Overclocking the i7-4710HQ on a G751JY and issues

[Edweird]

WARNING! Image Heavy!

Hey, everyone.

I recently purchased this beauty of a machine - a G751JY. (i7-4710HQ, GTX980M, 8GB DDR3L, BIOS 205)
After rummaging around the forums and with you guys' help (thanks to ASUS Support as well) I have managed (for now) to get some extra juice out of the already almost unnecessarily powerful 980M using GPU Tweak for Graphics Cards with the 344.75 drivers.

Yay! This is just the +135Hz standard overclock and the memory running at 6000MHz effective. I haven't seen if it would go up but from what I've seen I should be able to hit 6400MHz no problem.



So that being done, I think it's appropriate to talk about the CPU...
Especially considering that I can imagine a fair few circumstances where the CPU would hold down this beastly GPU.

So, what does HWINFO86 tell my uneducated mind about this processor:



Okay, so I did some basic research and from I understand, this information tells us that this CPU can take 2 extra frequency multipliers - so technically, I should be able to get a free 200MHz overclock! That is significant on a mobile machine with the thermal headroom and the G751 is definitely that.

Okay, so I loaded up Intel Extreme Tuning Utility and noticed that the clock multiplier can be pulled up but only by a factor of 1...huh. Okay so our new multis are 36-35-34-34 and a cache multiplier of 36. (Worth noting that I'm a complete noob when it comes to overclocking.)

So I applied the settings and ran a benchmark to see what would happen.



Odd... the CPU hit 3.4GHz only once for a split second. I determined that, obviously, XTU stresses all 4 cores for this benchmark.
Looking down we can see something more odd - the CPU is being starved for current by XTU! I guess this is how XTU stresses the chip? I truly have no clue. However, my non-OC brain thinks that XTU does this to determine how effective the CPU is at blasting calculations when deprived of "food", which WOULD explain the lower clocks.

Moving on - I loaded up ThrottleStop because of its simple and effective threaded benchmark so that I could test individual cores.



This time, there is no throttling of any kind. We see that a single threaded test mostly works the CPU to about 3.4GHz, rarely peaking to 3.5GHz. XTU reports only 1 core being active. This is very odd. So we've set the multiplier to 36 for 1 core, the CPU has enough current but it barely makes 3.5GHz? Throughout my testing I have never seen this CPU go above 3.5GHz.

Next, 2 threads.



Hmm...same results. The CPU peaked at just over 3.5GHz - but this is expected as our 2-core multiplier is set to 35. But it rarely reached that speed. XTU jumped between 1 and 2 cores.

A solid 3.4GHz for 4 threads for a multiplier of 34 for 4 cores active.



Okay, so let's look at the benchmark results.
Again, we can see here that XTU starves the CPU of current and the CPU struggles to keep up at 3.1GHz.



The result:



No idea how this stacks up.

Now with the tweaked multipliers:



Throttling again, but this time it's a steady 3.3GHz.

Results:



...What? Same results? What's happening?

This is where I am stuck. First, as I understand it, the CPU should be able to do 3.7GHz. Second, it doesn't even do 3.6GHz through XTU. This is where I'm hoping some OC savvy people will jump in. Do I need to tweak something else to get an actual overclock out of this CPU? Could it be that ASUS has changed how the CPU behaves and limited the multipliers through the 205 BIOS release? And I'm not liking the dips on that blue curve down in the left corner...

Any and all feedback on the process and any discussion is welcomed and encouraged!

[NitroX]

Well, It seems that you've got yourself one hell of a laptop. It's the coolest one I've seen by now. I'm actually from Romania and I have the T7015D model (this is the cheapest one with Gtx980m) and my CPU temps tend to go much higher on stock voltage.

[Edweird]

NitroX wrote:
Well, It seems that you've got yourself one hell of a laptop. It's the coolest one I've seen by now. I'm actually from Romania and I have the T7015D model (this is the cheapest one with Gtx980m) and my CPU temps tend to go much higher on stock voltage.

Interesting...
The JT models that come with the 970 are T7009, T7010, T7012 and T7019 here. The cheapest 980 is this one, the T7065 and it is also the only one. The T7064 is the one with the 4860 CPU. I wonder if there is a correlation between performance/quality and these numbers. Perhaps there are subtle changes in the designs that would account for the differences we're seeing. Also, my G751 is the only one I've heard of coming factory sealed with 205 BIOS, which means it was built at the very least at the end of November.

EDIT: I checked the bottom and this machine was built this December. Could explain the decent sound quality im getting as well.

[Exostenza]

Regarding the CPU overclock you can check out my thread here which has some information for you.

[Edweird]

Exostenza wrote:
Regarding the CPU overclock you can check out my thread here which has some information for you.

Thanks, I have already read through that thread. I understand that newer BIOS cause issues - I'm trying to figure out why the CPU wont hit 3.6GHz. It does do 34x for 3 and 4 cores, though - but 0 performance increase!

[hmscott]

Edweird wrote:
Thanks, I have already read through that thread. I understand that newer BIOS cause issues - I'm trying to figure out why the CPU wont hit 3.6GHz. It does do 34x for 3 and 4 cores, though - but 0 performance increase!

Edweird, when testing XTU CPU improvements, don't rely on the XTU benchmark - if you look at the XTU usage graph you will see the test isn't a constant load, it is modulated / pulsed, so any change is going to get lost in that variance.

You should try some CPU benchmarks like Cinebench, or what ever other PCmark type Sandra, etc tool that focuses on CPU results. FPS in games may show a small increase - but the GPU tuning will show more - so focus on benchmarks that will use 1, 2,3,4 threads so you can see the improvement from single threaded and multi-threaded results.

It is about a 10% improvement for me, with a different CPU.

Unfortunately the Asus G751 BIOS update have de-tuned the top potential multiplier by 1x, so you won't see the max capable from your CPU. Complain to Asus via their Technical Inquiry and Feedback channels, and ask for a new BIOS with the full multiplier available. Asus is shipping Intel XTU with the OS, so they should expect people to be aware of the 1x reduction.

For games, get Asus GPU Tweak tool for Graphics cards (G751JT/JY) and increase the GPU +135 and the GPU memory for what you particular card will run stably and that is where the most measurable benefit will be for games - high CPU usage games will benefit from the CPU increase as well.

For now, just flip the sliders up to max available for cores and cache, and enjoy the most CPU performance available for your laptop. 🙂

[Edweird]

hmscott wrote:
Edweird, when testing XTU CPU improvements, don't rely on the XTU benchmark - if you look at the XTU usage graph you will see the test isn't a constant load, it is modulated / pulsed, so any change is going to get lost in that variance.

You should try some CPU benchmarks like Cinebench, or what ever other PCmark type Sandra, etc tool that focuses on CPU results. FPS in games may show a small increase - but the GPU tuning will show more - so focus on benchmarks that will use 1, 2,3,4 threads so you can see the improvement from single threaded and multi-threaded results.

It is about a 10% improvement for me, with a different CPU.

Unfortunately the Asus G751 BIOS update have de-tuned the top potential multiplier by 1x, so you won't see the max capable from your CPU. Complain to Asus via their Technical Inquiry and Feedback channels, and ask for a new BIOS with the full multiplier available. Asus is shipping Intel XTU with the OS, so they should expect people to be aware of the 1x reduction.

For games, get Asus GPU Tweak tool for Graphics cards (G751JT/JY) and increase the GPU +135 and the GPU memory for what you particular card will run stably and that is where the most measurable benefit will be for games - high CPU usage games will benefit from the CPU increase as well.

For now, just flip the sliders up to max available for cores and cache, and enjoy the most CPU performance available for your laptop. 🙂

Yeah, I assumed that XTU wasn't giving me the right results but I wasn't sure because I've seen bench results from XTU for this CPU that are 200, 300 points higher. You may have noticed in several topics that I have already posted about me using GPU Tweak with success - it's the first thing in this topic's OP. 😄

Also, I increased max TDPs to 50W and 60W accordingly. I'm still not seeing frequent use of the 36x multiplier, though. Successful undervolt at -90mV.

[hmscott]

Edweird wrote:
Yeah, I assumed that XTU wasn't giving me the right results but I wasn't sure because I've seen bench results from XTU for this CPU that are 200, 300 points higher. You may have noticed in several topics that I have already posted about me using GPU Tweak with success - it's the first thing in this topic's OP. 😄

Also, I increased max TDPs to 50W and 60W accordingly. I'm still not seeing frequent use of the 36x multiplier, though. Successful undervolt at -90mV.

Edweird, you didn't say how you came to the conclusion that the XTU changes weren't having an effect, I was trying to put in perspective that against FPS improvements, GPU tweaking will be noticeable, but XTU tweaking requires other benchmarks to see the improvement.

XTU isn't a reliable test - many have widely different results even when they should have the same - I have told people to not waste their time on XTU as a benchmarking tool because of this - don't sweat it.

If you are sliding up the settings, Applying them, and saving them as a profile that loads ok at boot, that is all I would use XTU to accomplish.

A parameter change is a parameter change. If you have set your Power Plan to High Performance, CPU 100%/100%, and run 1, 2, 3, 4 threaded tests to see the speed/x settings during load, that's all that matters. It is up to the individual application to use it effectively 🙂

What are you using to benchmark the before / after settings for XTU, to see the difference?

[Edweird]

hmscott wrote:
Edweird, you didn't say how you came to the conclusion that the XTU changes weren't having an effect, I was trying to put in perspective that against FPS improvements, GPU tweaking will be noticeable, but XTU tweaking requires other benchmarks to see the improvement, as even

XTU isn't a reliable test - many have widely different results even when they should have the same - I have told people to not waste their time on XTU as a benchmarking tool because of this - don't sweat it.

If you are sliding up the settings, Applying them, and saving them as a profile that loads ok at boot, that is all I would use XTU to accomplish.

A parameter change is a parameter change. If you have set your Power Plan to High Performance, CPU 100%/100%, and run 1, 2, 3, 4 threaded tests to see the speed/x settings during load, that's all that matters. It is up to the individual application to use it effectively 🙂

What are you using to benchmark the before / after settings for XTU, to see the difference?

Oh, by results I meant the actual score it's been giving me. I wasn't aware it was a non-effective benchmark before this because, as I said, other people with the same CPU have gotten much better XTU benchmark scores - so my ignorant and uneducated self assumed that there was something I could do within XTU to achieve those scores since there isn't much else you could do with these CPUs - and there were a significant number of people with those results that I would think it wasn't something more complicated like custom BIOS modding etc. (But to be fair, I've actually done a BIOS mod for my old N61Ja to enable hidden options in the BIOS screen.)

I will try Cinebench now. I was just using Prime95 and such to stress the CPU to see if it will hit the 36x and since it wasn't I assumed XTU was to blame. Everything you say makes complete sense, of course - I wasn't expecting anything more than these simple boosts from XTU but when a CPU heavy game that is actually FPS limited by the CPU such as Planetside 2 isn't making the CPU hit 36x it understandibly has me confused. Obviously what could be happening is the game never actually using just the single core - I know full well that game loves i7s for the core count.

Silly me! Will post results soon with appropriate XTU settings.

Oh, and since I've messed around with the power profiles, the CPU seems to now be hitting 3.6GHz basically at idle \o/

[Edweird]

Okay, Cinebench results:
Stock CPU score: 632cb (99.6% ref. match)
Single Core score: 126cb (MP 5.01x)
GPU score: 94.26fps

Multipliers set to 36x, TDP to 50W and 60W.
CPU score: 648cb (VERY close to i7-3770)
Single Core score: 130cb (never hit 3.6GHz)

Turbo Boost Power Time set to 96 seconds.
CPU score: 652cb
Single Core score: 130cb (all cores hit 3.6GHz at least once)
GPU Score: ~90fps (it actually went down...)

Max CPU Temp: 67
Max GPU Temp: 48 (lolwat)
Stock GPU score: 86.6fps (looks like Cinebench doesn't get you much of a difference - GPU usage topped out at 56%)

Final Cinebench results for CPU:
632 -> 652
126 -> 130

3% performance boost for 3% higher frequency.

[AQUASTEVAE]

Edweird wrote:
Okay, Cinebench results:
Stock CPU score: 632cb (99.6% ref. match)
Single Core score: 126cb (MP 5.01x)
GPU score: 94.26fps

Multipliers set to 36x, TDP to 50W and 60W.
CPU score: 648cb (VERY close to i7-3770)
Single Core score: 130cb (never hit 3.6GHz)

Turbo Boost Power Time set to 96 seconds.
CPU score: 652cb
Single Core score: 130cb (all cores hit 3.6GHz at least once)
GPU Score: ~90fps (it actually went down...)

Max CPU Temp: 67
Max GPU Temp: 48 (lolwat)
Stock GPU score: 86.6fps (looks like Cinebench doesn't get you much of a difference - GPU usage topped out at 56%)

Final Cinebench results for CPU:
632 -> 652
126 -> 130

3% performance boost for 3% higher frequency.

which cinebench are you using, because i am not getting results in "cb", but points instead. so i have no idea how to translate those. i am using cinebench release 11.5.