Why shouldn't I just buy a somewhat used workstation and upgrade the GPU? This is $820

Why shouldn't I just buy a somewhat used workstation and upgrade the GPU? This is $820.

Other urls found in this thread:

pcpartpicker.com/list/hDXzBP
pcpartpicker.com/list/hDXzBP/by_merchant/
twitter.com/AnonBabble

I don't see any reason why you couldn't.

You would't get a somewhat used gf.

It's not a bad idea so long as you don't mind the power consumption. It'll be future proof so long as you are using multi threaded stuff but if not your money is better spent on a standard i7 or ryzen

Used Dells are an incredible value

Buy it for what?

I need a new workstation to replace my aging i5 quad core based system. Building a new machine with Ryzen and a workstation GPU is easily $2000+. This almost cuts my expenses by about 30% after getting a GPU. Already bought an AMD Radeon Pro WX 5100

I've done this, and I can tell you that this works incredibly well.

It's 6c/12t + 6c/12t at 2ghz each vs 8c/16t Ryzen which is running at around 4ghz all core before an OC, not sure what the end speed difference would be.

Not sure I would waste anything on an old workstation like that, it might be good for a NAS but likely isn't great for a daily driver system.

around $1500 to start looks like, at $2000 with a 1950X it's 100% going to be a faster system.

Would go for the X399 Taichi though

PCPartPicker part list: pcpartpicker.com/list/hDXzBP
Price breakdown by merchant: pcpartpicker.com/list/hDXzBP/by_merchant/

CPU: AMD - Threadripper 1900X 3.8GHz 8-Core Processor ($449.00 @ Amazon)
CPU Cooler: Noctua - NH-U14S TR4-SP3 140.2 CFM CPU Cooler ($79.90 @ Amazon)
Motherboard: Asus - PRIME X399-A EATX TR4 Motherboard ($331.49 @ SuperBiiz)
Memory: G.Skill - Ripjaws V Series 32GB (2 x 16GB) DDR4-3200 Memory ($299.99 @ Newegg)
Storage: Western Digital - Caviar Blue 1TB 3.5" 7200RPM Internal Hard Drive ($41.89 @ SuperBiiz)
Case: Thermaltake - Core X9 ATX Desktop Case ($147.19 @ Amazon)
Power Supply: SeaSonic - FOCUS Plus Gold 850W 80+ Gold Certified Fully-Modular ATX Power Supply ($88.99 @ SuperBiiz)
Total: $1438.45

What do you require a workstation GPU for?

Afaik Titan/Vega Frontier have their pro features enabled to specially for prosumers

Machine learning and image processing prototyping.

I have access to incredible hardware but it means possibly waiting for hours, and then getting a small amount of time to test my code. It's good to be able to get preliminary results without leaving your room or asking for permission.

It's like the days of the old timeshare mainframes here.

Good system for workstation, but for gaming, not so much.

They're also great for home servers if you don't care about power consumption since they are usually about as fast as servers, but are nowhere near as loud.
Plus you can actually put an x16 gpu in there without some stupid expensive adapter.

I had a shitty fx 6300 build from years ago but I refuse to upgrade until ddr4 ram prices drop or I can find enough used ddr4 ram for cheap.

So I got one of those dell optiplex with a i7 3770 8gigs of ram and added a gtx 1050ti.

Pretty gud.

Most people do to be honest

And why exactly cant you do that on several gaming cards?

The "workstation" GPUs have a very short list of highly specialized features that gaming cards do not, odds are you will never need them.

>Machine learning
Are you a rich fag?
Y -> DGX Station with V100s
N -> 1080 Tis

You can get a better prebuilt machine with new components today for 100 dollars less

Admittedly I am not very familiar with current trends in hardware, but when I built my previous machine, GPGPU work with double-precision floating-point maths were more highly optimized on workstation class GPUs. If this is different now I might need to reconfigure my build

just buy AMD FX

this hits too close to home

Not everyone wants to feed into the shit show that is CUDA, locking everyone's code to a single GPU vendor for eternity.

Not one had a choice when AMD does jack crap

>amd fx
that shit belongs in the garbage bin
fuck the fx series

Vega is pretty good on compute, and Vega FE offers 16gbs of VRAM at under $1000 right now, Nvidia ain't got shit on that.

I found one of those in the dumpster.

None of that matters when there's no downstream software ecosystem for AMD. No major machine library has actively maintained forks for ROCm

*machine learning

Which is why we need to build it, also Vega has HBCC to extend it's VRAM, from what I gather VRAM is pretty important for machine learning stuff, like some stuff requires a titan XP just for the 1GB more VRAM

>Abloo bloo no ecosystem

Then go and make one

Yeah who's paying? AMD sure isn't.

Welcome to the free software world, do not expect payment. You make your own way and figure it out as you go.

then don't expect your cards to be used

>Power supply: single power supply
It's shit, if they choose to hold out info, it's because it's embarrassing