After successfully installing the drivers, I first installed Unigine Heaven (a DirectX 11 benchmark) to see if the 3D acceleration really runs smoothly and also how the power-hungry 55-watt TDP card performs in the relatively small 15.6″ EliteBook 8540p. As always, I monitored everything with HWiNFO64.

First surprise:

The Quadro 2000M reveals a lot more than the NVS 5100M did, namely very interesting VRM values such as temperatures and current/voltage values:

After Heaven had been running for a few hours without any problems and I noticed that I still had about 10-15 °C of thermal headroom before it would become critical, the question naturally arose…

…if the NVS 5100M could be overclocked so easily and significantly with MSI Afterburner, can the Quadro 2000M do the same?

The answer is yes – and the performance boost is nothing short of incredible.
But first, testing is carried out using original clock speeds.

First verify test values, with obstacles:

To maintain consistency with PassMark scores for comparison purposes, I installed and ran Performance Test v11.1 on Windows 10.

The result was only 711 points with original clock speeds (550/900 MHz) – significantly less than previously listed in the technical.city table (this could be due to many factors, such as Windows, drivers, CPU performance, background processes, daily mood, etc.).

It should have been 765 points, but with 711, it is now 7% fewer points.
At first, I suspected that the CPU was not powerful enough or that background processes were preventing the boost to 3.20 GHz on one core from being achieved consistently, meaning that it could not feed the GPU well enough.

Then I thought that it could at least partly be the “penalty” that this benchmark imposes because it cannot run the intended resolution (Full HD, but the panel built into our 8540p can “only” run HD+, i.e., 1600 x 900). This “penalty” is appropriately 7%, so that could well be it.

I verified this by connecting an external monitor via VGA (I had the cable to hand, DP would also have worked) and lo and behold – now I have 740 points. Still not quite there, but better. Only a 3.3% difference, so I left it at that.

Of course, at higher resolutions, the load is increasingly shifted towards the GPU, which in this case could reveal a very slight CPU bottleneck.

Now it’s time to overclock!

After a few hours with MSI Afterburner (slowly increasing the clock speeds one after the other and repeatedly testing with Heaven) and HWiNFO for monitoring, I reached the maximum overclock, almost at the limit—I had to reduce the VRAM clock speed slightly because one of the two cards started to show artifacts.

Since the original NVS 5100M was presumably suffering from dying VRAM components, it’s probably a good idea to go a little easier this time, regardless of stability.

Of course, I wonder what clock speeds would be possible on the GPU core if MSI Afterburner didn’t limit it (I suspect that the real limit, as is always the case, comes from Nvidia’s VBIOS, which is located on the card).

In the subsequent rerun of the PassMark Performance Test on the external Full HD screen, the result was a smooth 1100 points – a whopping 49.1% increase in points!

I am very satisfied with the results achieved; the Quadro 2000M runs much better than expected!

It was definitely worth it – of course, the card now reaches 95°C under prolonged full load, whereas previously it was more in the range of 80–85°C. The VRMs then exceed 100°C significantly, sometimes reaching 115°C or more.

In our case, I can live with that, since the GPU will never be under extreme stress for very long, as in a benchmark. It’s better to have performance reserves that you rarely need than to constantly run at the limit.

Here are the overclocking changes in tabular form:

So, looking at the entire process, these were the increases:

Original GPU, Nvidia NVS 5100M with 1 GB VRAM:
199 PassMark points – that was the baseline.
100% performance

Overclocking to the limit of the Nvidia NVS 5100M with 1 GB VRAM:
240 PassMark points (the stronger of the two)
121% performance

Conversion to Nvidia Quadro 2000M with 2 GB VRAM:
740 PassMark points
371% performance

Overclocking to the limit of the Nvidia Quadro 2000M with 2 GB VRAM:
1100 PassMark points
552% performance!

So the whole operation more than quintupled the original graphics performance of our 8540p and improved a few other things as well (DirectX 11, WDDM, 2 GB VRAM)!

Brief explanation of overclocking graphics cards:

I hope that I have been able to provide some general insights into overclocking (Nvidia) graphics chips. The whole process works in the same way with all Nvidia graphics chips that allow overclocking – you can find this out by installing and launching Afterburner. If everything is grayed out, overclocking is not possible.

Up to and including the Maxwell architecture (GT/GTX 9xx series), you could even freely edit the VBIOS limitations by flashing a modified VBIOS onto the card, which would then remove or extend the limits in Afterburner.

Of course, this is more dangerous than doing it with running software like Afterburner, but it allowed you to do cool things like writing the overclock directly into the VBIOS and thus being able to use it under Linux / Macintosh without special software – or simply passing on “pre-overclocked GPUs” that anyone can use without special knowledge.

Increasing or removing thermal and/or performance and/or voltage limits, programming other VRAM modules, flashing them, then desoldering the existing ones and installing larger ones – all of this is possible. It’s pretty cool.

Or undervolting and underclocking to optimize efficiency, which is also very useful, especially with limited cooling options – but that’s a science in itself and a topic for another post.

The topic of GPUs is now complete, so let’s move on to other hardware changes I’ve made.

Der Beitrag <h5>RetroBook 8540p #9: </h5><h3><b>Overclocking the GPU</b></h3> erschien zuerst auf flohs blog.

Since the Quadro 2000M now appears to be running on the hardware side, the previously downloaded Nvidia graphics drivers must now be installed, matching the respective operating system.

However, these do not work out-of-the-box, but must be slightly modified.
In order for the modified version of the drivers to be installed, driver signing must be temporarily disabled in Windows 10 and 11. This should not be relevant for Windows XP, Vista, 7, or 8/8.1.

How to disable driver signing:

• Restart the computer by holding down the Shift key and clicking on “Restart” in the Start menu.
• Now “Please wait” should appear and, depending on the PC, it should restart.
• Now select the “Troubleshoot” button.
• Then click on “Advanced options” -> “Startup settings”
• A list of possible options should appear; select “Restart”
• When Windows restarts, the list should appear again and you can select the correct entry, “Disable driver signing” by pressing “7”
• Then simply wait until the operating system has completely restarted. The driver should now install without any problems.

The signature is only checked during installation. Once everything has been installed and the computer has been restarted, this check is automatically re-enabled and the driver still runs without any problems.

Let’s get started with the driver modification:

I will now show you the procedure for Windows 11 Professional. For Windows Vista, 7, 8, and 8.1, the procedure should be identical or very similar. For Windows XP, the procedure differs slightly as fewer files need to be edited. Fortunately, you can see this immediately by simply paying close attention to the file names and entries (just ignore the missing files).

I have successfully performed this driver modification on Windows XP Pro SP3, Windows 7 Ultimate x64 SP1, Windows 10 Pro 22H2, and Windows 11 Pro 24H2, and everything is running 100% stable.

Getting started:

Now, pretend you want to install the driver normally and click “OK” here:

Until the setup, which runs automatically after unpacking (with the usual administrator rights prompt in between), then reports “The graphics driver could not find compatible graphics hardware”:

Then finish the setup with “X” or “Close” and navigate to the directory
where the driver was previously temporarily unzipped. By default, this would be at least in Windows 10 and 11:

C:\NVIDIA\DisplayDriver\377.83\Win10_64\International\Display.Driver\

C:\NVIDIA\DisplayDriver\377.83\Win10_64\International\Display.Driver\

Find out the PCI ID of the installed GPU:

Now open Device Manager and find the graphics card that should not yet have a “real” name (“Microsoft Basic Display Adapter,” the standard driver).

Once you have found it, right-click on the graphics card entry and open “Properties”.
From there, go to the “Details” tab and select “Hardware IDs” in the combo box that appears.

Then copy the first entry with CTRL+C or the context menu, or simply compare it with the following entry and then conveniently copy everything that follows without modification if the string matches:

The PCI ID of my Dell MXM card is:
PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C&REV_A1

Note the vendor (manufacturer), i.e. Nvidia (“10DE”), which makes sense. The Quadro 2000M obviously has the device ID “0DDA”, so far so good. But here’s where it gets interesting: the Sub-ID appears to be “1521” and this is completely missing from the driver. That’s where the problem lies, so let’s change that now.

If the PCI ID copied earlier in Device Manager now shows a number other than “1521” at the location shown, this number must be replaced with the different number in all subsequent text (in which case it is a Quadro 2000M from another board partner, with a different internal model number).

Reminder:
This driver modification applies EXCLUSIVELY to this specific Nvidia Quadro 2000M type. Other models may be compatible, but everything that is now being changed would have to be adapted, otherwise nothing will work.

Next, files need to be edited:

Notepad++ is very helpful for this kind of thing, firstly because of the option to obtain administrator rights retrospectively (which you also need in a subfolder of C:\ to edit files) and secondly because of the convenient context menu entry for opening files. Nevertheless, I will show the process here using Windows’ own “Editor.”

The easiest way is to find the sections beginning with “Entry” using the search function in the editor you are using (CTRL+F). The following entries must be added, first in the file:

nv_dispwi.inf

In the entry (quite high up in the file):

[NVIDIA_Devices.NTamd64.10.0…14310]

[NVIDIA_Devices.NTamd64.10.0…14310]

Add at the end of the block:

%NVIDIA_DEV.0DDA.1521.10DE% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

%NVIDIA_DEV.0DDA.1521.10DE% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

Then in the entry (directly below):

[NVIDIA_Devices.NTamd64.10.0]

[NVIDIA_Devices.NTamd64.10.0]

Add the same at the end of the block:

%NVIDIA_DEV.0DDA.1521.10DE% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

%NVIDIA_DEV.0DDA.1521.10DE% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

Last entry in this file (at the bottom):

[Strings]

[Strings]

It is best to insert this under the existing entry “Quadro 1000M”:

NVIDIA_DEV.0DDA = "NVIDIA Quadro 2000M"

NVIDIA_DEV.0DDA = "NVIDIA Quadro 2000M"

…this ensures that the GPU is correctly recognized and displayed throughout the system.
Note that this entry is completely missing (as is the screenshot), but curiously, many other Quadros from this generation and model series are present. Perhaps the Quadro 2000M was only available to OEMs who received special drivers at the time?
The NVS 5100M is also missing here.

Next up is this file:

nvblwi.inf

In the entry (again quite high up in the file):

[NVIDIA_Devices.NTamd64.10.0…14310]

[NVIDIA_Devices.NTamd64.10.0…14310]

Add at the end of the block:

%NVIDIA_DEV.0DDA.1521.103C% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

%NVIDIA_DEV.0DDA.1521.103C% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

Then in the entry (directly below):

[NVIDIA_Devices.NTamd64.10.0]

[NVIDIA_Devices.NTamd64.10.0]

Add the same at the end of the block:

%NVIDIA_DEV.0DDA.1521.103C% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

%NVIDIA_DEV.0DDA.1521.103C% = Section003, PCI\VEN_10DE&DEV_0DDA&SUBSYS_1521103C

Here too, most recently in this file in the entry (far below):

[Strings]

[Strings]

It is best to insert this under the existing entry “Quadro 2000M” with a different ID:

NVIDIA_DEV.0DDA.1521.103C = "NVIDIA Quadro 2000M"

NVIDIA_DEV.0DDA.1521.103C = "NVIDIA Quadro 2000M"

Finally, it’s time to tackle this file:

nvdmwi.inf

Only the entry:

[Strings]

[Strings]

Also to be supplemented with “Quadro 2000M” with the other ID:

NVIDIA_DEV.0DDA.1521.1028 = "NVIDIA Quadro 2000M"

NVIDIA_DEV.0DDA.1521.1028 = "NVIDIA Quadro 2000M"

Next, the entire “NVIDIA” folder should be copied to a directory where users also have write permissions. The easiest place would be the desktop.

This is very important to note, as I otherwise received various error messages regarding file permissions and the installation was canceled!

Now we should be able to install the driver

You can then restart the installation with “setup.exe”:

…Great, it seems to be working!

Once the installation is complete, you can check that it was successful in Device Manager, preferably after restarting the system:

If everything works, the temporary “NVIDIA” directories and the previously downloaded driver package can be deleted.

Unexpected positive side effects:

As a bonus, you can now see in the Windows 10 or 11 Task Manager that the Quadro 2000M is listed as a GPU in the “Performance” tab and can therefore be monitored. This was not the case with the NVS 5100M, which was too old – none of the cards offered by HP as compatible should be able to do this.

This is because, unlike the NVS 5100M, the Quadro 2000M supports DirectX 11 and thus also a newer WDDM from Microsoft – this compatibility is a prerequisite for this functionality. We now have this virtually free of charge:

To ensure that everything really works, I put a 3D load on the card for a longer period of time (see next post). Then I tested the DisplayPort (on the left side of the notebook) and the VGA port (at the back next to the power supply connection), both of which work without any problems.

Let’s move on to what was inevitable.

Der Beitrag <h5>RetroBook 8540p #8: </h5><h3><b>GPU drivers</b></h3> erschien zuerst auf flohs blog.

After extensive research and many comparisons, I was able to find some newer GPUs in MXM-A 3.0 format in addition to those intended by HP for the 8540p and 8540w (summarized in a table in the previous post). (Points are again PassMark points from technical.city).

This list from TechPowerUp helped me enormously in finding the models that were of interest to us:

There also appear to be sister models of the NVS 5100M, which are also supposed to be in MXM format—I was unable to find out whether they are A or B. Confusingly, these are manufactured in a different architecture (the 5100M is based on Tesla 2.0, while the sister models are based on Fermi).

The models are called “NVS 5200M 1 GB DDR3” and “NVS 5400M 2 GB GDDR3”.
However, both are weaker than the Quadro 2000M, scoring 505 points and 617 points respectively.
I couldn’t find a single image of either of these in MXM format (only directly mounted on motherboards).

The NVS 5400M with 2 GB VRAM could have been the best choice from Nvidia due to its lower TDP of only 35 watts compared to the Quadro 2000M – if you can find one or purchase one.

Is an unapproved model really a good idea?

There is virtually no reliable information about which of these cards work specifically in the Elitebook 8540p / 8540w, at least not in German. Then there is the problem of the potentially existing BIOS whitelist – MXM cards are of course also available from various manufacturers – almost all of whom have installed them in their own devices.

This means that the vendor ID or PCI ID may also change, which could result in no image output on the internal display or on one or all of the external outputs (VGA, DisplayPort, …), no signal is coming through, it no longer boots (POST fail) because the BIOS does not recognize the device, the driver does not run or does not run correctly or stably, etc.

Apart from that, there are also hardware differences: is the die in the right place for the cooler, is the height the same or at least similar, what about the position of the VRAM modules, does the mounting system still fit?
If not, is it adaptable or completely impossible mechanically?
Does the newer module have the same electrical pin assignment? Will anything burn out?

There are many ways in which installing an MXM GPU other than the one intended can go wrong, especially since MXM was never a 100% serious standard that everyone adhered to exactly.

Until the very end, I wasn’t sure if what I had in mind would even come close to working.

Spoiled for choice – or not:

I wanted an Nvidia GPU because of my experience with the drivers and because I considered it to have the highest potential to run and fit mechanically (or be easily made to fit). After a long search, I couldn’t find a single comment or report that the AMD FirePro M5950 COULD run, which made the decision easier.

So, from the cards above, I had my eye on the Nvidia Quadro 2000M with 2 GB DDR3 in particular.
According to technical.city, it has 765 PassMark points, compared to 199 for the NVS 5100M – more than triple – making it significantly more powerful than any of the models officially offered by HP at the time.

I read somewhere in the depths of the internet that this card should work, but only with driver modifications. The specifications of this card fit well with the rest in my opinion – it has to work somehow, and 2 GB VRAM is also nice compared to 1 GB.

That would also be a significant upgrade to the NVS 5100M in every respect.
Unlike the NVS 5100M, the Quadro 2000M and Quadro 1000M are based on the slightly newer Fermi architecture, so they support DirectX 11.0, OpenGL 4.6, Shader Model 5.0, CUDA 2.x, and OpenCL 1.1.
No sooner said than done – this is what I want:

⧉ Nvidia

Found and purchased.

I purchased the cheapest Nvidia Quadro 2000M with 2 GB VRAM on eBay Germany for less than €40 each, including shipping, after accepting the suggested price. It was definitely worth a try. The two cards were new products sold as used stock, presumably for warranty reasons.

All the better, as silicon and electronics wear out over time, especially under the high thermal load that a graphics card is exposed to, particularly in a notebook.

However, there was one downside – they were OEM models from Dell – which was dangerous due to possible incompatibilities. HP also had this specific card at the time, but even used, it was significantly more expensive at the time. So it was a full risk, “no returns” (click here for the item on eBay).
The bracket, i.e., the mount that later connects the card to the cooler, looked a little different in the pictures, which was not a good sign.

Two impatient days later, the cards arrived in perfect condition, very well packaged.
They even had brand new thermal pads* on the VRAM modules.
It couldn’t have been better, considering all the circumstances and the age of the hardware.

Now we prepare for the GPU conversion:

First, download the appropriate graphics driver for the Nvidia Quadro 2000M for all existing/desired Windows operating systems (do not run it yet):

Windows XP x86:
https://de.download.nvidia.com/Windows/Quadro_Certified/307.83/307.83-quadro-notebook-winxp-32bit-international-whql.exe (archive.org)

Windows 7 x64:
https://de.download.nvidia.com/Windows/Quadro_Certified/377.83/377.83-quadro-grid-desktop-notebook-win8-win7-64bit-international-whql.exe (archive.org)

Windows 10 / 11 x64:
https://de.download.nvidia.com/Windows/Quadro_Certified/377.83/377.83-quadro-grid-desktop-notebook-win10-64bit-international-whql.exe (archive.org)

If you want a different operating system, you can manually check for it at Nvidia:

Then download DDU directly and use it to completely uninstall the graphics driver on all desired Windows operating systems (button: “Clean and shut down – To install a new graphics card”).

Of course, this is only necessary if a graphics driver was installed in the first place.

Now the installation / conversion can begin:

The service manual also provides a detailed description of how to remove the existing MXM GPU, in case I have forgotten anything. All you need is a PH1 Phillips screwdriver* – and possibly something to pry the touch bar out without damaging it. However, you can also do this with your fingernails.

Here is a summary of the steps:

• Turn off the notebook, unplug the power supply, and remove the battery.
• Loosen the keyboard screws (3 pieces); these remain in place (thanks HP, good design!).
• Loosen the screws on the touch bar (usually hidden in the upper area under the battery).
• Turn it over and open the notebook.
• Push the keyboard locks on the upper edge downwards
• Carefully lift out the keyboard; a plastic spatula or similar tool will help
• Carefully disconnect the keyboard ribbon cable and remove the keyboard
• Lift the touch panel sideways towards the keyboard, again preferably using a spatula
• Carefully disconnect the touch panel ribbon cable and remove the whole unit

Now it’s time to remove the cooler and GPU:

• Loosen the 4 Phillips screws on the cooler attached to the CPU (left).
• Loosen the 4 Phillips screws on the cooler attached to the GPU (right).
• Very carefully disconnect the cooling fan (very fragile connector!).
• Remove the screws from the cooling fan. There is another screw hidden under the ribbon cables in the upper area; loosen this one as well.
• Now the cooler unit can be carefully tilted to the side and removed.
• The GPU is held in place by two Phillips screws. Be careful—the MXM module folds up gently like a RAM bar.
• Done! The GPU is now removed.

Of course, everything should now be cleaned (preferably with isopropanol*, but carefully!).
If necessary, now would also be a good time to dust the cooling fan, as well as the cooling fins directly in front of it.

Is the Quadro 2000M suitable?

Upon closer inspection, I noticed the following things:

• The position of the GPU die and VRAM modules is unproblematic and fits quite well.
• The thickness of the thermal pads for the VRAM looks good, but rearranging them makes sense in some cases.
• The position of the thermal pads for the power supply is roughly correct, but not ideal. Acceptable.
• The plastic bracket attached to the new Dell GPU with double-sided tape is too thick and prevents the card from being fully inserted into the slot!

The significantly thinner metal bracket of the original NVS 5100M GPU should also fit the Quadro 2000M, right?

If you use a Quadro 2000M variant with a metal bracket, you can save yourself the following modification, which is not entirely risk-free due to the force required to press it in. I was simply “unlucky,” if you want to call it that. But the modification was easy to do—it could have been worse.

Modification of the cooler bracket and thermal pads:

The original bracket of the NVS 5100M fits the Dell card, but is attached to the circuit board with very strong double-sided adhesive tape. This is probably due to years of hot-cold cycles.

I carefully pushed the threaded sleeves through the holes in the circuit board from above using a Phillips screwdriver and, when there was enough space, pulled further on the bracket from behind. If it is very difficult to remove, bend or press a little more. Heat definitely helps enormously, but I managed without it.

After successfully removing both brackets, inspect everything carefully and bend the metal bracket back into shape; the threaded sleeves fit very precisely and can easily become jammed. I actually prefer the original bracket because it will be more stable and heat-resistant.

Be careful when reattaching it, as it only fits properly in one direction.
Note the distance between the SMD components in the middle of the “square” and the bracket!

The original adhesive tape survived cold removal on both brackets in my case and will therefore simply be reused.

When rearranging the thermal pads, I used their previous position on the NVS 5100M as a guide.
Before positioning the VRAM pads, I looked at exactly where the cooler was located and aligned the pads accordingly. That worked well.

Now it’s time to put everything back together again.

Now you can insert the Quadro 2000M diagonally into the MXM slot – just like RAM.
Then press down gently and carefully secure it with the two Phillips screws you removed earlier.

Then reassemble everything as listed above, and don’t forget the new thermal paste*.
For “direct die” cooler contact (i.e., components without heat spreaders), I always spread it thinly over the entire die to make sure everything is covered. Notebooks never really have heat spreaders, or at least haven’t had them for a long time, so I recommend spreading the thermal paste thinly.

I use Arctic’s MX-4* as standard, as with the CPU before.

Once everything is back together, you can turn on the notebook and boot it up.
And hopefully look at the display.

It’s working, and I’m getting a picture!

So I can conclude that:
Basically, an Nvidia Quadro 2000M with 2 GB VRAM, which even has Dell as its vendor and not HP, runs in an HP EliteBook 8540p and in 99.99% of cases also in an 8540w!

I estimate the probability that a closely related Nvidia Quadro 1000M would also run as extremely high.

As announced, the next part will continue with the necessary driver modification.

Der Beitrag <h5>RetroBook 8540p #7: </h5><h3><b>Upgrading – MXM GPU – Additional options</b></h3> erschien zuerst auf flohs blog.

The familiar service manual also helps, at least in part, with this latest “major” hardware upgrade: https://www.levnapc.cz/ProductsFiles/hp-elitebook-8540p-8540w-manual-en.pdf (archive.org).

On page 80, there are some interesting things to say about graphics cards, and in this case, really “-cards”:

All of the EliteBook 8540p models I purchased had the smallest GPU installed – an Nvidia NVS 5100M with 1 GB GDDR3 as VRAM. This is specified with a TDP of 35 watts.

Until a few weeks ago, we had been using them for years without any problems – simply because the performance was sufficient for the (very) old games we play semi-regularly (EE, EE-Zde, AoE I and II, AoM, SS TFE and TSE, Anno, RCT, VG…) at 1600 x 900 on good to maximum settings.

Can the smallest GPU – the NVS 5100M – do more?

Yes, these small GPUs could be noticeably overclocked with MSI Afterburner, giving them a little extra performance. Of course, every silicon die is slightly different and can be overclocked slightly differently (or not at all, or only slightly). In our two 8540p devices, an NVS 5100M could be overclocked a little more than the one in the other device. Here is an overview:

That’s very impressive, especially with the original voltage curve (which cannot be changed), and all that in a notebook!

Back then, Nvidia simply left more freedom and reserves in the silicon.
The overclocking was particularly noticeable in areas with low FPS (e.g., EE with a large number of units…).

Why upgrade if the performance is still not sufficient?

We were actually very satisfied with this GPU, even though it was very small.
But then… after years of loyal, maxed-out overclocked service, one of the EliteBooks started to shut down at irregular intervals with short black screens, often accompanied by driver crashes, which led to a permanently black window in 3D programs (games), even after the graphics driver had been restored, which was irretrievably permanently black.

Annoying, especially during LAN games.
And when you haven’t saved often enough. Which is basically always.

Reinserting the card, swapping them around, cleaning the contacts with 99% isopropanol*, applying new thermal paste*, even swapping the RAM between the systems (who knows if that would help?) – none of it worked. RAM tests over several days yielded no results. Even with the original clock speeds and even significantly underclocked, the dropouts still occurred, but less frequently.

The error also occurred independently of the operating system, so it soon became clear that it must be a hardware issue, probably one of the VRAM modules slowly giving up the ghost.

Replace or upgrade right away?

So there was a new goal that was never actually planned:
Replacing or upgrading the GPU – which, fortunately, was possible thanks to the MXM-A 3.0 slot.
Very few notebooks have something this great – if the GPU had been soldered in place, as is the case in most mobile devices, even back then (it’s simply cheaper to manufacture), the motherboard, including the GPU and VRAM, would have had to be replaced, which would have been completely unreasonable financially, and plug-in systems are also much better from an environmental perspective.

Especially since it would have been the same GPU again and again, at best a different one available from the factory, as can be seen above. However, MXM opens up significantly more possibilities – unfortunately, there is very limited information available on this specific topic about what is possible or, more importantly, what is not possible.

Let’s take a look at the options, starting with the most important one:
The EliteBook 8540p and 8540w have an MXM-A slot in version 3.0, which accepts cards with a maximum TDP of 55 W in the 82 × 70 mm format.

Now we can start our search.

The “safe” options that could also be ordered directly at that time:

Here are the GPUs that HP offered at the time, all in the appropriate MXM-A 3.0 format, of course, along with PassMark scores from technical.city for a rough classification of performance:

The ATI GPU seems to be relatively powerful, but for the reasons mentioned above, I would have preferred to stick with Nvidia.

What do we see?

All of the options offered provide “only” 1 GB VRAM – not particularly satisfactory; an increase would be desirable if you are upgrading anyway. All cards use 8 memory chips with 128 MB (1024 Mbit) each, Nvidia always GDDR3, ATI the faster GDDR5 standard.

The NVS 5100M and FX 880M are virtually identical in terms of performance.
At first glance, the Nvidia Quadro FX 1800M looks the most interesting to me, even though the ATI offers slightly more performance.

It should also be mentioned that all cards in this table use PCI Express 2.0, so there is no difference worth mentioning. The EliteBook 8540p always provides x16 2.0 lanes on the MXM slot.

All three models from Nvidia are based on the Tesla 2.0 architecture, which means that these technologies are supported:
DirectX 10.1, OpenGL 3.3, Shader Model 4.1, CUDA 1.2 / 1.3, OpenCL 1.1

During our extensive research, we frequently encountered references to BIOS whitelists at HP, Dell, and Lenovo, often accompanied by YouTube comments providing valuable information, albeit in sometimes poor English. Does this also apply to MXM GPUs, meaning that only the cards listed in the HP table above will function? There are numerous uncertainties.

In the next post, I’ll continue with what else I found out and how I ultimately decided.

Der Beitrag <h5>RetroBook 8540p #6: </h5><h3><b>Upgrade – MXM GPU</b></h3> erschien zuerst auf flohs blog.