Clevo Overclocker's Lounge

Nvidia 388.31, gskill running at 2800mhz, windows 10 1703 and stock bios. And yes all settings are the same.

I did the benchmark last week before I got the premabios. I will try again today once I get home from work.

For the meantime try to apply nvidia profile inspector for deux and also take note that I disabled all unnecessary services, antivirus in my os.


Sent from my iPhone using Tapatalk

 

Thank you for sharing I will try that Nvidia inspector soon. I am using the one just above that driver you are using, its 388.59 on this new video below:

Thank you for sharing. This video shows two benchmarks of DirectX 12 in 1080P and 1440P on stock speeds. This video has exclusive full screen enabled. I will do more Deus Ex: MD videos soon. With DirectX 11 on 1080P & 1440P with and without overclock.

Deus Ex: Mankind divided
2x Benchmarks
P870KMx-G Prema Mod BIOS
7700K @ 4.5GHz
2x 1080 GTX SLI Stock speeds and voltages
64GB 2666MHz RAM
Ambient room temp at 27c
1440P & 1080P benchmarks using DirectX 12
Nvidia Driver: 388.59
Exclusive full screen

 

http://forum.notebookreview.com/threads/clevo-inferior-gpus.809455/
You can use as reference^

 

Cheers pal. That is a good score. I just updated to the 388.59 drivers earlier. And now I can not run Deus Ex: MD in DirectX 11... weird.

Edit: a system restart solved this problem.

 

I have tried all sorts to achieve your FPS score at 1440P. I'm thinking if your score is from a 1080P benchmark because I am achieving the same as you in 1080P DirectX 12. Did you use Nvidia inspector then? If so please share your nvidia inspector profile that you are using. These are some tests I have done. The first 4 did not have Nvidia Experience recording or working in the background. I thought this might change the final result.

Deus Ex: MD benchmarks:
1440P Exclusive full screen with everything maxed out, except for MSAA, which is off. Stock speeds on 7700K and stock speeds on 1080 GTX SLI unless stated otherwise. Using Nvidia drivers 388.59

DirectX 11 (Stock speeds)
Average: 79.6 FPS
Minimum: 50.0 FPS
Maximum: 105.3 FPS

DirectX 11 (7700K@5GHz overclock, no GPU overclock)
Average: 79.9 FPS
Minimum: 52.1 FPS
Maximum: 105.1 FPS

DirectX 12 (Stock speeds)
Average: 74.6 FPS
Minimum: 55.6 FPS
Maximum: 111.7 FPS

DirectX 12 (7700K@5GHz, no GPU overclock)
Average: 75.0 FPS
Minimum: 54.7 FPS
Maximum: 116.0 FPS

DirectX 12 (7700K@5GHz 1080 GTX SLI at 2100MHz overclocked with borderless option on and exclusive full screen off at 1440P)
Average: 75.2 FPS
Minimum: 53.9 FPS
Maximum: 93.8 FPS

DirectX 11 (Stock speeds on 7680x1440)
Average: 36.6 FPS
Minimum: 28.3 FPS
Maximum: 43.3 FPS

DirectX 12 (Stock speeds on 1080P)
Average: 98.5 FPS
Minimum: 80.6 FPS
Maximum: 120.2 FPS

You were 100% right. Bordless option lowers performance. Thank you for informing me +rep.

DirectX 12 (7700K@5GHz 1080 GTX SLI at 2100MHz overclocked with borderless option on and exclusive full screen off)
Average: 75.2 FPS
Minimum: 53.9 FPS
Maximum: 93.8 FPS

 

Interesting results.

In this video (DX12 exclusive fullscreen) you're getting about the same performance as the previous video (DX11 borderless), but GPU utilization is much lower for the most part. Especially noticeable on the main menu, and throughout most of the benchmark.

However, during the one brief period when you were getting full GPU usage in DX12 at the very beginning of the 1440p benchmark, the performance was much higher than in DX11.

Aside from the GPU usage drops, the fact that performance barely changed when resolution was lowered from 1440p to 1080p, tells me SLI is not working well if at all in this game. I would also imagine that overclocking the GPUs would not really increase performance either as long as SLI scaling remains problematic.

One question. Is 1440p the native resolution, or is it a 1080p screen but you're using DSR?

You should run the SLI benchmarks again with GPU bus load on the OSD, so we can monitor if there is any possible PCIe bandwidth limitation being hit. Furthermore, try disabling Temporal AA as this is a known culprit of PCIe bus saturation in SLI. Lastly, benchmarks with SLI disabled would be very useful to establish a baseline comparison for what kind of SLI scaling you're actually getting.

Side note, near the end of the benchmark sequence, there's a lot of shadow flickering on the wall in the room with the man and woman and candles. I don't get that in single GPU, so it seems to be an SLI related visual glitch.

 

I think Deus Ex; MD is very poorly optimized. I have been on this, all day testing lol. I think I have had enough. I'm back on driver: 378.49 now.
I do not use DSR. The benchmarks listed have been done on an external 1440P screen, even when I run it on 1080P. These are the two resolution on DirectX 12:

DirectX 12 (Stock speeds on 1080P)
Average: 98.5 FPS
Minimum: 80.6 FPS
Maximum: 120.2 FPS

VS

DirectX 12 (Stock speeds on 1440P)
Average: 74.6 FPS
Minimum: 55.6 FPS
Maximum: 111.7 FPS

 

Sorry I didn't elaborate more. For vrm cooling, do you just thermal adhesive on some mini copper ram sinks? What kind of adhesive should I use?

Sent from my Moto G (5) Plus using Tapatalk

 

Hehe, such is the SLI life.

BTW was G-Sync active on that external 1440p? Reason I ask is that G-Sync has additional PCIe bus overhead in SLI, and since you're only in x8/x8 those bandwidth consumers do add up.

The chokes and FETs should already have thermal pads attached to a copper plate which is passively cooled by an air opening on the bottom cover. At least that's how it is on my P650SG.

 

No, mine are just barely with a vent over them

Sent from my Moto G (5) Plus using Tapatalk

 

G-sync was turned off. I always turn G-sync off while benchmarking or testing. G-sync needed to be off anyways, as any driver above 378.91 causes frame drop or stuttering when connected to an external G-sync screen.
Here's a thread about this problem: http://forum.notebookreview.com/thr...-0-freeze-stutter-temporary-fix.805220/page-2

 

Yeah in that case thermal adhesive and copper sinks would be a good idea.

Strange that those components are left uncooled in yours, but they're cooled on mine. I guess they figured 970M vs. 980M didn't need it or something.

 

Ah. I was told that problem was fixed as of the last few recent drivers, but it seems there is some conflicting information.

 

Alright any thermal adhesive suggestions? Or should I try to piggy back something on to the gpu heatsink?

Sent from my Moto G (5) Plus using Tapatalk

 

Maybe fixed using a certain method. The latest driver that I used today was 388.59 and it still had the G-sync freeze on Deus Ex: MD

 

Driver problems seem so abundant especially with G-sync!

 

Spoiler: Older Clevo 1080 vBIOS is Better

With the older vBIOS from the DM3...

https://www.3dmark.com/3dm11/12541378

https://www.3dmark.com/3dm11/12541406

https://www.3dmark.com/sd/4919940

https://www.3dmark.com/fs/14391170

https://www.3dmark.com/fs/14391232

https://www.3dmark.com/fs/14391264

 

BUT, BUT, BUT the latest is always the greatest.

 

I found the newer vBIOS to set the name nominal clock as the older 86.04.56.00.2D. Some software misreports lower boost clocks (Nvidia Inspector), but there is really no change. I did find the older vBIOS has its throttle points offset around 5C higher, so it runs the core 1 13 Mhz notch faster around half the time.

 

You really do need to graph average clocks over a benchmark to be sure what kind of clocks you are getting these days. The average clock over the run should be the guide.

 

I use the benchmark score as the guide. Higher scores win every time.

 

Well a driver may improve per clock performance but other issues may lower your clock. I guess there is no magic bullet for all issues because frame timing is a thing too.

 

Has anyone else ever wondered why 3DMark, being a GPU benchmark, doesn't have a GPU clock graph but only a CPU clock graph?

 

Generally it's a little tricker to code I think die to all the brands out there.

 

Well freeware like GPU-Z does it like forever without even requiring an update for new GPUs to do just that: Simply reading the GPU core frequency.

 

I think the reason is they do not want people to see how spastic the GPU behavior is. I run OSD all the time and see it in real-time. Not pretty. I love that I have K-Boost still working with EVGA Precision OCX on my 1080 Ti. It sucks that this feature is not available for mobile GPUs. I wish there was a hack or tweak available to force it to work.

 

100% simply this.
People like consistency. Seeing numbers jump around can cause even the most brain dead person to ask what's going on?
If people actually saw 'POWER LIMIT' and "throttling" and their brand new GPU's not running as fast as they are supposed to and the FPS drops corresponding with those cancer restrictions, then those sheep would be smart enough to start questioning stuff. And maybe actually might go buy a console or other abortion that at least gives you performance you would expect that is consistent on all systems (Because, LCD), rather than give Ngreedia more money.

 

And, for less than the cost of a good GPU if playing games is all they are after.

 

Nvidia is part of the <team>. They pay for Futuremark's software. The money team Or maybe hidden shareholders?

 

And here naive me was thinking not monitoring the GPU clocks was maybe to avoid a score loss due to the overhead...

 

Brother Prema, that's for the learned and the wisened like us here. Not for the sheep out there!

 

You can also K-Boost by using the EPX default skin in MSI Afterburner:



Or without tools, by modifying the Registry

 

Nice tweak. +5 Rep.

That worked, but you have to use the skin from an old version of EVGA Precision X (pre-Pascal). The Precision OCX skin for Pascal will not work. I have attached the file that works so nobody has to install an old version of Precision X on a Pascal system. > LINK<

 

Ah yes, sorry I should've clarified that I was referring to the default skin for Precision X 16, not the skin for XOC.

 

No worries. Using EVGA Precision X adds those registry entries when K-Boost is enabled and removes them when K-Boost is disabled.

 

All Kboost does is lock the clocks at the set boost clocks as if you were on the curve, except they're locked at the default 3D clocks for the offset (e.g. +200 mhz I think would be 1860 mhz on a GTX 1070).
It's basically no different than locking the clocks with control L on the graph at a certain point, except that point is now at the boost clocks default speed. At least that's what it did when i tested Kboost. e.g.
default clocks are 1443 mhz with boost clocks at 1643 mhz. If you enabled Kboost here, it would start at 1643 mhz in 2D desktop at all times, rather than having to find the 1643 mhz voltage step on the graph manually.

If I'm wrong, tell me. But it just seems to be an automated Control L.

 

Yup, it's the same tweak, K-Boost just automates the process.

 

Yes, pretty much. Simply stated, it forces the GPU to run at full 3D boost clocks instead of the typical dynamic behavior crap. It does nothing (sadly) to eliminate the spastic Pascal room temperature throttling issues. It will still slow down as the temperatures rise toward normal.

 

Uplay is giving Assassin's Creed IV: Black Flag away for free this week (posted on Pcworld.com DEC 11, 2017)

http://hwbot.org/submission/3732779_papusan_wprime___32m_core_i7_7700k_4sec_8ms?recalculate=true

 

Here's something just for fun. Only trouble is, now I am wide awake and need to go night-night.



Got to love those 58-60°C CPU and 38°C GPU temps. Especially nice considering it is running a rock solid 5.2GHz with K-Boost enabled. Even the video rendering of the final product only maxed out the CPU at around 71°C.

 

Be prepared for new possible screw up from Futuremark... 3DMark v2.4.4163 (December 12, 2017).

Spoiler: ChangeLog

https://www.futuremark.com/support/update/3dmark
What's new in 3DMark 2.4.4163
This is a minor update. Benchmark scores are not affected.

Improved

• Benchmark loading screen logos and labels are now consistent across all tests. You will need to update your DLC files, but this is purely a cosmetic change. Benchmark scores are not affected.
• Use the new "Update All" button to update all DLC files to the latest version.

Fixed

• Fixed an issue with the Vulkan part of the API Overhead feature test that was caused by a change in the Vulkan specification.
• Custom Run looping now works as intended again.
• Restored the missing sub-scores on the Fire Strike result screen.
• Fixed a rare issue that could cause Fire Strike, Cloud Gate, and Ice Storm tests to fail on a few specific Intel processors when using integrated graphics.
• Fixed an issue that could cause 3DMark to hang on the splash screen.

All older 3DMark versions can be downloaded from techpowerup.com (if you deleted or didn't backup your prevous or current 3DMark Suite version).

 

Lol, welcome to the dark side Mr. Fox, reminds me of my water cooled titan-x

 

Check out Ted running a benchmark!

http://www.3dmark.com/fs/14404725

 

Strange... The name says Johnksss

 

Nah, that was him on my machine. . He can't submit it with my license number.


Edit:
But here goes a few New World Records with stock bios and stock vbios Though.

 

Hey Brother @Ted@HIDevolution - you look good behind that keyboard, bro.

 

@Ted@HIDevolution looks like someone I would be happy to be friends with. Sigh. he reminds me of doublelift

 

Congrats on the new EVOC monsterbook, Brother John. You did get SLI, right?

 

Congrats on new monsta

From Ted/Johnksss first run https://www.3dmark.com/fs/14404725

 

Thanks!

Yep. Just individual card testing at the moment.

@Papusan

Thanks sir!