*** Official Clevo P870DM2/P870DM3 (Sager NP9873/NP9872) Owner's Lounge! - The Phoenix 2 is here! **

I sent with a private message.

 

Just an update to the room. It looks like I have @otto_kahn pointed in the right direction.

 

@Johnksss worked hard, and solved the problem. Thank you very much.

 

@Papusan Thank you same to you. You are very interested.

 

Hey @Papusan , He was missing all of the basic stuff that we all tend to take for granted since we install it by nature (part of our factory installs of any os). Once he understood the rest was easy. And to answer another question. For what it's worth.....His paste job has no spikes in temp. They hold pretty damn even across the boards.

 

I haven’t posted here in a while but I caught up on this saga. I must know how you fixed this! LOL. I’m hooked on this like that fake airplane love story on Twitter.


Sent from my iPhone using Tapatalk

 

In his case...
LM + correct voltage settings + Cooler room temp = better temps.

 

Cooler room temp sounds nice around now. At least I have a window now

 

Is a mild buzzing from the 1080s at full load normal? I have recently done a cooling upgrade that quietens everything considerably and am hearing things I've not heard before...

Almost destroyed the VC heatsink while soldering - even with low temp paste - but luckily only a few internal posts popped and bulged in a non critical area that was easily fixed without any negative effect on temps or alignment, but I had to stop and there's probably partly cold joins everywhere... Will not be doing that again.

Still testing and checking for leaks every half hour but max gpu temp playing doom 4K at average 160W single gpu load was 44C... with no fan noise (auto). Kombustor did spin up the fans to nearly max but settled on 78C both gpus with 225W limit

 

You may want to redo your testing....

 

I'm by no means finished just basically using it normally for the moment with some a/b tests along the way

Same exact workload, 46C with it attached (higher ambient) and on, with no fan noise, vs 71C with it disconnected on a 'medium' fan speed/noise.

Its mainly a temp and noise mod, rather than outright performance, since with tweaking max fans can handle the components at maximum load.

That's why I asked about the gpu (vrm?) buzz, I just never heard it before due to the fans.

 

Make sure all your VRMs are making good contact but it could just be coil whine.

 

That's just it. You were able to obtain this only by throttling. You need to get past the throttling first. That will give an accurate account of the cooling power.
It's only a suggestion though
An example:
Max fans and laptop cooler are used.
Benchmark preset used.
2560x1440P
Start and finish temps.
Real time picture.

Spoiler

This is not about noise levels, only the throttling.

You have a very nice idea going, just want to see if it can hold up to the pressure....

 

240W per GPU limit
It was running for a couple minutes on max fans before the monitoring was started. 81C and still slowly rising for the GPUs.
Then the pump was turned on and dropped GPU temps to 53C. Loop was still slowly heating up so it would have risen a few degrees more from there, but even with the aborted soldering effort all the cooling effectiveness I could have dreamed for is there.
Drops to 4.0GHz with "VR THERMAL" lit up which I've never seen before, need to check some thermal pads.

This with 190W limit on the GPUs and auto fans and loop on from the start, this meant very little fan assist so most of the cooling was done by the loop


And a good 10+mins of graph history of Crysis 3 (maxed 4K no frame cap)


Full imgur album which is totally disorganised atm https://imgur.com/a/TTyAJ

 

The VRMs will get exponentially hotter as you push them harder.

 

HWInfo shows it as "RING: VR Thermal Alert"

But it only shows up when the GPU temps reach 81C under a 225W per GPU power limit on furmark. A 190W limit results in GPU temps of 75C and no thermal alert. Within seconds of turning the WC pump on and GPU temps decreasing, the alert disappears.

I don't understand the link between increased GPU power draw and a hotter VC heatsink, and this thermal alert. The CPU is under low stress, ~40W usage, ~1.3V, ~60C. I have overclocked my memory to 2800 CL 16 at 1.30V, and raised a bunch of voltage and current limits trying to get 4.8 @ 1.4V working without any throttling (unsuccessful)

and now, this particular ring thermal throttle shows up whether at 4.6/adaptive +45mV/x43 cache or 4.4/static 1.175V/x41 cache which leads me to wonder why on earth the uncore/ring part of the CPU is triggering a thermal alert under heavy GPU load but not heavy (>110W) CPU load?

edit: could it be PCH ???

I think I'll revert to BIOS defaults and start again. Firestrike standard score is through the floor (~22000) with the GPUs working only at about 70% usage in GT1 & GT2, no throttles or limits or frequency dips or stutters. That's lower than the stockety stock 24k I ran when I first got the thing. But Timespy is right up there where it should be (~12300) on exactly the same settings. Strange



Am I right in identifying these -
CSD87350Q5D
http://www.ti.com/lit/ds/symlink/csd87350q5d.pdf
as the CPU Vcore MOSFETs in a 4 phase doubled arrangement?
Is the rated instantaneous maximum of that part of the VRM 320A (8x40A), or 160A since they are doubled?
Of course real world balancing and efficiency losses and all those other things will lower it but TL;DR my question basically is outside the voltage/temperature of the CPU, are keeping these VRM components cool the main thing to worry about when heavy overclocking

 

VRMs waste more power the more your draw from them and the hotter they get. Not surprising as you push it up they may be getting hot.

 

... well aware of this which is why I have extra cooling over the vcore vrm sections in every heatsink mod I've ever done.

What and where is the "Ring VR" and why is it now tripping thermal alert when the GPU heatsink gets hot (despite it being hotter many times in the past), but doesn't throttle the GPUs?

I can't run a cache and/or memory test (e.g. AIDA64) to induce it as it throttles with "EDP OTHER" (TS Limit Reasons) a.k.a. "RING: Max VR Voltage,ICCMax,PL4" and "IA:EDP/Other,ICCMax,PL4,SVID,DDR RAPL" (HWInfo) before any thermal alert occurs which is the bane of my bloody existence since it happened on the other 6700K on the other P870DM-G on basically stock settings, and now on this too. Only thing common to both is this 2x16Gb 2666CL17 Ripjaws 4 kit...... hmmm


Which of these could be overheating? The one north of the GPU AUX power connectors, does that regulate the power going to the GPUs? What's the one circled just north of the slave GPU slot that would be under the heatpipes of the VC heatsink?

 

Speculation here....You may not have enough watts or amps unlocked. Or you may be giving it too much voltage...It's hard to say which way it actually is.

 

@bennyg on the TM when i had stock bios without having access to VR IC w/e options, those are the option throttles me the most. especially when I up the ring bus. upping the value fixes most problem and i throttle way less to the point of thermal limitation.

 

Also furmark usually will cause power issues as it is a power virus.

 

If anyone needs these items, they are available...

FS: Dual Clevo/MSI AC Adapter Convertor (AC-100) with Extra Long Cable
FS: Clevo/MSI 330W AC Adapter with Extra Long Cable
FS: Clevo P870DM-3 GTX 1080 Heat Sink (For Single 1080 GPU)

 

So even though I'm selling my NP9873 to make way for a newer one I've been traveling with it lately.

First image is from the Apple store while I was picking up my new phone and watch (I use iOS and Android). People had no clue what to make of my laptop which was pretty funny.

The rest is from the coffee shop this morning. Again, people have no idea what is going on when I pull this out of the bag...

OneDrive NP9873 Travel Album

 

You should have seen the reactions when I had the old P570WM on the train/airport.

 

I can only imagine


Sent from my iPhone using Tapatalk

 

It was cool to see the xray image however They would not let me take pictures though

 

9900K in DM3

 

Nice. How are the temperatures?

 

Not too bad, since stock is power gimped to 92W/138A Iccmax. About 80C on stock auto fans, default voltage, with a decent but not spectacular tube of some cheap EK paste I have.

I will report back with a variety of tests of that power level at stock, at maximum undervolt, under max fans, and of course with moar powah. It runs a lot cooler than the liquid metal delid + on top 6700K at 95W (makes sense given the decreased thermal density/lower clocks per core, higher mounting pressure too?)

 

80 C is not bad for stock auto fans and default voltage. Even with the power gimp, could it run the all core boost to 4.7 Ghz at least?

Also super surprised that it's running cooler than liquid metal delid since I've seen some reviewers get high 90C with even with 280mm or 360mm water cooler. de8uaer even put out a video on the topic, so maybe there's still a lottery.

 

It's absolutely power gimped compared to when it's in a desktop setup allowed to draw about 150W for 4.7 or 200W for ~5ghz, that's why it runs cool. These are just the very conservative stock settings in the bios.

I'm getting similar results (about 4ghz all core) to the 95W limited testing here:


Edit sorry it was hwunboxed previous vid before that.



This is what I was referring to

 

There is no magic bullet here, its the same silicon with 2 more cores, you should expect the maximum all core turbo sustainable in the chassis to drop.

 

True, and I went in with my eyes open here expecting about 8x4Ghz on the stock heatsink. I'm surprised it gets that without needing to undervolt it. It is an expensive upgrade though, but I got really really lucky just happening the check my emails at the right time to see almost straight away a notification for a second hand 9900K listing for under RRP, with Ebay's please-don't-be-tempted-by-Amazon-AU sitewide 5% discount code after that.

The stock DM3 CPU heatsink is not the final endpoint here either. My goal is 150W and 2000cb r15.

Random run - while watching derbauer delidding and destroying a 9980XE on youtube with 40+ chrome tabs open on the other monitor. No tuning at all yet. Beastly

 

If you need those threads it's still going to come out ahead and if you dont mind dropping a few 100 mhz and you do bring the voltage down it's going to help big time.

 

Congratz on the 9900K!!
Now to the reality. Your score and temps are based off a clock speed of about 4.0 ghz max.

I kind of hate to say this, but stock is 47 all cores and your system is no where near this unfortunately...
Here is about where stock should sit at 4.0 ghz@ less than 95W.

 

Package power numbers for 4.3 and 4.5 ?

 

Or rather
50
50
49
48
48
47
47
47

Hummm, not really. The lower it's set the lower everything goes. If that makes sense?

 

I think I understood what you meant. You tested 40x and used a bit below 75w in Cinebench. I wondered more about expexted power consumptions numbers if you put 4.3 or better 4.5GHz for 8 cores. This to see if this is an option instead for 8700K for those who want upgrade their older models. With unlimited power caps. 45 instead for 47 on 7 and 8 cores.

 

Ill check...One sec...
but my numbers will be based off of -125mV

 

4.3 Ghz

4.5 Ghz


I locked them at 95W's

 

Thanks. Maybe I wasn’t clear. Without power caps. Let it takes what’s needed.

 

That is really based off of your bios.
I posted that in another thread what it draws from this machine. And the watts can be manipulated so that is really no longer a valid way of testing true watts.

Side note:
In all honesty, stock is 4.7 on 8/16 so anything lower than that is kind of a problem?

 

Intel has blurred the lines on what "stock" is and people's expectations.

 

4.3 Ghz


4.5 Ghz

 

I know. Best would be to compare the wattage numbers directly from the wall (UPS or Kill-a-watt). Lets say if you run same setup and switced between the cpu's and tested.

But if the watts you show is comparable vs. what people run today (old setup DM/DM2). It will be too hot with more than 4.3GHz all cores.

Do you have any idea if the bigger 8 core die is able to run cooler than 6 core with same wattage? (6 core delidded with LM)

Thanks.

 

@Johnksss are these voltages any different from what you're seeing?

Give it a 150W PL2, -90mV undervolt and ...:



This is on stock auto fans which doesn't spin beyond 50% even when its up there thermal throttling...

Changed board UUID in the Obsidian tools has nuked my fan control licence. Wonder if it's anything to do with this:

 

The watts are not the problem. It's the voltage it takes when jumping around from 4.7 to 5.0 ghz, And messing with some stuff will brick your board. Most are recoverable with a programmer, but some require a chip change.

If that were true, you would be at 98C and you're only at 96C. It does not temp throttle at that point. However, you can power throttle at any range you do not have enough amps/current voltage or voltage for. So would need to see what those settings look like before we can start to compare

Side note:
-90 PL150W is a throttle point. It is closer to 170+W with that setting.

 

Yes you're right there was occasional clock drop throttling still occurring but most of the time it was 47.00 multis all round
I didn't show it on the screenshot but hwinfo did show thermal throttle flags on two cores and ring on the 3rd run, but not power limit or critical temp flags. All 3 runs 1990 points though... I saw a few thermal limit options in the BIOS so will delve into that a bit more
The temp readout is Package - that was always a C or two hotter than max core on the 6700k but maybe not on this, I should check the core maxes more carefully
I also didn't raise what Throttlestop says is ICCmax of 138A nor have I looked in the BIOS to see what PL3 and PL4 are set to. 150W PL2 with raised turbo duration, speedshift on, -90mV offset, and cache iccmax from 20A to 25A are the only changes from whatever the default values are set to in this BIOS.

 

1990-2000 is about right for a 100% run at 47 on all 8 cores and 16 threads

 

Can you post some temps as well for those speeds and such?

I am beginning to think that 9900K is not coolable by normal humans at 4.8GHz....