Clevo Overclocker's Lounge

That is definitely impressive, specially since the heat sink makes contact with the dual GPU heat sink.!

 

Well when it comes to TSBench, that's an advantage, really, especially since it's impossible for the GPU heatsink to screw up the alignment of the CPU heatsink. Plus heat dissipation through the thermal pad is probably pretty mediocre at best. I've also got CLU for both thermal interfaces, so that helps quite a lot.

 

I have to add Wprime 1024M stress test for processor. More like a real use of CPU.

[email protected] is impressive and you got a nice silicon I need 1.342V for same clockspeed.
What max temperature have you in Wprime 1024m and Cinbench R11.5? People complain about the 12v fans. But better with colder hardware than silent fans and more heat

 

I was curious if anyone has updated to the latest NVIDIA Graphics Driver v372.90 (Release: 9/21/2016)

If not, which driver are you on and sticking with?

Also, Fire Strike keeps saying that my graphics driver can't be validated? (This was before on an older driver.)

I have yet to test it now with the new driver mentioned above....just finished installing and restarted....I'll report back if this corrects the Fire Strike issue...

Update:
Installing the new driver fixed the Fire Strike issue. It now registers/valid.

 

Too lazy to install older Cinebench.
Think I need to lap my CPU and heatsink.

Spoiler

 

That really wont help much, since end of the day heat sink is shared and then there is the matter of the interface between the Vapor chamber heat sink and the CPU heat sink. Maybe stick FujiPoly Ultra thermal pad between the 2.

 

If 85°C is your max temp running wPrime 1024M I doubt you'll benefit much from lapping. There is nothing wrong with that temperature being the max after such an intensive benchmark. Almost anything else you do is going to be under that temperature. I'd call it good and just enjoy it. If it ain't broke, don't fix it.

 

It's not the max temp that I'm concerned about, it's that one of the cores is much lower than the others. Doesn't that usually indicate a warped heatsink?

 

Too funny.

 

I've had people do that to me quite a few times. I've even had a couple of submissions reported a while back because they did not "believe" my scores were possible and they "suspect" that I cheated somehow, LOL.

 

It's not people it's your team member. I said i would change it when i got back to it. Which was today. I just deleted it now.

I could have just had all his removed for being submitted wrong and not explain how to fix it. Which is what i did.

Wont happen again though

 

I think i have seen that name here on the forum. Is he just fooling around with you or something ?

 

Don't think so.
I would never report a TI team member. Just not my style!

 

He probably thought it's just like a PM or comment to correct the typo?!

 

Hi @johnksss,

I owe you an apology. You did say that you'll change the scores and I'm sorry. And yes, @Prema is right, I though it was just like a comment.

I hope we're good

Regards,
Mark

 

That may be a common misconception. One of the people that did that to me thought the same thing. I messaged him and asked why he did that and his response was he thought the message would come to me so I could fix it.

I have reported posts that are HWBOT database errors, but never submission errors. I always contact the submitter privately if they made a mistake or typo.

As an example, I reported one recently that the same exact benchmark submission was being counted twice. Same benchmark validation link, same screen shot and same score and same date, but for some reason HWBOT made it 1st and 2nd place, LOL. But, that was a database error, not a simple user error.

 

Ok guys....No where in here does it say send private message to user or pm. I have been doing this a very long time...


You get a link saying some one reported you, but it goes to the mods for review.

As far as it's concerned though, we are good. I just found that very odd That it would even go down like that after he sent me a message in PM on here about it a few days ago. And that report was filed 7 hours ago.

 

Oh, I know exactly what you're saying, and agree... and that message is why I don't do it when I find simple mistakes like transposed numbers and typos.

But, we both know many people don't slow down long enough read stuff. Happens all the time here, too. TL;DR seems to be status quo. Sometimes even I am guilty of that TL;DR thing.

 

All good though Brother Fox.

 

Yes, we're all good. But, it is good that you brought it out to everyone's attention. I'm glad you did.

The last one of mine that got reported was a 200W 980 Notebook benchmark that displaced a guy's ranking with a desktop powered by 6700K. He was upset and found where I accidentally did not select the "Notebook" version of 980.

 

I did that a few times with the 1080.

 

Nice!

Sent from my LG-H901 using Tapatalk

 

ALL RAM sticks must be exactly the same model to work their best?

I'm eyeing something something HyperX 2133 + CL14 16GB / Stick RAM, but also consider upgrading them later on, but want to not remove a keyboard from a P775 since it is a painful process.

This would be good to know for peolpe who don't want to remove keyboards and heatsinks since for P775 accessing the area under the KB is a bit more complicated operation than it is on P870..

 

My understanding was that they should be paired up (same mem) where banks are used in tandem. However, I've seen different RAM used in 2 or 3 single, different so-dimms. IIRC, they had a hecuva difficult time making sure timings were right and it required a lot of finagling to make it work.

Because of this, I think most ppl say forget the hassle and make sure mem matches. Take this for what it's worth , just one person's recollection. Also, remember, like you, I'm a software, guy.

 

Oh look, you've finally managed to draw your rated TDP of 94W on a 6700K.

At 4.8GHz.

With 1.3v

Why did Haswell exist again?

 

I write far too much in most posts to even begin TL;DR-ing =D

It can mean a bad paste job. When I was redoing my CLU application on my CPU a couple months ago, I kept getting one or two cores either way hot or way cold. I eventually redid it enough times and managed to get all the cores as cool as the lowest core temperature I found across about 4 or 5 applications.

 

I heard similar stories but wanted to confirm xD


Since this is an OC thread, how do we OC the RAM? Is it possible? How do we do it? What are the risks? How do we monitor speed and latency after OC? Any risks?

 

It's a big thread, but search for some posts by @Phoenix. IIRC, he had some questions about OC'ing RAM, and received some feedback on how to do it.

Also, I think due to silicon lottery and finickiness of RAM, he may have just decided to buy 3000 MHz modules instead

 

Very recommended to get a same-kit. Higher clocked RAM, lower latency RAM, and both in one, larger numbers of sticks, higher density of RAM, all contribute to compatibility. So the higher the RAM speed, the lower the latencies (especially outside of JEDEC), the more sticks used, and the larger the amount of memory in total all reduce the compatibility with other kits. 2133MHz 15-15-15-35 JEDEC DDR4 2 x 4GB you can use pretty much any stick with any stick and get to 4 x 4GB. 3000MHz 16-18-18-43? You probably don't want to buy 2 x 16GB now and 2 sticks later.

As for how to do it, you adjust the timings, voltage, etc in the BIOS. It's kind of like a table. The risks are stability; unstable overclocks can cause slower data transfers (ask @Mr. Fox about how some RAM he had running at higher speeds benched lower than when he reduced the speeds because it likely was experiencing some slight errors, even though it didn't crash him outright), lots of crashes of the OS, or even failure to boot. The speed and latency will be reported to programs that know how to read it, like HWiNFO64 or CPU-Z.

 

Yes, uneven temps across the cores 9.9 out of 10 times mean that there is not enough paste used causing it to have thin spots where the die or IHS is not making proper contact with the heat sink.

As I've mentioned earlier, @Q937 ...you'll have to be a little more generous with the paste or liquid metal application than normal...this doesn't mean a ton more, but if you're using liquid metal, make sure to tape around your gpu with electrical tape and put enough LM on there to where it is almost a slight shallow puddle. I'm on my phone right now, but here's a pic...hope you can make out the amount of LM...it's a nice mini shallow pool...

Also, after applying LM I usually take a hobby knife and trim the electrical tape, but it doesn't hurt to just leave it on. Up to you.



NOTE: Before anyone says anything ROFL, please let's prevent any comments of saying that there's way too much LM etc....trust me...this isn't my first time around the rodeo and I know what works and what doesn't for me at least. Putting too little of LM will not do much justice unless you have a perfect contact between the chips and HS. This isn't a desktop. The tolerance on notebooks is much less than desktop counterparts, which is why more paste is needed than normal, most of the time. I've been 10 for 10 so far so I'll go with those statistics.

 

overclocking RAM has never been a good experience for me. Yes it ran stable and passed all stress tests in my testing/benchmarking....but when it came to real world usage, I kept getting random errors while installing software, sudden crashes in explorer.exe, games like Need For Speed would dip to 3 FPS all of a sudden after a few mins of gameplay, I was lost...until I decided to revert back the overclock and everything worked perfectly.

 

The amount definitely helps if the heat sink is warped.

But please be very careful if you move the system around, no sudden bumps. I have had blobs of CLU slip around over the electrical tape.

Also from the thermal pads, seems like the mounting pressure good on the new systems.

 

I usually leave the electrical tape on as a security blanket, but I must say that I've had zero issues with leakage or pump out...knock on wood.

::iunlock::

 

After memory oc, I always run TM5 for 20 times.
Once I had error at 13rd run. I think you as a prime OCer should try again

 

The way you apply CLU is pretty much exactly the same way I do it on Clevos. It takes a little more to ensure good coverage even with good fitting heat sinks. I did not use as much on the Alienware M18xR1, M18xR2 or Alienware 18 because the heat sink fit was much better and they had higher contact pressure than what I have seen on any Clevo. I have also had zero issues with CLU moving like I did with Conductonaut and Phobya LM. The electrical tape is a good extra precaution. I have done the same thing on GPUs and it works great. I applied the electrical tape and left it there. Three years later it looked the same as the day I applied it.

I wish Pascal GPUs still used the support rim around the die. But, below is a photo of how I taped up the Dell 780M cards in the Alienware 18. The Clevo GPU already had a plastic shroud to cover the surface mounted components. If I remember correctly it was @Donald@HIDevolution that suggested using electrical tape during a phone conversation I had with him a few years ago.

Before | After ..
​

 

Yes Indeed. Electrical tape is a must and has been a common practice for me since day one even with traditional paste due to my major OCD. It's a free insurance policy IMO that should be had when using LM.

@Q937, here are some more pics showing the amount of LM that I usually apply to my CPU and GPU. Yea yea, this is from my old AW haha...but for the sake of showing the LM ... It's a nice shallow yet generous amount of pool...

I have posted a thread a while ago showing the taping and LM process on the old AW. Here are some pictures to compliment the method as shown in your pics as well.





For any future readers, it's important to use a good quality electrical tape (3M to be specific) and not some cheap dollar store bubble gum tape. I prefer and highly recommend the Super 33+ as it is thinner / stretchier than the 88, while having the same heat rating of 105C.



@Mr. Fox , BTW, thx for the recommendation on imgur...it blows away NBR's broken pic madness LOL.

 

The bigger die GP102 do I believe. The smaller GP104 dont it looks like.....

 

Hmm...I might have lost in the GPU silicon lottery. Can't seem to OC it to match other fellow 1080's...

The CPU seems to be a winner, but the GPU is gimping my scores...


http://www.3dmark.com/fs/10274130

 

What are your clocks like during the test? Have you tried leaving something like afterburner running in the background?

 

I'm monitoring that to see where it's being limited...hmm. ...a lot of it looks normal for the most part....I'll keep you all posted.

::iunlock::

 

Mine does not like going above 2101. If it does, the benchmark crashes. Try +200 on the core and adjust as needed

 

Better than my samples then I have been hovering around the 2050 region.

 

@Meaker@Sager , what is the name of the thermal pad that Clevo places on the vrms for the p870dm2/3?

The ones I have on my p870dmg are different.

 

I don't think the clevo ones are particularly named, they are the options for pads of the heatsink manufacturer.

 

I see, my concern is that if there is a re-paste needed the pads tear off quite easily. Any suggestions for a good replacement for these?

 

I've always opted for Fujiploy Thermal Pads...the 17.0w/mk ones.

::iunlock::

 

Grrr...my GPU is frustrating me. Lol...

What are your OC on core and cache?

+200 core and +??? Memory...

If I OC too high Fire strike stops after a minute or so and displays a score of zero.

Mind you this is all without Prema Magic.

::iunlock::

 

You mean memory? My max is +200 core/+600 mem...above +600 mem I get diminishing returns but the benches don't crash.

We do need @Prema magic because the gpu core clocks do not stay in one place ;-)

 

Oh wow...my GPU is majorly gimped. I can barely OC the memory and the core hangs on a thread at +200. Yikes...



::iunlock::

 

It's a little too early to think that yours is gimped, have you checked the re-seating of the heatsink?