HP Elitebook 8570W - is it still good ? Hardware modifications

[Daedalus]

On 7/16/2025 at 3:16 PM, AstroMan said:

You don't have to go with a large MXM-B with GTX 980m or mod interface to 3.1 to have RTX 3000/5000 go fast.
NVidia made Quadro T1000 (Turning) which are MXM-A and go for about 250$ a pop.
From what sellers tell me, they require eDP, so that's a thing.

i think the HP 8750W is limited by standard LVDS screen to use 980M max, anything newer seems to push the need to fit the DC screen or swapping in a new screen from a newer model.

i saw the T1000 in the Wiki however whilst a newer card, doesnt seem to have decent resolution output. Is the t1000 actually a more powerful card though, specs seem to say 980 is more gutsy.  whilst im still on the fence on best card to get (not ordered the card yet) still keen to take recommendations from your good selves. 

[AstroMan]

On 7/18/2025 at 9:38 AM, Daedalus said:

i think the HP 8750W is limited by standard LVDS screen to use 980M max, anything newer seems to push the need to fit the DC screen or swapping in a new screen from a newer model.

i saw the T1000 in the Wiki however whilst a newer card, doesnt seem to have decent resolution output. Is the t1000 actually a more powerful card though, specs seem to say 980 is more gutsy.  whilst im still on the fence on best card to get (not ordered the card yet) still keen to take recommendations from your good selves. 

I made a mistake in my post, what I meant to say is that you don't have to choose "big" (B type) cards if you want to go with Tesla architecture gpu.
I guess there is some benefit with encode/decode, maybe efficiency and some open source drivers but that comes at a steep cost of DC screen at the moment (and no guarantee it will work btw), oh and heatsink mods too.
But yeah, with less TMUs, cores, rops, memory and TDP, I don't think T1000 will be quite the... showstopper, even compared to 965m.
as always, the answer is: it depends 

[Daedalus]

On 7/19/2025 at 1:08 PM, AstroMan said:

I made a mistake in my post, what I meant to say is that you don't have to choose "big" (B type) cards if you want to go with Tesla architecture gpu.
I guess there is some benefit with encode/decode, maybe efficiency and some open source drivers but that comes at a steep cost of DC screen at the moment (and no guarantee it will work btw), oh and heatsink mods too.
But yeah, with less TMUs, cores, rops, memory and TDP, I don't think T1000 will be quite the... showstopper, even compared to 965m.
as always, the answer is: it depends 

Thanks for the response. indeed, im sure the t1000 isnt far off the older 965M and will use the same heatsink size. (happy to go "big" if it enables more card options (albeit with more risk of working). 

I see the newer 1070/1060 cards are options but are newer generation and give MXM-Type A capability. That said more work to enable (comparable to 980M "big". Happy to heatsink mod (ex-Engineer 😉 )  I think for me the less "driver mod" options which will probably push me to 965M/980M anyways.  I notice alot of the MXM "B" boards have unpopulated sockets/solder locations for stuff like RAM and caps / fets etc. why is that?

 

For the 980M was there a specific part / serial number board that worked?

 

[AstroMan]

18 hours ago, Daedalus said:

I notice alot of the MXM "B" boards have unpopulated sockets/solder locations for stuff like RAM and caps / fets etc. why is that?

When it comes to unpopulated components, its usually RAM chips, related power circuits (coils/inductors/chokes, mosfets and caps) and related caps and resistors. Really its just reusing (cost saving) the same pcb for different products.
A good example in Maxwell is GTX 970m 6GB vs GTX 980m 8GB.

In 970m you have clearly missing one power supply (among other things) and 4 GDDR5 512MB RAM chips, making it... surprise surprise 6GB.

 

If one is to try and add such components to juice up the gpu, you would also require modded vbios addressing the change. I have attempted this about a decade ago on 860m 4GB but sadly without success. At some point I was left to my own devices and learning everything was very time consuming for a young lad like me so I de-soldered my mods and left it back as it was. Just one issue less to deal with haha
I don't mean to discourage of course, but there is an art to vbios modding and memory timing tables and all that jazz.

I've done my fair share of modding but I'm by no means an expert, so if someone could chime in with juicy details, that would be great.

[Daedalus]

5 hours ago, AstroMan said:

When it comes to unpopulated components, its usually RAM chips, related power circuits (coils/inductors/chokes, mosfets and caps) and related caps and resistors. Really its just reusing (cost saving) the same pcb for different products.
A good example in Maxwell is GTX 970m 6GB vs GTX 980m 8GB.

In 970m you have clearly missing one power supply (among other things) and 4 GDDR5 512MB RAM chips, making it... surprise surprise 6GB.

 

If one is to try and add such components to juice up the gpu, you would also require modded vbios addressing the change. I have attempted this about a decade ago on 860m 4GB but sadly without success. At some point I was left to my own devices and learning everything was very time consuming for a young lad like me so I de-soldered my mods and left it back as it was. Just one issue less to deal with haha
I don't mean to discourage of course, but there is an art to vbios modding and memory timing tables and all that jazz.

I've done my fair share of modding but I'm by no means an expert, so if someone could chime in with juicy details, that would be great.

That is helpful, not yet at the crazy modding MXM boards yet.  I noticed there was a 980M board earlier in the thread that additional 220nf caps and FETs were added to improve. whilst 980M i think (please correct if wrong) is final incarnation of the 900 series maxwell board at 8GB, so odd they leave unpopulated, that said its an MXM B so might be reused in the 10 series before the card design went all weird. I'm comfortable soldering those on (dont have hot-air station) but couldnt see orientation in picture (didnt want to solder a cap backwards and wonder why card wont work). 

Ideally for me id like to minimise the Vbios work and fingers crossed its just new Nvidia drivers loaded with H'sink mods to make a card work in my 8570w

Looking at Mobile PCI Express Module - Wikipedia itwould appear the mighty RTX 4050/4060 (lovelace generation) is the final "type A" to be able to fit and use the standard H'sink in the 8570W. However i suspect drivers & Vbios might scupper that

[AstroMan]

9 hours ago, Daedalus said:

Ideally for me id like to minimise the Vbios work and fingers crossed its just new Nvidia drivers loaded with H'sink mods to make a card work in my 8570w

Looking at Mobile PCI Express Module - Wikipedia itwould appear the mighty RTX 4050/4060 (lovelace generation) is the final "type A" to be able to fit and use the standard H'sink in the 8570W. However i suspect drivers & Vbios might scupper that

One thing to also consider is power draw. MXM-B is much more power hungry.
If I remember correctly my Clevo GTX 980m drew about 125W maximum from the slot, which is quite a lot to ask, even from alienware laptops (I think it was its actual limit in 17 R1 Ranger stock). It's definitely a lot of stress for the motherboard and PSU (if it isn't upgraded).
 

I want to also point out that with those chokes on the side of newer gen gpus (or top of the card if connector is pointed downwards), there are heatsinks made for AMD which accommodate for them as AMD FirePro M4000 also has the big chokes in similar spot. I guess would be a good spot to start, just be careful as to not short things out as you will kill motherboards power rail and gpu itself

Edit: upon closer inspection, the placement of the chokes on the new cards seems different so it's a moot point

Edited Tuesday at 05:47 PM by AstroMan
Incorrect information

[Daedalus]

On 7/22/2025 at 4:28 PM, AstroMan said:

One thing to also consider is power draw. MXM-B is much more power hungry.
If I remember correctly my Clevo GTX 980m drew about 125W maximum from the slot, which is quite a lot to ask, even from alienware laptops (I think it was its actual limit in 17 R1 Ranger stock). It's definitely a lot of stress for the motherboard and PSU (if it isn't upgraded).
 

I want to also point out that with those chokes on the side of newer gen gpus (or top of the card if connector is pointed downwards), there are heatsinks made for AMD which accommodate for them as AMD FirePro M4000 also has the big chokes in similar spot. I guess would be a good spot to start, just be careful as to not short things out as you will kill motherboards power rail and gpu itself

Edit: upon closer inspection, the placement of the chokes on the new cards seems different so it's a moot point

That is all valid, i know other users have pushed the mobo to its limits running the hungry cards and limited the mobo longevity. Does anyone know the sensible limit they were designed for? I am running a 200w PSU as the 75W unit was getting toasty. So i have headroom, but 200w into a mobo and DC jack not designed might fail elsewhere. 980M as you say might be doable (with H'sink from 8770 and DVD port mod / select bay delete) trying to source the micro JST cables for fan splitter is proving fun as they seem too big a connector) Anyone know their name?.

 

The Type "A" journey ends at 965M it seems as i dont have DC screen and scouring Ebay etc, only finds a type B 965M card (which i dont think will fit the 8570W without clipping the DVD port. 980M's seem more plentiful. The only 965m "A" card i found on Ali was N16E-GR-A1 

Did you have any issues with your 980 clevo card? i see some list HP/Dell/Clevo etc so its difficult to identify correct card part number to get hold of.  do you have a part number sticker or N16 chip marker in the 980?