Anybody got a Skylake HQ/HK CPU yet? I want guinea pig-- I mean testers.

As in the title. Some of you may be aware of my Mobile i7 CPU information guide (which is slowly expanding), and in there I have found and stated enough proof that 99.9% of mobile HQ CPUs cannot hold above their base TDP for more than 2.5 minutes regardless of the laptop and BIOS options (with one or two exceptions that seem to have been re-purposed MQ chips from the Haswell line), and I want to test if Skylake suffers from the same design defects. ESPECIALLY the HK chips, because an "overclockable" and "TDP-limited" CPU is basically a very unfunny joke.

Spoiler: Preferred laptops:

Clevo P6xxRx series (especially if from one of Prema's partner shops, like Mythlogic)
MSI GT72 and GT80
ASUS G750 whatevers
Alienware's AW17 R3, AW15 R2

Gigabyte, Razer, slim MSIs, Lenovo and (also gigabyte) Aorus laptops I fully expect to overheat or encounter some other limitation long before we manage to find TDP as the root cause of a throttle.

Spoiler: Testing setup

If you volunteer, I'll have you make sure your cooling is up to snuff first and foremost. If you're unable to complete Throttlestop's TSBench using 1024M without overheating, you probably won't be able to complete the rest of the tests properly. We'll be using HWiNFO64, Throttlestop 8 Beta 4 and possibly HWMonitor to determine package TDP. The "Limit Reasons" program that @Dufus wrote that's included with the TS8 betas may be very useful in determining throttle reasons (as current limit can be a reason); but I don't know if it works for Skylake or if it needs an update.

First, we need to make something clear. We need to run HWiNFO in sensors-only mode, HWMonitor and Throttlestop 8, and look at the TDP used in TSBench. The "package power" in HWMonitor, HWiNFO and Throttlestop should all read mostly the same thing. It's fine if they don't refresh at the same speed, but maximums in HWiNFO and HWMonitor should do just fine for seeing limits. Once we see that they've lined up, I'd like the Limit Reasons program (bundled with TS8) and either HWMonitor or HWiNFO to be running. You don't need both. As long as we have a clear indicator of throttle reasons and max TDP. Throttlestop can be used for average TDP if you prefer to read it, but HWMonitor or HWiNFO needs to be running. Now, we're ready to run the tests.

Spoiler: Testing programs required

Throttlestop Beta 6
HWiNFO
HWMonitor
Prime95
Handbrake
Shadowplay (not required, but it's the best way to capture some video to use in Handbrake)
Intel XTU (may not be necessary; I have trouble finding a download link that works for me, you may be luckier)

Also, if you feel like it:
You can run the tests in this post by @moviemarketing
You can run Linpack locked to four threads. I refuse to link this; you need to find this on your own if you're wishing to abuse your system with it. For what it's worth, my current laptop can handle it after using Liquid Ultra thermal paste and elevating the back with max fans, but unless you've gone to similar measures in a thick notebook, you will likely not be able to handle it. No, the GT80 Titan does not have good enough CPU cooling for this according to users on this forum.

Spoiler: Testing methodology

First, if shadowplay is owned, simply run some game in fullscreen at 1080p (if your screen resolution is larger than 1080p, please run the game in 1080p for the purpose of this test) and record a ~10-15 minute video. This will require Handbrake to render the video for quite some time, pushing the likelihood of your machine remaining over the TDP limit. The reason is that low compression and high bitrate are usually processed very quickly in handbrake. Even under 1v at 3.5GHz I was using well over my 47W limit on my CPU. Higher resolutions render slower however, so 1620p and 4K resolutions are not the best. Make sure to increase your TDP limits via XTU or your BIOS to at least 100W if you can.

Handbrake test. Copy the settings you see in the screenshot, which also gives an example of how Handbrake is meant to break your 45W limits. Render a long video from Shadowplay (or find a low compression video elsewhere; quicksync videos are fine as well as long as they are high bitrate, etc).

Prime95 test should just be a CPU stress with all workers. I've never run P95 (never needed to) so I have no screenshots, however it is well known to use quite a lot of TDP.

TSBench is self-explanatory. Simply launch it, change it to 8 threads (or 4 if you have an i5, as those are quadcores now) and click the 1024M button. Note that due to my CPU's exceedingly low voltage at stock speeds, in order to surpass the TDP limits for TSBench I overclocked my CPU to 3.8GHz and returned it to default voltage, unlike in other screenshots.

XTU benchmark is also self explanatory. There's an option to run it. DISREGARD THE RESULTS OF THE BENCHMARK ITSELF. We're only interested in how the CPU behaves during the test. The results of the test rely far more on good RAM timings and speeds over the rest of your system, so the result is pointless. We're after the stress it puts on the CPU.

Simply note the trend in the CPU speeds according to Throttlestop, and what lights up in Dufus' Limit Reasons. PL1 and PL2 are power limits. EDP is current limit. We can worry about current limit later, but tell me if current limit throttle is showing up before power limit throttle has a chance to show up and we'll see what we can do. When you throttle, note the TDP limits you throttle to. Remember, a yellow light doesn't mean anything. Solid red means it's throttling for the limit reasons program.

If your BIOS is unlocked in some way, try using BIOS settings or XTU settings to raise your turbo boost power limits and turbo boost short power limits and seeing if the CPU listens to the settings, and how long it lasts for under loads. If you can overclock it too, even better.

Post any results here please and thank you!

I also wanted to write " guinea pigs testers" in the title but strikethrough wouldn't work D=

 

Would love to see benchmarks for creative applications as well...

SPECviewperf 12

https://www.spec.org/gwpg/downloadindex.html​

Adobe Premiere

PPBM7 benchmark
​

After Effects

AE Raytrace/CUDA Benchmark Test
Teddygage Raytrace Benchmark​

Photoshop

Gavtrain Photoshop Speed Test
Diglloyd Photoshop Benchmark
KitGuru Photoshop Benchmark
Photoshop Speed Test
​

Encoding

x264 Encoding Benchmark
​

Cinebench

MAXON | 3D FOR THE REAL WORLD: Overview​

PCMark8

Creative Test
Applications Test​

AutoCAD

Cadalyst Benchmark​

Blender

BMW Test Scene
Blender Render Benchmark
Blenchmark​

 

Skylake results are so bad that owners are ashamed to post.=p

 

I don't think the 6820HK notebooks are shipping yet... Should do so by the end of the month!

 

Some Alienware 15 R2 / 17 R3 owners have the i7-6820HK in their laptops. You might want to check the 2015 Alienware sub-sub forum, in the refresh thread and owners lounge, if so desired.

 

I have to go alienware hunting T_T.

ok

 

Heh, sorry for pointing towards your pain.
I'll help out with this project over at AWA. I've got someone with an i7-6820HK and an R9 m395x willing to do some initial benchmarks.

 

Which programs are you going to run as benchmarks?

If y'all wanna walk me through the process step by step, and depending on the quantity of benchmarks, I'd love to help when I've got time between work and school.

Packing a 17 R3 w/ 6820hk

 

Fantastic!

I'll be having you record a video with Shadowplay using 1080p, 60fps, 50Mbps bitrate, and then running it through Handbrake to see if I can force TDP over the consumption. If we can get TDP to crack your 45W limit doing this, I'd like to try rendering a video that will take about 10 minutes to render. This should prove whether or not your CPU can handle sustained TDP drain.

I'd also be using Throttlestop's TSBench since TS8 beta 4 seems to support skylake a lot better.

Finally, something like intel's XTU benchmark (not stress test) would be a good hit. Especially if we can induce current limit throttle, and then I could have you raise your current limit to see if it'll be fixed (that's another thing I'd like to know about the CPUs).

Maybe some Prime 95 as well, but I don't know if that'll be necessary. If you're up for it, Linpack as well is good, but I'm not forcing that on anyone due to the extreme heat and power draw it uses.

All of this would be best if done while you've overclocked the CPU. The more power we can make it draw under a consistent load, the better. For all I know, the 6820HK might top out at 70W over a long period of time rather than 45W, for example.

 

Brilliant!

I'm aware of many of these programs, but I've either not used, or used very little of them before. With that in mind, let me know what the best way for you to give me instructions is, then I can go from there.

 

You can do it here so the results are here. The main thing is we need HWiNFO and TS8 beta 4 (and HWMonitor too if you want) to see if TDP usage is consistent. Once they are, we can use them to check the programs. Note that Throttlestop's TDP is HWiNFO and HWMonitor's "package power", which is a combination of DRAM and IACores and Uncore etc. This is the TDP we're going to be referring to.

 

I'll go get those installed. I'll be able to run any tests when I get back home from work in about seven hours.

 

D2 Ultima be like:

Spoiler

 

Believe it or not, that's what my hair often looks like when it's long. Just a bit more curly.

 

Spoiler

 

Spoiler: More like this

 

Spoiler

I don't see the difference

Spoiler

Ima duck outta your thread now, good luck catching more alien guinea pigs

 

Can you try some of these? http://forum.notebookreview.com/thr...inea-pig-i-mean-testers.782689/#post-10114060

 

oooh this should be fun

D2, the first users have received their clevo P6xxRx machines, could be worth posting this in the owner's lounge

Sent from my Nexus 5 using Tapatalk

 

Let me get through D2's benchmarks first; if I can find the time later, then absolutely.


D2, what process do you want me to use to check TDP consistency between TS8 and HWiNFO? Just encode something from ShadowPlay via Handbrake to see if the TDP will crack the 45w rating?

Also, you mentioned I should do this OCed. Would you like me to apply a manual OC (in which case, I'm going to need a tutorial recommendation), or do it straight through the Alienware bios presets for OC?

Finally, were you looking at doing a comparison between stock and OC, or just doing one or the other?

 

I want to see http://www.putme.net/wPrime155.zip benchmark with the maximum default 4.1ghz overclock in Aw17R3. The same also with a maximum stable manual overclock on all 4 cores. Don't bother with stock clocks. i7-6820Hk is created for overclocking. Wprime 1024(8 treads) will run for over 2.5 minutes so I think this stress test is well suited for Hk/Hq chips. You can also run xtu with monitoring besides the wprime stress test. Remember to put high performance in win power plan. Not balanced performance.

 

Anyone know where I can find the download link for Throttlestop 8?

 

http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

Take a look in the Throttlestop guide in this forum . Page 1 have the link. http://forum.notebookreview.com/threads/the-throttlestop-guide.531329/

 

Download link didn't work, unless I'm missing something.

 

I'm using my phone nd can't test now. But that link with Throttlestop 8 beta 4 should work. I know. Don't use chrome for downloading. Use rather Ie or other browser.

 

The link in the OP is for the deprecated beta 2. TS 8 is up to beta 4 now. Here is the link for it. It can also be found in unclewebb's sig.

 

Worked perfectly! Many thanks.

 

Remember to check out for more Oc settings in xtu utilities before you put best Oc settings in Throttlestop. Afterward disable startup for Xtu. Only use Throttlestop at startup on logon.

 

I had AW15 R2 on order which I would have arrived 6 days ago with i7-6820HK, but Dell cancelled it with promise that they would upgrade SSD, unfortunately they stop shipment day before it arrive at my house. Dell wanted to build an entire new unit rather with a 4-5 week delay rather than just give me the SSD. Ended up getting a refund.

I aint to fussed about not getting it. I read about someone already getting their i7-6820HK. Dell do not overclock it factory as in their specs of 4.1 GHz, simply isn't true. There are 2 overclocking profiles in the BIOS. User of one reported when overclocked on profile 2, this allows for 4 cores to run at 3.8 GHz, but said it was not stable. Due to 3.8 GHz not being stable I didn't really see much point in this processor, might be better in a different unit, or different samples. Pumping extra voltage into it using Intel XTU seem rather pointless due to out of control thermals. No doubt this thing is going to rocket super hot if so, and start to throttle just as fast. First thing to throttle will be the temps limit. Would expect that the TDP would have limits for short duration, who know for thin one?

Personally I am think might be ok with decent cooling. In the case of the AW15, would be ok if using the external GPU, rather than sharing with on board GPU to lessen the impact of thermal share on the heatsink. Interested to see how people get on with it.

 

Okay guys, I was trying to update the OP with more clean information about what tests I'd like running.

Unfortunately however, after the following happened yesterday:
my mouse died
my router died
stabbed myself in the foot
was way too hot to cook any food
got no response from my ticket for my plextor drive dying

the power transformer near my house blew out around 8pm, and my house along with about 8 houses near us had no power. Across the street did. I still don't have any power. I'm at my mother's house, where I've updated the OP (see this is why laptop is good). I've had the laptop sleeping the whole time with the un-edited post ready, as I was just about finished editing the post when the power blew out, but I couldn't save. Glory be to my luck.

Anyway, TL;DR? Post is updated.

I have not slept yet.

But post is updated.

 

I'm currently awaiting my SAGER NP8678-S (Clevo P670RG) with an i7-6700HQ. Should have it by next weekend if you still need some testers.

Sent from my SM-N900V using Tapatalk

 

I will definitely need testers =D.

The absolute more the merrier.

 

Some info about the new Aw17R3 and i7-6820Hk. This was not unexpected. http://forum.notebookreview.com/thr...nd-15-r2-refresh.780681/page-60#post-10107512

 

I listed everything you'd need to know with visual guides in the OP

 

So... no results yet?

 

http://imgur.com/a/c7kFY

Too bad. It won't go over 45W so......

 

You seem to be TDP-throttling at 40W. It seems Dufus' limit reasons needs to be updated, as PL1 or PL2 should be lit right there.

If you have Intel XTU could you try raising the TDP limits, and overclocking the CPU? It should allow you to set 3.3GHz on all 4 cores going by Haswell and Broadwell's chips, and that should cause it to draw more TDP during handbrake renders etc. I know you're only at 3.1GHz but that's a surprisingly little amount of TDP for handbrake.

Thank you very much for testing!

 

I guess I cannot overclock.... Too bad.

 

now for some data on the 6820HK! i wanna know whats up with that thing...

Sent from my Nexus 5 using Tapatalk

 

If I am not wrong the voltage is unlocked, can you under volt the processor, and if so how much can you reduce it by?

 

*gulp* pretty sure you don't wanna mess with this. :x
If you don't mean this, the default clock is already the lowest.

 

Well that sucks. Can you raise your TDP sliders for me though? Set them to 100W and set current limit to 256A and see if you throttle in P95? It's sad that you can't access your turbo sliders there. I know you should be able to, because you can't even lower them and you should have access to lowering them. So that's ASUS' locked down BIOS to blame (which is why they weren't a preferred manufacturer except the G750 line =D).

And keep it running for a few minutes again to make sure. If all goes well, you'll go into thermal throttle this time.

 

Was hoping at least to reduce voltage with the 6700HQ, as will be getting thin and light platform

 

Nobody knows if undervolting is a thing. Just because a negative offset isn't present in XTU doesn't mean it's impossible. For 3.5GHz on my 4800MQ at stock voltage I'm at 1.0474v or so. I could use an offset of -50mV (-49.8mV to be exact) to get to 0.9976 like I currently do, or I could manually set 0.9976v and not use any offset.

What needs to be seen is if the default voltage can be reduced at all by adjusting the slider from its "default" position. But that's not what this thread is meant to test.

 

@DivineAura Could you run a Throttlestop 1024M bench or Wprime 1024 stress test and check whether throttling is still happening around 40W? It is bad if the TDP throttling occurs at 40W and not at 45W as expected. Fine if you run a Cinebench 11.5 test also.

 

Running TS Bench 8 Thread 1024M only uses 31~32W. If I run Prime95 then it starts at 53W but slowly drops to 45W.

 

And did raising the power sliders in XTU do anything?

 

http://www.putme.net/wPrime155.zip - http://downloads.guru3d.com/CineBENCH-11.5-download-2475.html

 

Nope, applied 47W TDP, ran Prime95, still capped out at 45W.

 

Cinebench 11.5 @ ~37W.