Linux support

Unless wine/proton/dxvk/whatever added DirectML Support, VEAI won´t use the GPU in Linux, since it uses DirectML. Last time I checked, it was not implemented in the DX-feature subset.

This certainly could explain why Gigapixel works now.

We optimized the library used for NVIDIA GPUs from DirectML to TensorRT.

Though funnily enough its Face Recovery that crashes it on mine, works fine with that turned off. Though as I use it A LOT its a lot less useful.

So you have TensorRT installed then? I for the life of me cannot get it installed as Nvidia have made it one of the most painfully convoluted processes I have ever seen. Will keep trying though.

I use Nvidia because Nvidia tends to beat out AMD on performance every time. Basically when it comes to specs, Intel wins on the CPU and Nvidia wins on the GPU. AMD does both but does both poorly. The only thing AMD had going on is that it offered more cores on the CPU than intel for years, but… now intel is adding more cores and AMD isn’t holding up that selling point anymore. Price wise anyways…

Debian based Linux distros are my preferred choice. I like Parrot OS, Pop OS, Debian, Pure OS…

That would be cool if there is a GUI at some point, but yeah, for now still cool that you are working on it.

You must have been asleep the last few years as AMD was beating Intel until 12th gen on CPUs.

Even now, Intel have better single-threaded performance but AMD are beasts for multi-threaded workloads, so its not clear cut and depends on the workload.

2 Likes

Thinking about it distributing VEAI as a Flatpak would be the best option in my opinion.

I’m still on Windows, but would prefer to switch to Linux if I can use Topaz. I use AMD. (6900XT)
Does ffmpeg have anything to do with VEAI?

For what it’s worth Topaz products ARE usable with Linux using WINE. The main issue being discussed in this thread is that they only work with cpu, not gpu.

And to update this little Bottles guide as it was originally for the .deb version which is no longer developed. It now seems to require the dll override “concrt140”. This can be done easily in the bottles settings. I have mine set to “Native, then Builtin” and all is well. Also the flatpak version needs access to system binaries and paths since it is sand-boxed by default. These permissions can be managed easily with the flatseal app.

Even an unsupported version, or supported only through the community forum, a “best effort”, provided for Linux, would probably be a great improvement compared to using Wine or a VM.
Regarding the votes in here: I have huge doubts, that people would exclude Topaz (because they want a Linux version but it does not run on Linux), then still connect to the forum just to vote for a Linux version …
Votes are relevant for features that concern people who own the software already (I use it through wine but it’s a horrible experience). So here, I think votes are just used to confirm what the development team already decided. Well of course, it’s ok, it’s their software after all, but it’s a bit frustrating :slight_smile:

linux support would be great! a .deb or appimage would be nice but having a version that works with wine with gpu would be great also.

I’d like to use VEAI on remote GPUs so I don’t need to keep my laptop running, but my GPU servers only run Linux. So it would be great to have it on Linux

1 Like

A little late to the party on this one but wow, what decade are you talking about? AMD doesn’t do anything “poorly” against Nvidia or Intel. At best 3000 series Nvidia GPUs are slightly better than the equivalent AMD GPU at quite a bit more money. Last I checked a 6900xt beats a 3090 in VEAI. I have a 6900xt, it’s a beast with VEAI.

As for CPUs, no idea what you’re talking about there. It’s constantly a game of catchup between AMD and Intel since Ryzen launched. Ryzen 1st to 3rd Gen couldn’t match Intel in IPC so they loaded the CPUs up with cores. 4th gen, (5000 series) began to beat Intel in IPC, price and core count so for a solid 2 years nobody was buying intel for performance at all. Intel may be catching up or pulling ahead at the moment but AMDs refresh will halt that. And keep in mind, When Intel was stomping AMD, their better CPUs were over $1000. When AMD started really competing Intels prices dropped and their chips got substantially better

It’s OK to be an AMD hater but one thing to keep in mind is that Intel was completely stagnant for about 10 years because AMD couldn’t catch up so they were the “budget” option. Once AMD started to make Intel sweat, then actually beat them, Intel stepped it’s game up. Pretty much the same deal with Nvidia on the GPU side. So one thing to remember is that if either Nvidia or Intel are actually improving substantially each cycle, you have AMD to thank for being worthy competition.

2 Likes
  1. Does it matter which decade? (current)

  2. The Nvidia RTX 3090 benches roughly 30% higher than the AMD RX 6900 XT. I wouldn’t say that’s a small difference.

3 You can compare the price of both GPUS on amazon and the price difference is $71.64. To pay less than $100 for a 30% higher benchmark I’d say that’s money well spent.

  1. AMD had a cpu years ago called the AMD FX-8320 it had 8 cores and that was its huge selling point compared to intel. It was still slower than intel on core speeds… and it wasn’t ridiculously priced either. It took Intel years to start upping their core count from the typical 4 cores. Eventually Intel started upping their core count and AMD started decreasing their count and coming out with models that supposedly compete with intel… Well… at that point why even go AMD if it’s core speeds are on average lower, priced bascially the same, and their new line up mimics the same amount of cores as intel. If you want to mention threadripper, that’s not even in the same ball park price as intel at that point, might as well get an intel xeon server cpu that is used and overclock it for cheap.

  2. I don’t hate AMD, I wanted, like really was rooting for AMD to beat out intel one year when I was building a pc and was really let down, the hype of their cpu’s was suppose to blow intel out of the water… They decreased the NM size of their cpus, this was suppose to give them a competitive edge and Intel couldn’t keep up… Intel got stuck, on their bigger NM chips and didn’t have a way of catching up to AMD… and when I checked the benchmarks… Intel still won…

The only part you and I can agree on is that I like that there is competition, if AMD wins based on price to performance great! That’s great for us as the consumers. But consistently, from what I have seen over the years. AMD has been known as the budget option, and yes now they are trying to be an apple to apple comparison (have been for quite some years now). Intel and Nvidia just happen to be the sweeter apples consistently. I hope that changes as it would lower prices for us and get us better gear. But it is what it is.

So bottom line. No, I do not hate AMD. I hope they kick intel or Nvidias butt one day and until then my comment about AMD tending to do things poorly (Maybe I could have chosen my words more carefully, tends to be the underdog) still stands based on price to performance. I do like that even if they are benched lower, they still like you said prevent Intel and Nvidia from being cozy/lazy as they are always inching up on those two companies. I wouldn’t want Nvidia being the only GPU company in town or Intel being the only cpu company in town. My loyalty is to the best price to performance during the year I am buying, not to Intel or Nvidia, they just happen to consistently be winners in that regard.

The 11th gen Intel I7 11700K(Q1 2021) still is faster than the AMD Ryzen 7 5800X3D (Q1 2022) cpu.

Lastest AMD cpu only beats out the 10th Gen i7 10700k, but the performance difference is so small and since the 10th gen is older its considerably cheaper and intel is still the better option.

Faster at WHAT exactly?

Passmark for example shows the 5800X (Q4 2020) as comparable single-threaded and 12% faster multi-threaded, the 5800X3D is 1.8% faster multi-threaded, 9% slower single-threaded. That’s ignoring the 5950X which is 46% faster multi-threaded. Its still 9.7% faster than a 12900k multi-threaded, though that is 17.7% faster single-threaded than the 5950X.

Now obviously I don’t take benchmark as gospel, it depends entirely on the workload these benchmarks are representing and how that compares to your real-world workload. But it shows how its not that simple, you can’t just blanket claim one CPU wins over another without context.

There’s also the fact those AMD CPUs work on ages old motherboards vs Intel which needs new motherboards. That is probably the most impressive feat, that they can push this performance on an old platform. (moving so much into CPU certainly paid off)

2 Likes

I am referring to userbenchmarks you can see the testing methods there:

As for AMD typically lasting longer on a motherboard, that is nice and all, but honestly, even AMD doesn’t last on every board. Last time I checked AMD lasted like 3-4 generations on a motherboard and intel 1-2 on a motherboard. Well, by that point, I imagine you would probably need to replace your motherboard anyway, unless you’re someone who has to upgrade more frequently than every 4-5 years. Which some people do and depending on the board, it maybe worth it for someone to choose AMD over intel. I don’t upgrade cpus that often. I dont have that need, I wait until their is a big enough performance gap between my old hardware before I consider upgrading to new hardware. So for me, intel is worth it.