Gaming on a Laptop while on Battery

Discussion in 'What Notebook Should I Buy?' started by PJPeter, Oct 11, 2018.

  1. Blacky

    Blacky Notebook Prophet

    Reputations:
    1,951
    Messages:
    5,131
    Likes Received:
    893
    Trophy Points:
    281
    It's a sound logic.
    I searched for laptops with low powered processors and gaming video cards and there aren't many options: noteb search.
    The best one I found is the surface book with GTX 1060, the rest are GTX 1050. There is also the option of going with an Intel+Vega setup.
    The Surface seems to have good battery life while gaming: https://www.notebookcheck.net/Microsoft-Surface-Book-2-15-i7-GTX-1060-Laptop-Review.284622.0.html
     
  2. luke1333

    luke1333 Notebook Guru

    Reputations:
    0
    Messages:
    61
    Likes Received:
    12
    Trophy Points:
    16
    XPS 9570 maybe?
     
  3. Blacky

    Blacky Notebook Prophet

    Reputations:
    1,951
    Messages:
    5,131
    Likes Received:
    893
    Trophy Points:
    281
    That thing throttles pretty badly, but it might be OK on battery. It's not a bad suggestion.
    I looked at reviews and in games it manages just 1 hour of battery life, vs the Surface which gives you 2 hours+.
     
  4. luke1333

    luke1333 Notebook Guru

    Reputations:
    0
    Messages:
    61
    Likes Received:
    12
    Trophy Points:
    16
    Yeah don't know how intensive games OP is going to run but I wouldn't think xps or surface would be good for many games
     
  5. PJPeter

    PJPeter Notebook Deity

    Reputations:
    110
    Messages:
    1,122
    Likes Received:
    35
    Trophy Points:
    66
    Thanks very much, I really appreciate the noteb search link - that customizable filter database/search is great, thanks.

    I came across that while doing some futher research over the weekend.

    I actually purchased a Y740 15" RTX2060 with hopes of fair performance while on battery.

    I noticed a lot of stuttering and variability in performance - basically the game (Mass Effect Andromeda) was unplayable at Ultra settings in any sort of battle when on battery. Even High wasn't much better - Medium with custom scaling disabled was Ok (I checked the performance guides for Mass Effect Andromeda, as well as checking into the nVidia GeForce Experience "optimized settings" - though I only found the latter today and didn't test them previously)

    To get some numbers on it, I decided to run 3DMark 11 runs on and off power.

    Results were not good - even excluding outliers, these are average runs:

    On Battery:
    P9804 https://www.3dmark.com/3dm11/13220955

    Plugged in:
    P17105 https://www.3dmark.com/3dm11/13220878

    I re-ran the tests a bunch of times to ensure - the average on battery was approximately P10000, while plugged in was P17000 (therefore ~59% performance on battery).

    I confirmed that I was on Max Performance for all these tests (including the top one - though thankfully that was an outlier).

    As a comparison, here's a Surface Pro 4 I had lying around:

    On Battery:
    P1591 https://www.3dmark.com/3dm11/13220847

    Plugged in:
    P1662 https://www.3dmark.com/3dm11/13220924

    Therefore that system (using an integrated Intel 520) runs at ~95% speed when unplugged.

    So back to looking for lower power MaxQ GPUs, I did some more research:

    "As with the CPU, we could detect a loss of GPU performance when the X1 Extreme ran on battery-power. Instead of scoring 8752 points, the X1 Extreme reached just 6606 points in the 3DMark11 Performance GPU test." - https://www.notebookcheck.net/Lenovo-ThinkPad-X1-Extreme-i5-FHD-GTX-1050-Ti-Max-Q-Laptop-Review.339576.0.html (75% performance on battery)

    "On battery, the G7 hits about 90-95% of its "plugged-in" graphical performance." - https://www.notebookcheck.net/Dell-G7-15-i7-8750H-GTX-1060-Max-Q-Laptop-Review.317686.0.html

    Combining all that with the details I found before (such as https://www.techspot.com/article/1571-gaming-laptop-battery-performance/) shows that the higher power GPUs are not the way to go - too much stuttering from power constraints on battery.

    I'm now boxing up the Y740 to return (hoping to drop it off for shipping about 2 hours) and I've ordered a Razer Base Model GTX1060MaxQ (heavily discounted now because of the new models coming out) to do further testing.

    If anyone has any suggestions on tests to do on the Y740 RX2060 in the next couple of hours I'd be happy to try.

    Seems like whatever the successor to that card (GTX1660 MaxQ maybe?) will be the sweet spot. Because these GPUs are close to the power limits of what the battery can provide, they can provide better performance that the top line cards and much longer battery life.

    So basically the summary of my thinking is: Since gaming on battery is my primary use case - why get something that gives P10000 on battery when I can get a cheaper and longer lasting card that gives approx P13000-P14000 both when plugged in or not?

    Thanks,
    Peter
     
    sa7ina likes this.
  6. Blacky

    Blacky Notebook Prophet

    Reputations:
    1,951
    Messages:
    5,131
    Likes Received:
    893
    Trophy Points:
    281
    Thank you for the feedback. The GTX 1060 and the RTX 2060 have the same power requirements. So going with the GTX 1060 Max-Q should perform the same as the RTX 2060.
    Now it might be a model/firmware issue, not necessarily an HW problem. I do agree that the upcoming GTX 1660 Max-Q will be the best for what you want.

    Anyways, my advice on your issue is that it is HW related or maybe it could be fixed by further dropping down the clocks only on battery, say by 10%, using MSI afterburner or similar software.

    I have an old GTX 280M, it has 75W TDP, which is roughly the same as the RTX 2060. I had this card for over ten years and have gamed many times on battery with it, and yes, on battery, it will drop to around 60% of its performance. All laptop video cards will downclock themselves, and you will get around 50-70% of their typical performance while gaming on battery. The lower powered video cards, like the GTX 1050Ti, will downclock less, but I think overall, the RTX 2060 will still give you better performance on battery vs the GTX 1050Ti on battery. I think notebookcheck has tests for how gaming video cards performance on battery. Just check for instance their review on the XPS 9570 to see what to expect from a 1050Ti on battery.
     
  7. PJPeter

    PJPeter Notebook Deity

    Reputations:
    110
    Messages:
    1,122
    Likes Received:
    35
    Trophy Points:
    66
    Here are my results with a Razer Blade 15 RZ09-02705E75-MSU1 Laptop (Base Model) w/1060 Max Q GPU on clearance for $1799 CAD+tax.

    I have the system configured using an undervolt through ThrottleStop of -0.1357V on the CPU Core/Cache and -0.0508V on the iGPU. This resulted in about a 4*C drop in temperatures, which is very much appreciated - and hopefully longer battery life too.

    Razer Blade 15 GTX1060 MaxQ Plugged in:
    https://www.3dmark.com/3dm11/13255291
    SCORE
    P13 130 with NVIDIA GeForce GTX 1060 Max-Q(1x) and Intel Core i7-8750H Processor - Plugged in with Whispermode disabled (through GeForce Experience Settings) and Max performance
    Graphics Score 14 163
    Physics Score 10 809
    Combined Score 10 723

    Razer Blade 15 GTX1060 MaxQ On Battery
    https://www.3dmark.com/3dm11/13257512
    SCORE
    P9 169 with NVIDIA GeForce GTX 1060 Max-Q(1x) and Intel Core i7-8750H Processor (with Battery Boost Disabled through GeForce Experience Settings)
    Graphics Score 10 820
    Physics Score 8 150
    Combined Score 4 686
    Result: 76% of Plugged in Graphic Score while on battery

    Compare that with the on Battery Lenovo Y740 Score above:
    SCORE
    P9 804 with NVIDIA GeForce RTX 2060 (Notebook)(1x) and Intel Core i7-8750H Processor
    Graphics Score 11 508
    Physics Score 9 805
    Combined Score 4 647
    Result: 1060 Max Q is giving 94% the battery performance of a RTX2060 laptop with much better battery life and most likely less heat and stuttering.

    And with the Plugged in Y740 Results above:
    SCORE
    P17 105 with NVIDIA GeForce RTX 2060 (Notebook)(1x) and Intel Core i7-8750H Processor
    Graphics Score 20 410
    Physics Score 11 910
    Combined Score 10 967
    Result: Plugged in, the 1060 Max Q is giving ~70% the plugged in performance of the plugged in RTX2060 laptop.

    People complain that the new RTX20 series and the upcoming GTX16 series aren't as much of a leap as the GTX9 to GTX10 were - they are right. But that's because nVidia isn't screwing around with increasing the TDP as much as they did then. GTX 860M/960M Laptop cards had TGP of 60W - GTX1060 jumped that to 80W. Of course it is faster and closer to desktop spec - it's eating juice much closer to a desktop card. But battery life suffered.

    So unless you want something like an ALIENWARE AREA-51M which is really a desktop card/system put in the general shape of a laptop, then it's a good thing they didn't just keep increasing the wattage. I'm willing to do with a bit lower performance if it means I can use this as my daily driver and play games on the go.

    And Battery Boost is great for longer battery life on the go by locking to max 30fps - which is fine for me in most instances - I just turned it off to get a feeling for the true potential in the benchmark (since it locked the score lower there as well).

    I did a lot of mucking about with the undervolting - I got it to the point where I found my Fn keys didn't work (-0.147V) and I had random restarts when running 3DMark 11 in full experience mode. I tried out Superposition, but I found even on Extreme it didn't stress the system as much as a 3DMark 11 run did. I also stress tested with XTU and it passed.

    One additional tweak with Throttle Stop - not disabling CPU Turbo Boost Mode when on battery results in a higher Physics score, but lower battery life and slightly lower graphics score:

    On Battery with Turbo Boost Enabled:
    https://www.3dmark.com/3dm11/13257007
    SCORE
    P9 425 with NVIDIA GeForce GTX 1060 Max-Q(1x) and Intel Core i7-8750H Processor
    Graphics Score 10 683
    Physics Score 10 041
    Combined Score 4 774

    Let me know what you guys think and if you have any questions - I didn't post any of this really to help myself, but more to help the community and others who have the same needs as me.

    And if anyone has any way to actually undervolt the nVidia GPU on this laptop I'd love to try it - I haven't found a way, I tried both nVidia Inspector and MSI Afterburner and both seemingly had voltage settings locked down.

    Thanks,
    Peter
     
    Last edited: Apr 13, 2019
  8. PJPeter

    PJPeter Notebook Deity

    Reputations:
    110
    Messages:
    1,122
    Likes Received:
    35
    Trophy Points:
    66
    Analysis comparing GA502 to Razer Blade Base 2018 for battery gaming:

    The GA502 is giving much lower performance on battery than I had hoped. I don't know if there is some other setting to prevent it, but as it is I'm guessing it's stuttering a great deal because of lack of power.

    I also tested a Razer Blade RTX2070 in person at the MS Store and I found it was able to consistently give 56-60% of plugged in performance (without stuttering) even down to 14% battery while in 3DMark11 which was 60fps locked on battery (and therefore gave lower results).

    My Lenovo Y740 tests though showed massive stuttering and as the battery drained the performance would get significantly worse as the available battery voltage dropped (it only has a 11.52V 3 cell battery as opposed to the 15.4V 4 cell in the Razers and I assume in the GA502). So I'm wondering if there's something different that Razer is doing to allow consistent high performance on battery since I know too that with other laptops I've owned battery gaming is like watching a slideshow because of poor power management.

    I didn't include Turbo runs in my analysis of either system - just the normal plugged in results. And I didn't enable throttle stop/undervolt on the Blade Base since he didn't do any such tweaking on your GA502.

    Analysis of Performance/Power Results Comparing Plugged In/Battery from TimeSpy Runs:

    Razer Blade Base 2018:

    Max GPU Power Plugged In: 73W
    Max CPU Power Plugged In: 43W
    https://www.3dmark.com/spy/7157396
    SCORE
    3 717 with NVIDIA GeForce GTX 1060 Max-Q(1x) and Intel Core i7-8750H Processor
    Graphics Score 3 550
    CPU Score 5 080

    Max GPU Power on Battery (17%): 55.7W
    Max CPU Power on Battery (14.8%): 41.1W
    https://www.3dmark.com/spy/7156524
    SCORE
    3 306 with NVIDIA GeForce GTX 1060 Max-Q(1x) and Intel Core i7-8750H Processor
    Graphics Score 3 110
    CPU Score 5 143

    89% of Plugged in Score while using up to ~83% of max power (Slight Power Throttling)

    ASUS Zephyrus GA502 (2019):
    Max GPU Power Plugged In: 61.7W
    Max CPU Power Plugged In: 10.8W
    SCORE
    4 552 with NVIDIA GeForce GTX 1660 Ti(1x) and AMD Ryzen 7 3750H with Radeon Vega Mobile Gfx
    Graphics Score 4 932
    CPU Score 3 170

    Max GPU Power on Battery (27.6%): 31.9W
    Max CPU Power on Battery (26%): 10.6W
    SCORE
    1 785 with NVIDIA GeForce GTX 1660 Ti(1x) and AMD Ryzen 7 3750H with Radeon Vega Mobile Gfx
    Graphics Score 1 657
    CPU Score 3 179

    39% of Plugged in Score while using up to ~59% of max power (High Power Throttling)

    You can see from the results that CPU is not throttled on the GA502 - it always uses about the same amount low amount of power for the same score on battery or not. The GPU though is heavily throttled.

    The result is very unfortunate to me since I had such high hopes for this low power CPU and high performance GPU combination, but I'm very glad I found this out before driving all the way across the border to buy one...


    Peter
     
  9. PJPeter

    PJPeter Notebook Deity

    Reputations:
    110
    Messages:
    1,122
    Likes Received:
    35
    Trophy Points:
    66
    Comparisons:
    Blade Base 1060MaxQ Battery / Adv Blade RTX2070MaxQ Battery / Blade Base 1060MaxQ Plugged In / Adv Blade RTX2070MaxQ Plugged In
    https://www.3dmark.com/compare/3dm11/13300097/3dm11/13299795/3dm11/13300088/3dm11/13299802

    Y740 RTX2070MaxQ Battery / Adv Blade RTX2070MaxQ Battery / Y740 RTX2070MaxQ Plugged In / Adv Blade RTX2070MaxQ Plugged In
    https://www.3dmark.com/compare/3dm11/13288206/3dm11/13299795/3dm11/13286059/3dm11/13299802

    My Blade Base on Battery at ~50% vs My Blade Base Plugged In vs ASUS GA502DU Plugged In: https://www.3dmark.com/compare/spy/7118508/spy/7111916/spy/6932283
     
    Last edited: May 12, 2019
  10. PJPeter

    PJPeter Notebook Deity

    Reputations:
    110
    Messages:
    1,122
    Likes Received:
    35
    Trophy Points:
    66
    Plugged In TimeSpy on Lenovo Y740:
    https://www.3dmark.com/3dm/35951519

    6 623 with NVIDIA GeForce RTX 2070 Max-Q(1x) and Intel Core i7-8750H Processor
    Graphics Score 6 572
    CPU Score 6 931

    Max CPU Power: 64.877W, 4.26W Min
    Max GPU Power: 81.151W, 7.999W Min
    Total System Power: 168.431W Max, 24.859 Min

    Battery TimeSpy run Y740 2070 After Hotfix (starting at 48% battery):
    https://www.3dmark.com/spy/7122324
    1 808 with NVIDIA GeForce RTX 2070 Max-Q(1x) and Intel Core i7-8750H Processor
    Graphics Score 1 641
    CPU Score 4 272

    Max CPU Power: 27.8W, 4W Min
    Max GPU Power 56.2W, 11.16W Min
    Total System Power: 69.5W Max, Min 27.6W

    Down to 30% battery at end of run (70% brightness throughout)

    27% of Plugged in Performance Score while using 57% of combined max power on components, and 41% total (Severe Power Throttling)

    Conclusion:
    The Lenovo Y740 with it's limited 11V 3 cell battery is not the one to buy if you need high performance while away from a charging outlet...

    Peter
     
Loading...

Share This Page