Recoil II RTX In Exploration

[bana845]

Today I received the Recoil II RTX 17'':
- NVIDIA GeForce RTX 2070
- SSD PCIe WD Black™ SN750 M.2 NVMe 250 GB
- 32 GB PC4-21300 DDR4 SDRAM Corsair CM4X16GE2666C18S4
- COOLER MASTER MASTERGEL MAKER

As initial impression it seems a very respectable laptop, the temperatures in idle seem excellent, the cpu is around 40° (18° in the room).
Arrived with a 50mV undervolt.
GPU in Idle stands between 40-45°.
Chassis and motherboard are TongFang models.

As for the mechanical keyboard I can not give an absolute opinion.
At home I have a MSI GS70 Stealth and what I can do is a comparison between these two models.
The MSI keyboard is clearly superior in quality, softer, silent and with a good touch feel, very high quality.
Compared with MSI, Recoil II RTX keyboard seems to be in medium-low quality.
Surely it is unusable for writing, often the keys must be repeated twice.
Also as regards the audio I am not satisfied, I expected more power.
To my surprise I found an excellent touchpad, very wide, responsive and precise, really a pleasure to use it even without a mouse. in this aspect superior to my MSI.
Even the display was a real surprise, the quality is very high, a beautiful color rendering and excellent brightness, higher quality than my MSI model.


I have performed a benchmark with PassMark Performance Test on all components.
the CPU has reached a maximum temperature of 80° and the GPU of 65°.

I have also performed a benchmarkwith 3DMark Time Spy with a result of 6300.
I've read in other thread that such result is lower than expected for a RTX 2070, I'll try to investigate it.
In this benchmark CPU has reached 90° and GPU only 65°.

 

[bana845]

Hi, I would need some opinions about the performance of the RTX 2070 during the 3dmark time spy test.
As already specified in previous post the total score is around 6300 (the last 6100), much lower than expected.
During the test the FPS are around 30-40, too low I think.
the maximum temperature of the GPU reaches about 60°, even this value seems too low under stress test.
The configuration of the nvidia driver (from the control panel) is the default, I have not changed any value, it does not seem to me that there are settings for which could increase the performance, but tell me if it's right.
No thermal throttling of cpu.
Is it possible that there is a CPU bottleneck? Any Suggestions?
Uploaded CPU and GPU statistics during test with HWInfo.
Thanks

 

[roakes]

I just did a run and got the following scores:

Overall: 7607
Graphics: 7826
CPU: 6567

Screenshots from HWinfo below:







Just out of interest, what does the hardware button next to your power button do? Does it start Turbo Mode or switch been Office and Gaming mode?

 

[aleksanderzakrzewski559]

Hi bana845, First of all, thanks for sharing.
Now as you probably noticed you're not the only one here with this problem, but in my HW results is see throttling everywhere.
Please do the following:
Update all drivers, especially nvidia (easiest way is through geforce experience)
Set laptop to gaming mode in gaming center software
If your fans don't go turbo automatically when needed, the button next to power should trigger it, switch it on
Run both Time Spy and FireStrike
Take snapshots of HWinfo results and benchmark results
Upload images to e.g imgur
Send message to PCS through main website explaining the issue,
There's really nothing more I can advice to you, I did the same, I suspect a faulty batch of units.
Out of curiosity, who build your laptop?

 

[roakes]

FYI. The button should switch between Office and Gaming as mine does, according to PCS support, so if yours works as Turbo then that's unexpected behaviour. I suspect both of your machines may be stuck in office mode.

 

[roakes]

I just did a run in Office mode:

Office Mode:

Overall: 7325
Graphics: 7849
CPU: 5318

Gaming Mode:

Overall: 7607
Graphics: 7826
CPU: 6567

So Office Mode dropped the CPU score, which I think is because it throttles the CPU to keep the laptop cool and use less battery. Lets try to isolate what's running slow on yours, can you run cinebench R15 and and R20 CPU tests?

Mine:

R15 Office Mode: 907
R15 Gaming Mode: 1268

R20 Office Mode: 1855
R20 Gaming Mode: 2980

 

[bana845]

Thank you all for the support.

1. My laptop was built by M. Bradley:
Builded by: M. Bradley
Tested by: L. Vaughan
2. The button next to the power-up changes the fan speed.

Following my assumptions (maybe wrong, tell me if you agree):
1. The laptop chassis should not affect the performance of a component.
Regarding the button, maybe only the function mapped is different.
2. I did an installation of all the software from scratch.
Windows 10 and all updated drivers including motherboard chipset (Intel HM370) and NVIDIA driver (version: 419.17, type: standard).
I have not installed any additional management software.
I have no software that can block me in office mode (I have only drivers installed).
if I'm locked in office mode it means that the block is encoded in the driver or in the bios (but looking at the HWInfo data both the CPU and the GPU reach the maximum frequencies as specified by those components, in this sense I see no limitations).
Or maybe office mode is actually the basic mode and gaming mode is a kind of overclocking. If so, then yes, I need the management software to pass in overclock.

One thing that I noticed looking at the differences between the screenshots of mine and those of @roakes, respect to the GPU, is that my GPU reaches higher peak frequencies and use more memory.
instead the average values ​​are lower than those of @roakes.
This leads me to think that is a diferent build of 3Dmark or is a different build of RTX 2070.

@roakes, could you please post the gpu-z screenshot to compare ours RTX 2070 specs and tell me which version of 3DMark you used in your benchmarks? Thank you.
I downloaded 3DMark from www.guru3d.com and it's the "3DMark version Download v2.8.6446 + Port Royale" with Time spy version 1.1.



Now I also try to run cinebench R15 and and R20.

Thanks

 

[Oussebon]

Thanks for posting the info

Sometimes, especially if you're new to the forums, your posts can get automatically held for moderation. When that happens, please don't post duplicates, you just need to wait for one of the volunteer moderators to release them

 

[roakes]

My 3D Mark and Time Spy are the same versions.

I may need to add a little clarification here regarding my HWinfo GPU stats. I have an undervolt set on the GPU, which caps the maximum clock at 1545mhz at .725mV, so it will never boost above 1545mhz. If remove the undervolt then it boosts much higher, but then runs hotter and I get bigger fluctuations in GPU clocks, sometimes dropping under 1545mhz leading to inconsistent frames, but mostly it's to keep the temperature down (the specified boost is 1440mhz so I'm happy with never going higher than 1545mhz)

In regards to your stats, the averages are what you should be looking at. The problem with maximum values is that they could have been for 0.5 of a second and it will still register as the max, even if it was for a moment. If your averages are low then this will be the clock speed that your component was running at for most of the time, make sure you reset the values immediately before hitting the button to run the time spy benchmark, so you don't have idle clock values pulling your average down too much.

My understanding is that the performance of both components is dictated by various throttling caps, but the most common to hit are temperature and power. Temperature seems not to be an issue for you, but power may be. This is where the Office Mode theory comes in, I think the Office Mode caps the power limits much lower than Gaming, this is either a TDP or Watts value cap but whichever it is, will result in allowing your hardware to boost as high as it can, but then it throttles it back to meet the specified cap.

As for the hardware button, it is not configurable in the Gaming Center software, so must be a toggle for functionality in the BIOS and then Gaming Center merely shows you the status. I just closed Gaming Center and the button still functions in the same way, reducing my performance in Office Mode (LED turns off).

I'd be interested to see your Cinebench scores in both modes that your button manages, to see if they change like mine. Also, if your Cinebench scores match mine, then we know that the GPU clocks are the issue, not the CPU, so we can focus on that component specifically.

 

[bana845]

my CPU undervolt : -115mV

Cinebench R20 : 2977 (without fan turbo)
Cinebench R20 : 3023 (with fan turbo, switched from button)
in both R20 benchmarks, thermal throttling occurs with cpu reaching 92°


Cinebench R15 : 1246 (without fan turbo)
Cinebench R15 : 1259 (with fan turbo, switched from button)
No thermal throttling in R15 benchmark

Instead regarding the management of the gpu in the timespy benchmark, perhaps it is your undervolt that allows you to have stable frequencies and therefore a total of higher FPS because there is no power throttling.
In any case I'm trying to figure out where to download the Gaming Center ... I do not see any download section from the pcpecialists site and I have not been given any indication about software to install.

Firestrike benchmark : 15154

Other differences from GPU-Z:

mine: PCIe 3.0 x16 3.0 @ x16 1.1
your: PCIe 3.0 x16 3.0 @ x16 2.0


Repeated 3Dmark time spy test:

Graphics score: 6 183
Graphics test 1: 40.08 FPS
Graphics test 2: 35.62 FPS
CPU score: 7 056
TOTAL: 6299

 

[roakes]

Definitely looks like CPU is not the issue then.

That GPU-Z Bus Interface value will change depending on the power management of the GPU, press the question mark button next to it, read the information then hit Render Test, mine jumped straight up to PCIe 3.0 x16 3.0 @ x16 3.0, make sure yours does the same.

Gaming Center should be available in the downloads section of tech support when logged into your PCS account. If not then I've uploaded it here: https://drive.google.com/open?id=1pRyOt1ZpO8H9TZi1wTSaNg1j3LLNppvY

 

[bana845]

GPU-Z show me "PCIe 3.0 x16 3.0 @ x16 3.0" when i execute render test as you said, so this is fine, are the same value.
I downloaded the Gaming Center software, set game mode, but the result is the same.

Now are remained only two cases to try:
1. Install the video driver provided by the pcspecialists support instead the official from NVIDIA
- Which driver did you install between them?
2. Try the undervolt as you did
- I tried with MSI afterburn but it does not show me voltage curve, which software you are using for undervolt?

thanks for your patience

 

[bana845]

ok, give up

I tried setting the frequency/voltage curve in MSI afterburner like you did, with a cap of 1545mhz/725mV with the exact same result as the default.

1. If I launch the benchmark without underfrequency/undervolt, HWInfo shows that the frequencies in boost are high, the GPU reaches 1905mhz / 1.069V with the same FPS and same result (HWInfo show power limited for maximum values)
2. With frequency/volt cap to 1545mhz/725mV, different frequencies but same FPS and same result (HWInfo show power limited for maximum values)
3. If switch in game mode, same FPS and same result (HWInfo show power limited for maximum values)
3. Even you have power limited with the voltage cap you have set (see HWInfo screenshot you've uploaded in previous post)
4. The temperature never goes above 62 degrees in any benchmark until now.

Power limit is a must!
But at this point power limit is not the problem since all the monitoring tools detect the power cap reached, even in your underclocked tests.
GPU-Z shows the exact specifications of the video card, same build, same bios, same everything.
CPU score are the same.

At this point I do not trust 3DMark anymore.
I think I will try other benchmarks to see what happens.

Thanks!

 

[roakes]

I have the latest Nvidia driver installed, updated using Geforce Experience, it's 419.35

All this talk of power limits made me realise something, in your HWinfo screenshot your GPU power value was missing. Did you remove it? You may want to check settings or reinstall HWinfo to get it back.



While we both eventually hit the power limit, mine reaches the specified 115W TDP. I wonder if yours is getting this high or has a lowered TDP set somehow? For example, the Max-Q variant of the 2070 has a TDP of 80W, so it limits performance to keep cool and run in thinner laptops. The average Time Spy score for an 80W limited 2070 Max Q is 6193 - https://www.notebookcheck.net/NVIDIA-GeForce-RTX-2070-Max-Q-Graphics-Card.386281.0.html

How about we try a different test to monitor the GPU performance. Download Unigene Heaven and then run it with the following settings:



This should push your GPU hard and will be a constant looping test so you can monitor easily. My HWinfo screenshot above was taken after running Heaven for 5 mins. Let's make sure that your GPU Power hits the Watts we expect.

You could also try the Furmark GPU Stress Test:

https://geeks3d.com/furmark/

Just install and hit the GPU Stress Test button, my GPU Power went straight up to 115W and stayed there.



Also, hopefully a silly question, but did your laptop come with a 230W power brick? I know a 180W exists as PC Specialist offer it incorrectly as a spare for my system

 

[bana845]

yes, I have a 230W power brick.
I tried Heaven benchmark, I reach 115W at 60°.
With configuration you indicated it generates many FPS, in some cases it even reaches 180 FPS.

I also tried the Basemark benchmark:

-test with Vulkan API: 5184
-test with DirectX12 API: 5059

The card in both cases reaches 116W at 56 °. HWInfo detects power limit cap (as usual) and voltage reliability (in Vulkan more stressed).









If I repeat the tests with undervolt 1545mhz/725mV the results are the same.
HWInfo no longer detects the voltage reliability limit (rightfully) but continues to detect the power limit cap, although the voltage is locked at 725mV.
It seems that the performances, although with lower frequency, are compensated by the lack of throttling on the voltage.

Some basemark reference tests: https://www.geeks3d.com/20181129/te...sed-with-opengl-vulkan-and-direct3d-12-tests/
MSI RTX 2070 Ventus reaches 7430 points in the Vulkan test and 7120 points in the DirectX12 test.
Obviously MSI RTX 2070 Ventus is a desktop card with TDP 175W, Texture rate 233 and Pixel rate 104.

RTX 2070 Mobile has a 70% TDP/TR/PR compared to MSI RTX 2070 Ventus.
In this test gets a score of 70%.(too low)

In any case, both with Heaven and with Basemark, the card produces a high number of FPS.
I think it has enormous potential, considering also the modest temperatures it maintains, almost never above 60 °.
Maybe The block is given by the power limit, 115W are too few for the potential of this board.
If we were able to bring the TDP max to 150-160W, with this margin of 15-20 ° of temperature it would be a impressive.
Then it is to be seen how far it is able to remain stable in terms of frequency and voltage.

 

[roakes]

Here are my results:







When running the Heaven benchmark, make sure you have a custom resolution set, if you run at low resolution then yes you will get high frame rates. The results from this benchmark were done at 3840 x 2160

Looking at all of the scores you've provided and comparing to mine, it looks like you're getting 25-30% less performance. I'm pretty sure that your clock speeds are getting throttled by something, you may hit max clocks and TDP but your averages are always way too low, unless you're not resetting the data immediately before the test and the averages include a long period of idle.

Unfortunately I'm all out of ideas, you will need to call PCS and point them towards this forum thread to see what they think.

 

[bana845]

Surely I am power throttling.
During the banchmark, GPU-Z / Sensors shows that they are constantly power limited.
In order not to reach the power cap I would have to go to frequencies lower of 1GHZ, but performance stil the same, even without power throttling.
However I wrote to pcspecialists, let's see what they say.

 

[roakes]

Picking this up from our PMs, with the drivers installed, could you pull up HWinfo and show your Video Adapter data:



If yours still says TU106M instead of TU106, then it could be that they fitted a 2060 chip instead of 2070, which would explain a lot. The motherboard BIOS could be hard coded as a 2070 but still operates if the 2060 is fitted.

TU106M is listed here as being both the 2070 and 2060: https://devicehunt.com/view/type/pci/vendor/10DE/device/1F10

 

[roakes]

Ah, quick update to Hwinfo 6.02 (I was running 6.00 before) and now mine says TU106M as well...

 

[bana845]

Which driver have you installed? notebook driver or standard driver?
However 10DE - 1F10 is a TU106M [GeForce RTX 2070 Mobile]