[BENCHMARKS] The OFFICIAL Thread

[D2ultima]

Just now, Mr. Fox said:

No, that won't do it. It needs more voltage to go past 1500ish on core. Around 1535 or so the driver crashes (TDR) about 5 to 10 seconds into a benchmark due to lack of core voltage when the GPU is still around 30°C. It's not having black-screen issues or lockups or anything like that.

Hmm... maybe it does. Maybe the lack of voltage only happens under LN2 temps. But okay. That's too bad though.

 

Question, does it happen with memory at stock as well?

[Guest]

21 minutes ago, D2ultima said:

Hmm... maybe it does. Maybe the lack of voltage only happens under LN2 temps. But okay. That's too bad though.

 

Question, does it happen with memory at stock as well?

Yes, no change in behavior based on memory clocks. The memory does not affect it. The voltage adjustment is also for core only. And, it takes the full 1.250V for 1500. If I drop it to say 1.225V the core max is around 1475. It's not a bad thing though. Nothing wrong with using more voltage as long as you keep it cool. Same applies for CPUs. You use what it needs, nothing more, nothing less.

 

The 4790K pulled 245W in the physics test in that 4.9GHz Fire Strike run, LOL. Almost exactly same power demand as the 4930MX CPU.

 

All of this fun is for AC cooled benching only. Trying to do this on air could potentially kill the GPU and certainly isn't "helpful" to the CPU.

 

The only sad part is the sketchy cooling system these Batman notebooks have. I see the new Pascal version of the same is having some thermal management issues as well. A unified heat sink is not a good design. They should have made the P7XX series about 0.25 inch thicker, with bigger radiators, bigger fans and discrete CPU/GPU heat sinks. That would have been much better.

[wersuss]

6 hours ago, Mr. Fox said:

Yes, no change in behavior based on memory clocks. The memory does not affect it. The voltage adjustment is also for core only. And, it takes the full 1.250V for 1500. If I drop it to say 1.225V the core max is around 1475. It's not a bad thing though. Nothing wrong with using more voltage as long as you keep it cool. Same applies for CPUs. You use what it needs, nothing more, nothing less.

 

The 4790K pulled 245W in the physics test in that 4.9GHz Fire Strike run, LOL. Almost exactly same power demand as the 4930MX CPU.

 

All of this fun is for AC cooled benching only. Trying to do this on air could potentially kill the GPU and certainly isn't "helpful" to the CPU.

 

The only sad part is the sketchy cooling system these Batman notebooks have. I see the new Pascal version of the same is having some thermal management issues as well. A unified heat sink is not a good design. They should have made the P7XX series about 0.25 inch thicker, with bigger radiators, bigger fans and discrete CPU/GPU heat sinks. That would have been much better.

So P775dm3 cooling is not perfect aswell, right? How is P870DM3 cooling, is it better? Or maybe p670 is best since heatsinks are not united?

Edited September 9, 2016 by wersuss

[Guest]

5 hours ago, wersuss said:

So P775dm3 cooling is not perfect aswell, right? How is P870DM3 cooling, is it better? Or maybe p670 is best since heatsinks are not united?

I cannot say anything about systems that I don't have available for my own testing except I don't know. P870DM cooling was superb. When I can secure a P870DM3, then I will be able to comment. If the P670 has a BGA CPU, it cannot be viewed as an acceptable product even if it has a good cooling system.

[wersuss]

8 hours ago, Mr. Fox said:

I cannot say anything about systems that I don't have available for my own testing except I don't know. P870DM cooling was superb. When I can secure a P870DM3, then I will be able to comment. If the P670 has a BGA CPU, it cannot be viewed as an acceptable product even if it has a good cooling system. 

 

But P775dm3 cooling is not good enough?  Because i'm planning to buy p775 model but it's late.

[Guest]

 

22 hours ago, Mr. Fox said:

Looks like 1500 or so on GPU core is my max with a 1.250V limit (driver crashes due to lack of voltage). @Prema I think it will take 1.275V or 1.300V to hit 1600 on core.

 

OK,  put your benches where your mouth is big boy!

[Guest]

1 hour ago, Prema said:

 

 

OK,  put your benches where your mouth is big boy!

I'll be testing 1.275V and 1.300V this weekend to see how that works out.

Wish me luck.

5 hours ago, wersuss said:

 

But P775dm3 cooling is not good enough?  Because i'm planning to buy p775 model but it's late.

I'm not sure, so I cannot give you an honest answer. It may be amazing or terrible, but I have no idea since I do not have one. The cooling system on my P750ZM is OK with the CPU running stock, but not very good with CPU overclocking. Even using AC cooling is just barely adequate for the CPU with a substantial overclock.

[Guest]

2000Mhz, DA FUCK?!:

http://www.3dmark.com/fs/10075267

@Johnksss
@Mr. Fox

@Khenglish

[wersuss]

13 hours ago, Mr. Fox said:

I'll be testing 1.275V and 1.300V this weekend to see how that works out.

Wish me luck.

I'm not sure, so I cannot give you an honest answer. It may be amazing or terrible, but I have no idea since I do not have one. The cooling system on my P750ZM is OK with the CPU running stock, but not very good with CPU overclocking. Even using AC cooling is just barely adequate for the CPU with a substantial overclock.

Thanks. Isn't p750 p775 cooling same?

Edited September 10, 2016 by wersuss

[johnksss]

5 hours ago, Prema said:

2000Mhz, DA FUCK?!:

http://www.3dmark.com/fs/10075267

@Johnksss
@Mr. Fox

@Khenglish

It is reading wrong.....

He has a 1070 also.

[Khenglish]

5 hours ago, Prema said:

2000Mhz, DA FUCK?!:

http://www.3dmark.com/fs/10075267

@Johnksss
@Mr. Fox

@Khenglish

 

The system also has a 1070, and that score is in line with a stock 1070. Maybe the 1070's clocks were read as 980m clocks? A 1070 at 2ghz should be significantly higher than 17600 though, and 17600 is what I would expect to see from a 2ghz 980m.

 

I can't figure this out. Its not a BGA swap to a desktop card or else the memory would be 4GB. It could be an mxm to pci-e adapter with LN2 pot, but that would mean someone made an LN2 pot just for mxm cards on an adapter. Also in that case this score is not uploaded to hwbot, which an LN2 iser would definitely do.

 

I'm leaning towards a 980m not actually being in the system and systeminfo just being very wrong.

[johnksss]

The score is definitely a 1070. 100% sure of that.

100% sure his 980M is not 2k/2k

And the machine in question is a 

https://www.google.com/search?q=Terrans+Force+X799&rlz=1C1CHZL_enUS707US707&source=lnms&tbm=isch&sa=X&ved=0ahUKEwiZtvaLloXPAhUQzGMKHbSsAmYQ_AUICigD&biw=2560&bih=1274#imgrc=qmeM6AgjWzp0lM%3A

So maybe a TB2/TB3 1070 is what he used. 

 

Their website is too damn slow to load. Seems they have a 1070 model as well, still...not sure how the 980M is in there unless they used maybe mini pcie to 1070

Edited September 10, 2016 by johnksss

[Guest]

6 hours ago, Prema said:

2000Mhz, DA FUCK?!:

http://www.3dmark.com/fs/10075267

@Johnksss
@Mr. Fox

@Khenglish

 

40 minutes ago, johnksss said:

It is reading wrong.....

He has a 1070 also.

 

35 minutes ago, Khenglish said:

 

The system also has a 1070, and that score is in line with a stock 1070. Maybe the 1070's clocks were read as 980m clocks? A 1070 at 2ghz should be significantly higher than 17600 though, and 17600 is what I would expect to see from a 2ghz 980m.

 

I can't figure this out. Its not a BGA swap to a desktop card or else the memory would be 4GB. It could be an mxm to pci-e adapter with LN2 pot, but that would mean someone made an LN2 pot just for mxm cards on an adapter. Also in that case this score is not uploaded to hwbot, which an LN2 iser would definitely do.

 

I'm leaning towards a 980m not actually being in the system and systeminfo just being very wrong.

 

30 minutes ago, johnksss said:

The score is definitely a 1070. 100% sure of that.

100% sure his 980M is not 2k/2k

And the machine in question is a 

https://www.google.com/search?q=Terrans+Force+X799&rlz=1C1CHZL_enUS707US707&source=lnms&tbm=isch&sa=X&ved=0ahUKEwiZtvaLloXPAhUQzGMKHbSsAmYQ_AUICigD&biw=2560&bih=1274#imgrc=qmeM6AgjWzp0lM%3A

So maybe a TB2/TB3 1070 is what he used. 

 

Their website is too damn slow to load. Seems they have a 1070 model as well, still...not sure how the 980M is in there unless they used maybe mini pcie to 1070

 

I saw this the other day trying to figure out how I lost first place to that machine. I agree with John and believe it is connected to an eGPU. If so, that would explain why 1070 is listed as "secondary" GPU. I think the 3DMark database has the results wrong from that machine for sure.

3 hours ago, wersuss said:

Thanks. Isn't p750 p775 cooling same?

As far as I know it is, but they may have done something to improve the fit or design of the heat sink. I don't know. I hope it is better than the P750ZM. If it is not, then yeah... cooling system blows unless you run everything stock. Even on AC cooling it struggles with high CPU clocks because everything is way too small to be effective. Another reason thin and light shitbooks are sucky.

[Guest]

Here we go, Brother @Prema

 

http://valid.x86.fr/u13ull

[Guest]

Well, Brother @Prema, I can't quite make it to 1550 on core, but the additional voltage did help. Here are some new single 980M records for the P750ZM.

 

Straight up tie in 3DMark 11 against P570WM with 4960X. http://www.3dmark.com/compare/3dm11/9208097/3dm11/11568175

 

Brother @johnksss still holds #1 with the Xeon beast.

 

3DMark 11: http://www.3dmark.com/3dm11/11568175

Fire Strike: http://www.3dmark.com/fs/10122415

Sky Diver: http://www.3dmark.com/sd/4296564

 

Brother @D2ultima - max GPU temp was 46°C

[Guest]

I think I am going to have to break my Windoze OS X RAID0 array and go back to Windows 7 to see how much the 3DMark 11 score improves. Those Windoze OS X physics scores are pretty damned shitty.

[johnksss]

Just load another ssd is all.

[dspboys]

21 hours ago, Mr. Fox said:

I think I am going to have to break my Windoze OS X RAID0 array and go back to Windows 7 to see how much the 3DMark 11 score improves. Those Windoze OS X physics scores are pretty damned shitty.

Here's my run using the same clocks http://www.3dmark.com/compare/3dm11/11567233/3dm11/11567165 

 

Win7 beats Win10 physics+combined scores by about 7-8.9%. However in the graphics test, it is losing at double-digit percentages except graphics test number 4. We need @Prema mod to make it work like the one below

 

gtx 980 Win7 vs Win10 http://www.3dmark.com/compare/3dm11/11251942/3dm11/11244048 

[Guest]

8 hours ago, dspboys said:

Here's my run using the same clocks http://www.3dmark.com/compare/3dm11/11567233/3dm11/11567165 

 

Win7 beats Win10 physics+combined scores by about 7-8.9%. However in the graphics test, it is losing at double-digit percentages except graphics test number 4. We need @Prema mod to make it work like the one below

 

gtx 980 Win7 vs Win10 http://www.3dmark.com/compare/3dm11/11251942/3dm11/11244048 

Windows 7 is a better OS than Windows 10, and that was a really nice run you posted.

[johnksss]

I found during my testing that it does still work in our favor for a variety of benches, that is for sure.

[coolane]

Not seeing 970Ms for awhile, here are my 7 x 24 clocks for every day use.

CPU and caches: 4.6 GHz x 4 @ 1.235v

Ram: 3000 MHz 15-16-16-35 @ 1.20v

970M_0: 1304 MHz @ 1.05v + 6300 MHz vRam

970M_1: 1304 MHz @ 1.037v + 6300 MHz vRam

Catching up default score of 980M SLI: 

 

Fire Strike: http://www.3dmark.com/fs/10155379

Fire Strike Extreme: http://www.3dmark.com/fs/10156851

3DMark 11: http://www.3dmark.com/3dm11/11576357

 

Thank you @Prema for fixing my throttling issue.

[Khenglish]

13 hours ago, coolane said:

Not seeing 970Ms for awhile, here are my 7 x 24 clocks for every day use.

CPU and caches: 4.6 GHz x 4 @ 1.235v

Ram: 3000 MHz 15-16-16-35 @ 1.20v

970M_0: 1304 MHz @ 1.05v + 6300 MHz vRam

970M_1: 1304 MHz @ 1.037v + 6300 MHz vRam

Catching up default score of 980M SLI: 

 

Fire Strike: http://www.3dmark.com/fs/10155379

Fire Strike Extreme: http://www.3dmark.com/fs/10156851

3DMark 11: http://www.3dmark.com/3dm11/11576357

 

Thank you @Prema for fixing my throttling issue.

 

You have a very good CPU. 4.6GHz at just 1.235V is excellent.

[coolane]

6 hours ago, Khenglish said:

 

You have a very good CPU. 4.6GHz at just 1.235V is excellent.

Really? That's the input voltage in BIOS, the voltage under load in realbench is about 1.285 ~ 1.295v.

Usually it fluctuates in 1.35v ~ 1.39v when idle witch c states off.

[Khenglish]

6 hours ago, coolane said:

Really? That's the input voltage in BIOS, the voltage under load in realbench is about 1.285 ~ 1.295v.

Usually it fluctuates in 1.35v ~ 1.39v when idle witch c states off.

 

Oh, ok so its not that great of a cpu then. I thought you had it locked to 1.235V.

[D2ultima]

3 hours ago, Khenglish said:

 

Oh, ok so its not that great of a cpu then. I thought you had it locked to 1.235V.

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.