Upgrading from 10+ year old Asus G750JM Laptop to Legion 7i RTX 5090

I am planning on upgrading my main laptop, which is a trusty Asus 750JM

My main use case is DevOps with MLOps

Lenovo legion 7i with the following specs is what i am thinking of upgrading to

https://www.lenovo.com/in/en/configurator/cto/index.html?bundleId=83F5CTO1WWIN2

🖥Intel® Core™ Ultra 9 275HX Processor (E-cores up to 4.60 GHz P-cores up to 5.40 GHz)
Operating System: Windows 11 Pro 64
🧠 Memory: 64 GB DDR5-6400MT/s (2 × 32 GB)
💾 Storage: 2 TB SSD M.2 2280 PCIe Gen5 Performance
🎮 Graphics: NVIDIA® GeForce RTX™ 5090 Laptop GPU (24 GB GDDR7)
📷 Camera: 5 MP with Dual Microphone
🔋 Battery: 4-Cell Rechargeable Li-ion 99.9Wh

Reason for going with a laptop is, I do need the mobility sometimes.
Also will be grateful, if anyone can check the corporate discount that can be applied to this

What do you guys think of this laptop ?
Any suggestions ?
what other options do i have ?

Getting this for 3.6L

Still need suggestions

1 Like

I don’t know much about laptop so can’t give you any advice just check gpu wattage and if laptop is able to cool itself under load. Battery can be increased i guess by turning of gpu and switching cpu to low power

just for sanity

why??? Why not build a system and get yourself macbook or something, I don’t think laptop gpu will be much use in mlops??or it is useful ??if so enlighten me please. I my self is unable to understand why would someone spend on 5090 laptop gpu as price to performance is very low!!

Thank you for replying l, much appreciated

Some of the reasons of going for a laptop but not s desktop, at least for now are

  • portability - although this laptop is heavy , it’s still portable and can be carried around with ease
  • Desktop gpu needs more power, we do have solar at home, but it’s not always sunny and also not an option if solar is down for some reason
  • Currently desktop rtx 5090 is overpriced and most unavailable
  • I plan on inferencing only locally, training will still be in the cloud and mobile 5090 has bandwidth that is similar to that of desktop 3090, which should be good enough i suppose
  • Getting this at a good price offline compared to what is available online ~ 90k- 100k INR diff

Here are some of the dilemmas i have - prefer not to have a things that will be not used

  • Should i get a cheaper laptop and remote into the said desktop - but then latency is horrible when not in the same network
  • I could setup a less beefier desktop, to avoid power concerns
  • Get a surface pro 3 - will serve as a laptop and tablet both, since my Samsung tab s3 is slow and hardly useable for note taking

It is better to get desktop which are better than laptop

since u can get desktop and laptop which can been use both for adjust

i7 ultra with second hand 4080 or 5080 rtx gpu pair 32 gb ram or above with case that will cost around 150-200k depend on parts and detail

and good laptop i7 13/14 gen or amd ryzen 7 7-8 gen with 4060 RTX gpu around 85k-100k

why not inference on cloud for testing? is it because you want to run model in model at the end?
inference on cloud is expensive? it is my preference if it is work do it on work resources

yo can do cost calculation on inference and compare it to either desktop+ laptop price annually
or compare it to laptop price (on which you are getting deal for 3.6L )

desktop+ laptop you can use 4090 for 165K and rest you have to struggle to find it
7900x cpu → 35k
b650 mobo → 15k
64gb ram → 18k
2tb storage → 15k
psu 1000w→ 10k
cabinet → 5k
cooler → 5k
dummy hdmi → 250
remote in using tailscale, twingate and there are other free( for now ) product might charge fee later
lower end → 2.8l higher end→ 3.2l added 5k to each part

surface pro 1l something it will give you better performance and you can travel light

my work setup is 5090 7970x with two cable and dummy hdmi plugged in i remote in with my zephyrus g15 using tailscale i can also do remote desktop if i need to but very laggy using tail scale i don’t have to expose port through ufw (depends individually)
i game on my zephyrus and remote into home and office servers

if you can do used parts and laptop it will be much cheaper for better performance and if you game you can also go 4060 laptop new and also game and divide your work and personal things

just for your info https://mdcomputers.in/build_your_pc/build_picker/735542

1 Like

Those are good suggestions, i do already have 3 x RTX 3090, that are currently not in use.
Also have a hx1500i psu

So, i might get surface pro and build a SFF pc (only issue is hx1500i might need a bigger case and RTX 3090 is, well a room heater, so that needs good airflow as well)

That should give me a convertible + a descent desktop for < 350k INR

Any suggestions for the above SFF build ?

How about this:

getting double the ram - 64 GB, double the storage 2TB nvme, with windows pro for 3.6L in Lenovo legion 7i

@bssunilreddy what are your thoughts on whether to get a laptop with 5090 or build desktop with one of my rtx 3090 + buy surface pro as a convertible ?

If you want mobility going with the laptop is the easiest way out otherwise desktop is the way to go if you are not concerned with mobility that much.

Lian li a3 good enough for your PSU ALSO 3090 will fit

3090, Winter is coming so good.

use liquid cooling for cpu matx board are readily available.

You can use ipkvm for remote access, if you are conscious about power there are other option and I haven’t tried ip kvm

otherwise you can use any zero tier To remote into your pc/inference server But you have to keep your pc power on 24x7

And if you ever feel generous and want to give me 1 out of 3 3090’s I would gladly accept it
:backhand_index_pointing_right::backhand_index_pointing_left::pleading_face:

hey! upgrading from your old asus 750JM to that lenovo legion 7i looks like a beast for your devops and mlops work. the specs are crazy good—like 64GB RAM and that RTX 5090 gpu will handle heavy stuff no problem.

just keep in mind, it might get a bit heavy to carry around and battery won’t last super long under heavy use, so you’ll probably need to stay near a charger.

about corporate discounts, lenovo sometimes gives like 10% off if you qualify, or check if your workplace has any special deals.

maybe look at asus rog or msi gaming laptops , they’re also good for heavy workloads.

hope this helps!

buy a 1L oled thinkpad and make a multple RTX5070ti homeserver , so that you can use it as vGpu from anywhere.

Thank you for replying

Do you mean to say, I should get an oled ThinkPad for ~ 1 L and then set up a homserver separately with multiple 5070ti gpu’s ?

What hardware config would you recommend for the homeserver that can host multiple GPU’s ?