PDA

View Full Version : LOW SCORES and HIGH TEMPS in Cinebench and Realbench with G751JM (4710hq)



Sagnik87
04-08-2015, 07:52 PM
Hi people,

I bought the G751JM a month back and and I have been breaking my head over the performance of the CPU. I have not been able to perform every stress test but I have been doing a few tests now and then on Cinebench R15 and off late in Realbench. I am not much of a gamer so I am really not bothered about the GPU much. I primarily use my machine for rendering in Cinema 4d which is why my focus is totally on the CPU part. Note that the GTX 860m is doing pretty well till now with viewport rendering.

Problem 1 :
I doubt my CPU is performing as it should be. Even on stress tests like CInebench, the maximum clock speed remains below 3.3MHz. I havent yet played with the OC setting so you can say this is on stock speed but I am of the idea that when stressing on all cores, the 4710hq can give out atleas 3.3 - 3.4 Mhz. This results in the lower CPU scores in Cinebench and overall score in Realbench. As I said, I use my notebook for rendering, even the slightest improvement in speed can do a lot in overall rendering time.

have gone through forums and found out that all the G751 series that uses the i7 4710Hq processors have relatively higher CInebench or Realbench scores.

Attaching screenshots
48266

Above you can see that my Cinebench scores(orange ones) are all below 631cb which is below the avg score of 657cb I saw with other G751 series with 4710hq. Also on the right side the you can see the max clock speed never even touches the 3.3 mark.

Another from Realbench
48267

You can see on the top left the my Realbench scores varied a lot and the last test gave out even lower score of 51169 which is way below than the avg Realbench scores of 57000 from the G751 series notebook. Also in the Realbench window the CPU speed is recorded as 798.1Mhz which is really annoying.

Problem 2 :
I am getting very high temperatures whenever I am doing those stress test. In the Realbench screenshot the processor core's max temp recorded 92C which is quite rare. I have gone through all the forums and hardly found anyone experiencing that kind of temps during Realbench test.

So I am a bit perplexed with the scores and the speed of the processor, not to mention the high temps along with it. It would be great of you guys can help me out.
I went through the forum and this is one thing i found useful suggested my Boom_Austin https://rog.asus.com/forum/showthread.php?55984-Overclocking-the-i7-4710HQ-on-a-G751JY-and-issues/page3

P.S. sorry for the long post. :)

arrtaas
04-09-2015, 12:02 AM
To begin I'll post my results to compare to what you have. The ambient temperature was around 21.5C while doing these tests. Also my system is a bit different there i would expect it to perform better.
G751JY - I7 4710hq, gtx980m and 32 GB stock ram.

Here are the pics.

48268

48269

As you can see for Cinebench the results were pretty close (20 points difference), although your temperature was around 12 C higher.

Real bench on the other hand again, results were kind of close (within couple thousands I'd say within couple percent difference) The bigger difference where at applications that use more RAM or GPU. So me having 32 gb ram helps, by making it dual channel, GPU ofcourse is better so no questions there. However again we can see that my temps were around 6 C less (I experienced max temp for around 2seconds, then the fans kicked it and droped it to 80-82, for the rest of the test part, I think multitasking that was).

Now the reasons of these differences are as follows:

Performance difference - better platform: Dual channel ram, better GPU. CPU score difference is negligible
Temperature difference - more important question, might have different reasons:
1. Might be bad thermal paste application job
2. Bad installation of heatsink
3. I do not know if G751JM has the same heatsink as G751JY has. Since the GPU is 860M, they might have installed a smaller/cheaper version of heatsink, therefore resulting in higher temps.

To conclude - if the heatsink is the same, then you should troubleshoot the first two problems. However if you dont have enough knowledge in laptops I would not recommend it, since it is complicated design and might cause you some troubles. Your option would be to send it to RMA.

Sagnik87
04-09-2015, 08:07 AM
Hi arrtaas,

Firstly thanx for the insight. It was much clear to compare with your results. Like I said your results are at par with the other G751 series laptops atleast in the Realbench test.

For the Cinebench test, the test only put stress on the processors (othere factors like GPU and RAM dont count I believe..please correct me if I am wrong) so should it matter wether it is a G751JY or JM when both have the same CPU ? Are these clocked on different frequencies ? Because my 4710hq can be overclocked I believe.

For the realbench test, it might differ as both the GPU and the RAM are taken into account when testing. Also talking about RAM, it is 24GB dual channel as well. But then I dont understand why the CPU is lagging behind. If you see my first post.. Note that the Realbench window(from my first post)shows my CPU to be only 798.1 Mhz whereas yours shows 3391.9 Mhz. I dont quite understand this. How come mine is only stuck at 798.1 Mhz. Should I look into other solution for this ?

I just ran Realbench again and this time I got a score of 55k which is higher than the previous two test I ran. The only change in the situation is the laptop is now in my office plugged into the power socket unlike previous times where it was plugged in to my home power socket. What I mean to say is, does a wrong supply of power affect the speed of the processor ? I think it does..not sure.

Now for the Temperature :
I'll look into the thermal paste matter again. While doing the Realbench test again, one core hit the 91C mark again for a few seconds and then went back to 85-87C mark during the multitasking test.

What I am concerned about is there is not much difference between the JY and JM except for the GPU. But when stressing only the processor it should give out almost the same result and also the same heat signatures I believe. Or are there major differences as well that I dont know of.

arrtaas
04-09-2015, 10:09 AM
Just a quick reply.

I do think that 4x8 performs better than 3x8 - at least i saw small difference and read somewhere that 3x8 does not work in dual channel.

What is your power management setting in windows? Make it 100% minimum on power. Maybe that is why it was showing 800 mhz on your part.

What power supply are you using? did it come with the laptop? How much W does it have?

And yes processor wise i think we should be scoring the same. It overclocks automatically based on demand and temperatures. So if yours is heating to fast maybe it throttles, but your pics show it does not.

Sagnik87
04-09-2015, 01:38 PM
Hi arrteraas,

Firstly THANX for the quick tip. It worked. Changing from the 'balanced' to' High Performance' (quite self explanatory..why didnt i think about it :/ ) mode definitely helped with the Realbench test. The processor performed as it should. The Realbench window showed the processor speed of 3292 Mhz instead of the 798 Mhz. However I also noticed this time the temp signature gave out a whooping 94C on core #1 for a few seconds although in the high performance mode the fan speed are set to Active on minimum as well.

On the Cinebench test, setting the power option to 'High Performance' didnt make any difference though. Pretty Weird..should have worked. It was same as a 628cb score. Temp reached to a maximum of 87C.

So now the focus is on the temperature part. Think it is a case of bad thermal paste.

Yes the notebook came with a AC/DC adapter. The output is 19.5 V ~ 180 W

Thanks again :)

arrtaas
04-09-2015, 03:33 PM
Please post a pic of the heatsink if you gonna change the thermal paste. I would like to confirm if it is the same heatsink as in G751JY, if not then maybe your temperatures are normal.

Sagnik87
05-04-2015, 11:56 PM
Hi arrtass,

I havent yet decided to go with the thermal paste till now when I saw this show up on the hardware monitor as I was rendering on Cinema 4d.

Check the TEMP on Core#1. Room temp is around 32 C. Its summer :/

48912

hmscott
05-05-2015, 12:48 AM
Hi arrtass,

I havent yet decided to go with the thermal paste till now when I saw this show up on the hardware monitor as I was rendering on Cinema 4d.

Check the TEMP on Core#1. Room temp is around 32 C. Its summer :/ ...

Sagnik87, the G751's have a shared heatpipe/heatplate for GPU/CPU cooling. If you have a high load on GPU it can bring up the temp on the CPU.

G751JY
48913

There is no need to re-paste. :)

Use hwinfo64 logging to get a temp reading every few seconds, then do the run, and then export the file changing ;'s to ,' and read with Excel. Then you can see how often the Max temperature was actually reached.

If you find that 99% of the time the temps are much lower than the peak, then there is nothing to worry about. Check the Thermal Throttling entries too, how often did it get past 93c and actually trigger Thermal Throttling.

The trick with G751's is to use Intel Extreme Tuning Utility to detune Core 3,4 multipliers down a couple of steps, and increase the Core 1/2 multipliers to Maximum. That way you get best performance for all 1/2 threaded tasks (most of the time), and reduce temps for 3 and 4 threaded or more tasks.

For example, if the maximum settings for multipliers are 36x,35x,34x,34x Cores and 36x Cache - set them to 36x,35x,30x,30x and do the run again. Adjust Cores 3/4 multiplier to not cause Thermal Throttling under rendering load.

Intel Extreme Tuning Utility - 5.1.2.2
https://downloadcenter.intel.com/search?keyword=xtu

Here is a sample view of XTU settings, your CPU multiplier slider may go higher
48914

Please give it a try and let us know how it works out :)

Sagnik87
05-05-2015, 01:20 PM
Hi hmscott,

Thanks for replying :) My next step would have been to directly ping you and ask for solution :P Anyways

So I have been trying to get my head around this problem of overheating. Just to let you know that Cinema 4d only uses the processor when rendering the final image..there is no GPU involved. I also checked the GPU temps and it was around 56C. However the the GPU was a bit warm as it was being used when working with the viewport before hitting render. Could that be a reason ??

When I render in my 3d software I usually keep an eye on the hardware monitor to check the temps. Till now before it hit 98C yesterday...my G751Jm reached a maxium of 95C on one render but there was no case of thermal throttling. The pattern I observed is..it is only core#1 that hits maxium temp above 90s when under load. Rest of the cores usually remain below 90C. It was the same case yesterday and as it reached beyond 95C, I saw that Core#1 was throttling on the hardware monitor.
I can understand hitting 93C on occassions bt 98C...thats really high for this series of laptop that boasts about dual cooling dont you think ?
I'll check the logsheet as you suggested.

Also since I use my laptop for 3d works, the more threads I get to use the better. For example, it is better if I get to use 4 cores running at 3.3 Ghz rather than using two cores running at 3.5 Ghz.


However, I'll post my results tweaking the XTU as you suggested.

arrtaas
05-06-2015, 03:09 AM
Sagnik,

98C is really bad. But, 32 C room temperature is definitely a factor to your temps. You laptop should start throttling when hitting 94C and it should never go to 98C (theoretically). Since your 4D only uses the CPU, I don't think it would should go above 85C at any point.

With 21 C room temp, running prime 95, my cpu gets up to around 78-80 degrees. in this case Your should go to around 90-92 (based on room temp).

If you are not doing repasting (which is ok). My suggestions are:
1. Undervolt (-80mV).
2. Get a laptop cooler where you can put the fans where you need them. I use this : http://www.coolermaster.com/mobile/notepal-series/notepal-u3-plus/
3. Find a cooler room/air conditioned.

Also how long does a normal non stop rendering session take?

Sagnik87
05-07-2015, 08:27 PM
Hi arrtaas,

Yes it is. I wonder why havent Asus put much bigger grills on the underside of the notebook. That would have surely improved the ventilation.

I'll probably go for the cooler pad. Tell me how well it works the 751 series..since the design is different.

hmscott
05-08-2015, 01:00 PM
Hi hmscott,
Thanks for replying :) My next step would have been to directly ping you and ask for solution :P Anyways
So I have been trying to get my head around this problem of overheating. Just to let you know that Cinema 4d only uses the processor when rendering the final image..there is no GPU involved. I also checked the GPU temps and it was around 56C. However the the GPU was a bit warm as it was being used when working with the viewport before hitting render. Could that be a reason ??
When I render in my 3d software I usually keep an eye on the hardware monitor to check the temps. Till now before it hit 98C yesterday...my G751Jm reached a maxium of 95C on one render but there was no case of thermal throttling. The pattern I observed is..it is only core#1 that hits maxium temp above 90s when under load. Rest of the cores usually remain below 90C. It was the same case yesterday and as it reached beyond 95C, I saw that Core#1 was throttling on the hardware monitor.
I can understand hitting 93C on occassions bt 98C...thats really high for this series of laptop that boasts about dual cooling dont you think ?
I'll check the logsheet as you suggested.
Also since I use my laptop for 3d works, the more threads I get to use the better. For example, it is better if I get to use 4 cores running at 3.3 Ghz rather than using two cores running at 3.5 Ghz.
However, I'll post my results tweaking the XTU as you suggested.

Sagnik87, that's the nice thing about XTU, you can make Profiles for each load you need to tune. :)

If you have a load like 4D that favors 1 Core over the others, and it is overheating, then make a Profile that de-tunes Core 1, but keeps the other 3 cores and Cache up high.

Then you can select that 4D Profile before starting those jobs, and swap it back to Max Multiplier Profile for general use. :)

Have fun!

Sagnik87
05-12-2015, 08:17 PM
Hi hmscott,

I did as you suggested to check the log file from the HWinfo monitor during a render in Cinema 4d. The colored portion is the render session and the red ones are the temps.

49144

Tell me if it looks normal to you.

The method you suggested using the XTU, detuning the core#1 for keeping all others high, i tried that yesterday and I got a lower result in the cinema 4d benchmark. It was much lower.

Update: These are my XTU settings. They are default settings.
49145


Thanks for all the help :)

hmscott
05-13-2015, 09:44 AM
Hi hmscott,

I did as you suggested to check the log file from the HWinfo monitor during a render in Cinema 4d. The colored portion is the render session and the red ones are the temps.
...
Tell me if it looks normal to you.

The method you suggested using the XTU, detuning the core#1 for keeping all others high, i tried that yesterday and I got a lower result in the cinema 4d benchmark. It was much lower.

Update: These are my XTU settings. They are default settings.

Thanks for all the help :)

Sagnik87, hmmm... you are sustaining >93c for a while during the run, but the Thermal Throttling columns all have a "No" in them, so you aren't Thermal Throttling.

Your 4710 will clock higher, as you can see you are losing 2x in all 4 Cores right now - the slider will go up more.

From your text you indicate this data is from a normal speed run, using the XTU settings you are showing? And, when you ran again with Core 1 detuned - how much 1x, 2x? What were the temps then? Can you please log those results and post them?

The idea is to use the most performance that the CPU will sustain without Thermal Throttling. So you want to put all 4 Cores up to maximum 37x, 36x, 35x, 35x and log temps during a run there, looking for Thermal Throttling - temps 93c and above - see what you score is there.

You can also undervolt the processor to reduce temperatures. My G750JW did -125mv CPU/Cache and my G750JX did -75mV, but my G750JH would only do -20mV out of the box new but does -50mV now.

You can also set the Power Plan to Balanced which has the CPU performance set to Min/Max 0%/100% which will downclock and upclock as needed which can allow the processor to cool down in between load spikes. That will reduce temps a bit, but may shave a couple of % off performance.

If the temp's are kept below Thermal Throttling your overall performance should be better - time your actual render runs - not just scores from a benchmark.

After thinking about it some more, the stock speeds might be your maximum, you aren't Thermal Throttling, but all 4 cores are getting close to that point. It will be interesting to see how full multipliers run, and if you Thermal Throttle.

BTW, is your GPU set to Intel? Even if you aren't rendering on the Nvidia GPU try setting the App to prefer the Nvidia GPU, it might reduce the heat generated by the Intel GPU - reducing your CPU temps.

That is the problem with enabling the Intel GPU - it adds heat to the CPU die - reducing head room on the CPU temps.

If you really need high CPU performance, I would return the G751JM and get a G751JT - it doesn't have the Intel GPU enabled, meaning your CPU temps when doing pure CPU work will be lower. On the G750 series the Optimus models run about 10c hotter CPU temps than the non-Optimus models under 100% CPU load like you need.

Sagnik87
05-15-2015, 10:51 PM
Hi hmscott,

Thanks again for all the possible solutions :)

I am really doubtful about the undervolt process since I have no knowledge about the way it is done. Like the logic behind it.
I have my power plan set to 'Balanced' all the time. Also, I have my 3d programs set to run on my nvidia GPU in the settings as you will see in the following test results.

I ran three test based on different XTU settings. 1. Detuned 2. With highest multiplier 3. Default setting. And i got some pretty random results. Here they are

Test 1 with Core #1 detuned by 1x in the XTU.
49283

Ran Cinebench. The bright orange is the current result. With previous results marked with faded orange mark. Also note the GPU temp in the HWmonitor
49284
49285

Test 2 with high multiplier setting in the XTU.

49286

In Cinebench, I expected better result with this setting but it was almost similar to previous results done with default setting in the XTU. Also a surprising factor was the temp. Didnt go beyond 92. But have a look at the last pic in the post. It will show some differnt results with the same settings.
49287
49288


Test 3 with default XTU settings.

49289
49290
49291

Now here is the random result I found running hi multiplier setting with the XTU.

The score dropped from 633 to 619 in Cinebench.
49292

While doing the benchmark test in XTU, there was thermal throttling with a temp of 96C
49293


So there it is...I am not able to lock on a particular solution. Any thoughts regarding the results :/

hmscott
05-15-2015, 11:00 PM
Sagnik8k, before getting into it, the undervolting has a simple justification.

Less power used / expended means less heat introduced into the system. Less heat generated, lower temperatures overall.

Same for lower speeds, lower multipliers, less heat generated - less heat dissipated - lower temperatures.

Whether or not it shows in a particular benchmark(s), lower voltage, less current draw with lower speeds, all reduce the heat generated and dissipated - lower temps.

Every little bit helps.

hmscott
05-15-2015, 11:03 PM
Sagnik8k, for the first run with the multipliers decreased by 1x - relative to your previous non-maximum setting, you see slightly less temperature with slightly less performance, as expected.

For the next run with maximum multipliers, the reason you didn't get much more of an improvement is that you didn't increase the Turbo Boost Power Time Window - run again after increasing to 28 seconds from default of 8 seconds to start - you can try higher numbers later. The default for the 4710 is 8 seconds, the 4700 is 28 seconds but not adjustable.

Also make sure you are using the High Performance Power plan with Processor speed set to 100%/100% Min/Max.

You can also try increasing the Turbo Boost Short Power Max by 10w, same for Turbo Boost Power Max, doesn't affect my results, but coupled with a longer Turbo Boost Power Time Window, it might help - might take a lot of playing / tuning - only to find it doesn't help.

As far as random runs showing low numbers, that happens because there are other things running in the background on the laptop - lots of background processes can go off at random times and affect benchmarks and screw up games - lag / glitches / etc

Make sure you have disabled all OS/app/anti-virus background updates, background OS/media indexing updates, disk optimizer idle time optimizations - idle keyboard registers as idle system to a lot of background programs.

It takes some time to get a handle on all of this, but once you do, you will be in control of your laptop, that is until you update or install something else ;)

Have fun!