*** Official Sager NP9877 / Clevo P870TM-G Owner's Lounge! - Phoenix 4 ***

Sort of I suppose, but we all wanted the new desktop level chips in the same chassis with no changes to noise or heat and that's not how physics works.

Then when the max-q hits it's TDP is lower to match the smaller chassis better and everyone goes a bit weird about it.

 

So as a note to self , for the 9k and 1180 single gpu if clevo is using the same chassis with same dimensions of P870 (am already discarding p775 etc as options )then I should consider twice before getting it esp if I have hopes of doing even a little oc in the future. Right everyone ?

 

I'd like to add something to the conversation. Have you noticed that's almost impossible to reach the Max turbo clock, given when only one core is active? That's because Windows 10 by default uses all available threads, so you actually never face the situation where only one core is active, hence the condition for the Max turbo speed is never actually fulfilled.

I tested using prime95 fft test, with 1 thread. The cpu goes yo 8,33% of utilization but none of the 6 cores goes to 100%. The load is distributed across the 6 cores

Enviado desde mi G8441 mediante Tapatalk

 

Not so much on the p870, I am voltage limited on the gpus which already have a 200w ish tdp, we already paid the tax as it were.

 

You can reach maximum turbo clock *IF* you enable only one core in the Bios.
Then only one core is active.
But not many laptop bioses allow you to set the number of CPU cores.

 

That's crazy impractical, Intel has to give another thought to turbo boost, probably by total CPU load. Within full turbo at 5% usage or less.

Enviado desde mi G8441 mediante Tapatalk

 

After watching Nvidia live stream - I'm hoping RTX 2080 is <20% performance increase compared to GTX 1080, so I don't get tempted to upgrade

I can live without beautiful reflections in games.

 

Put all cores equal

 

Are you sure that 20% is compared to the 1080 not the 1080 ti? I'd guess at least a 20% increase in performance 1180 vs 1080, otherwise is a fail

Enviado desde mi G8441 mediante Tapatalk

 

I did that for a while, all 6 cores at 4800 MHz, and it kinda works, for light loads, @100% load the frequency go down itself at 4000 MHz, the temperature was around 90c at that speed, so no need to try yo get any higher to me

Enviado desde mi G8441 mediante Tapatalk

 

Or you could try with msconfig > boot > advanced options > set number of processors to 1.

 

Is your CPU delidded?

 

Mine? Yes, I got the laptop from HID. why?

Enviado desde mi G8441 mediante Tapatalk

 

Odd then that 4.7Ghz gets so hot in normal gaming.

 

I think @Tyranus07 talk more about running 48x all cores in P95. This means +-160w constantly heat. And P95 is heavy enough on air in desktops.

And who knows how good his chips is. Or if he has tuned in voltage for his chips.

 

I have some p95 screenshots in http://forum.notebookreview.com/threads/another-round-with-hidevolution-p870tm1.815954/ running at 4.8GHz Throttling saved my CPU.

 

4.7GHz (5GHz with -3 AVX offset) Small FFTs... Expect higher Power consumption with 4.8GHz.

 

Try installing XTU (Intel extreme tuning utility). Warning though, don't ever have it installed at the same time as Clevo Control Center. It will cause neither program to work and both must be removed before then installing just the one you want. Also a simple bios change with both installed caused a BSOD for me.

When I first got my laptop it would not turbo well at all. It rarely would show any core having used 5.0Ghz. And all 6 would always show as 4.3 (as they should). Now this is going to sound crazy, but hear me out. Without making any changes to bios, multipliers, everything, stock. I installed XTU, (which requires restart). On restart 5.0Ghz turbo started being attained. Which is normal, but here is where it gets a little nuts. My 6 core speeds started performing @ 4.5...

My current setup has a multiplier curve between 45-51. I still get the same kind of turbo performance. In the picture below all 6 cores are at 4.8Ghz even though my six core setting is 4.5Ghz:



I showed as many as 4 cores running at full turbo (5.1Ghz which only had been set for 1 core) with the other 2 at 4.8Ghz (which is my 2 core setting):



And shortly after that, you can see all cores reached full turbo within a short time of each other:



These were all just taken now, on a reboot when the cores go through a clock ramp. So there is no load like the example you are talking about, but they mimic this same turbo performance under load as well.

Now, it makes no sense that XTU will turbo past what you have set. And maybe i am wrong, but I think what is happening is that it is cycling turbo speeds among the cores more quickly and efficiently. So quickly that the polling times to the sensors don't have time to differentiate? IDK, that doesn't make much sense either, because I imagine polling is done at exactly the same time to all sensors, so it seems like it would have to be from 1 point in time, thus accurate in number of turbo-ed cores.

What I can say though, is that stock my computer scored under a i7 8700k on CPUz, but with the only change being to install XTU; I scored over a 8700k on multi-core.

Something else I noticed, when I had a multiplier curve of 4.5-5.0, I did not have very good turbo performance at all. when I changed it to 4.5-5.1, suddenly I had very good turbo performance. So the stock curve may not be the most efficient. It has a range of 700Mhz (4.3-5.0), a 500Mhz range did not work very well for me, but maybe the 600Mhz range curve is the sweet spot. But definitely under stock settings, turbo performed better with XTU installed, and also turbos very well with my super stock settings.

 

^^^ get throttlestop and get rid of XTU lol.

And having XTU installed can possibly reek havok on the fan tables . Not only by itself but definitely installed any where near CCC . A re-format and a prayer is about the only way to fix a XTU fan issue. Lol been down that road ... Several times lol

 

I'm willing to give it a try, but I don't have throttling . XTU also seems to have some magical power, lol. That said, you have to be careful with it. It has a mind of it's own sometimes. Make sure you check all settings before applying, and it really should only be used for changes if you don't have the setting in your bios. So for you, don't use it for changes at all.

TY, for the warning on the fan tables, I had not heard that. I like to know what to look out for. thanks!

 

That's exactly what I was talking about. Running Prime95 (12 threads) at stock voltage with all 6 cores at x48, would give me 3900 MHz real performance and around 92ºC of max temperature. If I undervolt my CPU offset -65mV I get around 4000 MHz on Prime95 and 85ºC max temperature, when gaming I was avarage 4700 MHz but near 90ºC max..Now I set my CPU as follow:

*1 active core x48
*2 active cores x46
*3 actives cores x45
*4 actives cores x45
*5 actives cores x44
*6 actives cores x43

When gaming my avarage clock speed is around 4.3 GHz max temperature of 80ºc and when idle I get 4500 to 4600 MHz.

checked that running at 100% usage? If I set all cores at x48, when full load my clocks go down to 3900 MHz and stay there. When idle they stay at 4800 MHz

How you get so good temperatures with so high clocks? please details about your voltages and cooling system or any other tweak.

I really like the colored keyboard and the macro keys, and I think I can't get that to work without CCC? right?

What do you call the multiplier curve? I have only 6 points to adjust ( 1 active core, 2 actives cores, 3 actives cores,, etc)

What's your clock under full load? at what voltage? and what max temperatures you get?

 

Good temps. Share me your secret.

 

If you look into the screenshoot you would see TT https://www.tweaktown.com/guides/8481/coffee-lake-overclocking-guide/index5.html

Just to show the power consumption and max temp with 4.7GHz in a Desktop build. P95 Small FFTs is an unrealistic load in any computer.

 

Yeah without CCC you only are left with blue keys, although I did find a way to have the boot key colors (Random multi (3) color breathing), in windows. Just by booting into bios. And only macro key that will stay is the max fan Fn+! toggle.

The multiplier curve is just my name for it for any setting with different multipliers for different cores. Stock is [email protected], [email protected], [email protected], [email protected], [email protected], and [email protected].
I set it to [email protected], [email protected], [email protected], [email protected], [email protected], and [email protected]. (When I am not running at all cores with a common multiplier). This is the same setting in the screenshots.
I undervolt -25mV with these clocks, and I also raised power limits.

Under full load my clocks will be at least 4.5 (depending on the testing software), and sometimes the 4.7-4.8, and if it is a single core test; reaches the 5.1.
Temps are awesome in all testing I have run (never closer than 9c to throttling, and that is rare). Voltage is same as what my idle ss shows (1.2-1.3). I have not used Prime95 yet. I will check out that and the one kfxsti suggested tonight though. During normal gaming and other normal use it is always at temps lower than when testing and with clocks the same. I only push it hard at night though, lol, that sounded so wrong. But yeah during the day it idles in the 40s, at night I can idle in low 30s. Starting tests from that lower point always do better. So Ill share my results from the same testing software you're using tomorrow.

EDIT* I should clarify,,, the only hot keys that stop working are the ones set in control center. With the exception of the max fan toggle, which will still work. But the macros set by windows Fn+(F1-F12), all still will work. So things like volume, mute, lcd off, brightness, mirroring options, camera, airplane, and sleep; will all still work... And I imagine you mean macros you have programmed? I don't use this keyboard for gaming, but if I did; I would definitely want to program macros as well. I bet there are several macro editors that would recognize this keyboard easily though, but IDK. Lighting control definitely will have more limited options.

 

There is a fix if I'm not mistaken that allows you to change the color then uninstall CCC.
As for throttlestop. It's not just for stopping or taking care of throttling issues. You can create profiles for your overclocks, voltages. Change said overclocks, voltages -+ for adapative. The configuration is damn near endless. I have a 5ghz oc setting with the -voltage offsets like I want. A 4.5ghz profile with the - offsets. and a battery profile that keeps the core speed super low.

My CPU idle temps are 35-37~ (ac setting permitting lower idle when the wife isn't home to fuss about me loving 68-70f in the house as oposed to 72-75 because NC sucks wet dog fur with heat and humidity) max on the 4.5 profile is 60+~ while benching, BF1 and Warframe-
The 5ghz under benchmarks and or BF1 and Warframe the absolute max I have seen has been 74-77 core varying and thats atleast 1-1.5 hours of solI'd gaming. But I am delidded. Shimmed heatsinks(followed @Mr. Fox 's guide on YouTube, then made a few adjustments to the process- I ended up with a few heatsinks to test with lol), and a different fan shroud , a custom copper IHS and 1-2 other things I'm not allowed to mention as of yet.

GPU temps have never seen above 66-68c (messing with the voltage curve in afterburner helped alot on temps for me with the two 1080's) as of me getting everything cleaned, dented heatsink un-dented (thats a story for another night) . New pads, polished contacts, releveled for flatness,( still working out the the shimming process for the big ass dual GPU heatsink, as it's very thin in spots) and testing a new mounting method to allow for a bit more pressure on the gpu die-which has netted a bit better temps. But I'm trying to make it a bit more user friendly before posting how it all works lol.

 

Well Prime95 small FFTs test is the worst case your CPU is going to handle, so if your overclock/overvolt/undervolt survives that test, you're mostly safe. It's crazy the ammount of power that x50 (real 4.7 GHz with AVX load) needs. There is almost zero chance a laptop could handle that kind of power in the CPU.

From time to time I use Handbrake to re-encode videos, and that software takes full advange of all cpu cores at 100% load, pretty similar to Prime95 behavior.

I tried to use Intel XTU for a while but it didn't seem to work well on my system, I guess because the CCC. I use a macro key to get the "@" key with the Re Pag key, and other macro to turn off display. Plus I just noticed we have diffenrents cpus, I guess the new 8086k runs cooler that the 8700k? You should try Prime95 small ffts test to compare apple to apples though and check how hot your cpu gets, plus if your CPU downclocks itself

 

If you want minimal Keyboard controls here you go:
http://obsidian-pc.com/LaptopToolsFull.zip

Extract all files to C:\Program Files\LaptopTools

There´s 4 apps included, use the ones you want... KeyboardLed is for the keyboard... and its free, under initial development so its very basic!!
For it to work please run Control Station at least once and allow it to install a single dll, then you can delete all the other EXE files you do not want.
None of this will add trash to your unit, it will not create rogue files or reg entries...

 

Oh, I don't get running prime95 all the time.

 

If you installed XTU while CCC was installed, you should open CCC and check your CPU overclocking tab. If it acts like mine, you wont see any freq or temp readings from the cpu. The only way to fix it (that i know) is to have both uninstalled, restart, then only install the one you want to use. not sure about the macro for @, but the display you can turn off with Fn+F2, or dim and brighten with Fn+F8 or F9. Yeah I do have a different cpu, and that's why I mostly was making suggestions on how to deal with efficient multiplier curves. Those are more likely to act similar than the temps. And I am not sure I would say the 8086k runs cooler, as much as I would say it will run at the target temps into higher multipliers. So I guess that means it runs cooler, lol, but only at the clocks you shouldn't probably be running at anyway.

So I said forget it for P95... I have infinite more experience in encoding than I do in overclocking, I am very much a noob overclocker. And I will need to learn a little more and read up on the product before I even start using program like ThrottleStop. But since you mentioned Handbreak I took a look at that. And if you can find a better option for encoding I strongly recommend it. Handbreak will burn up a machine (that has power limits lifted) pretty darn fast. And again, maybe I just need to learn to overclock a little better. I have not even messed with AVX offset, as mine only seems to go in a + direction, lol. If anybody can tell me why that is? I probably need to use ThrottleStop to access that setting correctly. But Handbreak had my machine at temps I had not yet seen on it. I attempted to encode a 4K HDR .ts file into H.265 4k 60p .mkv. After just a few mins, my temps got up to 93,94,95,96,,,and I aborted the encoding...

Encoding the same file and settings in After Effects, and or Media Encoder, finished the encoding, but only utilizes about 50% cpu at full clock:

I mostly use After Effects when encoding (or at least rendering frames, for encoding later), so rather or not it is faster than Handbreak is nothing I can do anything about, it's the tool I need to use, at least to get my frames. I use Divx or Media Encoder to then encode my frames.

Divx encoder with same 4k file and encode output settings, does do 60-70 cpu utilization. It also is a faster than Adobe encoding, and possibly faster than Handbreak. And doesn't heat my cores up:


Since I really don't want to use programs that take all the Watts they can get, instead of use Watts they need; I had to stress my cpu out a different way to show you results with 100% utilization.

This is after CineBenchR15 had been run over and over, then captured during the last test:


The program that over utilizes to the most amount I am comfortable with is CPUz. This is after a 5 min stress test, and had my highest temps (besides Handbreak):


XTU stress test utilizes what it needs the most efficiently. And I would feel comfortable leaving this stress test running for days (this particular test was 10 Min). For me this shows me what I need to know, and hopefully close to what you where wanting to see. If an application is taking more power than this to obtain the same speeds, then I probably don't want to be using it. (IMO any power draw beyond this= pure heat, with no benefit) :

I completely agree (see above, lol). for me it does not replicate any condition I want to experience, lol. I will run a burn in application in short bursts, so I know how my system will react. But I don't expect nor want to have the same burning type activity from my tools. When an application just continues to pull all it can get, throttling or meltdown is imminent. So trying to bench with something like that, (right now anyway), is not what I want to do.

 

I was limited on attachments in last post, so here is one more .
It shows the graph during 10 min stress test zoomed out so you can see the whole run:

CPU (cyan)was @ or near 100% full time. Freq (dark orange) was solid flatline @4.5. Slight voltage variations (light orange). And most modulation was being done with TDP (red), and of course temp (green).
Note: The height that cpu shows at 100% in graph, is the same location for 100c (throttling) if the temp where to reach that. So you can see how far the temps are from critical. I am going to run same test again with all my cores at 5.1 tonight. I expect to see my TDP go up, but my temps should not raise by too much, maybe a few degrees.

 

If your CPU went over 90ºC with Prime95, I'll keep an eye on temperatures when playing games. Leave HWinfo running in the background for a while. Some games are CPU intensive, plus the whole system gets hotter when the CPU and GPU are running at high load at the same time.

Handbreak is free, I guess that's why I use it over other encoders. Plus it supports a lot of codecs and stuffs, including DTS and DD.

Well, Prime95 is just what it is, it is a stress test for the CPU, I like it, it gives some options for the quantity of threads, and what kind of test you want to performance. The odd thing is when running Prime95 my clocks goes somewhere 4000 -4100 MHz, but when running Cinebench my clocks performace as planned at 4300 Mhz... (I have set 6 cores at x43), I think it has something to do with AVX even thoght I have set AVX disabled on the local file of Prime95. On both tests my temperatures never goes over 82ºC which is ok to me.

 

I ended up just using Handbreak as a barometer of what to expect out of P95, so did not end up even downloading it. I am sure I will at some point, but I have to get to know how to handle what it will do. My guess is I could get an idea of what a specific frequency needs in terms of Watts (in my 4.5Ghz test It looks like I may actually want to turn my PL1 down to maybe around 90W, as it never went over 88W in that test). Then just let Handbreak or P95 power limit throttle. That is probably what is happening to you if you have not raised power limits. My raised limits I have now are probably the reason for the heat from those Amp suckers, for me. My games actually do fine, I only get heat spikes (similar to the CPUz ones), right at game start. I have started going full fans at start up and that fixes it. Then I Alt+tab out of window, open RLECCViewer, and set my fans to run at least 25% full time. It then also ramps up if needed. I do leave HWiNFO running, I even like to use RivaTuner Statistics viewer to monitor everything in game (I'm such a nerd! lol ). Haven't had heat issues yet, but I definitely will keep watching it anyway. And I never rely on RLECCViewer during full utilization, but it seems to do a good job for the games I am playing right now.

My guess is to get better performance out of Handbreak you need to raise your power limits, (if your are at 82c now). It seems to take as much as you're willing to give it. Just do it until you start to get the freq you're after. That is my guess anyway, and do at your own risk.

 

What's the option to modify the power limits? you do that just by rising the voltage?

 

Current limits and max TDP for the CPU.

 

@4mgin3 and @Meaker@Sager thanks guys, it was definitely the Turbo Power Limits (in ThrottleStop) the solution to my issue of not reaching turbo speeds on Prime95 Small FFTs test. The "Package power long" was set by default at 95 W, then I increased that value to 100 W, then to a 103 W, and finally to 105 W, and with 105 W my CPU is able to deliver 4.3 GHz on all six cores at the same time on Prime95 Small FFTs test. The downside is the max temperature went to 86 ºC, with an avarage temperature of 84 ºC on a 20ºC room. Where I live my room goes as high as 28-30ºC in summer, so that would be a problem.

Anyway I'm pretty happy that the laptop is able to handle 105 W in the CPU and keep the temperature under 90ºC with Prime95 Small FFTs test. By the way if you test Prime95 "In-place large FFTs" the CPU goes at 100% usage, but it draws a lot less power, something like 95 W, so the max temperature is 80ºC in my case.

No doubt small FFTs test is the king of tests when it comes to struggle your CPU

 

That's awesome! Glad I/we helped! Congrats, sounds like you got a nice build!

I would be happy with 86c I am usually between 87-89c max temps. And my room is usually between 25-27, if it gets much hotter than that Ill turn on the AC and bring it down to about 22c.

Ty, when I try Prime95, I definitely will start with large FFTs and check temps before moving on to small.

 

Hey can someone please run me through power limit 1 and 2 in bios? I saw it mentioned a while ago but mine are set to 510000 and 514000. Is that 510 Watts? I thought I saw somewhere that stock bios TDP is capped lower than that...

Thanks in advance...

 

The bios entry description gives the formular, divide by 256? It's the short and long turbo power max.

 

Ok. It certainly doesn’t say much in terms of how to work it out other than it’s milliwatts. I don’t have xtu anymore but it was the one thing xtu didn’t seem to be able to change, which increased my confusion.

On another note I’ve attached aluminum heat sinks to the flat area over the cpu and found a 4-5 deg drop in heat although only short term (1 ish hours). I’m in the midst of making a custom heat sink out of copper or silver for it and then try a push air flow specifically through it using a u3 pwm fan setup...

 

Yes, it's a power limit, so times milliwatts by 1000 to get the power limit in watts.

 

Ok cool, but there’s a cap for total tdp? Is that different to 514? And also I guess a more important question is does it make any difference to anything anyway with a stock bios...?

 

That limit is for the CPU only, there is no direct control over total system power limits.

 

"Technically" there is but that's an artificial TDP, although Clevo uses the same name for it as what others use for "Power limit 1" and "Power limit 2"(CPU package).
It's called "Platform Power Limit 1" and "Platform Power Limit 2", which are not the same as Package Power Limits 1 and 2. (of course Clevo calls them that; what clevo calls platform power limit 1 and platform power limit 2 are actually the package values). They're designed to be used with "TAU", or basically in conjunction with "POWER OF SYSTEM" (PSYS_PMAX). End users have no business messing with this. But systems integrators can install their own power limit overrides by setting PSYS_PMAX to be a certain value (example: 150 watts), and then setting the "Platform power limits" to be a percentage (still in watts) below the PSYS_PMAX total watts value (usually this raw value is also on a divider, e.g. 2700 is usually 337.5 watts).

Then if you exceed the platform power limit 1 and 2, your CPU will throttle, and if you exceed the total system power draw listed in PSYS_PMAX, your CPU will also throttle.

Changing these settings will not affect the cancer Embedded Controller enforced power limits (PECI) or the system throttling limits for total power draw with 1070 or 1080 video cards+CPU power.

Note:
I have no idea if the "platform" power limit 1 and 2's (actual MSR registers according to the intel documents) overrule or override the "package power limits" or not, or if they even function if PSYS_PMAX is set to auto (0).

 

http://forum.notebookreview.com/threads/p870dm-watercooling-mod.822390/

 

So with the rumors surfacing about the GTX1180 , it is not going to be much improvement over the 1080 (ie roughly 1.5 times pascal performance) but the price is going to be DOUBLE or more? Does it even make sense to buy 1180 with these numbers ? Also, ray tracing has a total of 10 games support that nvidia has promised and I have no clue if it will fail or run the previous year/generation games or not! What do people think here ?

 

Depends on what the 11 series for notebooks actually is which I don't think anybody outside NDA knows right now.

"50% over pascal" is cherrypicked Nvidia marketing benchmark... wait for indys to do their thing to verify.

 

It's going to take something more than Bf5 ,Bf1 and Warframe to make me jump ship from SLI 1080s. Anthem looked neat . But given the premium price we already pay for our hardware 1.5 performance gap isn't enough to warrant 1k+ per card for me. Aslong as it's running my fav games @ 1440p -120fps with the eye candy I prefer, I'll be holding out for something abit more impressive lol

 

I’m interested in knowing how hard it will be update both the CPU with 9th generation and the 1180 (or whatever it will be numbered). Supposedly the 9900K is compatible with the Z370 motherboard. Does that still require a BIOS update? Who pushes that update out (how does the end user download it)? We don’t know any specifics of the upcoming GPU, but will that be as simple as plug and play or will there be a need to flash VBIOS?

Lots of questions and possible hand-holding needed.

 

Too soon to make those questions.
Only few might know the answers, and those who might do should have NDAs hanging on their heads

 

Although the context of the question was in relation to new hardware not yet released, my main thrust was asking for guidance on how to actually replace those components. There is a huge flurry of posting activity related to the benefits of LGA components. A huge seller for me on the EVOC line is the ability to upgrade and replace components. The more I read in the various threads, the less sure I am that such upgradability truly exists for the laptop form factor.

I’ve seen threads related to upgrading CPUs from 6700K upward; 980 GPUs to 1080. BIOS mods, but not necessarily official software pushes. For example, I just watched a JayZTwoCents video on building a desktop. That exercise included updating the motherboard’s BIOS to support the latest chipsets. Even though the EVOC has a Z370 motherboard, I have not seen any discussion that leads me to believe upgrading these laptops is for the feint of heart. Cutting heat pipes? I cringed when I saw the heat pipes on @Mr. Fox ’s P870DM. It did not look right TO ME. Sure it was. But, am I willing to do such surgery on my own machine that I’ve paid in excess of $5k?

This is a fascinating area, and one that has me hungering for answers. Just don’t know where to look. Don’t really care about videos changing the CPU on old HP laptops. I’d like to get a feel where I would be heading if I were to purchase an EVOC.