M4800 Owner's Thread

[Parad0x]

Hi All.

 

I know these questions are likely covered in the previous pages, though i'd prefer to get the answers now, order the parts, then read this whole thread.

 

I became a new owner very recently, completely unaware of the lack of M2 drive onboard - It's just become so ingrained in everything i own, that i never expected a fullsize laptop to not have one... It does somewhat explain the low price, i got a reasonably well specced machine at, but honestly i don't know if stepping backwards in storage technology is something that i can stomach.

 

What method has come out on top in terms of providing as much bandwidth as possible? (and what is the limitation?)

 

Just before we dig in - There's three pressure "pads" on the inside of the laptop undercarriage. Can anyone tell me if these are meant to be padding/damping points/pads, or whether they were once intended to be thermal transfer points from the chassis to the wide surface area of the baseplate? I cleaned them away and i have my suspicions, but i wish to be sure. I cleaned out the radiators and pulled apart the fans when i got it, but i didn't repace any thermals yet, as there's this last job which seems like it might be a total strip down.

 

I know little about ExpressCards - This was how the thread opened, but eBay shows a numer of EC - USB3 ports, which gives me a glimmer of hope - theoretically operating at 5gbps+, but i don't see the EC being USB3.2 Gen 2x2 (20gbps) overall, giving the full throughput of 4 lanes of PCIe3 as M2 is capable.

 

There's mSATA (too slow, essentially USB2.0), and there's the WiFi slot - Likely using PCIe A & E - Would be my first choice if i chould live without WiFi/BT. Then there's the the 1/2 Height mPCI slot, likely a type A slot. The final, drastic measure is to remove the nVidia GPU and connect via the MXM (which to me doesn't actually bother me too much, as the only GPU i've owned in the last 10yr is a GT710 PCIe x1 card - for when you get caught without iGPU on a homeserver. It's served me faithfully, fits anywhere, runs on anything and gives me 3 display output types).

 

So, before i banish my GPU to the e-cyclers, what throughput does the ExpressCard provide? (5gbps duplex as i'd imagined?) Has booting from this storage been resolved at all?

 

Next step would be, sacrifice internal WiFi and make use of the A&E slot (1 PCIe lane... 2.5gbps duplex in theory) - Anyone been down this path? Did this exceed base USB3.0 speed or was the difference negligible?

miniPCIe, also single lane....

 

Last, and step i'm happy to take, as it will reduce the power consumption, and heat output of the laptop overall, and no actual loss to me - Remove the nVidia GPU and unlock proper, full potential NVMe - TRIM, SMART, should also be bootable... Who has taken it this far? This is likely where ill actually start... It's only benefits for me, as the GPU is of little use - iGPU can provide me enough!

 

Whilst it's entirely possible to get 8x SSD drives, without touching USB or eSATA, entirely internal

• B+M key SSD to 2.5" SATA (#2)
• optical 2.5" adapter (#4)
• WWAN mSATA 2 M2 (#5)
• (WiFi Slot) A+E Key 2 M2 (#6)
• miniPCIe 2 M2 (#7)
• MXM to M2 (#8)

Most of these wont be extraordinarily fast.

 

It's merely displaying that there are many ways to go about it, depending upon your needs. Like most, I'm not interested in migrating my fileserver to my notebook.

 

Last i heard. Brocade had produced an enterprise backplane which addressed 30/32 NVMe drives... I also imagine this would have been quite large for practicality sake, but realistically 240TB high speed, low power would take up less volume than a new mobile phone box! Times are a changing.... Again!

 

Spoiler

Slightly OT, but discussing the future of flash storage.

 

I've had exposure to numerous different NGFF/Occulink adapters, converters etc building custom routers from N100 era HW, by utilising the various high-speed, onboard connections for things like 2.5Gb, 10Gb, 6x2.5Gb outputs. I suppose you could say this education first began when i was mining Ethereum, and (I can't perform something without understanding how it works) the need to have a much higher density of GPU's / PC was desirable. I stopped at 10, possibly brcause there wasn't a single GPU remaining in Australia (or the world who were months ahead of us) at that stage. This along with exploration of the Xeon architecture & chipsets and how they differed from their retail counterparts. Whilst i never stepped away from consumer grade HW to complete my build, the lessons stayed with me. Without moving to server grade HW, where $/PCIe lane is both cheaper (in reference to older enterprise HW ig. Intel skt2011, e5-26xx v1-4 CPU's) These are cheap as chips, newly released consumer boards are being released to take advantage of the silicon overflow... Someone like Supermicro or such simply needs to follow in the Chinese motherboard manufactueres pioneearing footsdteps, and release a modern day,  affordable board utiising the likely much more controllable C4-600 series chipsets - It's only then will we see high numbers of NVMe drives each accessing 1 or more lanes per drive, both through bifurcation, PLX if you dare, but also due to the increased number of lanes in the Xeon architecture and the custom configuration that can be undertaken by the motherboard manufacturer to allocate those specifically where required.

 

Getting way ahead of myself here though, because really, who needs access to 8x NVMe PCIe 4.0 drives (80gbps PER drive), in a home, enthusiast or SMB market? Just the sheer amount of data needed to saturate a single link/drive for any noticeable period is today, unimaginable on a scale beneath the cloud, massive amounts of data manipulation and generation (AI), and as always, Enterprise hardware is one long jump, and a few hundred thousand dollars away from our realities. I just wanted to talk about this, because in another forum there's many users asking why they can't get 4x NVMe > x16 PCIe card working, or if they do, they lose another x8/x16 entirely and a x16 gets reduced to x8 etc. There's no need for people at home to worry about stuffing as many NGFF drives inside their file servers just yet. However, if you do, there's absolutely no need for them to require anything more than a single PCIe lane. This is the minimum without the use of a PLX chipset, though we might see a bit more use of these to allow addressing massive amounts of NVMe drives in a single server. Whereas a x8 HBA & 20-port SAS expander - each connected to 20TB HDD's, created alot of Storage Space, flash isn't quite on par.

 

[Aaron44126]

Pretty sure ExpressCard is one lane PCIe 2.  Maybe two lanes?  I remember looking into it for an eGPU solution and it looked like the bandwidth was going to be quite limited.  (Edit -- Wikipedia says one lane.)

 

This system has SATA3 for the mSATA slot, so it should be a fair bit faster than USB2.  It would be on par with other (2.5") SATA SSDs.

I think the other mPCIe slots available are PCIe2 (up to four lanes), except for the MXM slot which is PCIe3 (16 lanes).

 

I have not tried installing an NVMe drive into one of these systems, but I did install a NGFF M.2 Wi-Fi card into one of the mPCIe slots using an adapter.  There is vertical clearance even with the bottom case on, but you have to be careful because the adapter can push down and scratch the motherboard.  Put some electrical tape or something underneath.

You can fudge a NVMe card into this system (using other PCIe slots with an adapter, as you have noted) but it won't be directly bootable.  (The BIOS doesn't know to look for it.)  You can boot it by chainloading with the OS on the NVMe drive, but the boot loader on another drive that the BIOS will check.

 

I have not heard of someone using an NVMe drive in the MXM slot on this system, but I've seen it done in Alienware systems.  There can be battery/power issues with the NVMe drive drawing power even when the system is powered off.

[KMikhail]

Howdy,

 

I have mSATA SSD for OS/ScratchDisk and 2x 2TB Samsung 2.5" HDDs in a ReFS mirror installed. Honestly, if we're talking about I/O subsystem, the old 4xCore processor and old gen video card (M2200 in my case) do not warrant much more throughput. However, if only 8TB SATA 2.5" SSD (Samsung QVO) were cheaper (as I'd hate to throw a lot of cash into the older tech), I'd be happy to get two of them, split them 1+1 Raid0, 7+7Raid1, and enjoy 1Gbyte/s for programs and scratch space, and 7TB of bit-rot-protected mirror for my data.

 

There's a better compatible WiFi card (6/7), but again - what is it really providing in everyday computing? At some point I connected my NAS over the WiFi for portability, but it was a) slower than 1Gbe b) not secure. Internet speeds where I am are lower than what original WiFi provides.

 

There's a "5GBps" Ethernet adapter, that provides about 3-3.5Gbps, in case one has a sophisticated NAS (i.e. Flashstor Gen 2 from ASUS, that has 10Gbe/5Gbe/2.5Gbe ports). That'd be a nice boost over 1Gbps indeed! One may bypass upgrading internal I/O and go straight for a really fast external one, which will be future proof, but not portable, of course.

 

In raw performance department there's potentially an opportunity to punch real hard with 4050 MXM, but only if it works. The M2200 is significantly faster than original K2100, the 4050 MXM will more than triple the performance again (+250%). But, alas, it is a 6GByte card, so not as cool for acceleration of video/photo editing.

 

Regards

[jerboy04]

 

On 11/16/2024 at 7:12 PM, Aaron44126 said:

 

Do you know what Wi-Fi card it has installed presently?  Most cards should work out-of-the-box on Linux.  In any case, you can find an Intel 802.11ac mPCIe card for cheap (if that's not what it came with) and that will definitely work fine in Linux (Intel contributes drivers for their cards to the kernel).  You can even install a newer NGFF Wi-Fi card in the WWAN slot with an adapter, if you don't want to put a mSATA drive there.

 

If you can get NVIDIA Optimus working properly, this thing should have respectable battery life.  (That would be Windows only.  You need newer GPUs to use Optimus on Linux.)

yeah i took a look, and it is the dell 1601 card. there are drivers for it, but for ubuntu 12, and i tried to brute install it on ubuntu to see if theres any signs of life, alas theres nothing, so the next best thing i could do right now is harvest the wifi card out of a latitude e7240 i have lying around, and probably try to see if it works with this machine (i know it runs well in linux because it has debian 11 on it). i also tried to check out optimus to see if it would work, this is the error i get:

 

i have the latest drivers for this machine, and the nvidia card is currently disabled as it is, but even with the graphics enabled, i still get the same error. heres a picture of the driver:

 

its not bad right now, if anything i have access to an outlet, but it would be nice to have better battery life on this machine, its really growing on me to be honest.

 

as of right now, this laptop is doing great. only have a couple of gripes still, like with the lower ac adapter warning, since i am using one of those amazon GaN chargers + usb -> barrel converters, and still struggling with battery not going to lie to you. I also found a japanese keyboard off ebay, but its for the latitudes of this era. might try to see if i can retrofit it in, since i did test it, and it surprisingly works with this machine. considering swapping out the dvd drive with a hdd sled for more storage space, but i think 1tb off 2 drives is more than enough. repaste with ptm and keyboard mods after im done with my finals for this sem.

[Aaron44126]

Pretty sure the keyboard is the exact same part as the one used in 15" Latitudes of the day.  If it "looks the same" then it will probably fit and work fine.