The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.

Possible GPU upgrade for Precision M4600

Discussion in 'Dell Latitude, Vostro, and Precision' started by garmac, Mar 6, 2015.

  1. newbie.545

    newbie.545 Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    1
    Trophy Points:
    6
    I cleaned the whole machine and reapplied Arctic MX4 when i install the card. The issue only happen in Ultra Performance power plan though. Change to any other power plan fixed the problem for me. Actually I dont think the Ultra Performance power plan get me any additional power.

    Btw, I had my chipset reball like 3 years ago, a mosfet near the DC input blown up and cant start the machine, the repair shop told me the chipset was shorted or something. Was charged $25 for that. I dont know if every Precision M4600 get hot at the bottom around the docking connector or my particular unit is defect. Some people say this issue will definitely come back but it has been 3 years now.
     
  2. haceru

    haceru Newbie

    Reputations:
    0
    Messages:
    4
    Likes Received:
    0
    Trophy Points:
    5
    Hello everyone,

    First of all I would like to start by saying a big "Thank You" to all contributors to this old, long but amazing post, full of very valuable information that is hard or impossible to find anywhere else.

    Secondly, I would like to share my experience with the M4600 and the upgrades I performed, maybe it will help others in the same way this thread helped me. I will include more information than just the GPU so please excuse me for any off topic details. So, here it goes:

    I purchased a Dell Precision M4600 from the official Dell Auction website in early 2015. The original configuration was: Intel i7-2820QM, 8 GB 1333 Mhz DDR3, Nvidia Quadro 1000M, 7200 rpm 320 GB HDD, DVD Rom, Windows 7 Pro x64, non-IPS FHD 1920 x 1080, manufacturing date October 2011. It didn't include the webcam or the bluetooth module. The first upgrade performed was to add a 128 GB mSATA card SSD in the WWAN port as I didn't want to lose the optical drive unit in order to have a SSD primary drive for the OS and I kept the HDD as a secondary drive. Then I added the Bluetooth module and the webcam (it took me a long time to research the exact part number for the camera and at the time, the only way was to see the part number of the microphone-only board the laptop came with and see which Latitude E-series models will have it compatible so to look for the camera part number on those). After those hardware changes were done, the next logical thing to do was to update the BIOS and fresh install Windows. All good so far, Windows Experience index gave me 6.5 with the lowest score being the GPU, of course. It didn't bother me much for a long time until I started to play World of Tanks and crave for a better FPS performance and higher overall quality of the game play (the only acceptable setting for the graphics was Low). This is how it all starts, right? :biggrin:

    So here I am looking into possible GPU upgrades and happily finding the M5100 being the best possible option. However, reading this amazing thread, I noticed the reviews about the non-compatibility between the GPU and the thermal sensors of the Dell BIOS causing the GPU fan to not spin regardless of how high the GPU temperature went. So, the first thought for me was to update the BIOS to the latest version. And now it comes the big big BIG disappointment: the particular BIOS version I have has a major problem - it can not be updated anymore. The version is A08, first release (somewhere in 2012) and the problem lies with the "Signed Firmware Update" option. On that particular version, it seems Dell messed up the crypto-algorithm of the signatures and it signed it with an empty key. Of course any subsequent BIOS version would be looking for a valid encrypted key but since they only find an empty one, will fail the BIOS update process. This was such a messy BIOS version that apparently Dell technicians were sent to customers to have the entire motherboards replaced for the machines still in warranty in order to perform the updates. So, I was out of the option of updating the BIOS and thus this road with the M5100 was completely blocked for me.

    The next best option, considering the situation above, was to chose between the "official" supported GPUs, which are only three: the Quadro 1000M, 2000M and the Firepro M5950. After a long and painful decision process / time, I finally decided to go for the Quadro 2000M. All good, got the GPU, installed it but when running the HWINFO I noticed some high temperatures; also the Windows Experience Index was only 6.8 as opposed to 6.9 that I was expecting. The Furmark test brought the GPU temp to 88 C in less than 2 minutes so at this point, I decided to open up the laptop, clean the dust and re-paste both the CPU and the GPU. There was one piece of information I still didn't have, the thickness of the thermal pads. Since I still had the old 1000M, I measured the pads and to my surprise, they are thicker than the "standard" 1 mm ones most of the GPUs are using. As such, I ordered the Kryonaut 2 mm pads as well as the Kryonaut Grizzly thermal paste; they advertise 8 W/mK thermal conductivity for the pads and 12.5 W/mK for the non-conductive paste. One detail I noticed when working on the CPU, the heatsink has a thermal pad for the North Bridge, a few inches above the CPU; that pad is thinner, I would say the 1 mm one.

    After all those troubles, I run again the tests and Windows Experience Index finally reached the expected 6.9 and the Furmark could not overheat the GPU past 84 C. However, what I noticed is that at idle, the temperatures of both the CPU and the GPU are higher than with the previous configuration. They rise together slowly from 44 to 60 (CPU) - 33 to 59 (GPU) at which point the fans start working at 2500 RPMs and cool down the system to 50 again in about 10-20 seconds. Then the fans stop spinning, and the cycle repeats itself. I run the High Performance energy plan with the 90 W wall adapter (travel kit). Another thing I noticed was that now the heat under the upper keyboard bezel is evenly spread as opposed to before repasting, when it was concentrated around the letters O-P (where the GPU lays).

    One thing I forgot to mention is that I also upgraded the RAM, from 2x4 GB 1333 Mhz to 2x8 GB 1600 Mhz (even though Dell says "not supported" - contradicting Intel which for the said CPU and chipset says "supported" :wacko:). I used a pair of HyperX Impact CL9 and to my surprise the HWINFO says they even work at 1.35v as opposed to the 1.5v I was expected

    All in all, I am happy so far with the upgrade and with the learning process involved in the thermal pasting. The only thing I am still concerned a bit is the fact that now the fans are needed even when idle-ing; what i've read about this particular setup (Quadro 2000M and i7-2820QM) are some mixed reviews - some say they run cooler, some say they do run hotter. I am guessing that the power plan has a big role to play here but for now I am ok with it. If anyone cares to provide some feedback, it would be greatly appreciated.

    Thank you for your time reading this long post and I really hope someone might find it useful at some point.
     
    Last edited: Feb 15, 2019
  3. Aaron44126

    Aaron44126 Notebook Prophet

    Reputations:
    874
    Messages:
    5,548
    Likes Received:
    2,055
    Trophy Points:
    331
    The fan cycling behavior annoyed me to no end in the M6700. See my signature for information on a manual fan control setup. (Should also allow the M5100 GPU to be installed.) I just have it never cut the CPU fan below 2500 RPM, but you also have the option to run them at 0 RPM and increase the threshold at which the fans should kick in.
     
    haceru likes this.
  4. haceru

    haceru Newbie

    Reputations:
    0
    Messages:
    4
    Likes Received:
    0
    Trophy Points:
    5
    Wow, that is an amazing job you did, thank you for sharing ! For now I am not so much annoyed by the cycle, I am more concerned about the CPU / GPU reaching those temps while idle (13-15% CPU usage). I will wait to see how the system behaves back home, on the 210W "brick" while playing WoT during summer at the room temp of 30-35 C and then decide if I should re-paste or not. But thanks again for your reply, will keep it in mind as well.
     
  5. newbie.545

    newbie.545 Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    1
    Trophy Points:
    6
    I just upgraded to M5100 just a few days ago, and everything are fine.
    My machine originally come with Quadro 2000M. A mosfet on the mainboard exploded and kill the card. I fixed the mosfet and upgraded to Firepro M4000. Work fine until a month ago the card is dead again. Initially i suspected that was a mainboard problem, so I ordered a brand new mainboard from Ebay (its actually pretty cheap). After replaced the mainboard, I found out the card is dead and then buy a firepro m5100 at a local store for like $50.
    The mainboard come with BIOS ver A08. It might have sit on the warehouse for so long that even the bios battery run out. I was able to update the bios to A16 without any problem. I used the Bios from Dell support website.
    The fan seems to work alright, but not optimal. It will maxed out at 2500rpm unless the CPU is being stressed, at which point it will speed up to about 3500rpm. Its alright for me so I dont use any fan control software.
    I used the heatsink from the Firepro M4000. It only has one heatpipe going to GPU fan unlike the heat sink of the quadro 2000M (2 heat pipes to both fan). It does interfere with the fan a little bit because of the larger heat pipe, but the fan can be secured just fine. I'm pretty sure the CPU temp is lower because of this (compared to when I use 2000M heatsink)
    The temperature under full synthetic load from Furmark is about 80~82C for CPU (i7-2760QM), 78C for GPU. CPU clockspeed at 2.7GHz all core. If only GPU is stressed, GPU temp is about 82C. When play Starcraft 2 (medium setting, ~65fps), CPU and GPU reach about 75C max.
    Btw, I use a standoff to elevate the laptop about 25mm from the table surface. This help lower the temp by 2~3C.
     
    haceru likes this.
  6. haceru

    haceru Newbie

    Reputations:
    0
    Messages:
    4
    Likes Received:
    0
    Trophy Points:
    5
    Thank you for your reply. I am very glad it worked out for you to upgrade successfully to the M5100, I bet there is a noticeable improvement compared to the original GPU. Your temps are similar to mine so I guess this is the range / behavior to expect from the machine when reaching the power limits. Also, that is exactly what I was thinking about the original GPU heat pipe - it touches the CPU one next to the fan and it heats it up, reversing the heat flow towards the CPU (not much, but still). By any chance, do you know the part number of the heat-sink you are using now, the one with just 1 pipe?

    For now, I won't be pushing my machine any further but maybe I will start a new project with a M4800; at the Dell Auction site I saw some nice ones for just under $400.
     
  7. newbie.545

    newbie.545 Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    1
    Trophy Points:
    6
    I think the original gpu heat sink just block some air flow so it raise the cpu temp.
    The gpu heatsink I use is the one came with firepro m4000 in the dell m4700. You can just search for some image of it. It fits m5100 as well.
    A fully upgraded m4800 is still a very capable machine. $400 is a good price. I live in South East Asia so unfortunately I dont have access to dell auction. Might go for a desktop pc next time
     
  8. haceru

    haceru Newbie

    Reputations:
    0
    Messages:
    4
    Likes Received:
    0
    Trophy Points:
    5
    So ... for the past 10 days or so I have been running all sorts of tests on the machine, in order to see if I can get the "perfect" temp setup and I thought of sharing some interesting findings.

    The "methodology" I used for some of the tests was to time how long it takes between waking up the computer and the moment the fans kick in. But before getting to the actual findings, a little background on the settings, as it seems they affect a lot the behavior of the computer. First of all, the charger I use for now is way too small for what the laptop can do - 90 W. The actual power rating of the laptop is 180 W and back home I use the "brick" of 210 W for full power. This is the limitation number one which causes either CPU or GPU throttling. The good news is I could test separately the CPU while on battery - it powers it up to full 3.0 Ghz due to TurboBoost and separately the GPU - it powers up to full voltage / clocks while the AC charger is plugged in. Could not test them both full power due to the said power input limitation. Another setting that affects the way the fans work (among other things) is the C-states option from BIOS. If the C-states is enabled, than the fans start to spin when the CPU temp reaches 64-65 Celsius; if C-states is disabled, then the fans spin constantly at 2600 RPMs keeping the temps at 45 / 42 (CPU / GPU respectively). The clock at idle is 798 Mhz. I didn't change the Windows power plan but I left it at High Performance, with minimum / maximum processor power setting at 100%.

    Now, for the thermal paste ... I re-pasted 3 times both the CPU and the GPU as I wasn't sure if the paste was correctly applied and I also wanted to test different application methods. Both chips (CPU and GPU) do not have an IHS so they make direct contact with the copper plate of the heatsink, so obviously we need the entire chip covered in paste for the best results. To add to this, the i7 Sandy Bridge has the cores on the side, not in the middle so we definitely need them covered. I tried the spread method, the line method on the CPU and the "pea" method on the GPU. As reported on various videos and forums, the difference between them is of maybe 1-2 degrees C, thus irrelevant. The paste - Thermal Grizzly Kryonaut (even though I have my doubts about the quality of it and / or if it is the "real" deal or a fake). Also, one note ... when carefully examining the copper plate of the heatsinks, they seemed chipped and scratched so I took a 3000 grit polishing sandpaper plus some cardboard and got them as close as possible to a mirror-like polish (my knife sharpening skills came into play .. LOL )

    Also, I added a metal net / mesh stand to which I manually attached an external fan that blows air from one side to the other (not up the bottom cover but parallel to it); power for the fan comes via USB from the back eSATA / USB 2 port via a split adapter.

    And now the best results so far:

    - without any external help (no stand, no fan) it takes about 13-14 minutes at idle (798 MHz CPU clock) for the internal fans to kick in
    - with the stand but no external fan it takes about 18-19 minutes at idle
    - with the stand and the external fan on at half the RPMs (around 1800) it takes about 30 minutes at idle

    Those cycles are for the following temps variations: CPU - 40, GPU - 38 and ambient - 27 until CPU - 64/65, GPU - 58/59 and ambient 44/45; room temp was constant at 23/24 Celsius. There is another thing I noticed while assembling the laptop which would explain the high ambient temps - the GPU heatsink makes tight contact with the HDD caddy (metal enclosure) via a captive screw so it seems the laptop was designed to spread the heat as much as possible; also I believe the faster RAM (2x8 1600 MHz) is playing a role for the CPU idle-ing at 798 MHz (the clock speed of the RAM) - maybe using the "officially" supported 1333 MHz would have clock the CPU lower at 666 MHz (just a hunch)

    Hope this helps others having the same "quest" of finding the best temps / performance / stability of the systems (my wife actually calls it "obsession" but ... well ... you know :) )

    ***Update****
    A week later my suspicion regarding the quality of the paste was confirmed. I repasted using TF8 from Thermalright and totally different story: without any external help, it took 21 minutes for the fans to kick in and not due to the CPU or GPU but because the internal ambient temp rose to 46 Celsius. The CPU didn't go above 58 and the GPU 59. So ... either I got a fake or a bad batch of Kryonaut ... the paste was the one to blame. One suspicion I have is that Kryonaut does not hold well above 80 C, despite the manufacturer claim / rating of up to 350 C. When looking at it after use it looked dried out in certain spots, due to the GPU / CPU reaching 84-85 during tests / benchmarking.

    Some test results: on battery, with max. power usage of 28.2 W the CPU scored 439 / 117 multicore / single core in Cinebench R15 with a max temp of 79 C and with the AC charger plugged in the GPU averaged 13 FPS in Furmark for 2 minutes at 1280x720 with a max temp of 75 C. Not bad I would say ...

    And for the cherry on top, I was able to squeeze a very nice 495 score in Cinebench R15 , on battery with the C-states disabled and with the active cooling provided by my stand (CPU power consumption 54.7 W). This concludes my tests and I can leave the system in peace ... until the summer WoT marathon when real life stress test begins :)
     
    Last edited: Feb 28, 2019
  9. Ulysses Isambert

    Ulysses Isambert Notebook Enthusiast

    Reputations:
    0
    Messages:
    13
    Likes Received:
    0
    Trophy Points:
    5
    Anyone have any issues with color accuracy with the M5100 GPU upgrade on the M4600 machines? I definitely see an increase in performance in lightroom (develop Module) and Photoshop but the color is desaturated. I have the AMD Radeon Pro R9 200x Driver installed and its just working right. Any thoughts? As a photographer, color accuracy is important so using the main laptop screen is not needed but was hoping to externalize to at least two monitors @ 2k/60hz and 4k/30hz.

    Thank You for your help.

    Ulysses
     
  10. NevynPA

    NevynPA Newbie

    Reputations:
    0
    Messages:
    6
    Likes Received:
    1
    Trophy Points:
    6
    Hello all -

    Thanks for this thread and for keeping it alive! I recently picked up a few HP EliteDesk 800 G1 USDT ("Ultra-Slim DeskTop") micro-towers. They are basically "laptops in a box" - They use laptop RAM, a laptop power brick, but use desktop CPUs - however, the 'S' designation ones (lower TDP/wattage; mine has an i5-4570S). They also have an MXM 3.0a slot for a dedicated video card (OEM option was a Radeon 7650a that had about the same performance as the Intel HD 4600 integrated but had 2GB of VRAM).

    Thanks to this thread and some others, I picked up an M5100 card for in it and will be working on making a custom cooler for the card out of an all-copper Socket 370 low-profile server cooler.

    I've also upgraded from the stock 130W power brick to a 230W model to guarantee that nothing is power throttling.

    Hopefully I'll also be able to overclock and flash the card thanks to @valuxin and the others in this thread:

    http://forum.notebookreview.com/thr...locking-vbios-mod-dell-precision-m4800.789891

    I hope to have partial results later tonight, and full results Friday evening.

    Also, I have both a Samsung memory card and one with SK Hynix memory (oops!) so I plan to do some test comparisons between them and then I will have one for sale.

    [​IMG]
     
    Rippchen likes this.
Loading...

Share This Page