Clevo P751DM2-G GPU upgrade feasible?

Hello there,

Has anyone here been able to upgrade the GPU on this system to a GTX 1080?

The only selections available from the likes of OriginPC and XMG (Schenker XMG U507) are the GTX 1060 and the 1070.

Thanks.

 

http://forum.notebookreview.com/posts/10328921/

 

Nice, let me follow that up.

@Prema would the current BIOS version on this system work with the 1080?

 

I also came across this: http://www.eurocom.com/ec/components(416)

Eurocom offer an upgrade kit up to the 1080. Will inquire with them shortly.

 

Excellent, keep us updated. For me, I'm waiting for Volta, if Clevo stay with their current MMX standard.

 

check this
https://www.ebay.com/itm/Nvidia-Cle...119148?hash=item3d5790e6ec:g:teQAAOSwIjJZVu2c

 

One question: On this laptop, I have one fused heatsink, i.e the CPU and the GPU share the heatsink assembly.

What I'm seeing here (in regards to the heat sink pictured above) is similar to what you'd see on larger systems (such as the SLI-based laptops) where each GPU has its' own heatsink.

Would such a heatsink assembly work on my SKU?

 

that heatsink is for a p870. may be @tanzmeister can help you

 

Thanks for the clarification.

 

you are welcome!

 

The 1080 is identical in layout to the 1070 so the current heatsink will swap over to the 1080 in your system.

 

Did you ever get this upgrade to work? Did you need to modify the BIOS in any way to do so?

 

No.

I decided to get a second laptop instead.

 

Fair enough, was that because it didn't seem possible? Did you find out more info about how to upgrade other than what's on here?

 

Just get the 330W AC Adapter and pop the 1080 in there, only real issue will be proper cooling, but it works.

 

Okay thanks, I'll let everyone know how it goes (and get my hands on the best paste/LM I can)

 

So I installed the card (after struggling with a stripped screw for 2 hours, thanks PC Specialist) and it's all working fine. Only issue so far is the Nvidia drivers refuse to install due to a "lack of compatibility". Do I need to mod the BIOS/drivers in any way?

 

Use a modified INF file from the likes of laptopvideo2go with driver signing off.

 

https://premamod.wordpress.com/2017/08/15/j95-nvidia-inf/

 

Once you mod the inf and get the driver installed, lemme know how the temps with the 1080 are under max load. I can also offer some tips on undervolting the GPU.

@Prema do the P751DM2-G and P751TM1-G share the same heatsink design?

 

Thanks, got ahead of myself though and modified my own .inf file by clicking on the first laptopvideo2go link I found... Doh! Seems to work though!

I just ran a unigen Heaven benchmark and it managed to keep it at a steady 90, though I imagine that's with a fair bit of throttling! Even so though, that the 1070 jumped to 86/87 so it doesn't seem too bad, although my poor 1070 was probably hurt by whatever horrible paste has been stuck to it.

As for the heatsink, if it's any help the CPU and GPU share one assembly.

Any tips on undervolting would be much appreciated though!

 

A comparison of the benchmarks for those interested, without the 1080 being fine tuned yet.

The 1070 reached temps of 87 where as the 1080 reached 90.

 

91C is the thermal throttle point so it'll just sit at 90C. I'm pretty sure every 1080 gets up there at max load as the 15" was really only designed to handle the 1060 and 1070. Even the old 17" P775DM3 had the same problem with the 1080.

Yeah I know that all P7 models except the P775TM have unified heatsink, but some of the older ZMs and DMs had different looking heatsinks and heatpipes. Anyway I just checked and the internals on your P751DM2-G are identical to my P751TM1-G.

Undervolting is really simple. Install the latest version of MSI Afterburner. When you open up the main interface, press Ctrl+F and you get the voltage/frequency curve editor like this:

Spoiler

Now go back to the main interface and either slide the core clock slider to the right, or enter a value in the box. For example I entered 160. Then click apply.

Now bring up the V/F graph again and it should look like this, with your new curve being a copy of your old curve shifted up on all points by exactly the same amount you just entered (160 in my case):

Spoiler

Now to actually undervolt instead of overclock, which is what you just did, you have to find the voltage you want to run at under load (I chose 912mV) and click the box corresponding to that. Then flatten out every box after that so it's just a straight line from that point on. Use the frequency on the left side as your reference. For example I clicked each box after 912mV and then used the mouse or arrow keys to drag it down until it matched 1885MHz on the left side:

Spoiler

And that's it. Fire up a GPU load and you should experience not only lower temps but also higher clocks and performance.

On my stock curve I was power and thermal throttling in Heaven within minutes and clocks dropped to low 1700s or even 1600s:

Spoiler

With the undervolted curve, it maxes at 170W and never power throttles. I can run Heaven overnight and have it stay at 83C and maintain a solid 1847MHz at 912mV:

Spoiler

With the frequency locked like that, a further benefit is that microstuttering due to constant clock speed changes from hitting various limits like power, thermal, and voltage regulator is completely eliminated. The PerfCap Reason box in GPU-Z is gonna be constantly lit after you flatten the curve but you can just ignore it.

Try running Heaven in DX11 (better performance than OpenGL). Here's what I get after undervolting my 1080 and disabling G-Sync:

Spoiler

 

I may have missed it or forgotten: Which compound are you using? (For posterity's sake.)

 

I did forget to put it, I am using Thermal Grizzly Kryonaut.

 

Thanks very much for the informative post, I'll experiment with various voltages when I get the chance and post the results. I haven't actually received my 330W charger yet so until then I could well encounter power throttling before heat throttling anyway.

 

So I tried throtlling the card, eventually ending with a completely curve at about 1650MHz and it certainly had an effect, it took a lot longer to reach 91 and had almost plateaued by the time it got there. However, it still ended up heat throttling the card.

I'm curious @yrekabakery, which vendor did you get your machine from? Is it one of the vendors that modifies to the cooling system to improve it?

 

Nope, just Kryonaut on the GPU originally. I replaced that with Prolimatech PK-3 when I took off the heatsink. The paperclip mod increased pressure but did not prevent the card from overheating eventually on the stock curve.

1650MHz is really low and would put you below 900mV which should not overheat. Are you using max fans? What's your room temperature?

 

Hmm okay, do you know of any good guides for performing the paperclip mod? I'm not able to make any sense of the first results on Google.

You caught me out there though, room temp is probably 19c so nothing to worry about but I hadn't installed the Clevo Control Centre so the fans were doing their own thing. After setting the fans to max manually it levelled out at about 75c which is much better (although the noise is unbearable) so I'll have to get my girlfriend a good pair of headphones if I want to be allowed to do that!

 

It's the toothpick and paperclip fulcrum mod for the CPU and GPU heatsink: http://forum.notebookreview.com/thr...ctive-cooling-mod-for-p775dm2-p775dm3.803626/

It didn't really help me at all. Yes pressure increased slightly as verified with pressure paper, but the GPU was still overheating within minutes unless I undervolted. So I don't think it's worth doing.

You don't actually need Clevo Control Center to use max fans. Just press Fn+1.

 

Does not seem worth the hassle for the amount of improvement.

 

+60 FPS seems like a noteworthy bump IMO.

 

On Max frame comparison. Avg FPS comparison only about 10 FPS, not too worthy.

 

You always pay a lot more for just a little extra gravy at the top.

 

This is why I love the NBR community.

You fellas are pretty awesome.

 

It's also worth pointing out that was before I had underclocked the graphics card so it was experiencing pretty serious throttling. I also have only just received my 330w charger too so have been reluctant to push it until now which the reason for the lack of updates. Unfortuantely my new charger appears to be broken, when I ran a benchmark using it my laptop powered off and it stopped charging, it then wouldn't turn back on until I unplugged the charger. After an hour of waiting it did work again but the same issue reoccurs if the machine is run to maximum for a couple of minutes again, I assume I should send it back for replacement?

 

So I finally got a coursework deadline out of the way so I could take a look at some of the issues I've been having over the past couple of weeks. I switched the new 330w charger back to my 230w charger and that fixed the turning off issue. However, I'm now instead getting a very strange display issue after a couple of minutes of benchmarking, I've attached pictures of what the screen shows. The only way to recover the laptop at this point is to hard reset it.

I've already tried updating the drivers, removing any overclocks, removing underclock, reducing underclock and no matter what I do the issue still occurs. I really hope it's not another dodgy graphics card, has anyone got any ideas of what it could be?

What the screen shows:



I've also had a wider variety of error logs shown in event viewer, interestingly enough the last crash didn't have any but the one before produced this:

If anyone has any suggestions they'd be muched appreciated.

 

That's generally the sign of a failing GPU.

 

Sadly that looks like a dodgy graphics card. The Event Viewer error is a graphics driver crash. Have you removed the heatsink and ensured there's good coverage of thermal paste on the GPU die and thermal pads on the various PCB components?

 

My guess would be bad video memory, sadly. Definitely GPU failure in some variation if not that, specifically.

 

Not the responses I was hoping for I must admit... Looks like it's time to start my third RMA of the fornight! The temperatures are all staying pretty good so I doubt it's that. The video memory one sounds quite possible actually, it crashed a while back saying it had run out of video memory.

 

@Meaker@Sager
Another dead GTX1080 from Clevo.

 

Considering I was the first one to advise there might be an issue with @makejason 's card then yes I know.

We are almost 2 years in the number of failures compared to other cards has been totally normal at this point.

 

Hi,
the problem is driver 397.64. My friends and i killed 2x Gtx 1070 and 1x P5000 with this driver in 3 different notebooks. Looks like Nvidia raised the core voltage.
Try an older Version like 391.35. Maybe the Card will work again normally.

 

It didn't raise the core voltage.

 

Not you, Nvidia did it with the latest Driver.

 

The latest driver did not increase GPU voltage. I game with an OSD and my voltage has stayed the exactly the same on every driver since 391.35.

 

Impossible. The drivers cannot raise the core voltage on a card where the vcore is locked by the video bios!
Even the unlocked voltage slider in MSI afterburner is ignored because the video bios won'to accept the voltage increase.

But you killed 2x GTX 1070 with the Nvidia driver somehow.
How come I have yet to kill one GTX 1070 in a MSIbook that is HARDWARE MODDED to 210W TDP ? Explain that? your card is IIRC, 125W.

No difference in temps or power draw with latest driver.

 

That happened, while running firestrike with the new driver.

 

Looks like a GeCube/Eurocom GTX 1070 that got the magic smoke while running at 115W.