Are you poor or something?

Are you poor or something?

Other urls found in this thread:

youtu.be/xBtlJBUbjV0
twitter.com/NSFWRedditVideo

Just 15k USD.

>not buying 2xTR 1950X for 64C/128T
kek

I heard threadrippers had problems with some heavy loads and would crash.

>Not being rich enough to have a super computer with 500K cores that can run Crysis @ 60fps
You rich-poorfags are the worst

source; Intel Damage Control Team

Supercomputers only run Loonix though, so no games. :(

or graphics card drivers for that matter

>that can run Crysis @ 60fps
HAHAHAHAHA OH MY GOD

dunno, running dual xeons right now

Or any useful software.

Do devs ever test with retarded setup just to see if their engine dont break in there real world if it has too many core to work with?

Or should a game that works with 4 cores automatically work with more than that?

>205W
No matter how much money I have, I have no justification for wasting so much power.

The usefulness of such a test would be future proofing to make sure the game doesn't crash on future consumer CPUs with more cores.

what hash does it get on monero?

>too many cores to work with
In an online problem where the amount of data is unknown, there is no such thing.
You can horizontally scale for as long as your data throughput can saturate each core.
In an offline problem where the amount of data is always more-or-less known (i.e. in a video game), there are diminishing returns on using more cores. You have more overhead handling task and data distribution, meanwhile certain algorithms (the ones that matter in video games) are known to be very serial in nature. This is Amdahl's law.

>crash on future consumer CPUs with more cores.
Impossible. The only reason an application would crash on a newer CPU is the deprecation of ISA extensions. It's more likely that an application crashes because an operating system's kernel and drivers became incompatible with the target operating system that the application was compiled to.

uhh each 1950X is only 16c/32t

a hash rate which is enough to pay itself for 200 years

I forgot to mention: Developers (for video games) don't scale to as many cores as you have available. Partially because of the diminishing returns I mentioned, but mostly because they're lazy and care more about the game than they do about performance. So a game will for the most part only use a fixed number of threads.

my bad I meant Epyc 7601 kek

Crypto mining hardware generally pays for itself in 3-6months, barring any major market fluctuations

Most of the problems games deal with are trivially asynchronous up to the final data exchange point. Game devs are just poorly paid overworked code grunts solving difficult problems with poor tools.

As for games that were very poorly tested for increasing core counts, fallout 3 comes to mind.

That is for 7980XE when using 18 cores and the game only sees 16

Most of the problems in games aren't asynchronous at all, which is why so few games make effective use of multiple cores. About the only part that's "trivially asynchronous" is the final rendering, which is why it's done on a GPU.

>fallout 3
The fuck is this shit? Does it just deadlock with more than two threads?
>writing your own multithreading libraries
>not even once

I don't own a server machine or something?

Right on brother!
youtu.be/xBtlJBUbjV0

i only have a consumer-grade motherboard with only 1 desktop CPU socket and desktop chipset?

On single core processors with or without Hyperthreading, it behaves as expected. One multicore, it crashes at launch

I/O is a trivially async task
AI is a trivially async and threadable task
Physics can operate almost completely independently of graphics frame preparation up to the final matrix result for each hull, which is then the only synchronization point before shipping the frame and its uniforms to the GPU, at which point, every other process can start async again.
Frame preparation is trivially async and threadable on dx12 and vulkan and only needs matrices from the physics iteration

>tfw komplett

/thread

It's not fast enough.