Hello guest, if you like this forum, why don't you register? https://fanrestore.com/member.php?action=register (December 14, 2021) x


Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
AMD new ZEN2 CPU and NAVI GPU
#1
AMD come back with great CPUs, and quite good GPUs...

If someone here thinks about getting new CPU and/or GPU, I'd take AMD in serious consideration - actually, I'd buy their CPU for sure, best price/quality ratio at the moment, while I'd wait for few weeks for GPU.

Opinions?
Reply
Thanks given by:
#2
I am not sure about CPU, but one major argument for NVIDIA GPUs is CUDA for me. Sure, machine learning theoretically works with AMD, but many applications, especially open source ones, use CUDA and AMD cards don't support that.
Reply
Thanks given by:
#3
Agree that, if you use CUDA accelerated software, the only way to go is Nvidia. But, we are a restoration forum... one of the most used software here seems to be DaVinci Resolve (I do not take in account avisynth, as it usually not use GPU at all).

Well, with Resolve, AMD GPUs are even better than Nvidia (at comparable price, of course).

About CPU: still not 100% sure about single core supremacy (even if pretty much they are there), but more than sure that multi core performance is better than Intel - without talking about vulnerability fixes that should slow down AMD around 3% at most, while much more for Intel (IIRC up to 38% in multithreading, for full mitigation)
Reply
Thanks given by:
#4
Fair. But the future holds so much for machine learning. Think of what DrDre did with it, I think he used it to create an even better color matching algorithm. Dunno if that was CUDA, but I think it might be. Also, upscaling, denoising, all things that can profit from machine learning potentially. Still a bit hard to use, but what in 5 years time? Smile

I have heard of those security vulnerability fix slowdowns, it's quite annoying to think of. Also Intel has that Management Engine afaik which is basically a secret hidden operating system that nobody knows what it does. But I'd bet AMD has something similar, so there's that.
Reply
Thanks given by:
#5
Yes, AMD has an equivalent of that, but it's less advanced that Intel's IIRC.

Unless some catastrophe happens, Zen 2 will destroy Intel and their Skylake rehashes
Reply
Thanks given by:
#6
The new 16 core Ryzen is looking very strong for a home user, as is the more affordable 12 core, both would give very good performance for the price.
Machine learning isn't unique to Nvidia by the way, we are doing ML on AMD hardware, they have a robust toolset for it.
Donations welcome: paypal.me/poit

bitcoin:13QDjXjt7w7BFiQc4Q7wpRGPtYKYchnm8x
Help get The Original Trilogy preserved!
Reply
Thanks given by:
#7
(2019-06-13, 06:23 AM)poita Wrote: Machine learning isn't unique to Nvidia by the way, we are doing ML on AMD hardware, they have a robust toolset for it.

I know AMD supports it in principle, but the few Github repositories I saw in the past typically required CUDA. Might just be confirmation bias I suppose. Of course this wouldn't matter if you write your own algorithms but I don't quite have enough programming skills for that.
Reply
Thanks given by:
#8
Today is the Zen2 and Navi launch date... some thoughts? Someone grabbed them, or think about it?
Reply
Thanks given by:
#9
My company got a 12 core Zen2 and a 5700XT card in for testing.
So far it easily outperforms the 10 core intel machines we have here, by quite a margin.

For compute, the Navi card is substantially faster than the 1080Ti in the intel machine, but we have only done a few test renders so far.

Will have more details, but if you don't need CUDA, then the new Navi based card screams along for compute tasks.
The new 12 core is very, very quick. We will be getting a 16 core in September when it is released. They make for a very nice workstation, particularly with the ASRock boards supporting thunderbolt 3.
Donations welcome: paypal.me/poit

bitcoin:13QDjXjt7w7BFiQc4Q7wpRGPtYKYchnm8x
Help get The Original Trilogy preserved!
Reply
Thanks given by:


Forum Jump:


Users browsing this thread: 1 Guest(s)