this is not meant to be a discussion of fanboys. I bought a GTX 1080 Ti Founders Edition when it came out and build my Ryzen R7 1700 based setup. Over the course of one year I used the GTX 1080 Ti for gaming and testing WPA2 hashes with Hashcat. Since I run Arch Linux installing the proprietary nvidia driver is easy ‘pacman -S nvidia’ so using it under Linux is easy for me. When I told a friend of mine that I get around 620kH/s on WPA2 he told me he has about 100kH/s less. since then I try to figure out if there is something the RX Vega 64 can do better than a GTX 1080 Ti.
I’m comparing those two cards because they cost about the same right now where I live.
So can you help me find outr what a RX Vega can do better than a GTX 1080 Ti?
Not really. It comes close for stuff like what you’re doing and machine learning, but the V64 is basically as performent as a 1070ti. So really if they’re the same price then just pick your favorite, but if its between that and the 1080ti nothing beats it right now. Except maybe the Fury X2, but that’s a dual GPU card.
Mods, stop editing my posts. It’s getting rediculus. Or at least pm first, if you have any reason to edit them. Talking about keeping the discussion “open” when i’m literally edited by the minute. And @wendell it’s been a week since i asked for deletion of my profile. Could you please answer my pm? Thanks
And open source i presume (Arch)? Then Vega, kernel 4.15 & 16 is out so the support is there and getting better, literally by the day. also for reference
well the proprietary nvidia driver works exceptionally well. it’s just not free. I use Linux because it’s an awesome OS but not because there is a lot of free software around it. I’m a bad person I guess
i don’ care why you use it. I’m just glad there are more and more linux users in general. More users, more support, etc.
Everyone is a bad, bad person. (shun upon the other) Now back to topic. If you don’t care about free software and nvidia works so well. Why question it?
Supporting and having interest in technology, goes together at least for me. Thinking it does for others also.
So if you want the absolute best performance in most scenarios and a Titan V is not an option the GTX 1080 Ti is the way to go. Otherwise the RX Vega 64 and RX Vega 56 look like really good GPUs to me. I wish they were all available on MSRP. The current prices make my 700€ purchase of the GTX 1080 Ti from last year look cheap now lmao…
Why question it?
When I tried to figure out how much performance the Vega series delivers most of the metrics I found were gaming benchmarks on Windows 10. You could even make my overclocked R7 1700 look worse than Intel CPUs if you only look at gaming benchmarks on Windows 10. I was kinda hoping for Vega to have it’s place in the market for me. Also Nvidia did some things in the last couple of months which make me want to not support nvidia anymore.
Probably correct. All the support and tweaks are there for the nvidia, intel combo. I had the FX at one point and it wouldn’t even run cs:go properly, who knows how many other titles. BSOD’s etc. Nothing wrong with the computer itself and at that time windows 10 wasn’t even out. Point being, you may have use cases that differ, so instead of one vs another. Get them both, when if/finances are there. Having both would likely mean your actual opinion on both cards
Personally more support for vega encoding, codecs etc. On linux would be awesome
I assume I’d need to upgrade my PSU then. I currently have a be quiet! Dark Pro Pro 11 750W and I guess if I’d want to use a GTX 1080 Ti and a RX Vega 64 in the same system I should upgrade to a 1000W PSU?
I could get a RX Vega 64 for about 900€ … which I don’t want to buy it for. Maybe Navi will be more affordable and maybe even highly competitive. Given that the AMD graphics driver on newer kernels are pretty good I don’t have any problem as an Arch user. But what are Ubuntu users or even Debian users doing when they get a Vega card?
Depends if they’re on the LTS branches, which run the LTS kernels. The unstable/testing run the latest stable kernel, and ones on newer hardware stick as close as they can to upstream. If you buy hardware from two years ago, you really won’t have any problems.
I would not run them both in the same system, rather put them to the test one by one. If you’re going to anyway, make sure your psu supports the pins needed, i would think 1000watts is enough support for them both under non-heavy loads. (if gold certified or higher) I doubt that Navi will be different from either polaris or vega which the market used to shovel loads of cash, the market will probably do that with navi to.
Whats not to do for Ubuntu/Debian users? The power of vega is amazing for many use cases on linux, even gaming and getting even better.
Are you saying power efficiency (certification) has nothing to do with how much (clean) power makes it through or it can withstand delivering in percentages for example? I thought higher certification meant, lower power loss for example among other things i’ve read at johnnyguru way back
Take a car engine for example.
100HP are 100HP, question is how much fuel it burns to get those 100HP
Same for PSUs, the secondary voltages and power is fixed, so the input is what varies.
For example: 1200W nominal power at 80% efficency equate to 1500W draw from the wall.
if you can use cuda for your tasks get the 1080TI if you dont use cuda get the vega 64. personal preference is the vega 64 but that has a lot to do with voting with your wallet and not backing bad consumer practices of Nvidia . if you dont care and about what any company does then anything works as long as you have the budget.
cuda to something that AMD can use is/ was being developed but it is unknown for how far it got or if the nvidia legal team managed to put a stop to it. have not heard about it in a while so i dont know.
ROCm is the more interesting platform from AMD for CPU + GPU compute. They also have a tool called HIP to “hippify” CUDA code into a C++ like portable language which can then be compiled for either NVIDIA or AMD hardware, thus potentially breaking vendor lock-in.
The big question is whether it catches on and whether frameworks, such as for machine learning, end up being supported well on ROCm. There is some progress, but it’s an area of active development.
While waiting for video card prices to come back to earth, I’m looking to see how things turn out for ROCm before deciding what GPU to get for compute.
