90K+ Video content creation + AI Local LLM model

B

bekz

Beginner
Hello Everyone,

I'm building a configuration that can last for 5+ years. Could you please help me balance my needs? I'm focusing on 1) good performance within budget, 2) finding reliable parts with excellent after-sales support, as I can't afford a downtime of more than 3 days for my work, and 3) Looks aesthetically nice.

Current configuration (pcpricetracker) costs 1.85L excluding few components!

What is your budget?Upto 1.5Lakhs
What is your existing hardware configuration (component name - component brand and model)N/A
Which hardware will you be keeping (component name - component brand and model)N/A
Which hardware component are you looking to buy (component name). If you have already decided on a configuration then please mention the (component brand and model) as well, this will help us in fine tuning your requirement.
  • CPU - AMD Ryzen 5 7600
  • Motherboard - GIGABYTE B650 AERO G
  • GPU - RTX 4070 12GB
  • PSU - NZXT C850 80+ GOLD Fully Modular
  • Cabinet - NZXT H6 Flow ATX White
  • RAM - G.Skill Trident Z5 Neo RGB 16GB x 2 DDR5 6000MHz
  • Internal SSD - Samsung 990 Pro 1TB
  • Internal HDD - Seagate Baracuda 2TB 7200 RPM
  • Monitor - LG 32QN600 - 32"
  • Keyboard - Corsair K60 Pro Mechanical
  • Mouse - Logitech G402 Hyperion Furry Wired
  • UPS - APC Backup Pro 1500G
  • Speakers - Sonodyne studio monitor
  • Cooler - Not planning. Feels not needed. Not an expert!
Is this going to be your final configuration or you would be adding/upgrading a component in near future. If yes then please mention when and which component
  • No compute upgrades for 5 years
  • HDD will be added later upto 4 nos with the goal of configuring RAID 1 disk mirroring to saveguard from harddisk damages. No plans to purchase external NAS storage.
Where will you buy this hardware? (Online/City/TE Dealer)Online
What is your intended use for this PC/hardware
  • Video content creation for Youtube Tutorials, Instagram reels etc.
    • Daily content creation where video will be proccessed from Sony A7r4 14BIT RAW, XAVC S 4K: 3840 x 2160 (30p, 100M)
  • Creating bootcamp & full stack development content for AI LLM like Mistral, Llama etc.
Do you have any brand preference or dislike? Please name them and the reason for your preference/dislike.
  • Do not want to buy products with bad after sales service in India
  • Prefer made in India in to save import duty money
If you will be playing games then which type of games will you be playing?
  • Fortnite
  • Need for speed
What is your preferred monitor resolution for gaming and normal usage
  1. Gaming - 1280x720
  2. Desktop - 1440x900
Are you looking to overclock?No interested in overclocking. If any configuration can be adjusted to save money, would be super interested.
Which operating system do you intend to use with this configuration?Windows 11 64 bit
 
Last edited:

1) Intel performs better for productivity workloads.. since it has much better single core performance and more cores.. since you dont care about upgrading every year intel is a better choce.
2) Zotac has 5 years warranty and RMA'ing is done via kaizen which is miles better than RPtech.
3) Deepcool's RMA is also pretty good.
4) This Deepcool case has 3 x 140mm's front and 120mm rear included.
5) You'll still need to add your peripherals.
 
  • Like
Reactions: bekz
A intel build will be better for productivity (13600k). Used RTX 30 series GPU will also help you save money, unless you plan on doing AV1 encodes then better to get RTX 40 Series.
 
  • Like
Reactions: bekz
Thx, Will compare Intel i5-14600KF & 13600K!
HE-MAN said:
A intel build will be better for productivity (13600k). Used RTX 30 series GPU will also help you save money, unless you plan on doing AV1 encodes then better to get RTX 40 Series.
Click to expand...
Encoding wise fine with 30-series! But now confused between 3060@16GB, 4060@16GB, 4070@12GB. Since 12GB to 16GB VRAM helps in AI LLM, optimizing it towards 16GB!
 
your config looks good stick with AMD , after five years you would be changing the processor and not the the whole rig .Nvidia GPU with larger VRAM will help you for quick transcoding especially AV1 encoding support on RTX4XXX series
 
  • Like
Reactions: bekz
Reworked on this based on the suggestions and to bring down the cost from 1.85Lakhs to 1.6Lakhs.

https://pcpricetracker.in/b/s/2083e3f6-8e6b-4273-9cad-b6e8aa56a616

ComponentModelConsiderations made
GPUZotac Gaming GeForce RTX 4060 Ti Twin Edge 16GB GDDR6 PCI-e 4.0 (₹44.2K)
Notes: Ti provides extra VRAM of 16GB suited for AI LLM. Power consumption: 165W, Recommend PSU: 500W.		Inno3d GeForce RTX 4060 Ti Twin X2 16GB GDDR6 (₹42K)

4070 is costly & needs more power.
MonitorLG 32QN650 - 32"(₹25.4K)
CPUIntel i5-14400F Processor – (₹20K, 10 Core, 16 Trd)

Note: Opted due to price + cores + DDR5 support
  • AMD Ryzen 5 7600X – (₹20.9K, 5.3GHz, 6 Core, 12 Trd)
    AMD Ryzen 5 5600X – (₹14.7K, 4.6GHz, 6 Core, 12 Trd) - doesn’t support DDR5.
  • AMD Ryzen 7 7700X (₹32.9K, 5.4GHz, 8 Core, 16 Trd)
  • Intel i5 – 146000KF (₹ 29.9K, 14 Core, 20 Trd)
UPSAPC Backup Pro 1500G (16K)
MoboMSI B760M Bomber Wi-Fi (₹11.7K)
Note: Opted due to price & DDR5.
  • AMD: ASUS TUF B550 PLUS (₹14.9K)
  • AMD: ASUS PRIME X570 PRO (23.9K)
  • AMD: ASUS PRIME X670 P WIFI, DDR5 (29.9K)
PSUNZXT C850 80+ GOLD Fully Modular (₹10650)
Note: Kept 850W for accommodating any future needs.		Deepcool 850W (10630)
RAMRIPJAWS DDR5 6000MHZ CL40 16GBx2 (₹9630)
Note: Went with performance than aesthetics.		ADATA XPG Lancer DDR5 5200MHz RGB 16GB (2x16GB) (9200)
Internal SSDSamsung 980 Pro 1TB (₹9825)
Note: Went with safe brand Samsung
Internal HDDWD Blue 2TB 7200 RPM WD20EZBX (₹5449)
Note: Opted due to price.		Seagate Baracuda 2TB 7200 RPM (6550)
CabinetNzxt H6 Flow Atx Mid Tower White (₹9999)
Note: Aesthetics of 2 side tempered glass.		DeepCool CH780 ARGB (15.2K)
CoolerDeepcool LS520S Liquid Cooler White (₹7.5K)
Note: Opted due to prices & better RMA		NZXT Kraken 240 RGB Liquid Cooler LCD White (₹13.5K) – LCD ₹500 compared to x53
NZXT Kraken X53 Liquid Cooler RGB (13K)
 
Those telling intel is better for productivity have certainly never used it for AI and other machine learning workloads.

Ryzen 7000 architecture is actually good for CPU offloading and CPU acceleration in tensorflow and PyTorch.
It has to do with the cache size and architecture.

For AI, especially Local LLM inference, VRAM is king.
Anything above 8B params, becomes difficult to run.
You could run heavily quantized Models, but you get unbearable quality in inference.

You want to run LLMs at the cheapest?
Get dual rtx 3060ti or 4060 ti 16 gb models.
Keep in mind, it'll take a bit (or a lot) of effort in getting them working, but it's the only cheap option available.

And don't ignore CPU offloading.
It can actually make things cost effective.
 
Top Bottom