*OFFICIAL* Alienware Area-51M R1 Owner's Lounge

Discussion in '2015+ Alienware 13 / 15 / 17' started by ssj92, Jan 8, 2019.

  1. QUICKSORT

    QUICKSORT Notebook Evangelist

    Reputations:
    206
    Messages:
    671
    Likes Received:
    525
    Trophy Points:
    106
    Okay let me explain it to you as simple as possible.

    So basically indeed that curve shows at what GPU core frequency how much voltages to use. And these are the four base rules:
    - Moving the whole graph up is an overclock. (without changing voltages, so forcing the GPU to run at higher clock frequencies at same voltages)
    - Moving the whole graph down is a down clock (same as above but lowering the frequencies)
    - Moving the whole graph to the right is an increase in voltages. (to make your gpu use higher voltages on same Frequencies)
    - Moving the whole graph to the left is a decrease in voltages. Basically Downvolting. (To make your gpu use lower voltages on same frequencies)

    Of course it's not possible to move the nodes to left right. But you have to adjust them just like as if you moved the whole graph to the left/right.

    Before we get into it. At a given Frequency, the GPU needs a certain amount of voltage to be able to do all calculations at said Frequency correctly, Give it too much voltage, you can physically harm your machine. Lower the voltages too much, it won't have enough voltages to reliable do all calculations correctly at that frequency. How much exactly it needs depends on whole lot of factors, where "silicon lottery" one of the most "famous" one.

    Now. Unfortunately downvolting isn't as simple as moving the whole curve to the left in this case. It could work for certain machines but not all. Definitely not on this Alienware Area 51M, if your target is to get lower temps. Because there is the boost frequency that lowers itself the higher the temperature goes. And there is the constant power limit throttling that you get (I didn't try 200 watt vbios, don't want to increase the GPU power limit, essentially allowing more heat).

    So there are a few things you should know. changing the frequency without changing the voltages can affect temperatures, However changing voltages affects temperatures significantly. So you have to know that both can affect temperatures where voltages of course the main source of heat.
    Besides that. When the GPU is at 100% usage, for the Area51M It CANNOT sustain mv of >= 1000. Due to the power limit I reckon (perhaps also the temps a little but I will get there).
    And then there is the behaviour of the Nvidia core clock Boost, So as you can see on the curve. Even though the GPU is rated at 1590 MHz, you can see the maximum frequency is around or above 1950 on that curve, so there is that first, (of course you can also see to sustain those boost frequencies, you have a significantly higher core voltages in use compared to like 1590 MHz). But that is not only it when it comes to boost frequencies. When your temperatures start to get I think above 50-60 degrees, (not sure but I'm pretty positive it's around there), nvidia actually downclocks the whole GPU, And I mean actually move that whole curve down little by little, the higher the temperatures get.

    So even though I selected 1875MHz on 875mV. If temps get high enough, you might get results such as 1860MHz or 1845MHz while you are still on 875mv, even though 1860MHz and 1845MHz should run at lower voltages according to my curve.

    So knowing all this, let us get to my curve.
    So I monitored my gaming, During the heavy games, where gpu usage is 100%, (which is where downvolting matters the most), My GPU was around 1780-1860 MHz, fluctuating roughly within these values. And the votages were fluctuating within roughly these values: 925-975 mV. Especially when reaching 82 degrees, it wouldn't even go higher than 925mv.

    So I realized that basically 925mv is still too high for my GPU. And I basically don't want my GPU to reach those mV. So as a starting point for my first real attempt of undervolting after doing tests and looking stuff up myself, instead of mimicking other people their curves.I decided that I want to be at 875mv MAX at any given time.

    Now looking at the stock curve of course what you could do is, Flatten out everything to the right of the node which is on 875mV. So on the stock curve you see the node of 875mv on 1725MHz.
    So when you flatten out everything after that node. The highest point of your curve is now 1725Mhz. There is no other node that is on a higher frequency, which basically means, that you put a Manual frequency top limit on your GPU. So your GPU should never go above 1725Mhz if configured like that (it can still go a little higher depending on temperatures and behavior of boost frequency. But It should never like reach 1800Mhz anymore)

    HOWEVER, there is also this. When there are multiple nodes on the same frequency, only the node with the lowest mV will be used, the rest is ignored as if it doesn't exist. So the node for 1725MHz with the lowest voltage is 875mV, So all my games, even at 100% load, should never go roughly above 875mv (again it might vary a little due to boost behavior, but that should be marginal).

    Now doing this means that you lower the performance of your gpu. you simply did put a limit on your gpu. Yeah sure gpu will get lower temps, but it will also get slower. To compensate for that, we are going to overclock what we have. So we now have a curve with a mv limit. which is at 875, as there is no other node that has higher mv with any frequency that is higher than the max frequency.

    So I said overclock was moving the curve upwards right. So that's what I did. Moving the whole curve, causing the frequency on 875mV to be higher. In my case 1875MHz. So it's an overclock of 150MHz. This basically compensates for the loss of performance. Now we basically overclocked our GPU. But set a limit to how high the Clock frequency can get, and how high the miliVolts can get.

    Now when you change that curve and hit apply, some nodes shift a little, not sure if that's done by the gpu or msi, but yeah it's there. And when you hit apply at higher temperatures like e.g. when above 70 degrees. You see immediately the whole curve jump down, essentially seeing the current state of nvidia boost performing downclock. Don't worry though, what you applied is still roughly correct. Even though you see at that moment a complete different curve, when you save that profile and reselect it, you still see roughly your original curve.

    To be quite honest, I didn't do a whole curve increase of 150Mhz, only like 105Mhz, but I increased the 875mv part 150Mhz, although when hitting apply it was changed on its own to 135Mhz. also allowing the gpu to get to 881mV at 1875MHz.

    But I haven't tested this enough yet. I will have to do more testing, and see if there is a better max voltage / frequency combination I can use while sustaining low temperate and stable frequencies.

    Hope this helps you understand a little more. and sorry for the long post. I wish someone explained it to me like this when I initially asked.
     
    Last edited: Apr 27, 2019
    SlickDragon, raz8020, Darkhan and 5 others like this.
  2. QUICKSORT

    QUICKSORT Notebook Evangelist

    Reputations:
    206
    Messages:
    671
    Likes Received:
    525
    Trophy Points:
    106
    Thank you for pointing out OC Scanner, never noticed that, or even heared about it. I ran it and got this result. Though I'm not familiar with what this means. I reckon it means, yous system is stable enough, but not 100%. So glitches and issue may occur?
    [​IMG]
     
    Vasudev likes this.
  3. Korben_Dallas

    Korben_Dallas Notebook Consultant

    Reputations:
    40
    Messages:
    124
    Likes Received:
    128
    Trophy Points:
    56
    also to me has reported 90% confidence but overclock on the core of + 124MHz
     
  4. Korben_Dallas

    Korben_Dallas Notebook Consultant

    Reputations:
    40
    Messages:
    124
    Likes Received:
    128
    Trophy Points:
    56
    if you need 5 degrees less on cpu and gpu without touching anything thermaltake massive 20 rgb does a good job
     
    Fire Tiger likes this.
  5. Korben_Dallas

    Korben_Dallas Notebook Consultant

    Reputations:
    40
    Messages:
    124
    Likes Received:
    128
    Trophy Points:
    56
    Is it normal that 3dmark says that the drivers are not approved? my gpu drivers are stock, does it have something to do with the problem that I can't update the vbios?
     
  6. DannyB513

    DannyB513 Notebook Evangelist

    Reputations:
    136
    Messages:
    315
    Likes Received:
    367
    Trophy Points:
    76
    I don't think it will show over 90% confidence level.
     
    raz8020 and Vasudev like this.
  7. freeekychakra

    freeekychakra Notebook Enthusiast

    Reputations:
    5
    Messages:
    29
    Likes Received:
    0
    Trophy Points:
    5
    After using Alienware M17xR3 for 8 years without any major issues whatsoever, just placed an order for Area 51m. Was quoted for £4300 on the website. Got it at a stellar price of £3500(I am in the UK).

    Configuration and add ons -
    i9-9900
    NVIDIA 2080 8GB
    17.3 FHD 144Hz screen with G-sync and Tobii eye tech
    512 M2 SSD (will add another 512 M2 and a 1 tb 2.5 Samsung Evo/Pro)
    8gb RAM (lowest since upgrading to 64gb is half as expensive from Amazon compared to Dell)
    Mcafee for 36 months
    Windows Pro
    Elite Gaming mouse
    Large Alienware mouse mat
    4 years of Premium support plus with onsite service
    4 years accidental damage protection
    - the last 2 was alone was quoted for 738 quid.

    Can I ask what make of the SSD does the system come with? Is the heat spreader included(for the occupied and empty M2 slots)? Any thoughts?

    For the 2.5 1tb SSD I am leaning towards Samsung EVO as I cant justify the higher pricepoint of the PRO.
     
  8. Predator2181

    Predator2181 Notebook Enthusiast

    Reputations:
    21
    Messages:
    48
    Likes Received:
    64
    Trophy Points:
    26

    RESOLVED THE CASE FINALLY .... FOR INFO OF THOSE WHO FACE SAME ISSUE HERE IS THE SOLUTION
    ISSUE WAS MAINLY EXCESS VOLTAGE WHICH WAS NOT GETTING CONTROLLED EITHER VIA THROTTLE STOP OR INTEL XTU

    FOR THOSE WHO MAY FACE THE SAME ISSUE ..ALL THAT IS NEEDED TO BE DONE IS TO MAKE AN OVERCLOCK PROFILE IN COMMAND CENTER AND FIX THE VOLTAGE TO 1.2 AT 5GHZ

    SEE THE ATTACHED SCREENSHOTS AND BEFORE AND AFTER 3D MARK RESULTS ...

    FINALLY NO MORE SHUTTING DOWN DURING POWER FAILURE

    1.jpg

    2.jpg

    3.jpg



     
  9. TRANCE1

    TRANCE1 Notebook Enthusiast

    Reputations:
    6
    Messages:
    48
    Likes Received:
    61
    Trophy Points:
    26
    Did you guys know there are quad speakers in the front of this thing? That's a first and pretty darn rad.

    51m speaks.png
     
  10. Alan Fagner

    Alan Fagner Notebook Enthusiast

    Reputations:
    0
    Messages:
    24
    Likes Received:
    15
    Trophy Points:
    6
    Does anyone know of any site that I can leave company claims? in the case I will leave a complaint from HIDEvolution
     
Loading...

Share This Page