3770k IHS Removals - CPU temp dropped from 79C to 71C

[dr_dx]

I wonder if Liquid Ultra is just about as good of an option for IB die as Liquid Pro is.
The reason being that its structure resembles paste and, thus, is more likely to work better in cases with uneven surfaces and poor contact, while Liquid Pro makes a very thin layer.
Or am I wrong in this assumption?

No. As far as an IB die is concerned, Pro will be better (if properly applied). The die surface and underside of the IHS are flat, and after removing the black adhesive, you will find that the IHS will rest on the die without touching the PCB - so you know there is no gap between the 2 flat surfaces. Just apply a very thin layer to each (use a very small drop, smoothed out to a mirror finish on both) and you will have the very thin layer that makes Pro so effective. Ultra, which has less than half the thermal conductivity of Pro, won't be as effective. Pro is pure liquid with no bulk. Ultra uses a graphite copper matrix (small particles of graphite and copper mixed together) to add bulk to fill in gaps. Pro works best on flat, even surfaces, Ultra best on uneven surfaces. Regarding safety, they both contain gallium and are harmful to Aluminum, so I don't consider Ultra any safer than Pro.

[dr_dx]

At the moment I don't have time for this, but in a while I will try lapping the inner side of IHS. The reason for this is this "stain" of residue (I guess everyone observe on their samples after the IHS removal) which might affect the overall thermal conductivity.

I wouldn't. Yes, I had the "stain" as well, but I cleaned the IHS underside with IX's cleaner (also 70+% alcohol works too) and removed any last bit of the stock TIM (and the stain was still there).

The underside of the IHS is flat, if you "lap" it to remove the stain, it won't be flat anymore and won't make as good a contact with the "flat" die surface. I think the stain is a result of a chemical reaction between the nickel and the stock TIM, not because there is any TIM residue.

[owikh84]

3770K @ 5.0GHz HT 1.44v

IHS off + Arqtic MX-4:


IHS off + Coollaboratory Liquid Ultra:


35C ambient Malaysia

[mav2000]

Make sure and "wet" both/all surfaces.

Thanks did that the second time around and even though the best core didnt move much in temps (as compared to wetting only the die and not the underside of the IHS) the worst core did improve in performance. I think I may have put a bit more than I required, but with 12 degrees in avg temp drop, I guess I am happy with the way it turned out.

[Zeus]

owikh84, are running without the IHS or did you place the spreader back on? 5GHz LinX is stable is what i'm after but i think 1.44V for daily use is a bit much. Pretty both my 3770K chips are gonna need that too for stable operation.

[owikh84]

owikh84, are running without the IHS or did you place the spreader back on? 5GHz LinX is stable is what i'm after but i think 1.44V for daily use is a bit much. Pretty both my 3770K chips are gonna need that too for stable operation.

It's without the IHS I put the TIM directly on the bare naked die.
This 3770K batch L220B196 is my best chip I've owned many pcs before all of them needs 1.5v+ for 5GHz.
My water cooling gears are HF + D5 EK Top + EK X2 250 Res | EK FC7970 | 2x RX360 + 12x GT
Ambient is hot here 35C sigh

[Zeus]

Thanks for the info man. I've been trying to find what it takes to get 5G 4c 8t on my chip but even 1.44V isn't cutting it for Cinebench 11.5 Cinebench 11.5 is a very fast method to determine stability and it's an absolute sucker for volts. I can run prime for an hour and even have CB 11.5 fail on me.

That is a pretty hefty watercooling setup you have there, i hope you don't have 12x GT 1850 rpm's at full speed running? If you have it will be noisy i guess. I am running a feser 480 with 4x GT 1450, the case is on top of my desk so the fans are pretty close to my ears and even 4x 1450's at full throttle are far from quite.

I have A 3770k 3218C176 Costa Rica here that runs awefully hot, i am still undecided wheter i will delid it or not. It seems a decent clocker but not sure if it is a 5G daily chip. I have to say that i did not see a lot of improvement in terms of overclocking headroom with the chip i'm running right now but the other chip might be different.

[owikh84]

Thanks for the info man. I've been trying to find what it takes to get 5G 4c 8t on my chip but even 1.44V isn't cutting it for Cinebench 11.5 Cinebench 11.5 is a very fast method to determine stability and it's an absolute sucker for volts. I can run prime for an hour and even have CB 11.5 fail on me.

That is a pretty hefty watercooling setup you have there, i hope you don't have 12x GT 1850 rpm's at full speed running? If you have it will be noisy i guess. I am running a feser 480 with 4x GT 1450, the case is on top of my desk so the fans are pretty close to my ears and even 4x 1450's at full throttle are far from quite.

I have A 3770k 3218C176 Costa Rica here that runs awefully hot, i am still undecided wheter i will delid it or not. It seems a decent clocker but not sure if it is a 5G daily chip. I have to say that i did not see a lot of improvement in terms of overclocking headroom with the chip i'm running right now but the other chip might be different.

A proper way to run prime95 is custom blend, min FFT 128k, full available mem detected in Task Manager. Not the small FFT s of course.

My GT 1850s are running at full speed but I'm used to the noise because I got another rig blasted with 16x GTs haha...

[Zeus]

Nice rig man. Why did you remove the pic of the benchtable? I liked that one too.

Do you prefer Prime over LinX? I think LinX makes the cpu run even hotter but maybe i'm still using an outdated version of Prime?

[owikh84]

Nice rig man. Why did you remove the pic of the benchtable? I liked that one too.

Do you prefer Prime over LinX? I think LinX makes the cpu run even hotter but maybe i'm still using an outdated version of Prime?

this pic is outdated because it was taken on a Arqtik quad TEC. The TEC is broken & it's under RMA right now so using back my trusty HF.
So here's the pic again


I prefer LinX as it detects errors faster & 1 hour is enough for 20 loops. For prime95 you need to use the latest version 27.7 build2 (of course on win7 SP1) then stress it for at least 3 hours. Anyway it's not like i'm gonna use my IB for 24/7 lol it's just for benchmark purpose so doesn't really need to run stress test anyways..

[OC Nub]

@owikh84 You do some of the cleanest, and best looking builds I have seen on any forums.

[mastabog]

I see some of you ran the cpu without the IHS. What heatsink/cooler did you use? I imagine not all all of them work because some assume there is an IHS 2-3mm tall and hence will never touch the bare cpu die. Am I wrong?

I just orderd the Noctua NH-C14. Will it work without the IHS? I noticed it has mounting brackets that seem to act as spacers ...

Cheers

p.s. I replaced the stock tim on my 3770K and even though temps dropped with 12C, I think I have one of the crappiest CPUs out there. On stock speeds, my load temps are about 65C (they used to be 75-76C). With the TIM replaced, I hit 90C quickly on 4.7 GHz @ 1.30v. I was so hopeful that after replacing the TIM I would be able to get at least 4.8GHz ... I have a Thermalright Venomous X and hope the NH-C14 will do a much better job.

[OC Nub]

^ You would need to remove the retention bracket in order to run it without the IHS.

[dr_dx]

Dr_Dx | Core i7 3770K [4C/8T] @ 5100.0MHz | 1.328v | ASUS Maximus V Formula| Water | L221B009 | LinX 0.6.4 AVX w/ Linpack 10.3.11.019 | Liquid Pro






5.2Ghz @ 1.4v LinX stable





5.3Ghz @ 1.488v LinX stable




32M SUper Pi @ 5.5Ghz

[M3kk]

Did this on mine too.

Setup:

CPU: 3770K @ 4500Mhz - 1.224v (1.2v DMM)
MB: Asus Maximus V Gene
Cooler: Antec Kuhler 920

Results:

IHS on with STOCK paste + Artic Ceramique 2



IHS on with Arctic Ceramique 2 + Arctic Ceramique 2



IHS on with Coollaboratory Liquid Pro + Arctic Ceramique 2



STOCK:
IDLE: 28 25 19 26
LOAD: 70 73 74 72

Ceramique 2:
IDLE: 24 23 16 21
LOAD: 62 66 66 64

Coollaboratory Liquid Pro + Ceramique 2
IDLE: 23 22 14 21
LOAD: 53 56 58 57.

[MacClipper]

STOCK:
IDLE: 28 25 19 26
LOAD: 70 73 74 72
Peak GFlops - 71.58


Ceramique 2:
IDLE: 24 23 16 21
LOAD: 62 66 66 64
Peak GFlops - 76.38

Coollaboratory Liquid Pro + Ceramique 2
IDLE: 23 22 14 21
LOAD: 53 56 58 57
Peak GFlops - 86.00

The difference is a ~20% jump in Peak GFlop output from 71.58 to 86.00 just by reducing the already lowish core temps (well below stated throttle triggers)... incredible huh?!

[M3kk]

Dont know why are that differences .. Dont think its about the throttling .. At 76 and 86 i had rams at 2000, at 71 dont know exactly, 1600 or 1800 or even 2000 too .

[dr_dx]

Dont know why are that differences .. Dont think its about the throttling .. At 76 and 86 i had rams at 2000, at 71 dont know exactly, 1600 or 1800 or even 2000 too .

While valuable for temp comparison, I don't see where you are getting 4.5Ghz - looks more like 1.6ish...

[cab]

Why do you guys insist on repairing defective CPUs? Why not use the forum to launch a class action suit? Kick some butt for a change..

[dr_dx]

Why do you guys insist on repairing defective CPUs? Why not use the forum to launch a class action suit? Kick some butt for a change..

While we all would love to get an IB using Intel's fluxless solder for the TIM, it would be very difficult to prove that the IB doesn't perform within it's designed parameters.

If you think about it, it's genius on Intel's part. They are willing to sell overclocking insurance to any sucker that wants it, knowing full well the IB can't be damaged by overclocking. In stock form, TJ will be reached long before any lethal voltages can be applied for any amount of time and if modified, the warranty is null and void, so it's a win/win for Intel. Brilliant!

[M3kk]

While valuable for temp comparison, I don't see where you are getting 4.5Ghz - looks more like 1.6ish...

LOL? Its with EIST and C1 activated, and leaved at idle 10-15 min after passing the test..

[dr_dx]

LOL? Its with EIST and C1 activated, and leaved at idle 10-15 min after passing the test..

Then that's why you have different GFlops... temperature alone won't affect that.

I did a before and after with only a TIM change. The after was, on average, -26C cooler, but the GFlops remained the same. All throttleing had been disabled on both runs:


Before w/ Stock TIM






after w/ Liquid Pro

[ChaosAD]

OMG, 5ghz @ 1.3v @ 51c under linx. Thats crazy. Is this water? I need this :-)

Sent from my GT-I9000

[MaxT]

I am now a very firm believer in Prime95 27.7 v2 (AVX) stability testing. I found I could pass AVX Linx or IBT for 30 max memory runs, pass superpi 32m and pass cinebench rendering, but Prime95 would fail within 50 minutes, especially when running 8 threads on 3770k. Had to bump vcore a few times to make it truely stable. So I no longer consider any screenshot valid unless it has a 16+ hour latest prime95 stable run in it.

[OC Nub]

I agree, P95 custom blend has always been a much better way to test stability. But I wont be doing anymore 16hr runs, too lazy for that now.