TR Forums

My final results: 4.9ghz @ 1.35v with the cache/uncore at 4.7ghz @ 1.35v, XMP enabled DDR2133.

I was able to get to 5.0ghz "stable" but it took 1.45v and the temps were in the low 90s, too high for me to do a full muti-hour run of OCCT, the overtemp triggered after 2 minutes, but I did game for a bit at this just for fun.
http://i.imgur.com/qFH4JeS.png

I also delided and it dropped my temps by 10c, I was later able to drop my voltage from the 1.375v pictured below to 1.350v once I got the vcache up to speed. I did not use liquid metal TIM, just some GELID X.
Stock intel paste 90/91/90/79 http://i.imgur.com/xwNvIQQ.png
GELID extreme paste 77/79/80/70 http://i.imgur.com/3Ig2upT.png

Thoughts: Improvement from 4770k, it took almost .1v less to run the same speeds, and where my 4770k hit a wall at 4.7ghz core and 4.2ghz cache, this one kept going until the voltage/temps were too high for me. The extra caps on the bottom of the chip and under the lid may be what is helping, the paste was still inferior.

Delid Pics:
Stock Gunk http://i.imgur.com/LPO7lC1.jpg
Stock Gunk http://i.imgur.com/TxIqgI6.jpg

The vise put some nicks in the copper so I went the old C2Q route and did some lapping, glad I did, it was not flat:
Here you can see the edge damage and the start of the sanding: http://i.imgur.com/C00oizd.jpg
Going http://i.imgur.com/tXlBCry.jpg
Going http://i.imgur.com/H5FJPea.jpg
Gone http://i.imgur.com/cu9nISP.jpg

All cleaned up:
Top http://i.imgur.com/8EsxsAf.jpg
Bottom http://i.imgur.com/rtWOuR8.jpg

Computer specs:
Corsair 800D
Asus Z97 Pro
Toshiba Q Series Pro 256gb
Asus GTX670
Corsair TX950
Dtek Fusion v1 Block
Swiftech MCR320-QP Quiet Power radiator 3x120mm w/Corsair Air Series SP120PWM High Performance Edition High Static Pressure Fans
DDCT-01s Pump Laing DDC-3.2
Tygon 3603 Tubing
http://i.imgur.com/56Bpdjs.jpg

OC notes:

Glad someone got some clocks out of one of these!


Though I don't think that my H80i would be up to the task- it's good for 4.8GHz on my 2500k (I run at 4.5GHz).

That seems nice. The 4790K seems well worth purchasing just for the stock speeds and not being feature cut. I'd be happy to get one and run it at stock speed.

flip-mode wrote:

That seems nice. The 4790K seems well worth purchasing just for the stock speeds and not being feature cut. I'd be happy to get one and run it at stock speed.

Agreed, during these summer months I may just run it at stock speeds to keep my office cooler.

flip-mode wrote:

That seems nice. The 4790K seems well worth purchasing just for the stock speeds and not being feature cut. I'd be happy to get one and run it at stock speed.

No kidding... especially when ASUS boards will even default to running it at 4.4Ghz with a minimal bump in voltages and still good for decent temps. My experience with a 4770K seems just silly by comparison.

Nice work and thanks for the log.

Thank you for taking the risk of removing the lid and doing the experiment for us. Very nice work!

Wow-- still that big of an improvement switching TIM's under the lid...

continuum wrote:

Wow-- still that big of an improvement switching TIM's under the lid...

That's cause it's not due to the TIM it's due to the better mating of the IHS and die after removing the excess adhesive (which has always been the case)

klepp0906 wrote:

continuum wrote:

Wow-- still that big of an improvement switching TIM's under the lid...

That's cause it's not due to the TIM it's due to the better mating of the IHS and die after removing the excess adhesive (which has always been the case)

Exactly. More pressure on the die by the IHS = better thermal transfer. The trade-off is that you risk cracking the die. The real risk is running without the IHS hough, even with the special accessories that some MB makers, EK etc supplies it is pretty easy to crack the die.

People have re-used the Intel TIM and seen the same improvements as with any other TIM.

klepp0906 wrote:

continuum wrote:

Wow-- still that big of an improvement switching TIM's under the lid...

That's cause it's not due to the TIM it's due to the better mating of the IHS and die after removing the excess adhesive (which has always been the case)

Too bad you did not use Coollaboratory's Liquid Ultra metal based TIM. I think you would have even better temps. just have to be careful since it conducts electricity and eats aluminum.

vargis14 wrote:

Too bad you did not use Coollaboratory's Liquid Ultra metal based TIM. I think you would have even better temps.

Didn't you read what klepp0906 wrote? The TIM makes very little difference, the main thing is that you get rid of the excess adhesive so you get better pressure/mating on the die. People have done the tests, expensive TIM, cheap TIM or reusing the Intel TIM, makes very little difference. You can simulate the stock situation with some paper or such in place of the adhesive on your own if you want. Same thing if you don't remove the excess adhesive, not much of an improvement then.

I would link the the thread over at Xtremesystems , but their forum seems to be down/broken.

Pholostan wrote:

klepp0906 wrote:

continuum wrote:

Wow-- still that big of an improvement switching TIM's under the lid...

That's cause it's not due to the TIM it's due to the better mating of the IHS and die after removing the excess adhesive (which has always been the case)

Exactly. More pressure on the die by the IHS = better thermal transfer. The trade-off is that you risk cracking the die. The real risk is running without the IHS hough, even with the special accessories that some MB makers, EK etc supplies it is pretty easy to crack the die.

People have re-used the Intel TIM and seen the same improvements as with any other TIM.

Yep. I had a 4770k bite the dust due to this. Purchased another after selling a display to fund it - looked for a "bright side" which I told myself was that my new chip would overclock better of course this was very much the opposite lol.

Either way, like the foolish enthusiasts we are, I proceeded to do the same thing to my 4790k (fortunately all went according to plan).

Fwiw results were pretty much the same as the OP. Which is the higher end on these chips. 4.7 is about average. 5.0 you need borderline extreme coolant or a willingness to run a pretty high voltage.

If I left uncore and memory stock I was able to hit 5.0 at 1.39v but it was only IBT stable. Prime and x264 locked up after some time. It was game stable as well which is my primary use but being ocd I couldn't live with it knowing that if I decided to encode a blu ray some day it may lock up lol.

That said, 4.9 took significantly less. Still I was looking for the best "all around" overclock with complete stability which ended up being 4.9/4.6/2133 @ 1.42v which is 24hr prime95/x264 stable.

Temps via IBT hit 85c and temps via aida64 FPU test hit 93.

What really surprised me was the degree that memory speed hindered overclocks at these clocks. The ram 1866 I've had up to 2300 on my 4770k and with this I can't even do 2200 and keep stability. I assume I could re stabilize by adding some voltage but I told myself this was it. I didn't want to throttle during avx/2 stressing and I've kept my goal.

Got my hyper PI 32m score under 10s finally so I'm satisfied.

So yea... Go red wings?