[BENCHMARKS] The OFFICIAL Thread

[coolane 69]

2 hours ago, D2ultima said:

I've never seen someone manage to lock voltage on a skylake chip. Manual/Static voltage doesn't work. The chips do what they want, though Manual voltage and offsets do have some kind of effect.

Yeah, seems like only desktop motherboard can do that. In laptop the closest thing to have the voltage fluctuates less I guess is increasing the PLL value. 

[D2ultima 324]

2 minutes ago, coolane said:

Yeah, seems like only desktop motherboard can do that. In laptop the closest thing to have the voltage fluctuates less I guess is increasing the PLL value. 

I was including desktops in my statement of "I've never seen a skylake chip lock voltage".

[coolane 69]

6 minutes ago, D2ultima said:

I was including desktops in my statement of "I've never seen a skylake chip lock voltage".

Though desktops can't lock the voltage, but at least it can get the value closer to what BIOS set, right?

The voltage in GPUZ of the last 3rd picture is pretty close with the voltage set in BIOS:

https://rog.asus.com/19262015/overclocking/guide-overclocking-core-i7-6700k-on-the-maximus-viii-extreme/

[D2ultima 324]

1 minute ago, coolane said:

Though desktops can't lock the voltage, but at least it can get the value closer to what BIOS set, right?

The voltage in GPUZ of the last 3rd picture is pretty close with the voltage set in BIOS:

https://rog.asus.com/19262015/overclocking/guide-overclocking-core-i7-6700k-on-the-maximus-viii-extreme/

Not really. It's the same deal, more or less. Didn't the other user say he put it at 1.235v or something and it sits around 1.267 to 1.295 depending on what's done? That's how the desktops act too. The skylake chip does what it wants, you can kind of just influence it a little bit. If you overvolt a lot, it'll probably stick. But since the point is to use as little voltage as possible here, the lower you try to set it, the more variance should show up.

[coolane 69]

2 minutes ago, D2ultima said:

Not really. It's the same deal, more or less. Didn't the other user say he put it at 1.235v or something and it sits around 1.267 to 1.295 depending on what's done? That's how the desktops act too. The skylake chip does what it wants, you can kind of just influence it a little bit. If you overvolt a lot, it'll probably stick. But since the point is to use as little voltage as possible here, the lower you try to set it, the more variance should show up.

Haven't touched desktop board for a long time, now don't know how the voltage works there any more. My last time overclocking a desktop chip it was the age of Pentium 4 and AMD Athlon 3000+. 

It would be cool if AMD can drop the 95W Zen in a Clevo laptop. Rumor saying that it has similar performance with 6900k clock to clockwise.

[D2ultima 324]

1 minute ago, coolane said:

Haven't touched desktop board for a long time, now don't know how the voltage works there any more. My last time overclocking a desktop chip it was the age of Pentium 4 and AMD Athlon 3000+. 

It would be cool if AMD can drop the 95W Zen in a Clevo laptop. Rumor saying that it has similar performance with 6900k clock to clockwise.

Said rumor was using Blender which was using GPU acceleration from the APU on it, is what I heard. So far it looks like AMD is playing the same game nVidia and Intel play, and not well enough.

 

I don't like that. If AMD is going to be the same kind of garbage company, then even more reason to simply buy the stronger parts.

[johnksss 720]

It's because on a laptop you can not control the load line calibration setting.

 

On most desktop boards you can control this thus locking the voltage in place.

[Guest]

New single 980M score! http://www.3dmark.com/3dm11/11583651

 

Beaten only by Brother [USER=128218]@[email protected][/USER]  with the P570WM with Xeon Physics, but the baby beast matched and beat the graphics score by a very tiny 0.3% margin... http://www.3dmark.com/compare/3dm11/11407903/3dm11/11583651

 

 

And, how 'bout that physics killing Windows 10 turd OS? 10% hit on CPU performance with that crappy new OS, LOL.

http://www.3dmark.com/compare/3dm11/11583651/3dm11/11568175

[johnksss 720]

There you go speed racer.

http://www.3dmark.com/3dm11/11456296

Edited September 17, 2016 by johnksss

[Guest]

Wait a minute...so @Mr. Fox has beaten @johnksss highest GTX980M 3DM11 GPU score?:

 

http://www.3dmark.com/compare/3dm11/11456296/3dm11/11568175

 

At least for me these scores are hard to search for, so not sure if John (or someone else, LOL) still has a higher one somewhere...

But looks like the crown-prince has taken his rightful place next to the King.  

 

If John still has his P570WMx, then this may turn into an interesting weekend...

[Guest]

15 minutes ago, Prema said:

Wait a minute...so @Mr. Fox has beaten @johnksss highest GTX980M 3DM11 GPU score?:

 

http://www.3dmark.com/compare/3dm11/11456296/3dm11/11568175

 

At least for me these scores are hard to search for, so not sure if John (or someone else, LOL) still has a higher one somewhere...

They are hard to search for because there are so many scores in the 3DMark database. Clicking the link to "fetch the next 1000 results" for single GPU scores frequently does not work correctly. But, the top one or two scores usually are revealed, then you have to manually select a CPU to display any additional search results.

 

As we can see, Brother @johnksss had another one that wasn't visible in the search results that came up for me. It's pretty rare to beat him at anything running the same hardware. But, it is still a superior single GPU 3DMark 11 and a great accomplishment on the graphics score for a 15 inch laptop sporting an extremely unimpressive, if not downright pathetic, cooling system and a blast furnace CPU. 

[johnksss 720]

At the end of the day....After the numbers are crunched....They are still the same. A half of a fps does not really settle things.

[Guest]

4 hours ago, johnksss said:

At the end of the day....After the numbers are crunched....They are still the same. A half of a fps does not really settle things.

You're right. And, it's all for fun anyway. I'm just excited to finally, after all this time, have one out of six 980M cards that actually works well and performs as nicely as yours. And, even that would not be possible without the modding talent of Brother @Khenglish. This same GPU that runs like a bat out of hell was a $700 piece of shit before he made things right on the hardware side. It is sad that it took two years and multiple GPUs to end up with just one good enough to place me at or near the top in 980M benchmark scores. But, it's nice to have a happy ending to 980M.

[johnksss 720]

Yep. This is true. You certainly did have a bad luck run on gpus, but it's always nice to actually get some good ones from time to time.

 

Side Note:

That score was hidden because it's over all score was lower.

Edited September 17, 2016 by johnksss

[iunlock 91]

Still tuning, but here are some benches so far:


http://www.3dmark.com/fs/10274130

 

http://www.3dmark.com/spy/496947




 

 

Edited September 25, 2016 by iunlock

[Guest]

Here are some new single 980M Fire Strike Extreme and Ultra benchmarks. 

 

http://hwbot.org/submission/3326152_ | http://www.3dmark.com/fs/10292693

 

http://hwbot.org/submission/3326137_ | http://www.3dmark.com/fs/10292537

 

[Guest]

Just like the good old days, benching the M18xR1 and M18xR2 on hotel room AC with Brother @johnksss on the phone teaching me all the ropes back when I was a noob.

 

Those were some good times. Hotel AC cooling works just as well for Clevos, LOL.

 

[iunlock 91]

Thanks @Prema, for the Team T|i hwbot invite.

 

I've officially joined. Submitting some scores now...

 

Update: I'm at 3rd Overall for Fire Strike and Time Spy. Please help me beat my PR... Just need a few more points to bump up to 2nd.

Edited September 27, 2016 by iunlock

[D2ultima 324]

5 hours ago, iunlock said:

Thanks @Prema, for the Team T|i hwbot invite.

 

I've officially joined. Submitting some scores now...

 

Update: I'm at 3rd Overall for Fire Strike and Time Spy. Please help me beat my PR... Just need a few more points to bump up to 2nd.

I've never touched a Pascal card or an unlocked Skylake, but I want to help. I wonder if I can do eet xD

[coolane 69]

Found out I got into the top 5 ranks with 970M. If excluding the first 4 which are using egpu, then I got the top place.

My new daily drive after some tuning:

Core: 1330 MHz; Memory: 6400 MHz

http://www.3dmark.com/fs/10329610

Single card for fun:

http://www.3dmark.com/3dm/15157037?

[iunlock 91]

Finally beat my personal record.

3rd place overall for 1x GTX1080 (Notebook).

1st place for 775 Class.





Another one down in the books for Team T|i

---

[Khenglish 1155]

P150EM direct dGPU output progress continues. 980 is still stuck in pstate 5 (3D battery power state), but with some mods performance from P5 is now acceptable to wait for Prema to do BIOS mod magic to fix standard 3D and boost power states.

 

Voltage hardmod done boosts voltage by 97mV over the pitiful .825V P5 voltage:

 

875/7012 3dm11: http://www.3dmark.com/3dm11/11635126

 

Stable prior to the voltage mod was 760 MHz. P5 just does not clock well at all, even considering the low voltage. This should have gotten me over 1GHz stable.

 

I reflashed to an optimus BIOS for a performance comparison test clock for clock at something P5 can do for a optimus vs direct comparison.

 

760/7012 direct 3dm11 gets 12090 GPU score: http://www.3dmark.com/3dm11/11635090

760/7012 optimus 3dm11 gets 11459 GPU score: http://www.3dmark.com/3dm11/11635075

 

So at a measly 760 MHz core, optimus is an uninsignificant 5.2% performance loss. I did a quick Unigine comparison and got 3.3% loss there. These losses also include the fact that P5 runs the PCI-E at 1.1 speeds, quartering the bandwidth. All optimus tests were done with an underclocked P0 running at full 3.0 speeds.

 

UPDATE:

 

It turns out my bad heatsink mount was to blame for the poor core clocks. Here's 1001 MHz core:

 

http://www.3dmark.com/3dm11/11635270

 

Also I noticed that 3dm11 seems to try to lock fps to roughly 10 Hz intervals when over 60fps. Ex I often see 60.2 fps reported, 90.1 flickering with 90.9, 100.0, and 80.0. This seems odd and I am wondering if this is normal for direct output.

Edited October 6, 2016 by Khenglish

[Papusan 167]

New Wprime 1024M score - 132.237 sec

http://hwbot.org/submission/3333272_papusan_wprime___1024m_core_i7_6700k_2min_12sec_237ms

 

[Guest]

57 minutes ago, Papusan said:

New Wprime 1024M score - 132.237 sec

http://hwbot.org/submission/3333272_papusan_wprime___1024m_core_i7_6700k_2min_12sec_237ms

 

Nice job, Brother @Papusan

[johnksss 720]

Decided to go ahead and bench the 6820HK/1070N. It would seem this little beast knocked out all the competition in it's class at the moment. Being BGA and all.

http://hwbot.org/hardware/videocard/geforce_gtx_1070_notebook_mxm/

Edited October 7, 2016 by johnksss