrees C and the fans ramped up a little, never ramped to max, and knocked the temperature back to about 33 degrees on average for what was for the most part a pretty silent run through the prime Test.

I also ran bout an hour of Crysis2, the machine handles it without the need to wake its fans up. In either prime or in Crysis when previusly cooling by Air, the old air fan started up near right away and was howling, so hte Antec Kuhler will not be loud in everyday applications, in my view compared to the Air method I was using previously this new System is Whispering Death to the Heat Gremlins.

Regards
Lex

 

[Lex_Sinistre]

Also to 4Ryan... your system looks Massive..by that I mean awesome.

regards
Lex.

 

[4Ryan6]

I'm presently running 8 gallons of water in the cooler, that much thermal mass allows a long time at controlled water temperatures, I have 4 3/4 filled gallon jugs of water frozen in the freezer at all times.

Normally I'll suspend one jug in the morning which allows below ambient room temperature all day, surfing or gaming, if I'm testing and need the water temperature lower, I'll suspend a jug for an hour then change it out to a fresh one, which will drop my water temperature much faster.

The 8 gallons of water not only takes a long time to warm up with the CPU trying to heat that much water, it also takes a while to drop the mass's total temperature, by just suspending a single jug, so I can use more than one jug of ice if I need to drop the total mass temperature faster.

Even when the water temperature reaches ambient room temperature, it still cools much better than the Radiator/Reservoir loop did, at least for myself it is the solution I was looking for regarding cooling the Sandy Bridge.

Additionally it's completely opened a new door to what you can get from a Sandy Bridge K series CPU, but with every great advantage comes side effects, that being having to be more conscious of condensation building.

So far I haven't had to run water temperatures on a regular basis that causes condensation, but I have been there!

Some don't like the idea because it is an open loop, but it's only open while accessing the cooler, it's completely sealed the rest of the time, it gives me the option to clean anything on top by using a aquarium skimming net (Bought just for that reason), and the Silver Kill Coils seem to be keeping anything from growing in the Distilled water, so far! :lol:

At least nothing has crawled out of it yet! :lol:

 

[Lex_Sinistre]

Best answer selected by lex_Sinistre.

 

[Lex_Sinistre]

Like the story 4ryan6... sounds like the begginning of a plot for a sci Fi horror Story, "The Monster from Sandy Bridge"..lol

 

[4Ryan6]

^ ROFL :lol:

With some of the crap on the SyFy channel lately, they should use it, at least it could be imaginable!

Thanks for the Best Answer!