LG Gram 2021 announced: 14, 16 and 17

Discussion in 'LG' started by RS4, Dec 16, 2020.

  1. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    And for GPU - yes for gamin the i7 is better too. Also at higher energy consumption.. Here I guess it´s again mostly the single core turbo that counts. Again if you need that performance - either use an eGPU or another notebook IMHO.
    3D Mark.PNG 3dmark11.PNG
     
  2. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    I always wondered how bright is the gram 16 at which percent...
    brightness_nits_gram_16.PNG

    vs power consumption (averaged 3 minutes on 10 second intervals to see where it stabilizes)
    Total system battery use at idle:
    0% 1.8w
    10% 2w
    20% 2.1w
    30% 2.2w
    40% 2.3w
    50% 2.4w
    60%2.5w
    70%2.65w
    80%2.9w
    90%3.3w
    100% 3.7w

    So 90% and 100% are kinda an inefficient overdrive. How much energz does it use at 0%. 200mw? So taht would mean at 270% this screen just consumes 1.3w. As we have a 80w hour battery - it is clear this screen is just damn conservative on battery power. If we had an M1 processor in this one - 30 hours battery life for surfing, working, and so on would be very likely... Just move your mouse and that laptop at 0% screen brightness uses as much as idle at 100% screen brightness. And if the 400nits 17" screen ramps up the same, it's clear that you effectively get better battery life on the 17". Size does not matter, it just matters to be in the efficient range. And likely this is more efficient than any oled screen right now, much more efficient. The first 200 nits increase needs just a tiny bit more energy than the last 100 nits increase. Battery life is about CPU, not about screen brightness.

    taken from here:


    Now that explains why the gram 17 is better on battery life. At 80% the gram 16 is already very power efficient vs 100%. Now at the gram 17 the 400 nits display will have a huge advantage - as it will work at a lower percentage of its max brightness, not getting into the top 20 percent where it becomes inefficient as often. And even in pitch dark room I feel the gram is really low on contrast at 20% brightness. 30% brightness however is much better. I always tend to avoid 0-25% brightness and if on battery above 80%..

    This is a comparison between Samsung Ion2, Galaxy book pro and Gram 16. That gram 16 had pretty bad contrast, as some other test had 1500:1 contrast. And yes below 20% that display becomes really low contrast - that is quite visible... As I usually use 80% - that is maybe 270 nits. I kinda guess the measurement device they used, or their room was at 0.04 nits. Because that OLED screen should be much better. Really strange for me that the energy consumption of the lg gram is in no way linear with the nit output.

    I think the gram 17 has the improved display if sold with bios version 300. Some gram 16 are now also on bios 300. Maybe they got better? They should really mention the bios version in each review as I feel the screens are constantly improved a bit. And so strange that even though quite a few European countries are just now getting the LG Grams in, my LG Gram from Spain is on January bios, meaning it was produced pretty long ago and just sitting around in some warehouses... On the gram 17 maybe you have better chance of getting a recent model.



    And you can see the actually biggest flaw of the gram 16 in this video. Absolutely no coating preventing reflections. The OLED screen on the Galaxy Book Pro is also glossy, but it cuts reflections sooo much better than the gram without losing sharpness. LG really has to put some sort of coating on it, and I think they did that on newer version of the 17". If you have the 300 bios you also have an anti reflective coating (while still being so slight to be considered a glossy display, like the Galaxy Book Pro). But yeah the reviewer said at the end he was surprised to personally end up preferring the gram over the book pro. Because of better brightness, better resolution, better aspect ratio and better keyboard and I think better speakers. But if he would be younger he would go for the Book Pro.


    Really LG needs to put in an 500 nits screen (best OLED), put on a light anti reflective coating, upgrade to Alder Lake ASAP, and maybe work a bit on the cooling. But with Alder Lake that should not be needed as much. If Alder Lake is not losing much battery life on 4K then go for 4K OLED.. Or give us an 2000:1 contrast Igzo panel and stick to factory calibrating it. Haha plus my personal opinion that not everyone will share, get rid of that numpad and center the keyboard plus spread it out a bit, Include a proper Enter key and move down the arrow keys a bit.
     
    Last edited: May 13, 2021
  3. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    oh yeah - and look at the charger of the Galaxy book pro. It's about half size, and one third the weight but charging quicker. Samsung really included latest technology there. Its a freaking 60w charger at the size of your smartphone charger. I would like to know how it stacks up vs the RAVPower 65W PD Charger 2-Port Foldable iPhone Fast Wall Charger, GaN Tech & PD 3.0 Compact Type C that is AFAIK right now the smallest/lightest 65w charger on the market not from Samsung. The LG charger is GAN, but kinda the cheapest they could possibly find in China I feel. Right into the rubbish bin with it or use as a charger to never travel cause its heavy and huge...Damn my 2 year old china GAN charger is soo much smaller than that LG crap. Lenovo X1 Carbon Gen9 now also comes with a super compact 60W charger..

    I need a two port USB-C charger anyhow - so while travelling can charge laptop and phone at the same time. And not carry two chargers with me. I still take a tiny 18w Samsung USB A charger for some older stuff without USB-C for now. Better than a 3 port 80w charger.
     
  4. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    Check this setting in the registry: Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power\PowerSettings\DE830923-A562-41AF-A086-E3A2C6BAD2DA\13D09884-F74E-474A-A852-B6BDE8AD03A8
    You can adjust it to 100% if you like full brightness with energy saver mode here. I wonder why yours must be set lower here. Either change it to 100 or set the attribute to 2 so you can edit this value in the windows power plan. Full brightness is possible in any mode.
     
  5. skipper63

    skipper63 Notebook Consultant

    Reputations:
    4
    Messages:
    110
    Likes Received:
    26
    Trophy Points:
    41
    It is at 100% set in the registry. Please try again. In battery saver mode put your brightness at 100%. Then move the slider to the right from battery saver to better battery and see if you don't see an additional increase in brightness. This is how it works with mine (obviously unplugged, if you keep it plugged it has absolutely no difference).
     
  6. skipper63

    skipper63 Notebook Consultant

    Reputations:
    4
    Messages:
    110
    Likes Received:
    26
    Trophy Points:
    41
    Yesterday I went out a bit for window shopping. In one Mediamarkt here in Germany (it is kind of the Best Buy equivalent) they had the new Samsung Book Pro in display, the 13inch. Looks almost fake, how light it is. But the keyboard is completely NOT acceptable for me, so I don't have any buyer's remorse with my Gram. I could never work with such a shallow keyboard. The flex on this machine was also too much. I didn't have to spend a single second more with this laptop. So for me LG is the winner by far in balancing all the options.
    I checked also the new M1 Macs, I had forgotten how horrible keyboards they also have... and how heavy they feel for their sizes
     
    RS4 likes this.
  7. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    I have 100 percent no matter if plugged in or not. There is a tiny change by deactivating the power saving options in the Intel graphics control center, also independent of mode. They do actually lower static contrast, as fully white display state is dimmed a tiny bit, but black not. Cannot really see much power use difference either.

    ANSI contrast is the same, so I keep it active. In general would be nice if contrast would be measured with checkerboard ANSI and not full black/white. Cause that is way more important.

    For me also keyboard and resolution disqualify the galaxy book pro. Also max brightness being very low. However that should equal out because of less reflections.
     
  8. RS4

    RS4 Notebook Consultant

    Reputations:
    32
    Messages:
    181
    Likes Received:
    43
    Trophy Points:
    41
    So, what is your final conclusion? for lighter workload get the i5 and then put in performance mode?
     
  9. hfm

    hfm Notebook Prophet

    Reputations:
    2,250
    Messages:
    5,264
    Likes Received:
    2,993
    Trophy Points:
    331
    Comparing the i5-1135G7 and i7-1165G7 on a single sample basis even in the same model laptop is pretty much useless. The ONLY difference between these two CPUs on paper is:
    • Intel's binning for stability at higher frequencies, depending on demand and revenue forecasts they could even just FORCE more i5 or i7 into the market by enabling/disabling parts of the package.
    • 4MB (8 vs 12) more L3 cache on the i7
    • 16 more EUs for the iGPU (therefore 128 more shader units as a consequence)
    Simply the variance for tolerance of all the components in the system could have an effect on exactly how much power is consumed, not to mention variance from CPU to CPU itself even among the same exact CPU model (say 15 samples of i5-1135G7). This is the reason some CPUs can overclock or undervolt better than others. The variances from heat management unit to heat management unit as far as contact area and airflow etc.. could also have an effect. Since these CPUs are so close to one another in architecture the i5 could easily outpace the i7, especially in workloads where the additional L3 cache makes no difference, just based on how well all the parts are working together as a whole (heat management, power delivery, the unique CPU itself, list goes on) in any one laptop that came off the assembly line.

    I could perhaps see a scenario where MAYBE the iGPU using more power could have an effect on the balance between the CPU cores and GPU power consumption in a combined workload situation, especially where the heat management is saturated in shedding the heat from the packages. But I would imagine if you bought 15 of the same exact Gram 16 units with the same CPU you would probably find that the idle power consumption and ceiling for a workload like cinebench would vary between them all.
     
    skipper63 likes this.
  10. extremecarver

    extremecarver Notebook Consultant

    Reputations:
    3
    Messages:
    126
    Likes Received:
    11
    Trophy Points:
    31
    I don't really think so. It's been for quite a few generations that i5 uses less power and runs cooler. And there is no more undervolting where you could get lucky. So far all tiger lake seem to be the same stepping. Usually new stepping equal lower voltage. When I compared both it actually seemed that the i5 uses a tiny bit higher voltages.

    Single core and graphics the i7 will be 5-15 percent faster. However at the expense of using way more power. Multicore the i7 uses only a tiny bit more power but is 5-10 percent slower due to heat once throttling starts. Maybe it's hotter because more things in the same space. And in my case the i7 was able to run at 1w more power, so the cooling definitely wasn't against it. I will at some point repaste the i5, maybe try to fix another heatpipe, but space seems limited for that.

    I looked a bit a some single core power usage - and it really seems anything above 3500 mhz or so really needs incrementally more power. I should have limited the i7 to 4200 mhz to have a fair comparison. However if you need to limit the i7 to 4200mhz, because it just needs unreasonably much power when going faster - there is not much sense in buying the i7. It really seems Intel can only reach speeds over 3500-3600mhz or so by using loads of power. The faster it is running , the worse the efficiency. However until 3200-3500mhz or so the efficiency is really good.

    If I set EPP to 180 running an installer that extracts many small files (5-10MB each) at 3500 mhz it uses around 8w. At EPP 128 and 4200mhz it needs around 14w. Even though this is single thread, I think because of many small files, it taxes the CPU more than simple CPU benchmark.
    So lets stop the time how long the installer needs at different EPP to get this correct:
    EPP 128 - package power around 13.5w - speed about 4000mhz, C0 around 16%: 2:00.63 minutes
    EPP 170 - package power around 9.5 - speed around 3500mhz - C0 around 14%: 2:19:56 minutes.
    EPP 80 - package power around 14w - speed about 4100 mhz - C0 around 16%: 1:59.08 minutes.
    As I already more or less knew - single core EPP 128 at least on the i5 more or less gives full speed. EPP 80 is more relevant when running two tasks minimum. EPP 10 would maybe shave another second.
    So 13.5w*121=1633.5
    9.5w*140=1330

    So increasing the turbo even more - will just run into much worse efficiency regions. If you are using power not battery, and your laptop has good enough cooling - great. The i7 is the way to go. The i7 would have likely used around 20w for this task at EPP 128. While doing it in 1:48 minutes or so (I'm just guessing from the numbers I know previously, I don't have an i7 to compare anymore - This is a very positive guess for the i7. As the speed increase usually was even less while it used around 6w more if not throtling.).
    That would be an hypothetical:
    20*108=2160....

    Now let's see EPP 210 and EPP 255.
    EPP 210 - package power around 5.8w - speed around 2500mhz - C0 around 14% 3:18.79
    5.8*199=1154 (however mind this is pure package power - the total consumption would add around 2-3w to all of this figures for screen, nvme, mainboard and so on). EPP 210 cannot increase the efficiency in reality anymore. (I do know 170-180EPP gives me best efficiency. Likely 180 is the best)

    EPP 255 (now everything is sluggish - if another task is running. It's painful to even type).
    package power around 4.4w - speed around 1800mhz, C0% around 14.5%, 4:30.80
    4.4*271=1192 - we are getting really inefficient here considering now even the CPU package power itself is less efficient.

    The sweet spot in speed for the tiger lake seems to be around 3000mhz. Both lower or higher the efficiency drops. Now the lg gram cooling on the i7 really is bad - because for all core power the speed will drop below those 3000mhz, even if only a little after some time. On the i5 this will only happen when not using the performance cooling mode.

    And this also explains why on all core turbo the i7 didn't doo as bad. Yes 5-10% slower. And 1w more power draw (that is only about 4% more power). It simply runs in less efficient speed range because it has to drag along the Cache and added execution units of the GPU. If the cooling was better - say 10w more cooling then the i7 would get much closer. If the cooling is soo good that neither processor throttles - the i7 will pull ahead. However in the latter case, get an i9 or a processor with more cores. The i7 U is the wrong choice.

    I am pretty sure, EPP values are relative to max speed. So if you want the i7 at the same speed as the i5, you need higher EPP values. EPP 170 on the i5 is likely equal to 185 on the i7 or so.. So the i7 has a higher EPP value sweet spot than the i5. I do not remember the 2core max turbo for the i7. The i5 has the same 4200mhz as single core turbo - and that is more than the LG gram can cool so it will throttle. Here the i7 disadvantage would be lowest - as the throttling is soo little, that it increases efficiency as it is not dropping to below 3500mhz.

    Oh and another thing - more threads = more efficiency. I just ran Cinebench R20 at 4 threads. It gave me 1513 points - vs 1920 at 8 threads. C0% was around 51.5 percent and max consumption around 30w. So it took longer until it throttled, but it still throttled. It did manage to keep running around 3350mhz vs 2900. If the cooling could get around 30w instead of 22w all core turbo - the CPU would run more efficient on all core load. It is really clear - to catch up performance vs AMD tiger lake U is actually optimised for around 28w...


    Cinebench R20 - 2 threads at EPP 170 = around 3480-3580mhz. = 15w = 847 points.
    2 threads at EPP 80 = around 4100 mhz = 22w - however soon throttling to 19w and 3800 mhz or so. 911 points. Also clearly much less efficient. If it would not throttle it would score a bit better, but be even less efficient.


    All of the above at 25 degrees room temperature. At 20 degrees scores would be better of course.


    But yeah of course at real idle - meaning 75% in C10 the difference in power draw between i5 and i7 is too small to measure. Now C8 is about 1w more vs C10 for 75% idle on the I5. Already at light load the CPU is spending kinda all in C7 or C8 and not C10. The difference between C8 and C10 is that the L3 cache is switched off. So switching off 6MB of cache gives you 1w advantage - that is the same as not having those 6MB. So just the L3 cache not being switched off 4MB additional is about 0.66w higher consumption. Then comes the graphic card and the higher speeds. For battery life the conclusion is dead clear - get the i5. If you do not care about battery it is a difficult question on the LG Gram. Because this additional single core speed is not helping you much, even more as it is only about 300mhz because of throttling. Then however you carry the additonal 0.66w around all time - shaving off your speeds - and worse making the PCU run less efficient if you load 4 threads or more (or 5 threads or more in performance mode). The 16 execution units will also carry some excess wattage..
    So if you want the fastest notebook for productivity/office work and do not care about battery life - that is the only workload where the i7 makes sense. It will be up to 10% faster. (productivity work does not need GPU much - because GPU is more like 15% faster). Games will run faster on the i5 as they are multi core load.

    And this also means - if you can improve the thermal solution of the Gram - it will run longer on battery! This will be negligible on light load (only the fan spinning less if you manage to get more heat to the fan without the fan spinning faster) - but quite noticeable on medium to heavy load.

    However even for office work the 1145G7 would likely nearly match the i7. So the i5 1145G7 which is not available on LG Gram would be the perfect processor for office productivity on AC power, while you can just throttle it down to same efficiency as the 1135G7 for battery use.

    The i3-1125G4 will be kinda the best for battery life. It's speed are so low that it rarely reaches the inefficient top speeds - and the much smaller graphic card will save some more. The graphic card should still be good enough for anything that is not gaming or needs GPU a lot. For videos it will be plenty fast. However intel prices the i3-1125G4 just a tiny bit below the i5 - so I do not think we will see this CPU out in the wild at all.
    For real improvement we will need to wait for Alder Lake. That CPU will be the first Intel CPU since years to really drop down power consumption. In the last years it was much more the components besides the CPU that helped increase battery life, but not the CPU (with Ice Lake being the worst).
     
    Last edited: May 15, 2021
Loading...

Share This Page