Do YOU have a good Haswell i7 4770k? (How to tell)

[CTBlankenship 36]

I'm doing a bunch of research on this overclocking art form and I ran into this ... apparently it comes from the ASUS OC Gods to their OC munchkins.

 

In the UEFI:

 

1.  Set your RAM speed ... (DDR3 2400)

2.  Set your RAM voltage ... (1.65V)

3. Manually set Vcore to 1.2

4. Set the multiplier on all cores to 46x

5. Every other setting at their defaults

 

If the system boots past the UEFI and either begins to load or ideally, makes it into the OS and is stable, you have a 50th percentile or greater chip on the Haswell overclocking-ability bell curve.  If it won't at least boot there and make it into the UEFI, you probably have less than a 50th percentile chip.  You can expect chips in the lower 50th percentile to top out in the 4.4-4.5 GHz range at 1.25V

 

If your chip will boot at 4.6 GHz and 1.25 V, that’s very good. It means you have at least an average chip.

 

If it will boot at 4.6 GHz and is stable there, then you may have an above average chip.

 

The best chips will be able to do 4.8 GHz stable at 1.25 V. Our sample did 4.8 GHz, but at 1.3 V and on a custom water loop. Using 1.3 V will likely put a chip out of the air cooling / AIO (All In One ... like the H100i) water cooling thermal envelope. Temperatures in all of these scenarios, from the dog 4.3 GHz chips up to the good 4.8 GHz chips, will always be in the ~90°C range.

 

That’s just the nature of Haswell. With the VRM on-die, think of Haswell as Ivy Bridge plus 10° C.

 

 

Shamelessly reposted

[JimmiG 244]

Mine boots at 4.3 GHz / 1.2V but is not entirely stable, it needs 1.218V to consistently pass stress tests for 24 hours. For 4.4 GHz it needs 1.24V and for 4.5 GHz at least 1.32+V.

[HLJames 2,568]

Haswell FIVR  (manual above 4.6GHz) stable FSX:

4.3 - 1.168

4.4 - 1.264

4.5 - 1.296

4.6 - 1.312

4.7 - 1.348

4.8 - 1.392

4.9 - 1.44

5.0 - 1.49

 

HLJAMES

[superglide17 11]

Do the good chips all come from a particular Lot number? Is there another way to tell if you are buying a top performing chip?

[CTBlankenship 36]

Superglide:

 

Sorry dude ... from what I understand ... it is a lottery ... you play ... either you win or you lose ... unless you want to buy another chip ... then you get to play again ^_^ ... $340 is an expensive ticket though isn't it?

 

HLJames:

 

Are those guidelines or are they what you have experienced with your rig?

 

Well, I'm getting ready to dive into the deeper darker side of over-clocking ... in these next few hours I'll be delidding and lapping the IHS (if I'm going to waste a warranty why not go the extra mile).  Instead of using a dry sandpaper I'll be using 3 and 1 oil ... just like sharpening a knife back in the holler.  I will take pictures for show and tell. 

 

After about a week of contemplation and research I've decided to go with the more lengthy razor blade method instead of the vice method.  As a former Electronics Technician all kinds of warning flags went up when I considered sending multiple shockwaves through a piece of PCB that is 160 square millimeters containing 1.4 billion transistors which are spaced 22nm apart. 

 

For comparison I manually set Vcore at 1.25 and the multiplier at 42 (not aggressive by any means) ... I ran Prime 95 for two hours and the temps were as follows (these are MAX temps ... the average is about 4 to 5 degrees lower):

 

Core #0:  84C

Core #1:  80C

Core #2:  77C

Core #3:  74C

Package: 84C

 

The cores are roughly at 33C with no CPU load and just hanging out at 4.2GHz.

 

Doing the same thing with the CyberPowerPC I bought, and quickly returned, sent temps into the lower 90s after 15 minutes of Prime95.

[AirborneZed 168]

I will try this right now on my new haswell build

[HLJames 2,568]

 

 

Are those guidelines or are they what you have experienced with your rig?

Yeah thats my experience with a retail chip (not delidded).

The water cooling industry understands what you are trying to do. They have already got the components you need to take Haswell / Ivy Bridge to 5.0GHz.

 

HLJAMES

[CTBlankenship 36]

OK here is a reality check for all of you lurkers out there ...

 

Assume you have a decent chip like mine and you are running 4.2GHz stable at 1.25V and temps ... even under load (and MSFX is not a load per se) are perfectly acceptable ... cool even ... and

 

1. You fire up MSFX, load the PMDG 737NGX at Imagine Sims KATL running Orbx weather and FTX Global and you are running 18 fps at the gate (the gate is the worst ... fps go way up on the way to the runway)... and that's just not good enough.  Oh yea, loading that plane up anywhere else (without airport scenery and the fps is pegged at 25 ... my .cfg setttiing is UPPER_FRAMERATE_LIMIT=25) ... I just did a test ... bumped the system to 4.4GHz and dropped back into KATL ... for 200MHz the frames per second increased by 2 ... TWO ... still wanna do this?

 

2. You are OK with throwing away a perfectly good 3 year parts and labor warranty (I unfortunately found out too late that if you prep the IHS to the point that you remove the lettering ... Intel refuses your warranty ... so I personally am not losing anything as I did this with CLP's abrasive prep pad ... documented in a previous thread.)

 

3. You know that lapping only gives 3 to 4 degrees of increased thermal dissipation capacity ... if that) ... but the sandpaper (600, 800, 1,000,1,500 and 2,000 grit costs around $35 to obtain at Advance Auto Parts)

 

4. You understand that delidding (no matter how you do it) carries a risk of throwing $340 away.

 

5.  You have two to three hours to burn (it is raining and ugly outside) 

 

AND

 

6.  You have the guts to push that power button after loading your newly prepped CPU into your computer and seeing NOTHING happen ... all the while knowing that the next chip you buy ... may be a CLUNKER ... all to squeeze another 400 to 500 MHz out of your system which, gives you at most another ... what ... I'm guessing here ... 1 to 2 fps per 100MHz ... to an already well performing simulator!

 

then ... for people like you and me ... the term is GAME ON!

[wanabflyer 53]

Kinda disappointing to me, reading threads like this one (and I have read a few on the HWL), here I am with what is now considered an old processor (2600k) and I really don't read anything yet, that would encourage be to go out and spend money on upgrading my CPU, further requiring me to she'll out for a new MOBO.

 

Seems to me that Intel made quite a huge advance in this technology when it announced the SB, but since then, nothing really ground breaking.

 

Speaking from a relative layman to some PC gurus.

 

Regards,

[CTBlankenship 36]

The deed is done ... CPU delidded and CLPd and lapped ... unfortunately ran out of the CLP TIM ... mounted the Corsair cold block on the lapped IHS anyway just to get on here and give an update, and the temps are running around 36C without any TIM ... don't get me wrong, I'm not going to run it until Wednesday when I get more CLP in but the results are encouraging.

 

Hi Rick:

 

If you have a rig that is allowing you to fly FSX in a way that is acceptable to you then upgrading your rig is really not required.  For me, it was time (three years) to build another rig for the family ... the last one is still going strong but it had to replace an 8 year old laptop that the wife was no longer pleased with using.  So, I wanted to put together something that would last another three years.  And, here it is.

 

The cool thing about this is that the lapping and the delidding are completed and I can now follow NickN's progressing articles about overclocking. 

 

Step by step this thing is coming together ... it has been a LOT of work though.

 

I have pictures ... will post them on Wednesday when the TIM arrives and I can complete this build.

 

36C with no TIM ... ha.

 

Just to let you know, mistakes were made that took some work to overcome.

 

1)  CLP is conductive ... I "always" packed the case with rags when working with CLP anywhere near the MoBo ... except this once.  My had came up, hit the case, syringe flew up and out of my hand ... (you know ... this was the kind of situation where time slows down to a crawl ... every rotation of the syringe was visible ... tumbling end over end) ... and landed smack dab in the middle of the motherboard.  Thank God the lucky stars were with me and it didn't land in the CPU socket or that would have been all she wrote.

 

2) I still TOTALLY recommend delidding using the razor blade method.  It takes a LOT of force to initially get the edge of the blade into the corners of the CPU ... but once there it is a simple effort to start working the blade slowly around the IHS.  Start with all of the corners first ... then progressively work the flat edges. 

 

WARNING:  When looking down on the CPU you'll see a row of rectangular solders ... the resistors of the CPU are directly down from those.  DO NOT let that blade go underneath that IHS any further than a 1/4 of an inch at that particular location or you risk damaging one of them.  If you do that ... you are done.

 

Even with all of the incompetence the PC is up and running.

 

Talk to you Wednesday.

 

C. T.

[CTBlankenship 36]

The process began with removing the cold block for the H100i and examining what it looks like when CLP is applied to the top of an i7-4770k IHS.  Notice that it spread nicely over the entire IHS.

 

Here is the i7 delided and ready to be cleaned with 91% isopropyl alcohol.  The best way I found to safely remove the resin from the PCB is to use the edge of a credit card and scratch away ... plastic on plastic causes no heartbreaks.  Notice that the razor I chose is not one that fits inside of a safety knife ... those are much thicker than the ones illustrated here ... in this situation thin is in.

 

Here and here is the i7 PCB after a complete cleaning ... notice my almost ultra boo-boo ... since those rows of metal circles where right there I lifted the blade ever so little so I would scratch them.  This was the worst possible location on the entire chip to do that ... keep that blade flat here and DO NOT let that blade penetrate more than 1/4 of an inch ... use a Sharpie to draw a line 1/4" back from the edge of the blade so you have visual clues as to how far to penetrate under the IHS.

 

Before cleaning the underside of the IHS scribe a rectangle inside the old TIM outline that is about 1/8th of an inch in from the old TIM outline.  This will come in handy later when you apply the TIM to both the CPU heat spreader and the IHS that sits down on top of it.

 

Here is the underside of the IHS with the scribing, all resin removed (as good as I could do) and scrubbed clean with the abrasive pad provided with the CLP.

 

And the lapping begins ... but why lap???

 

This shows how lapping supposedly improves the IHS thermal dissipation properties.  Notice that the top of the IHS was not flat ... as a matter of fact there was a depression exactly where you would not want one ... directly over the CPUs thermal plate.  This picture was taken after 240 passes with 600 grit sandpaper ... with oil as a lubricant.

 

IHS lapping before 800 grit sandpaper ... 120 iterations with oil ... up and back ... counts as one ... do it 30 times on each side, rotate, repeat ... until all four sides have been completed.

 

IHS lapping before 1000 grit sandpaper ... 120 iterations with oil ... notice the Band-Aid ... you might want to put one of these on your thumb before you pass it over sandpaper a thousand times.  In all truthfulness, you could probably stop here and save $10 in sandpaper ... but nothing exceeds like excess ... onwards.

 

IHS lapping before 1500 grit sandpaper ... 120 iterations with oil in a figure 8 pattern ... one figure 8 is an iteration.

 

IHS lapping before 2000 grit sandpaper ... 240 iterations ... dry ... in a figure 8 pattern.

 

IHS lapping completed perfectly flat and shines like a mirror.  Notice that there are indications that I did not keep the IHS perfectly flat ... there are rounded corners.  You can do better than this ... I was aware of this risk when I was doing the lapping process ... but it still happened ... concentrate hard with each iteration ... you can do it better.  Now is this really cause for concern?  Nope ... virtually NONE of the thermal dissipation occurs this far out on the IHS.

 

Here is the IHS ready to be mounted on top of the CPU PCB.  Notice that the thermal plate on the PCB is not centered.  Also notice the rectangle of CLP on the IHS it is exactly where is should be and is about an 1/8th of an inch smaller in size on each side ... flip the IHS on top of the PCB and you are done.

 

Cautions: 

 

1) Use electrical tape to cover the electrical components to the left of the CPUs thermal pad ... actually I used electrical tape to mask off the entire area so as to prevent the electrically conductive CLP from getting places where it shouldn't.

 

2) After mounting the IHS don't let it shift around ... ESPECIALLY from right to left ... get CLP on those electrical components and the game is over.

 

3) When you mount the CPU/IHS composite into the motherboard (remember there is no longer any resin holding the IHS in place) and lower the pressure plate, it is going to want to slide down, hold the IHS in position for as long as you can to prevent this ... and finally snap the lever into place ... breathe a sigh of relief (for now ... you still have to press the power button).

 

Use electrical tape to mask off the IHS (that is if you are using CLP on top) ... apply the CLP ... remove the tape and mount the cold block ... you are done.

 

Turn it on and smoke test it ... results on Tuesday afternoon.

 

At least MSFX still runs so I couldn't have screwed too many things up ... but it was not for lack of trying .

 

As a final note ... if I ever delid another i7 I probably will not even mess with the PCB on the critical corner and I wouldn't insert the razor blade anywhere over the rectangle of solder "squares"  I'd thoroughly unseat the IHS everywhere else and then use a twisting motion to break that critical corner away from the resin ... just my two cents. 

[CTBlankenship 36]

This is where I got the notion and the guidance to lap a CPU ... notice that I lapped the IHS after it was removed from the CPU ... in these videos the CPU and the IHS are still one unit ... I didn't think this was too wise.  I cleaned my CPU PCB and wrapped it up in the static free sheet that came underneath the motherboard and stored it safely away until it was time to remount the IHS.

 

CPU Lapping Guide (Part 1 of 2)

 

CPU Lapping Guide (Part 2 of 2)

 

Lapping is the easiest part of the IHS modification ... deliding and applying the TIM to the CPU's thermal plate and the underside of the IHS is what causes heart palpitations.  I was literally stressed to the MAX using CLP under the hood ... however, from what I've read, this is where the benefits of the CLP comes in handy ... in a previous thread and in other locations on the internet it has been expressed that CLP on top of the IHS only grants a few degrees of difference ... underneath the IHS is where the real benefits lie ... very scary though.

 

Enjoy ...

 

C. T.

[Tabs 2,790]

One of the stores should start a service testing these for overclockability. I'd pay a bit more to know for sure that I was buying one of the good ones.

[CTBlankenship 36]

Hi Ryan:

 

How much more is the question ... not sure that is a fiscally viable operation to undergo.

 

Glad to see you stopped by.  You had stated in a previous thread of mine that I would never get to 4.7 on a Haswell and you told another participant in the thread not to make such comments ... however, I'm sure glad he did and I'm glad you did ... I would have never delided and lapped if my goal were anything else but to hit 4.7GHz. 

 

Today I did a full test at 4.4GHz and the rig was stable through AIDA64, OCCT and Prime95 ... unfortunately my temps rose into the low 90s after about 20 minutes of Prime95.  That is when I made the decision to delid and lap.

 

So, I still might not make it ... depends on the CPU I guess ... but I sure am giving it the old college try! 

Thanks for the motivation!

 

C. T.

 

PS:  When, exactly, is the T7 coming out :wub: ?

[CTBlankenship 36]

Final thoughts for the night ... here is a good read from Coollaboratory about their Liquid Pro TIM.  Notice in BIG letters that warns not to use this product on any aluminum surface.  Since most aluminum surfaces are on heat sinks of low quality the chance of a reader of this thread running into one is minimal.

 

Also, that document gives an interesting read about the true purpose of TIM and why it has to be used on surfaces that have not been lapped.  Which is a pretty good argument for lapping.

 

Also, since there is a significantly tighter fit between my IHS and the H100i's cold block I decided to not use CLP but instead opted for ARCTIC MX-4 (4g) Carbon-Based Thermal Compound, Non-Electricity Conductive, Non-Capacitive ... as I don't want CLP squishing out from around the IHS and causing problems due to its conductivity. But, if what I read on the net is true ... one must still use TIM due to the fact that the H100i cold plate itself cannot be lapped ... but at this point I think I'm spitting hairs.

 

I guess this is why, when I mounted the cold block to the IHS, the temps remained relatively cool ... the connection was tight enough that the heat transfer was sufficient to prevent a overheat in a light use condition ... ha ... who knew.

 

Also, the pictures I took of the IHS after the 2000 grit sandpaper really didn't show the reflective properties of the process ... however, with the use of the MX-4 I don't think there will be a problem ... do you?

 

C. T.