NVIDIA GTX 680M Thread - M18x R2 and M18x R1 Upgrade Discussion

Thanks for the explanation, I will mess around with it some more tomorrow and see what I can come up with.


Sent from my iPhone using Tapatalk

Edit: So after some messing around, I stress tested the gpus for about 30 minutes with a 100mhz increase on the gpu cores. So each gpu is running at 819mhz, there was absolutely no throttling with me. It stayed at 819mhz throughout the entire stress test.

 

how hard is it normally to add new GPU's to the m18? i'm a total novice to gaming laptops period (and the last time i took apart a desktop was in 2007)

just got off the phone with xotic, the guy said i couldn't use the clevo 680m's as an option, while i'm not too surprised at this, i do wish that i could have worked something out.

 

for a matter of interest can you try this GPU memory benchmark tool for me and post results here?
Thanks!

OCCT PT 4.3.1 Available !

 

Isn't that a gpu and CPU stress testing tool? Where is the option to benchmark?


Sent from my iPhone using Tapatalk

 

u can do tests individually

 

I have used the gpu section before but it never gave me any "results" per say, just used it to stress the gpu and look for throttling. So what exactly are you looking for?

 

I had 0 experience with taking apart laptops when I did my upgrade. Following the right video helps a lot. Just don't rush anything, take your time, and it shouldn't be too difficult.

 

That one is pretty brutal, too. Stress testing is not really necessary. It's OK for a short burst, but letting it run for long periods of time is kind of silly and risks harm to your system if not monitored closely.

Why not just do this http://forum.notebookreview.com/ali...mperatures-game-benchmark-screen-display.html and watch the temps right on your screen while running Heaven, 3DMark11, Vantage, etc.

You can also have HWiNFO64 log the CPU and GPU temps to a CSV file on your desktop while you are playing games or running benchmarks. Temps during gaming and benchmarks are accurate representations of real-world results. Stress tests are just... well... stressful.

 

I agree, I don't use them that much as real world gaming never pushes the gpu as much as these stress tests do.

UltraGSM if what your looking for is me stress testing my system I would rather not lol.

 

stress-test means loading your GPU RAM to maximum with test-data to see max MB loaded to physically available GPU RAM, you can do it for 1 minute and even not, you can cut off test with ESC button at any time desired, you see realtime voltages and temperatures, and I in example I ran my 7970m test for 1 minute no hassle, gpu didnt heat up as much I managed to alocate 2081MB from my GPU, dont ask how should be 2048MB total, well test shows little more, if youre not willing to try that then dont having 8GB GDDR5 and not knowing if all is accessible its weird lookin seriously just not knowing the fact ...
If you decide to find out what are your 8GB GDDR5 is doing in your system you can select these options in the utility

You can select:
GPU: 3D
TEST TYPE: AUTOMATIC
DURATION: 0h 1min
IDDLE PERIODS: 0 / 0
DX version: either
FPS: 0
SHADER COMPLEXITY: 3
RESOLUTION: 10920x1080
FULLSCREEN +
ERROR CHECK +
and click [ON] button,

once its canceled or finished you will be presented with test results of your system and one will show your GPU MEMORY LOAD, something like this:

(7970m DELL OEM Card)








EDIT:

Ok may be wrong there it may be literally the physical DDR3 ram in my cart , im unsure how to findout loaded GDDR5 RAM chart... tests are unnecesarry

my bad

 

Lol yeah thats what confused me, I was certain it didn't output a chart or result or anything for GPU ram. Anyways man if there is anything else you need I am more than happy to help.

 

Yup, if still interested, you can use this one, I have just ran it myself and allocated and verified 2012MB of my 2048MB shown (used short test). You would need to alter the settings file to override 1024MB GPU RAM limit

(only if interested)
HOME PAGE OF MISHA CHERKES



EDIT: FAIL, it actually ran 1024MB while driver reported 2021MB... should read little more.

my bad x2

 

If all you want to see is whether the system detects the entire gpu memory I posted a screenshot a few posts up. And as Mr. Fox said, even if you have 8GB total you will still only be able to use 4GB as SLI does not combine gpu memory.

 

Hello everybody. I currently have an m18x r1 originally with 460m's in sli. I just purchased the 680m upgrade kit from upgradeyourlaptop on ebay and had it installed yesterday. The shop that installed it for me left one 460m in there and set it up as a physx card. Ever since the swap, I've been having horrible performance. Much worse than with my original 460m's in sli. I just pulled a score of 1142 in 32mark11. I am using drivers provided on disk by the seller. I tried disabling physx thinking that perhaps that was interfering with my computer using the 680m for all the heavy lifting but still had the same results. Everything in Nvidia settings shows that the 680m is being used primarily but the way performance has been it feels like an integrated card or the left over 460 is being used. I suspect that the 460 is the issue. Can anyone confirm this for me? Any input would be greatly appreciated! Glad to hear that others with a single 680m have been seeing great results and looking forward to share the same!

 

My best advice is to take the 460m out and test the system again. This is the first time I've heard of the use of a lesser GPU was a PhysX card, theoretically it should work as it would in a desktop, but the laptop drivers most likely can't support two different types of GPUs installed at once. Someone enlighten me on this subject, it is rather interesting.

 

Thats what I am thinking as well, I've never heard of this being done before either.


Sent from my iPhone using Tapatalk

 

I'm kind of surprised it does not work, to be honest. Are you 100% positive the 680M was installed in the left PCIe slot? If it's not, it's going to have to be.

Do you have SLI enabled or disabled? If not already, disable SLI.

Silly question now, but have to ask it... you're running with the AC adapter connected and using the "High performance" power profile, correct?

 

Thanks for all the replies gents. The 680 is definitely installed in the left port as I can hear the fan gets quite a bit louder from that side. I see no option for SLI to be enabled or disabled so I am assuming that it is disabled, and I always plug in with high performance enabled as well. It must be the 460 that is the issue. I'm ordering another 680m within the week to replace it with but in the meantime I will have the shop remove the 460 and test everything out again. I'll keep you guys updated! Thank you all for your input. These forums are great! +rep to all of you that replied!

 

Fan noise is not indicative of which slot your 680m in installed in. My left fan will spin up even if I'm running off the integrated graphics with the dedicated cards MUXed off.

You should be able to look at device IDs and PCIe port numbers to determine which slot each device is using. I can't tell you how to do it as I'm not near my own system, hopefully someone here can guide you.

 

Ditto what everyone has said. In your device manager under Display Adaptors the top one (PCI Bus 1, Device 0) should be the left graphics card / primary card. the next one down should be the secondary card (PCI Bus 7, Device 0). And the left fan will always come on first / work hardest no matter what is installed since it will always be used as the primary graphics card. Hope this helps!

 

I just confirmed that the 680m is in the left port taking Charlie and DVSman's advice and not relying on the fan noise. Thanks guys! In device manager I have the 680m listed under PCI Bus 1 and the 460m under PCI Bus 7 so I think I'm okay in that aspect of things. I still don't get why I'm getting such poor performance. I'm talking severe stutter on any game that I play that I used to have running smoothly with just the 460's in SLI. I updated the BIOS the other night from A03 to A05 using dell's driver downloads. I'll be taking everything in to the shop again in a few hours to have the 460 removed as that seems like the only thing left to do. I really hope that I don't have a dead card and that removing the 460m will be the end of my problems. If any of you guys can think of anything else that I can do in preparation before going to the shop please let me know! Thank you guys! I'll keep you posted.

 

I don't think you have a dead card - since if that was the case you wouldn't likely get video at all as I don't think you can run a card in the secondary slot by itself. Plus, your 680 is still showing up in the device manager and all. I'm not an SLI / physx expert but I've always thought the second card physx thing was only for desktops.

Also, if you don't want to wait for the shop to get around to doing it, you should just look at the youtube videos on the m18x teardowns - Mr. Fox's or the Official Dell support channel. It's actually not that hard to do. I've gotten pretty damn expert at doing it after having to do it so many times to straighten out my various problems post delivery / arrival (I bought from the outlet and alot of stuff that should have been QC'ed were not).

 

I just got it back with the 460m uninstalled and I think I see where the problem lies now. After running 3dmark11 and pulling the same 1133 score for my graphics, I noticed further down that while my memory clock is running fine at 1800mhz, my core clock is only at 324mhz! Normally for the 680m I believe should be around 720mhz. Is this why I am experiencing such bad performance?

 

It's normal for the card to downclock when it's not being used. Play a game or something and run gpu-z in te background. Alt-tab out of te game and check out what the gpu clock is running at during gaming.


Sent from my iPhone using Tapatalk

 

OR install hwinfo64 ......switch the sensors on and then monitor while gaming through MSI Afterburner ..this way you wouldnt need to alt tab out and it also shows you exactly whats happening while you are in your game..i was recently taught this by mr fox

Here is his link http://forum.notebookreview.com/alienware/655138-how-monitoring-cpu-gpu-temperatures-game-benchmark-screen-display.html

 

Thanks for the tips gentlemen. Downloaded GPUZ and ran it during HeavenDX11 demo. During the very beginning few seconds of the cinematic, everything runs quick and smooth. Clocks at this point are 718.5 for core and 1800 for memory. Then after those brief two seconds, the clocks drop all the way down to 135 mhz for core and 324 mhz for memory. The drop comes almost as soon as i hear the fan spool up. gpu load stays at 99% and memory controller load at 42%. I took a screenshot. Any ideas why it's acting like this? I feel like I'm getting closer to my goal of having everything smooth again. Thank you to all in advance.

 

Very interesting. I want you to try something. In nvidia control panel go under the profile section. Select the profile for the heavendx11 benchmark. If its not there create one it's quite easy. Anyways under the profile section look for power management and change that from adaptive to prefer performance or something along those lines, run the benchmark again and wait until its finished then check gpu z and see of the gpu has downclocked during the benchmark.

Also check out what the peak temperature was.

Sent from my iPhone using Tapatalk

 

I wasn't able to set a power profile for the heavendx11 benchmark as nvidia control panel only lets me select .exe files while my heavendx11 is a batch. But, I did do what you suggested for 3dmark11 and the clocks throttle down during that test as well. Max temp was 91c and average was 71c.

 

I'm wondering about your 91c temp . . . is it really hot where you are running? As a point of reference, I just did this upgrade as well 2 weeks ago (also an M18xR1). I am running overclocked at 915/1125 and through multiple 3dmark11 runs and a bunch of game benches, I hit in the mid to high 70s max. Maybe ask your shop to repaste? In the meantime, can you post one of your GPU-Z log files (or part of it)?

 

I live in Hawaii. Air temp where I ran is probably around 76-79F. I ran the benchmark again and pulled a 1210 graphics score. Log is attatched.

 

Ok, I'm in Los Angeles and temps are about the same right now (heat wave). I'm going to go back to your temps as the cause, but hopefully one of the true experts can chime in after looking at your log. Your card seems to be working normally until it hits 94c and then pretty much stops and limps along - I would assume this is the temperature based throttling I have read about, but haven't see it myself (and your card gets there in less than a minute). As previously suggested, take it back and ask them if they placed all the thermal pads and how they attached the heatsink to the GPU - they should reexamine it.

 

Any reason you can't just remove the 460M yourself? Doesn't take more than 10 minutes...

 

You sir, are a genius. I used Mr. Fox's teardown video and got into the guts myself. I compared how the thermal pads were on my card to the diagram on the ebay sellers page and it was as if they took a shot in the dark on placement. I shifted them around myself to match the diagram, close everything back up, boot up, open gpuz and 3dmark11 and pull a smooth P5485 score!!! constant 718.5mhz core and 1800mhz memory with temps between 60-76c. I am absolutely elated. It is such a relief to say goodbye to this issue. Thank you Mr. Fox for making an easy to follow video that even a person who has never opened a laptop before could follow successfully. Thanks to everyone on this forum who gave their input and helped me on my way to victory! :thumbsup:

 

Wow congrats! Nice job on the tear down as well. Not easy for a first timer with the M18x!

 

Thank you LannBot! I was very apprehensive before starting as I've never opened a laptop before. But my frustration with the downclocking finally pushed me to do it. Mr. Fox's video was really key and without it I may have never had this resolved! This is a great support community.

 

I am glad it went smoothly with you as when I installed my 680m's it was my first time as well. Its quite easy and its all thanks to Mr. fox's video.

 

Just placed an order for a brand new 680m heatsink and support plate for my card through dell. All for cooler operation and overclock in the future :] Now that I know how to take everything apart, I'm looking forward to getting the parts in so I can do the install myself! Thanks Mr. Fox!

 

Vicious Panda, how much did the parts cost through dell (if you don't mind my asking)? I need to get some new back support plates.

 

No problem man, I spent a total of $53.39 for both the heatsink and support plate. The actual parts themselves only cost like $29. Sorry I don't know the exact breakdown of the individual part cost. I was so happy that I could get them through dell rather than ebay. I've seen people ask close to $90 just for one heatsink.

 

Someone tried Clevo 680M 4gb version from upgradeyourlaptop in M18x R2 yet?

 

If it works in the R1 it will definitely work in the R2.

 

Dell's pricing on heat sinks is very reasonable. They come with all of the thermal pads and an excellent phase change compound for the core already applied. As inexpensive as they are new from Dell, it makes little sense to buy them anywhere else unless they are dirt cheap.

ViciousPanda, I'm excited about your 680M SLI upgrade. I'm sure it is going to be more awesome than you can imagine right now. I am glad to hear you found my video helpful.

Looking to upgrade? Have you had about all of the AMD problems you can stand?

These are all standard MXM video cards. They are all more or less compatible with the M17x R3, R4 and M18x R1 and R2. Some of the aftermarket (non-Dell) GPUs can be missing some proprietary functions (like HMDI in or out) on certain models. With the M18x R1 and R2 they seem to be fully interchangeable.

Even the M17x R2 is capable of running the 680M. (For the M17x R2 some tweaks are required for SLI function. The original AC adapter is also under-powered.)

At least for now, the only immediate source for them for an Alienware laptop upgrade is a Clevo vendor, such as Eurocom, Upgradeyourlaptop/Upgrademonkey (ebay), etc.

 

Just was thinking about geforces but i have decided to stay with amd till the end lol

 

Anyone heard any "official" news from Dell as to when the 680m is "validated" for the 18x R1 as an upgrade? - I recall that we were told on one of the conference calls that it had not been validated (not to say it won't work as it obviously does) yet and as such, it won't be offered for sale by Dell as an official upgrade for R1 owners. I keep asking my contact but as per usual, new sales take priority and he tells me that the 680m's are only available via technical support/replacement issues.

Would be good to get a response from one of the Dell guys here on NBR.....maybe Mark aka Alienware-Mark_O could advise.....

 

Good Question Stevie,

We do not typically validate new cards on older revisions. From what I have seen in the forum, folks are using them in the R1's, but with Non Dell OEMS.

You hit on a key point. Availability of the cards is the issue at the moment. They will likely be available for upgrade when stock is readily available.

Hopefully it will not take that long. Some good benchmarks on them.

 

Thanks for posting, Mark. It would be great to get official support to retro fit these babies to the older revisions, as they are indeed capable of working with older machines. Last thing I wanna do is mess with my warranty with non-Dell cards so I guess we will have to wait and see when they are readily available to order.

 

Hi Mark,

Thanks for the update, there. I was also guessing the issue was still the availability of the cards.

I'm not sure if this is the best place to ask about this, but is there going to be a BIOS update for the M18x R2 that enables overclocking on the 680m's without a vBIOS mod? I remember someone on the Alienware team was discussing this with us a couple months ago, around when overclocking was enabled on the M17x R4 (BIOS A05, I believe).

If anyone else has news on this, and I missed it, please chime in!

Thanks

 

Quick question - do I need to use thermal paste with the compound that comes on the heat sink? With the first card I ended up having to mount it twice, so the second time I cleaned it off and did use the paste. It seemed to be running hot the first way, but could have been user error since it was the first time I had done this. I have a second one almost ready for installation (waiting on SLI cable from Dell of course), and debating whether to use the paste or just the compound (or both even). I have a brand new 580M heat sink from Dell. Which way is better?

Thanks!

 

If you had to install it, remounting it later is not good. In that case using paste was appropriate. Either will work fine. If the phase change pad for the core is in good shape you can just use that and it will do great. Don't use both. If you decide to use paste, then I would clean off the factory material for the core and use only paste.

 

You are right. Availability is what we are looking at.

Good question on the BIOS update. Will ask engineering if this is being discussed.