Repasting with CooLaboratory Liquid Ultra, any tips before I start?

Well, let's find out together... team effort. You can do the tests with a fresh application of IC Diamond for us if you want to. I think it is going to be difficult to beat IC Diamond overall, because it is clearly superior to everything else I have used. But, I decided it was worth taking some time to investigate since I have not tried either of these pastes that I ordered. The kind of extreme overclocking that I am so fond of, every little bit helps... even if it is only 2°C improvement it could mean the difference between a new benchmark record or thermal shutdown. What will be most critical is durability. A paste that works great at first and rapidly degrades won't be worth buying. Durability is where IC Diamond has shined for me.

 

I started testing with your values:

1 hour XTU Stress testing:

 

32 & 1024M ThrottleStop 6.00 benchs:

 

this cant be right... Im in love with these temps

 

3DMark Vantage CPU Test:

http://i39.tinypic.com/e84llh.jpg

PS: I can't upload more here. So I used tinypic
PS: I have taken for granted that all this test must to be done with fans @ 100% right?

 

Yeah, I agree those are fantastic temps. Keep 'em coming Nikos... pretty sweet. The more corroborating results we can see, the better.

I received the Noctua NT-H1 in the mail last night, so I'll be able to start phase one of my paste comparison tomorrow or Saturday, depending on my work schedule.

 

I agree with this. Very good points here.

This "thermal compound" seems to really work pretty well, but durability is key. I'd hate to have to re-paste every month. ICD lasts at least 6 months (with some overclock), at least in my experience. In normal use on every day computers, I've seen ICD last 12 or more.

I'm glad you've decided to try this stuff out, Mr. Fox. I'll be happy to see your results as well, especially if you break new records with that M18x beast of yours. :thumbsup:

 

Tell me the test do you want, as soon as I can I will bring here the results.

 

I would like to point out that the max. power consumption recorded by Nikos during XTU and TS Bench were only 66W and 60W respectively, which would have led to sub-80 temps with pretty much any decent paste job. We need to see some results while doing Prime95 or Intel Burn Test that really ramp up the heat generation. At least show us some 75W results if not an all-out 90W run please

 

Ivy XM doesn't draw anywhere near that much power at a modest 4.3GHz overclock. Even at 4.8GHz the 3920XM barely hits 100W, unlike the 4930MX that easily draws 130W at 4.5GHz if you can keep it from shutting down. So much for Haswell being a more "efficient" product, LOL.

FYI - I am probably going to do similar testing on the Alienware 18 with Liquid Ultra to see what happens with Haswell as well (pun intended).

The same ones in these screen shots.

 

Sounds like a plan.
I did a re-paste only last Saturday. Do you think I should repaste again before the tests ? or just go for it as is?

Also whats everyone's BLCK set at ? I can't make out yours in your picture Bro Fox but it looks like 100.1

 

Here you have one more:

Intel Processor Diagnostic Tool:

 

4.4Ghz after 20min

View attachment 107263

 

3DMark 11 PT:

 

3940XM?
Liquid Ultra?
How long since you did the repaste?
100% stable without vcore bump?

 

100.3 in XTU. ThrottleStop will see it as slightly different, like 100.1 or 100.11 etc. If you click on the gear icon in the upper right corner you can enlarge the image full screen to see those values in XTU and ThrottleStop.

I would just go with your repaste from last Saturday for now. I expect new paste to work well and a week of normal use your ICD should be as good as new.

@[Nikos], those temps are super. It looks like you are getting some unwarranted power throttling that is going to hurt performance. Part of that problem is Windows 8, but you can improve it. Are you using ThrottleStop? If not, I would recommend it. Being crash-free without a minor vcore bump might not always deliver the ultimate results in a benchmark score.

Set your BIOS to match everything you see on these screens HERE and HERE and see if your 3DMark11 Physics score improves. It should be around 10.5K or more at 4.3GHz, even with Windows 8 holding you back a little.

@ht_addict - feel free to jump in with the same testing using the settings mentioned. The more, the merrier. BTW - I can't determine what your max temp was from the XTU screen. All I can see is a graph with wavy lines.

 

just beyond awesome ... having mental jawdrop since I started looking at this tread.... pure awesomness!

I honestly wont mind to repaste every 30days if its a must, but I cant imagine this being the case for some reason...

Running hard @ top GHz 24/7 breaking thermal limitations that we had before are totally worth 30mins effort every month or so if its the price to pay for the performance/stability potential (which like I said I doubt frequent repaste will be necessary)

Like one wise man once said - "performance and benchmark, higher score that matters regardless of conditions or efforts, if you can achieve it - you are on the top of leaderboard"

Shut up and take my money!

 

No way changed the values to match exact as yours but again 9700 points :/ don't know why.
Maybe the version of my 3DMark 1.0.2.0 and SystemInfo Version 4.17.0.0?
Right now I have 256 Pri Plane I'm going to raise it a little bit.

EDIT:

9939 Points after raising to 896 PriPlane.
I'm going to bump the vcore a little.

 

Oh, wow... 256 Pri Plane is no good. Set it to either 0 (which means "automatic") or 896. FYI - "automatic" is the same power value as setting 896 manually.

At 256 Pri Plane your CPU cannot hold the overclock. That shows in your minimum core speeds being so low. Your CPU is slowing down under load because of not having enough power. After changing this, please re-check your temps because they might be quite a bit higher than they were with 0 or 896 instead of 256. In XTU Pri Plane is Core Current Limit and 112.000A is what you will see setting Pri Plane to 0 (auto) or 896 in the BIOS. You may need more than 896 if you go much higher than 4.3GHz. You can set Pro Plane to 960 or 1040 and possibly not see any change in temps. Temps are affected more by voltage.

This is why I provided all the settings in the BIOS, XTU and ThrottleStop in post #47. If you are not using the same values it will not be a less accurate comparison of thermal paste effectiveness. If they are all over the board it will be more of a haphazard apples/oranges/bananas comparison.

 

Previous test was made with correct PriPlane, I will do them again anyway with this current BIOS config:

Sorry for the images :$

 

Here you have it again:

Intel Process Diagnostic Tool:


PS: Better temps haha this thing is funny

 

Here I come with more this time:

4.6Ghz!!!! My personal record!!! Yeah! I think ThrottleStop didn't get the temperatures right I'm going to run it again with HWInfo running too! (I closed it unfortunately before the test started).

 

download wprime95, ThrottleStop benchmark is not so consistent as wprime95

 

Again 4.6 with HWInfo:

I lowered the vcore into XTU a little bit and the time went up to 204 from 198
Also the temps where mostly 99ºC

 

99C @ 4.6GHz. That's pretty warm.

 

Was 4.6Ghz I'm playing with voltages and PriPlane with 4.5 I've got 91ºC you can see it going back to page (3-5).

 

please run wprime95 benchmarks

 

I will run it tomorrow morning if possible

 

PriPlane is only a maximum current setting so if you get improved temperatures from playing with it, you have just compromised your performance..

 

My bad I meant just voltage. PriPlane always between 896 and 960 as Mr.Fox said.

 

For 4.6GHz on nothing but laptop fans, the fact that it can pass a benchmark like that without a thermal shutdown is actually reasonable.

Here is another example of 4.6GHz that I just ran with the fresh Noctua NT-H1. It will be interesting to run the same test a week from today to see how the temps compare.



I am going to update the previous post with results from a fresh application of Noctua NT-H1. Results are quite good compared to 30 days of heavy overclocking with IC Diamond, but NT-H1 viscosity is thinner and creamier compared to IC Diamond. I am skeptical of how durable it might be in comparison IC Diamond.

Edit: link to previous post with NT-H1 results added

 

You're right. The temperatures you got are impressive for the same overclock. If this stuff is that good months from now, it may be worth switching over.

Is it just as easy to apply & remove as ICD?

 

Well, it was easy to get too much on. I was expecting it to be thicker, like ICD. When I first went to apply some it came out to fast, so I wiped it off and reapplied less. It wiped off really easy. Otherwise, no different. I used a thin line length of die as I do ICD.

ICD is this good at first, too... so, I am kind of excited but still kind of skeptical of durability. The consistency reminds me a lot of Tuniq TX-4 which was lousy stuff IMHO. It did not cool well and only lasted a few days.

I am going to push this NT-H1 stuff pretty hard with more extreme benching to see how it holds up while I am waiting for Liquid Ultra to arrive. After it has been brutally tortured a little more, I will post results for a repeat of the same series of tests before I repaste with Liquid Ultra. If it holds up well it will be a good option since it seems to be really easy to clean up.

 

CPU is a 3940XM(QS). Did a repaste about 3 weeks ago. 100% stable without vcore bump.

 

Tested with settings from page 5. Flex and Pri are a little to stingy for me at 4.3 though.



TS 32M have temps around 71 degrees - as did TS 1024 before it closed (1st time lack of Pri, 2nd time lack of flex)
Didnt go on with an XTU bench.

If I increase the flex it will ruin the control of the test...

 

No, bump it up a bit if you have to. Just keep it as low as you can for testing since more voltage makes more heat. 3DMark11 started crashing on me with 5 flex in Windows 8 and I had to bump it to 8 to make it finish on the last round of testing after repasting. At 4.3GHz, I normally run 15 to 18 flex because it usually returns higher benchmark scores for me than lower flex does. Pri Plane will not directly increase your temps. It will help the CPU not downclock under load. It only indirectly increases temps by the CPU running at a higher clock speed without dropping. Once you have enough to keep a steady clock speed under load you can stop because the rest is not providing any benefit. The CPU only takes what it needs from Pri Plane.

 

Ok I'll juice it up some as soon as I can test next.
I can't help but feel we should have gone for a higher flex across the board for a better comparison though.
Regardless this Liquid Metal is looking pretty good so far. From what I can see it's getting similar temps at 4.6 to what I was getting at 4.5 with standard fan profiles. I have a feeling that this material might have good durability too.

 

Well, we can certainly do some higher-powered testing. In fact, I plan to do some if for no reason other than to torment this NT-H1 to see how much it can tolerate. As a control point, I was shooting for something that more people could relate to in terms of clock speed and might have shot too low on the overclock settings. I was also thinking about the fact that not everyone lives in a meat locker like I do, so ambient temps might present a problem running higher than 4.3GHz for some folks. For people living in really hot places without a good air conditioned work environment, 4.3GHz might even be pushing it a little bit. And, for those with Haswell CPUs, achieving good results even with cool ambient temps is a serious challenge. Most are struggling with crazy temps at a humble 4.3GHz on Haswell.

 

I think the OC is a good number for testing, I just meant use a higher flex but still at 4.3 on all the machines. Mainly because some machines manage to complete the benchmarks at 4.3 on 5 flex, others can't, so to complete a bench the results are somewhat misleading and not on an even keel if a user has to dial in more flex (heat ).

 

Yep, whatever your machine is most stable at is fine. At 43x it is a little iffy at 5 flex for me... sometimes it does, sometimes it does not, but 15-18 is crash-free at 4.3GHz.

Here is some 4.6 and 4.8GHz overclocking torture for the Noctua to see how long it can tolerate temps. It's not too bad, but seems like it might already be starting to lose a little bit of its effectiveness.

Spoiler

Note: Vantage CPU Test 1 & 2 will not complete at 4.6GHz as the rest of these tests do. It is too intense for the fans to manage... hits 105°C and has a thermal shutdown using nothing but fans.
View attachment 107296 View attachment 107292

 

AnandTech Forums - View Single Post - Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

Someone else already tested all the TIM mentioned in this thread, on a 3770K bare die.

NH-T1 is good but it doesn't last.

 

Liquid Ultra and ICD seem to hold the champion flag so far

 

That's kind what it's looking like. It is now only a little bit better than my 30-day old IC Diamond that has been rode hard for longer. It seems too thin and wipes off easily, so I am guessing "pump out" will be an issue with it on a heavily overclocked CPU. I bet it probably works great for video cards, though.

I had seen most of the other tests of thermal pastes. Some of them are questionable, some are pretty decent. It will still be fun to test them for ourselves, though. I have always stood by IC Diamond and some folks don't like it. So, there is a chance we might reach a different conclusion.

 

always stay open-minded, the only way to discover something new

 

Exactly right... always keep an open mind and an open heart. And, the only way to get better is to keep trying.

Temps on those 4.6 and 4.8GHz runs I uploaded a few posts back aren't all that bad. I'd say very close to IC Diamond with durability being suspect. It did great with 3DMark11, but Vantage was a bit much for stock fans at that clock speed.

 

Here I come with wPrime (v.2.10) 1024M test results @ 4.5Ghz:


Since here is winter (center of Spain) my house has heating turned on all day and night is when it's warmer. So now the temperature is lower in the test.

 

Here I come with wPrime (v.2.10) 1024M test results @ 4.3Ghz:

 

I think I can say that this is my best repasting ever. I think it's time to move on into GPU's benchmarks with differents modded vbios. I rolled back to win 8.1 with no secure boot and fast boot to being able to use OV vbios.
If anyone want me to run other benchmark or stress test please ask me, I will do it asap.
thank you all for your support I will keep posting in this thread with GPU's results and some days from now I will redo the CPU's test to know the durability of this paste.

 

Your temps here are fantastically low! They are lower here at 4.5 than your more recent 4.3 bench!
If your still get these kinds of temps 1 month from now I think I'll have no choice but to get some.

 

How long are we to run the ITU benchmark?