Clevo Overclocker's Lounge

Yeah they transferred to FedEx yesterday, I told them to hold it at a store close to my place since I'm travelling right now.
But it was supposed to be here yesterday, stupid FedEx hit delays.

 

No bad dude. I'm Kryonaut and my temps are still solid after 2 months.
Though yes my heatsink and processor ihs are both lapped.

 

I replaced the Arctic mx4 with Kryonaut from a rig of a friend of mine and got a reduction of 17°. Not bad so far!

 

http://www.catzilla.com/showresult?lp=689713|*|Result Details
http://hwbot.org/submission/3232392_

 

http://www.3dmark.com/aot/141315

 

it should rather say "Fox Communications"

Sent from my Huawei Mate 8 NXT-AL10

 

Quick question, is the top surface of the Power Phases, conductive? I mean the square grey things on the GPU.
Im planning on sticking copper shims on my old heatsink with Artic Silver Adhesive, and then using normal TIM (ICD/GCE/TGK), as an experiment to see if helps with cooling.
Just dont want to short something.

@Papusan @Mr. Fox @johnksss

 

I do not think so, but I am not 100% certain. Their electrical connections are on the sides. I believe those are inductors and I am not sure if they are ceramic, graphite, a metal cover or what. Even if they are conductive, if the tops of them are not physically connected with anything else it may not matter. You might be able to test with an ohm meter to see if touching probes on opposite corners (not the contacts) has any kind of reading.

That would be an interesting experiment. Those suckers get extremely hot... with 780M and 880M they got super hot as well. If you don't have access to an ohm meter, maybe post something at Tech|Inferno and ask @Khenglish about it. I bet he would know for sure. He's the guru on that kind of GPU hardware stuff. He is the fellow that transplanted desktop GTX 970 memory to his 980M and can clock the GPU memory to ~2000 without any problems.

 

I see you don't use Profile #1 who is DDR4-3000 CL16. Why use CL18? Stability problems? Or maybe I'm wrong about this ram?
Ps.. Break 5000 posts with my <good>English

Edit. @johnksss nice score. +1800 is the new goal

 

So it would seem....

 

You have to loosen timings as you speed the ram up, the overall latency can't drop massively below stock or the sticks/IMC will give out.

 

Hmm, lets see if i can find one around (doubt it though) .
Yeah, i will make a post there, i read his thread on the transplant, and that stuff was just mind blowing.
My main object is to be able to hold high clocks for extended periods with sub 70C temps. Ill start a separate thread logging the whole process.
For the components on the GPU ill stick the shims to the heat sink using ASTA, and for the core will probably let it float with TGK between the shim and heat sink and CLU between the core and shim. Might just end up with TGK on both side, im curious to see how the former will perform.
My main concern right now is how to be able to stick the shims for the other components flat onto the heat sink. Though since ill be using TIM in between it wont matter as much.

Woot ?

 

I think, hes using the 3000Mhz kit. and that kit does have a profile with lower timings.

 

Congratulations on the 5000 posts, Brother @Papusan. You are a valuable member of the community, bud.

I was playing around with manual memory overclocking trying to set 3200MHz and had to reset NVRAM. I did not notice it was on XMP Profile #2. I'll switch it. Both profiles are very stable. Thanks for noticing.

 

Btw, how'd you unlock the multiplier?

Sent from my SM-G900P using Tapatalk

 

iirc there are indeed some xeon models with unlocked multi... @johnksss himself mentioned that over at TI i think i was surprised at that myself!

Sent from my Huawei Mate 8 NXT-AL10

 

I know that, like the 1680v2 which is actually a 26XXv2 locked down because of faulty cores or cache and resold as a different model. Either he unlocked the locked portions or got a locked xeon to use multiplier through an exploit in the microcode is my thoughts. Inquiring minds...

Sent from my SM-G900P using Tapatalk

 

Here we go... all better now.

 

All xeon 16xx parts are unlocked.

No need for modding, it comes unlocked. Generally the 16xx parts uses the same die as the EE CPUs so no extra cores but 1680 v2 actually did use the MCC die. I doubt intel really planned for the 1680v2 to end up in non OEM hands.

The xeons tends to overclock MUCH better than the EE CPUs as well. Much more careful binning etc.

 

Look at the cpu in the screenshot before commenting. It's a 2690v2 which is locked. Meanwhile, research some of the original strings on them locking down 269x chips similar to amd with their six core chips years ago.
Sent from my SM-G900P using Tapatalk

 

Yes. Much better. You should not run your ram under specs

 

Its a bug I believe.....

1680v2 I believe was made for Macs. So probably some type of bug?

2690v2 are a 10 core part anyways. 1680v2 should be made from the same MCC die that made 2690v2.

 

Don't know about faulty cores and being resold as different models or such. Although i think the 16xx series unlocked xeons showed up in hp servers first. And it was the most expensive 8C unlocked Xeon on the market. Normally Xeons are known for being locked.
And lastly....The chips is a 16xx series and not a 26xx series. People tend to make this mistake a lot over the years. Including cpuz.

I'll post this message since it best describes this chip...


From lutjins over at hardforums

1) Ivy Bridge-E/EP doesn't use an onboard Fully Integrated Voltage Regulator (FIVR) like Haswell and Haswell E/EP do. This onboard FIVR is a source of extra heat on the processor that can interfere with overclocks. It adds roughly 10W to the TDP of a given processor (TDP depends on bin level of the die as well).

2) The 10-core MCC (medium core count) die for Ivy Bridge-EP is a step up in size from the 6-core LCC (low core count) die of the regular Ivy-Bridge E processors. In order to have a 8-core Ivy Bridge-EP processor, the MCC die must be used. As the MCC die by default comes with 10 cores and with 2.5MB of L3 cache/core, this results in a total of 25 MB of L3 cache on the die. This cache can be left fully intact, even if cores on the die are disabled, which is what's done with the E5-1680 V2. Two cores are deactivated on the E5-1680 V2 to produce an 8-core CPU, but the L3 cache is left fully intact at 25MB (makes this chip quite unique). This is the only non-LCC die based chip ever to be released from Intel in an unlocked form and has the largest L3 die available on any unlocked CPU.

 

Intel Skylake-E, LGA 3647 socket, 6-channel DDR4

 

Damn that's big. (unfortunately thats not what she said )

Good thing im holding out on a desktop workstation until the end of this year.

 

Man, you are all over the place!
But with some very interesting news in all those locations.

 

Servers are my main interest for Linux/Solaris projects.

I need to keep up with cutting edge technology I can use in new projects that I can't possibly guess what they will require ahead of time. A detailed knowledge of new technology comes in handy when I need solutions for those projects.

 

Yeah the new socket was announced a little while ago, certainly fun but I bet it wont be cheap (in the "if you have to ask way")

 

That could be an over clocked 980. lol Speculation of course.

Also looks like i might have to skip the 6950X as well....

 

Was thinking that also, but trying to decide between a 5930k and 5960x for a water cooled project, or wait for a zen 8 core...

Sent from my SM-G900P using Tapatalk

 

If you already have the board, then 5960X before people find out that it's still better than the 6950X and go jacking the price back up. If it's a fresh build, then maybe wait it out a bit and see what zen has to offer.

 

It's a fresh build, but was thinking about trying to pick up a decent chip cheap before people read what i have. .. zen is supposed to compare with Haswell EE, although 95 tdp, 14nm process, and October release. Right now, just have my ZM...

Sent from my SM-G900P using Tapatalk

 

Key word: supposed

I definitely won't believe that until I see proof. AMD has given me no reason to have a positive outlook on their products, especially CPUs. I hope I am given a new reason to be positive, but will definitely take a skeptical 'show me' wait and see approach on Zen. Unless it is at least as powerful as 4.5GHz overclocked 4930K or 4960X it won't matter much.

 

I agree and that is my hesitation. With the smaller lithography, which should give a 20% improvement for two nodes and finFet, plus SMT over CMT, which is another 20%, you still have, at minimum, a 144% increase over their current "eight core" offerings, so long as they didn't completely f up. They said 40% without the die shrink, which should be 168% with the die shrink, of which they now say about 150% improvement. By my measure, that puts it easily at 5820k to 5930k, with above that being a far stretch. But, with that, compiler and instruction optimization for the chip alone could hold down performance, so, as you said, I'll see it to believe it, as they've already admitted leaving 18% potential performance on the table. ..

Sent from my SM-G900P using Tapatalk

 

It depends on your workload, I think he is talking about the performance in lower threaded applications where the 5xxx series is clocking better.

 

Was hoping for a crush all cpu, but it wont take all the test because in anything below 9 core use it can be beat out without using extra cooling methods, but the chip is rock solid. Finding a high overclocker might be quite hard to comeby.

 

Read what Fugger had to say about it....
http://www.xtremesystems.org/forums/showthread.php?296657-Intel-6950X-Overclocking-Report

 

Finally passed over 10500 in cpu-z bench with 4.8 GHz aka 10559 pts



My next best score with 4.8 GHz <10483 pts>

 

You guys should try come in to microcenter then buy 10 6700k and try to see if any of them will run 4.8 undervolt, keep the one that does 4.8 at lowest voltage and return the other 9 to micro center next week lol

 

Most of us have 6700k's that do 4.8Ghz easily. Heck i took mine to 5Ghz

Its runs perfectly stable at 5.1Ghz on a desktop board.

 

Ah yes, 4.8 is attainable but you want the chip that uses the least volts compared to the others.

 

They already do bebe.

But tbh, its too much work to test for a processor that magical chip, since those are rather rare. Might have better luck if someone is able to provide a batch no. or something.

Or maybe im just too lazy lol.

 

or just throw money at the silicon lottery and call it a day

 

That works, but i dont think promise low voltage ones, they promise you that the chip will work @ 4.8 Ghz etc.

 

At a specific voltage. It may be lower voltage, but they guarantee speed per voltage.

Sent from my SM-G900P using Tapatalk

 

Yeah, Futuremark managed to ruin the old better interface on 3DMark. But they are unable to validate new drivers. Yet another bad bench driver from Nvidia Yeah..., Pascal is on the way!!

 

So close, so close but not today either . A new Cinbench R11.5 at same 11.73 pts as before!!
Otherwise ok Now I miss only one bench over the damn 11.69. And I'm free for bench under 11.70 in the ranking window I will continue working for +11.75

 

Doesn't seem to do any favors for 1080 bench wise

368.25


368.39