Alienware 17 R1 - RTX 3000 issue

ÊtaPegasus · January 14

Hello everyone!

I'm a long time reader of NoteBook Review and NoteBookTalk as they are a great ressource for laptop modding.

I recently managed to grab one of the PNY RTX3000 on ebay for 100€ (a bargain!) to replace my aging Quadro P4000.

The installation goes fine since i have the 5 pipes full copper gpu heatsink from Cicichen and i carefully cover with thermal pad every part necessary.

After having some trouble modding the nvdmwi.inf file i finally get it to install properly and now start the trouble.

GPU-Zinfo.JPG.bb9fb9683bbf5bce504d0f35d4cdb1f6.JPG

My issue is the following :

- My GPU core won't boost higher than 945MHz (wich is base clock) and won't consume more than ~30W

- The PerfCap Reason is "PwrThm" while on battery (see the first image) and only "Pwr" when plugged in while doing the GPU-Z test (see second image)

GPU-Zsensor_battery.JPG.c04406ae03f3a3445ecd8395bfb68a94.JPG GPU-Zsensor_pluggedin.JPG.f4afb9fb7c33043c3e8ce224963ae2a3.JPG

- While gaming, the limit reason in MSI Afterburner is "No Load" and GPU-Z says the card is in "Idle" while reporting almost 100% utilisation.

GPU-Zsensor_pluggedin_TESOgameplay.JPG.9001ecd5824d4062ce3bd1c7dbd64214.JPG

I spend the last few days testing mutiple things :

- Revert the bios to default and revert my cpu overclocking --> no effect

- Tinker with the different driver parameter : prefer high performance, dedicate the graphic card to gpu intensive task etc --> no effect

- Activate/Deactivate GPU Hardware Scheduling --> no effect

- Windows 10 fresh install on another ssd --> strange results as the PerfCap Reason changed to "idle" everytime but same clock speed (~945MHz) and power consumption (30 to 40W)

- Plug out the graphic card, verifiy thermal pad placements, reseat the gpu in the MXM slot etc --> no effect

I carefully read all the other topic on this subject bt didn't find anything helpful for my case.

If anyone has any ideas i'm all ears, otherwise i might just put back my venerable Quadro P4000.

Thanks in advance!

Edit (25/02/2024):

After gathering the experiments from many people and making many test, the actual conclusion to this topic is :

- The PNY/Adlink Nvidia Quadro RTX 3000 doesn't work with the Haswell generation Alienwares (17 R1 nor 18 Viking)

- The same card works flawlessly on previous generation Alienwares M17x R4 and M18x R2 (for more info on M17x R4 performance, look into this thread and for M18x R2 look into @ssj92 threads)

- The issue doesn't seem power related (even though 240W is the bare minimum to make this card run, see my experiment below) as some user reports having the same issue with either 240W or 330W power supply.

- The resistor mod was tried on the M18x R2 by @ssj92 and didn't help in any way stabilizing the Thermal+Power throttling issue

- As the HP variant of the Quadro RTX3000 (and above) are reported to work flawlessly on either Alienware generation an attempt to flash the HP vBios was made by @ssj92. The vBios flash was unsuccesfull as the vBios chip wasnt recognized by neither Nvflash nor the external programmer.

So far, the best easy upgrades (without chassis or heatsink modifications) are Nvidia Quadro P4000/P5000 or standard Gtx 1060.

I know there are some standard form factor Gtx 1070 but their reliability seems questionnable.

Once again, thanks a lot to all the people who took the time to look into my issue!

Edited February 25 by ÊtaPegasus
Update

Jerryzago · January 15

I do have the same GPU on my M18x R2. It's performs excellent from day 1.

You could try something. From the control panel, go to Manage GPU utilization. Select Dedicate Graphics Tasks. Reboot. Go to control panel once more and select Use for graphics and Compute needs. Reboot and check again for utilization.

Though in your GPUz, Open CL and CUDA are unckecked. So it might be some driver conflict.

I stopped modding the inf file and started using NVCleaninstall application.

aldarxt · January 15

My m17x r4 I have 330W AC Adapter. My voltage is 1.0430v, yours is 0.6370v

ÊtaPegasus · January 15

Thanks for the replies.

I tried to change the nvidia control panel settings back and forth and no luck, it doesn't change anything.

I noticed that i was missing OpenCL and CUDA when i switch to"dedicate GPU to graphics only".

I tried to install the driver again via NVCleanInstall and no luck. Though, looking at your screenshot aldarxt, i see you have Physix enabled and i don't. Maybe i do not have the right driver for this card?

I am actually using this one "537.99-quadro-rtx-desktop-notebook-win10-win11-64bit-international-dch-whql" from Nvidia's website.

To those who installed a PNY/Aetina RTX 3000, what nvidia driver are you using? Are you also in SG mode? If, so what intel HD driver?

I am also using a 180W power brick, that could be the culprit?

aldarxt · January 15

6 hours ago, ÊtaPegasus said:

Thanks for the replies.

I tried to change the nvidia control panel settings back and forth and no luck, it doesn't change anything.

I noticed that i was missing OpenCL and CUDA when i switch to"dedicate GPU to graphics only".

I tried to install the driver again via NVCleanInstall and no luck. Though, looking at your screenshot aldarxt, i see you have Physix enabled and i don't. Maybe i do not have the right driver for this card?

I am actually using this one "537.99-quadro-rtx-desktop-notebook-win10-win11-64bit-international-dch-whql" from Nvidia's website.

To those who installed a PNY/Aetina RTX 3000, what nvidia driver are you using? Are you also in SG mode? If, so what intel HD driver?

I am also using a 180W power brick, that could be the culprit?

For some reason the 537.99 is not approved on 3DMark so I dont use it but it gave me good scores. The physx I downloaded from Nvidia

https://www.nvidia.com/en-us/drivers/physx/physx-9-19-0218-driver/

I like to keep it just in case cause this happens now and then but as for openCL and CUDA I would try the modding .inf method. I never tried NVCleaninstall yet so Idk how it works, but I think settings have to be set correctly and modding .inf has become 2nd nature for me

And Yes a 180w is not enough. A 240w at least but if you need to buy 1 get a 330w, that is why your voltage is low!

5 hours ago, aldarxt said:

For some reason the 537.99 is not approved on 3DMark so I dont use it but it gave me good scores. The physx I downloaded from Nvidia

https://www.nvidia.com/en-us/drivers/physx/physx-9-19-0218-driver/

I like to keep it just in case cause this happens now and then but as for openCL and CUDA I would try the modding .inf method. I never tried NVCleaninstall yet so Idk how it works, but I think settings have to be set correctly and modding .inf has become 2nd nature for me

WOW! just went through the version history of NVCleanstall, too much right now, but it has great options!

ÊtaPegasus · January 15

Thanks for the advice, i just found a 240W power supply for 15€ and i'll get it in the next couple days! I'll keep you in touch to see if that's solve the issue.

In the meantime, i'm thinking about the strange behaviour of my Quadro P4000, especially on the new games (Elden Ring and Resident Evil 4 Remake) where the card never stayed at 100% utilisation and was jumping up and down in power consumption. Could it be the power supply wasn't enough and with my 4940MX the card was starving?

aldarxt · January 15

1 hour ago, ÊtaPegasus said:

Thanks for the advice, i just found a 240W power supply for 15€ and i'll get it in the next couple days! I'll keep you in touch to see if that's solve the issue.

In the meantime, i'm thinking about the strange behaviour of my Quadro P4000, especially on the new games (Elden Ring and Resident Evil 4 Remake) where the card never stayed at 100% utilisation and was jumping up and down in power consumption. Could it be the power supply wasn't enough and with my 4940MX the card was starving?

Yes 180w is not enough for the XM and P4000. I have 6 laptops in my room my wife calls a man cave but I know its not, but a while back we came home from a little trip and I took my Laptops out of my gun safe, that keeps them safe when we go away. I dont remember how but my m17x r4 with i7-3940xm and GTX 680m was doing exactly what yous is doing and I had plugged it into my 180w. I just dont remember ever using the 180w, I just dont use it, but hooked it up by mistake, so even the GTX 680m needs 240w minimum

ÊtaPegasus · January 20

I just received the 240W power supply and you were right, this was the issue all along!

The GPU-Z test now shows the card boosting to 1800MHz sometimes and pulling 80W constantly!

Thanks again for the answers!

I'll post some 3DMark results comparing the Quadro P4000 et RTX 3000.

ÊtaPegasus · January 20

Well, another issue appears now, the card seems to power/thermal throttle to ~40W with the vRAM going as low as 200MHz.

Periodically, the card will throttle the vRAM then recover and again and again.

I don't see anyhting strange in the sensors' readings, albeit with GPU-Z, MSI Afterburner or HW Info : the card doesn't exceed 60°C...

EldenRingissue.gif.315f8f1cfe1e8b05047359e7bc118ded.gif

I remember seeing this strange behaviour in sliderfra's topic but i don't see any solution.

I did try to put a cooler underneath the laptop to see if it helps but it has no effect at all.

I took the card out the pc again and realized i don't have thermals pads on the pieces circle in red, could it be the culprit?

RTX3000.png.b0b179cef3e4e0663810365a527d8948.png

Jerryzago · January 21

The ones you have marked, need to be cooled. I added thermal pad to all those exposed chips near the VRMs, just to be safe.

ÊtaPegasus · January 21

Okay, i'll repad the card just to make sure.

In he meantime, i took a look at GPU-Z log file recorded while playing The Elder Scrolls Online in ultra settings then high settings and i plotted the results in the 3 graphics below :

From what i can see :

- The GPU temperature isn't the main factor here.

- Every time the power draw reach ~78W or above, there is Power+Thermal Throttle on GPU clock + vRAM Clock for ~6s

- The jumps in power condumption are linked to the loading of different area in the game.

- After i change the settings from Ultra to High, the power draw never exceed 65W in peak and 40W in average and no more throttling!

So i guess there is something wrong with the card cooling since every time i reach high power draw the card throttle, something overheats locally and recover very quickly afterwards?

Can someone point on the card which component are the VRM and what are the little component i didn't cover with thermal pads?

TESO log file.ods

Jerryzago · January 21

I asked pny for the product manual. It shows the thermal zones of the gpu that need cooling. The area above the Ram is the power supply area. I made sure everything makes contact (with pads) with the heatsink.

Do you monitor the thermals of the GPU hotspot? Mine during heavy gaming, reaches 90 degrees which is the limit the manufacturer states. What are your temps?

The 80w draw is propably dictated by the manufacturer to keep the gpu from frying itself. So it reaches that and starts throttling. Keeping the gpu lower, runs at full speed all the time.

I had two 980m with the prema no throttling bios. Being 100w gpu, mine draw almost 125watts during load.

ÊtaPegasus · January 21

Would you mind sharing those schematics? 😃 I'd really love to take a look to better understand how the card works!

As for the GPU hotspot, the log indicate a maximum of ~80°C just before throttling appears. Since the limit is set to 87°C according to GPU-Z i imagine this wasn't the cause of throttling.

Jerryzago · January 21

Here are the files. You can dig into and find any info you need. I might need to re pad my GPU too.

MXM_Specification_v31_r10.pdf EGX-MXM-RTX3000_50M-00021-1000_10.pdf

ÊtaPegasus · January 21

Thanks for the documentation!

Looking closely to it, i found several interesting things :

- 1st, the thermal policy says that after reaching 97°C, the card will throttle itself by 87,5%!! Wich is consistent with the throttling occuring on the memory side (1750MHz base vs 202MHz throttle). Though, the documentation talks about GPU clock and i don't know if it refers to Core clock or Memory clock? Other hypothesis, the throttle mechanism appears on wichever one exceed the temperature? In wich case, i should look closely to my vRAM pads?

- 2nd, the power management says that if the current exceed 10A, the card should throttle

- 3rd, the ambient measurement location are as follow in the figure below with really low temperatures threshold (65°C average on the 3 sensors and 75°C max on one sensor).

So now i have many things to reconsider :

- 1 or more vRAM chips report temp as high as 90°C and then throttle the card --> I have no sensor in GPU-Z or HWinfo to confirm.

- My battery (wich is suppose to be an OEM replacement with the right sticker) doesn't hold perfectly the power delivery and/or the 240W power supply isn't enough --> I rule out this power delivery issue since sliderfra has the same problem with a 330W adapter.

- The ambient sensors are detecting high temps locally that my heatsink can't evacuate (maybe the 5 pipe heatsink is counter productive here) --> This is coherent with the time needed to Pwr+Thm throttle on a fresh start (after 1 hour cooling period, take almost 1 min to throttle) and a pc used then stressed (take 15-20s to throttle).

I put pads on the pieces pointed in red earlier but that didn't change anything...

Any ideas?

aldarxt · January 21

I have to come in with a stupid comment! I went back to your first post and you paid $100. That my be the problem.

ÊtaPegasus · January 22

This is indeed what i came to think as well ^^

But as many others on this forum got the same card from the same reseller (PNY) on ebay and got it working i thought i was safe.

Maybe i got a bad sample....

For now, i took it out the Alienware, put the Quadro P4000 on again and it works like a charm!

What surprises me is the performance gap (in the wrong way) : even without throttling, the Quadro RTX3000 pulled almost 80W at full boost to keep 60fps on The Elder Scrolls Online at ultra settings and the Quadro P4000 pulls only 45W to achieve the same result on the same area....

I really don't understand, is this a driver optimization issue? Or my card is a dud?

Nevertheless, to figure this out, i bought another Alienware 17 Ranger for 130€ with an allegedly "non working internal display", i'll see if i have the same behaviour on the 2 laptops.

aldarxt · January 22

Just a long shot, but check your Nvidia settings

ÊtaPegasus · January 22

I already tried all those settings individually then all at once, and nope, no effect at all on the throttling behaviour nor the performance of the card....

Thanks for your help, but for now, i think the card will stay in a box until i have a second pc to try it ^^

aldarxt · January 22

So that means the P4000 is working as expected in the 17 R1? That eliminates the 17 R1 for issues, and RTX 3000 ???

My RTX 3000 works both in m18x r2 and m17x r4. I also have a RTX 4000 sitting in a box and might be my fault for swapping in different vBios. IDK if its bricked or just ruined. It will only fit in the m18x r2, it being a HP RTX 4000 and to get it to fit in the m17x r4 chassis grinding is needed and am not ready to do that. I love the MXM cards but having a box full goes with it

I noticed this thread might be helpful troubleshooting the card

gabrieloj · January 22

On 1/21/2024 at 1:59 PM, Jerryzago said:

Here are the files. You can dig into and find any info you need. I might need to re pad my GPU too.

MXM_Specification_v31_r10.pdf 749.51 kB · 3 downloads EGX-MXM-RTX3000_50M-00021-1000_10.pdf 1.04 MB · 2 downloads

Hi, would you have the schematics and/or repair guide for the GTX 980m? Mine is damaged and needs repair. My tech guy has identified a blown mosfet and will replace it, but the files would be helpful just in case replacing the mosfets won't work. Thanks.

ÊtaPegasus · January 23

This means that there might be an issue software side between my Alienware 17 and my RTX 3000 or the RTX3000 is defective.

I'll try to take the time to test it with a multimeter and report back.

As for @gabrieloj request, i'm sorry i don't have any schematics for the GTX 980m to help you.

Jerryzago · January 24

On 1/22/2024 at 10:48 PM, gabrieloj said:

Hi, would you have the schematics and/or repair guide for the GTX 980m? Mine is damaged and needs repair. My tech guy has identified a blown mosfet and will replace it, but the files would be helpful just in case replacing the mosfets won't work. Thanks.

No, I am sorry I don't have documentation for 980m. I also had a 980m go bad on me. I send sold is as parts to a user back in NBR (and still a member here) and never had a reply what was wrong!

I wanted to know, to prevent from happening again. I now have two 980m in my shelf with the prema vbios.

I bought the pny rtx3000 from ebay, and the seller send me the documentation for it to study. Hope you fix your 980m.

ÊtaPegasus · January 24

Maybe a long shot, are you using Optimus Jerryzago? If so, what Intel HD driver are you using (i use the latest one from Dell website)?

Are you on unlock bios (mine is stock Dell A14)?

I looked into older threads about upgrading to Pascal graphic card on the old NBR forum and TechInferno and i came across a few threads talking about random power throttling with high wattage GPU (1070, Quadro P5000 and even 980m). The solution was to run them in dedicated mode or to use a hibernation sleep bug to resumue power throttling...

Could it be a wrong combination of locked bios settings + optimus not working properly?

aldarxt · January 24

On 1/23/2024 at 9:04 AM, ÊtaPegasus said:

This means that there might be an issue software side between my Alienware 17 and my RTX 3000 or the RTX3000 is defective.

I'll try to take the time to test it with a multimeter and report back.

As for @gabrieloj request, i'm sorry i don't have any schematics for the GTX 980m to help you.

Just found an old thread from https://www.techinferno.com/index.php?/topic/11378-msi-gtx-1070-mxm-successfully-working-on-alienware-m17x-r4-another-socket-victory-against-bga-crap/&

He mentions disabling "adapter warning" in BIOS. He says it stopped the throttling of his GTX 1070. Maybe its the same issue

Alienware 17 R1 - RTX 3000 issue

Recommended Posts

Link to comment

Share on other sites

Top Posters In This Topic

Popular Days

Top Posters In This Topic

Popular Days

Popular Posts

Jerryzago

aldarxt

ÊtaPegasus

Posted Images

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Create an account or sign in to comment

Create an account

Sign in

Important Information