New Clevos with Coffee Lake-S are here, P870TM, P775TM, P751TM

It seems to be a GPU bottleneck. If you look at the GPU utilization, the F5 is running at 100% most of the time in that scene, while the gt73vr and the desktop is maxing at 99% utilization.
In the next video, the frame drop at 100% GPU (jump to 5:00) is ~ the same as aaronne's average frame rate in that scene:

Could it be that the reason for aaronne's FPS limit is the 50W TDP difference? If so, how can this effect be explained (since the clock difference is not that high and the GPU power draw doesn't appear to be at max for that TDP)?
Is it possible (I'm speculating here) that this behavior is somehow similar to the "pascal problem", where gpu is "starved" when the voltage is too low, but the clock remains high?

 

yes, apart hitting vigorously the power limit I also think that lowering wattage (150W vs 200W) maybe disable some cu, so going to test again with 200W vbios on it (poor Tornado F5 cooling system..you will be hit hard!!)

 

The performance however is. To make it worse it's also 60s vs 40s fps, which is a huge deal since that's when the game becomes less fluid. Hence we talked a bit in private and found out that his 150W card is underperforming quite a bit.

Well duh, of course the GPU is involved with this. The 150TDP does make the difference, according to notebookcheck there is a rather big difference between 150 and 200W GTX 1080 cards.

https://www.notebookcheck.net/NVIDIA-GeForce-GTX-1080-Laptop.171212.0.html

So yeah, the TDP is limiting the card by quite a bit. That's why the FPS on aaronne is much lower than the GT 73VR for instance.

 

@Danishblunt
I knew that the 150w RBP is considerably inferior in performance, but what I didn't know is how much did the lower TDP contributed (and in what way), since the RBP's cooling system is probably limiting the performance before it has the chance to reach the maximum power draw at that TDP level.
The RBP might not be the best reference when the purpose is to evaluate the performance difference between different TDP stages.

What it also confusing, is that the (highly tuned, where I'm referring to @iunlock 's work) AW 17 R4 with a lower TDP 1080 (I still don't know exactly if it's 170W or 180W) seems to be performing the same (or even better) in benchmarks (firestrike, etc) as msi gt73vr.
I know that benchmark scores don't always translate to the same performance in gaming, so it would be great to know if AW 17 is also limited in this gaming scenario.

There should be a sweet spot (the point of diminishing returns) for the gtx 1080 TDP, where a higher TDP wouldn't be worth (benchmark competitions would be the only exception) the increase in power draw and heat for a very small increase in FPS.

 

Your video posted of the guy talking about the pascal problem actually refers to the exact issue you're talking about. Clocks don't mean everything in pascal, apparently there are way more factors determing the actual performance, which are currrently unknown. It seems pascal cards to react to certain conditions differently and makes it really hard to figure out why a certain pascal card could underperform.

Also the sweetspot u're talking about is not a thing, every silicon is different, meaning the sweetspot is not the same for all cards. Finding the best clocks to voltage ratio without the card shutting some features off is apparently the new thing to do right now.

Personally I think the clockspeed might not even be correct, NVIDIA might have done something to fake the real clockspeeds in order to hide the fact that the cards might not run that well. I've heard and seen several pascal cards acting weird in many ways. Hell, even Jayztwocents has encountered something rather weird when testing 2 GTX 1050s and both had very different performance despite being GTX 1050s from MSI.

Here's the video:


So it seems that now silicon is more important than ever. Sufficient voltage is now equally as important as the clockspeeds, which makes it more of a pain to overclock. Powerdraw back then was important, now it's even more important.

 

wow couldnt say it better, now going way back to all those posts about bga being better than lga. lga has better silicon, overclocks better and runs cooler, this is now a fact, and is what me and @Papusan @Mr. Fox and many others been saying. either its doublestandard or just trollol

 

Some would actually bother reading the entire post, you however don't.

 

jokes on you however. if i dont wish to read the entire post, i wouldnt read the last part of it, unless plot twist i read in reverse.

 

The entire post was about GPUS and pascal being.. well pascal and you started talking about LGA vs BGA. So you're apparently an odd person that reads in reverse. (by your logic)

 

The only thing you can do really is skip cycles to cause that, which is not that common.

 

this is what u dont get. you think silicon only matters go GPU but that is not the case. its for both GPU and CPU.

 

Silicon quality has been at the heart of overclocking since the get go.

It's only adaptive boosting that has revealed some of this to the general public.

 

Silicon quality on a CPU only matters to a certain extent, that is what you don't get. Overclocking a CPU in 2017

@Meaker@Sager
Mind explaining the skip cycles thing?

 

certain extent. now you want to make exceptions and conditions. if it matters it matters, and no, it matters a lot otherwise there wouldnt be website like silicon lottery or overclocker uk selling binned chips. its okay to admit you're wrong from time to time!

 

Wrong or the lack of knowledge is normally two different things, bruh

 

I would like to have this GT 73VR video link, because my googlefu is weak and I cant find anywhere a gt73vr equipped with gtx 1080, because I would to check his settings, hardware and other things like driver version etc..

at the sametime, here my video:
gtx 1080 at 200W Prema's vbios instead EUROCOM std on at 150W.

150W minFPS 37
200W minFPS 47

 

Yeah so the performance is now normal up until your GPU hits around those 79-80c, then the performance drops around 10% lower. Wished you had made a longer video, but still, thanks for the update man! appreciate it

Are you really that desperate? What are you trying to prove? I literally gave you example as to what I mean and now you suddenly imply that intel is actually selling unstable CPU's to the public. I think you have a really poor undertanding of how the silicon quality actually affects the CPU.

If you buy a 7700K some have better silicon quality others have worse silicon quality, however all 7700K's have good silicon quality, otherwise intel wouldn't brand them as a 7700K. Every 7700K is overclockable and the differences in temp and clocks aren't that huge. Most of the temp issues are actually caused by terrible mounting, thermal compound, board not wanting to OC properly etc. Most people who struggle to get a 7700K on 5ghz are people with weak cooling who somehow want a kably lake to run 5ghz on a very mediocre cooling. Hence, silicon quality doesn't matter, because intel already sorted the cpu's with sufficient silicon quality to the unlocked CPU's while keeping the low quality ones to the low end CPU's.

 

Yo!
This 5min video is 1,97GB (and I've uploaded at 120KB/s so past 4hrs), on OBS if I choose NVEC hw decoder instead the x264 cpu decoder, it sucks 10W/25W (I dont get the percentage) from gpu power, lowering my FPS a bit. (next time I'll instal geforce bloatware shadowplay)
Going to re flash stock EUROCOM 150W vbios on my gtx 1080 because I like to have better temps!
Meantime.. no one orderered one Clevos with cooffelake cpus?

 

@aaronne
This might be the gt73vr video (4:07 for Crysis 3):

 

yes thanks, that is the video, I saw it but I lost then!
but it dont mention nor show some things (cpu clock, driver ver, no screens of graphical settings by games) so I still cannot compare at best, but again thanks

 

You could ask the uploader, leave a comment on YT and maybe he will reply in useful time.

 

If you want to be 0.1% better in benchmarks results in some OC competition, then silicon matters.

 

Well said. Nothing else has to be added.

 

That only applies to the masses that are not really serious about overclocking and benching. Silicon quality matters a lot when it limits your ability to compete because you drew a short straw that can't be overclocked as far as you need it to in order to be competitive. There are not many gamers or even power users that have a clear understanding of this simply because the concept is foreign to them and it doesn't apply to their hobby. But, that is OK because they don't really need to understand it. Likewise, I don't need to understand when two strangers in a public place are having a private conversation in Chinese.

 

And better silicon can't gain you more than 0.1% better performance in benchmarks? Dang What's the point with Silicon lottery? As all silicon is almost equal LOOL
+rep
@ole!!! Doesn't matter what silicon you get. All clocks almost equal. Save cash!!

 

Plz dont take this as offensive. I respect that everybody has thier own hobby OC-ing or get crazy about OC .
I mean that times when i searched best CPU/GPU stepping/series for max OC long gone.
When I realise that I have same fun playing, for example: Witcher 3 with avarage 40-50fps G-sync as 50-60fps with "mega ultra OC 67xxxGhz CPU".
Temps are normal (~65C) and fan noise is much lower (e.g.my signature)

When someone reaches the ceiling with extreme OC then this "0,1%" better in bench matters and u can find solusion in "silicon" .

 

Better silicon can mean lower temps and fan noise for the average user.

 

No offense taken. I totally get what you were saying. I just wanted to clarify what you had to say was accurate in the context of the hobby of playing PC games and not applicable when a person is serious about their overclocked benching hobby. There are many impressionable people in the community with limited knowledge, so my response was for the sake of clarity.

As @Meaker@Sager also pointed out, better silicon also means better thermals. On today's laptops that is critical. They make them run faster that they should for under-engineered thin and lightweight machines, and all it takes is 1°C too much to mean the difference between normal function and thermal throttling, which ruins everything no matter what one's hobby (or business) happens to be.

 

u sound pretty desperate trying to prove people wrong!

needless to say, i want every last drop of my dollar and performance. can't stand crappy 4ghz mobile bga any longer, need 5ghz. needless to say you know nothing, because unlike many others i care about the smallest performance gain but i dont benchmark

 

I guess people have never used their system for productivity tasks, otherwise they just might realize the actual difference between rendering/photo-editing @ 4Ghz vs 4.8-5Ghz.
But hey, who am i to talk, as already confirmed by multiple (2) people, silicon doesn't matter.

 

You talk about productivity tasks. I expect computer users want finish the task fastest possible. 54% slower ain't nice and will also bring less cash in your or the employers pocket.

 

[Admins, please let this live on this thread, it is freaking relevant!]

 

Oh wow, really? Two people? Well, I guess that settles it, LOL On a more serious note, there are, at minimum, 120,000 we can identify that disagree. But, what the heck do we know. The two that disagree are legends in their own minds.

 

Mr. Fox, you're slacking on your achievements and challenges! JK.

 

I will try harder going forward.

 

The desktop systems having that 6 core at the full TDP is currently going to crush anything mobile.

I've had a couple of the quad core ULV machines come through my hands and they are very decent for office use but they wont maintain a boost clock for long.

 

Let's be honest. There is literally nothing you want to be done with a ULV.

I usually have at least 10 explorer tabs, 10 chrome tabs to 50 chrome tabs, at least foobar2000, Photoshop, and MS office running, all at the ame time. Just for my basic tasks and just because my work requires me to.

I can't imagine how doing this on an ULV would feel like. Even if we take out PS, and leave only typical office, ULV does so poorly when it comes to anything multitasking.

 

Unless those tabs are active then switching between lots of tabs is exactly what an ULV can do as each switch itself requires short bursts of activity.

 

Active lel.

I think that most people are far calmer than I am when it comes to aggressive switching. Kinda sad that I will never be able to work with an inexepensive device so we..

Me wanted to get a Me tablet. Me thought about Asus something something because it has nice display. Me gave up when me found out that it lags and automatically terminates chrome in my typical usage. Me is considering Huawei Mediatab Young something something. Me remembers that this is not a tablet forum but a lappy forum. Me stops posting this.

 

you're gonna love optane SSDs, so far no worthy optane m.2 model though. 5ghz is a must its so much snappier than 4.2-4.3 ghz

 

The question is...
What can we expect with proper oc’d 8700K in P870TM1? Have you seen results or touched the new Clevo’s ?

 

lmao i tried to grab the only 5.3ghz 8700k there was and it was sold out while i was typing in password on my paypal LOL, then i instantly switch to 5.2ghz on SL then on checkout page also sold out. i refresh SL and then all sold out :/

well now, time for 8core CFL-S

 

Or wait for...
8th generation CPU family

 

Nope, those are just the support screws to even-out the heatsinks pressure to the LID, though only one of them is actually in use in the current iteration, so not sure what the original plan was. The CPU heatsink is now also unified with the CPU FAN.

I think modding original DM3 GRID to fit the TM1 CPU cooler would result in the best cooling solution.
They again cheaped out on the GRID and went with the FINs with reduced height like on the later revisions of the KM1 GRID.
Also the reduced heatpipes for the Slave GPU don't look sufficient for systems with no throttle firmware...

 

I mean I mentioned the reduced heatpipes for the Slave GPU before as a downgrade. We will have to see how this will fare.

The second small FINs/grill for Cpu... Is it in even high with the cpu bracket? And is there a possibility for Mod and place a 0.5/10 mm thin fan on top of the new cpu bracket for push air into the second small FINs ?

Hope you understand my Q. I have posted about this possible mod before.

 

It's the same height as the FAN, but what we can't see on those picture is that now the main FIN area is a lot wider and reaches deep into the FAN unit itself.
CPU cooler is looking great!

 

Thanks. Need more pictures to see if and what type Mod that can be possible.

 

wait for the review yo, once review comes we'll all see it

 

Though the heatpipes in this case are not doing the heavy lifting of heat, that's the vapor chamber.

 

The vapor is awesome but mostly taking care of the core, while almost all RMAed Pascal as have burned MOSFETs...