Coming Soon: The Relab 176 (Retro 175) Tube Compressor/ Limiter

[onerob]

Given that Relab provides a 4x OS option (default off), I don't see why having a multi-core option that adds a little latency (default off) would hurt - other plugins incur latency when enabling oversampling, for example. It just takes time and effort to code. It's clear that the path of least resistance for the business was to not code such a thing, so they made choices accordingly.

It's one thing for a synth to offer multi-core processing as it can distribute different voices to different cores, which makes sense. Effects processing, on the other hand, favours single core operation. I'm not aware of any multi-core effects plugins, but they may be out there.

I doubt Relab are being lazy or seeking to discriminate against Intel/AMD by not providing a multi core option.

This article is helpful: https://www.opus101proaudio.com/how-to- ... roduction/

[vitocorleone123]

Given that Relab provides a 4x OS option (default off), I don't see why having a multi-core option that adds a little latency (default off) would hurt - other plugins incur latency when enabling oversampling, for example. It just takes time and effort to code. It's clear that the path of least resistance for the business was to not code such a thing, so they made choices accordingly.

It's one thing for a synth to offer multi-core processing as it can distribute different voices to different cores, which makes sense. Effects processing, on the other hand, favours single core operation. I'm not aware of any multi-core effects plugins, but they may be out there.

I doubt Relab are being lazy or seeking to discriminate against Intel/AMD by not providing a multi core option.

This article is helpful: https://www.opus101proaudio.com/how-to- ... roduction/

Thanks.

CLAP offers better multithreading. The DAW and OS are more responsible and need the updating more than fx plugins.

That said, again, it's entirely possible to add a small latency hit and suddenly have vastly more performance with no audio penalty using Audiogridder. Curious.

There's still something odd, or more likely, something I've not learned, about how Apple chips are 3x - 5x better at single-threaded (as measured by this plugin) than the best AMD and Intel chips when, on paper and in general benchmarks, there's not a huge % difference - for example, in Geekbench, the M3 Ultra scores ~10% higher than my 9950X. Is it the chip? Is it the DAW and OS? It's interesting.

At any rate, I'm not getting this plugin, as the small but real audio difference I heard is too far outweighed by the CPU hit at this time. Again, I do quite like the way this plugin sounds.

The Intel/AMD market in audio can't be less than 20%, can it? Just thinking in terms of addressable market, I'd think Intel/AMD make 30+%, so everyone using one of those chips will be less likely to purchase this plugin and any others coded similarly unless and until either there's a better coding solution or faster CPUs and better DAWs, or all of the above.

[VitaminD]

vito, you've stated a lot of thoughts I've had too in that last post of yours.

I'm rather flummoxed at how the Apple machines seem to perform significantly better in practice even though the differences in CPU performance is extremely minor in benchmarks. I think it just must be partially down to the Apple OS doing a better job with scheduling / threading than Window does, with respect to audio. But also I think the DAW has some play in it as well...

Some users claim they're able to get tens of instances of this Relab compressor on their M2 when I can barely get a single one going on my 5900x using Windows 10 in an actual production. I can get around 8 on audio tracks with nothing else working. But that's not a normal means of workflow for me..

It got me thinking enough that I took a bedroom Mini PC that has windows 11 installed and then installed Cubase and some plugins on it today just to test if the system scheduler performs any better. My result was mostly inconclusive though. I think I see more activity per thread in task manager in w11 over my w10 machine. But the cpu in the w11 system is a measly 5700u so its low watt. It does seem to have higher utilization in w11 at least..

In any case the net result was the same, one instance of R176 installed and the playback halts.

But this plugin got me even thinking about a computer upgrade, despite otherwise not needing one, and looking into future CPUs. I don't think the next big AMD cpus are being released until late next year or possibly early 2027. Supposedly big per core performance gains and a large clockspeed (6.4Ghz possibly). So maybe then would be the right time to use these hyper demanding plugins. I think I won't make the same mistake though of getting the biggest core CPU I can buy and will instead opt for lower cores/threads but higher IPC per core. Windows and Cubase just don't seem to do a great job at it, than I think they could. Esp considering how much better this plugin performs in the same project if I wrap it in Audiogridder.. It's curious.

[fceramic]

I'm on a M1 Pro, and the Relay 176 uses about 20%, which is perfectly reasonable.

[Ploki]

I'm rather flummoxed at how the Apple machines seem to perform significantly better in practice even though the differences in CPU performance is extremely minor in benchmarks. I think it just must be partially down to the Apple OS doing a better job with scheduling / threading than Window does, with respect to audio. But also I think the DAW has some play in it as well...

Bigger cache

Much MUCH better memory latency when cache is full

[dermage]

I'm rather flummoxed at how the Apple machines seem to perform significantly better in practice even though the differences in CPU performance is extremely minor in benchmarks. I think it just must be partially down to the Apple OS doing a better job with scheduling / threading than Window does, with respect to audio. But also I think the DAW has some play in it as well...

Bigger cache

Much MUCH better memory latency when cache is full

Amd x3d processor should match that performance, at least cache. But what does a plug-in need memory for? It's usually just computation performance. Possibly context switching is better on Arm.

[kraster]

Also Apple, for obvious reasons, have invested a lot in making their chips run more efficiently so you get less thermal throttling.

They generally have higher IPS (firestorm and successors), unified memory architecture, large reorder buffers, advanced branch predicition and arguably most importantly unified stack. ie the vertical integration of hardware and software.

PC systems have to contend with a much broader array of hardware and their drivers etc. Apple control everything so they tend to have less abstraction layers. The OS and chipset are much more like one unified system.

Take core audio as an example. This is built in to Mac OS whereas ASIO is not native to windows and effectively bypasses the Windows audio stack. Core Audio on the other hand is fully intergrated into the system.

[Ploki]

I'm rather flummoxed at how the Apple machines seem to perform significantly better in practice even though the differences in CPU performance is extremely minor in benchmarks. I think it just must be partially down to the Apple OS doing a better job with scheduling / threading than Window does, with respect to audio. But also I think the DAW has some play in it as well...

Bigger cache

Much MUCH better memory latency when cache is full

Amd x3d processor should match that performance, at least cache. But what does a plug-in need memory for? It's usually just computation performance. Possibly context switching is better on Arm.

Everything that goes to CPU goes through cache, if cache is full, it goes through ram, if ram is full, it swaps to drive and goes through drive.

[dermage]

I'm rather flummoxed at how the Apple machines seem to perform significantly better in practice even though the differences in CPU performance is extremely minor in benchmarks. I think it just must be partially down to the Apple OS doing a better job with scheduling / threading than Window does, with respect to audio. But also I think the DAW has some play in it as well...

Bigger cache

Much MUCH better memory latency when cache is full

Amd x3d processor should match that performance, at least cache. But what does a plug-in need memory for? It's usually just computation performance. Possibly context switching is better on Arm.

Everything that goes to CPU goes through cache, if cache is full, it goes through ram, if ram is full, it swaps to drive and goes through drive.

That's quite simplified I'm afraid. I've studied that field, and not everything goes through cache.
You have to differentiate data from actual processing commands in the CPU.
There are registers for processing and the cache stands between memory and the CPU commands processing it. But if registers can be used, no cache or memory is used.
I doubt that most plugins use much memory actually, because they process numbers usually which fit well into cache lines (there's more than one cache of a CPU) or even registers and does not need much interaction between memory and cache.
Memory is important for sample based players and big libraries, not usually simple number solvers like this Relab plugin. If it was based on volterra kernels (like Acustica stuff) or convolution IRs, then maybe memory may play a role, but not that I know of, it looks like an algorithmic implementation which heavily relies on pure CPU processing power where IPC (instructions per clock) are important and possible parallel processing using SSE or AVX instructions may be used (although these mostly are used in synths with multi-voice filter processing etc.)

more info if interested:
https://www.geeksforgeeks.org/computer- ... -register/

[vitocorleone123]

The new benchmark includes a Mac Mini (they'll work on getting some higher-end ones in). And, while it gets crushed compared to the top AMD and Intel in multithreaded kinds of things, when it comes to the single-threaded test, even the lowly M4 chip dominates the next best, being the 285k.

I'm wondering if I made the right choice getting a 9950X when it came out (I needed a new computer), seeing as I mostly use it for audio stuff these days. I was going to get a 285k but Intel kept delaying and delaying and I had to send the motherboard etc. back that I'd purchased in anticipation before I couldn't. Ah well. I'll upgrade the CPU with the next gen and it'll be ok.

This would help explain most of the results happening with the Relab 176.

https://www.scanproaudio.info/2025/08/0 ... 5-edition/

[VitaminD]

Wow! I wonder how the M2 compares on the DAWbench type charts too?

Their write up is something I wanted to highlight, "...if the OS is managing the CPU load correctly..." this is a bit out of context from what they meant, but I do kind of wonder if Windows just stinks at this sort of thing? The audiogridder workaround seems to indicate there is more in the tank on Windows PCs, it's just it isn't using another thread/core when it could be..

[indigo76]

Wanted to see what the hype is all about and requested an invitation for the trial. Barely passed minimum specs -- I have a first-gen Macbook Air M1.

Put it on a Piano track (normally I do more ambient stuff, but I also occasionally dabble in neo-classical).

Jaw dropped. Okay, so that's what hardware sounds like.

No I'm seriously consider getting it, though I'm strapped for cash. Decisions, decisions...

[dermage]

Wanted to see what the hype is all about and requested an invitation for the trial. Barely passed minimum specs -- I have a first-gen Macbook Air M1.

Put it on a Piano track (normally I do more ambient stuff, but I also occasionally dabble in neo-classical).

Jaw dropped. Okay, so that's what hardware sounds like.

No I'm seriously consider getting it, though I'm strapped for cash. Decisions, decisions...

UAD also has a 176, and maybe you haven't tried DMG TrackComp... just saving you some money

[TMaudio]

Wanted to see what the hype is all about and requested an invitation for the trial. Barely passed minimum specs -- I have a first-gen Macbook Air M1.

Put it on a Piano track (normally I do more ambient stuff, but I also occasionally dabble in neo-classical).

Jaw dropped. Okay, so that's what hardware sounds like.

No I'm seriously consider getting it, though I'm strapped for cash. Decisions, decisions...

UAD also has a 176, and maybe you haven't tried DMG TrackComp... just saving you some money

The AB tests posted on gearslutz done on abx.funkybits.fr with relab 176, uad 176, mjuc, and dyna-mu shows people do not favor the relab one.

[indigo76]

Ha! Thanks guys for pouring cold water on my GAS, I think I needed it.

Haven't tested the UAD one yet (Might be worth a try?) but I have the Klanghelm. In direct comparison between the MJUC and the Relab I do notice a difference. The question is if this is worth 150 € to me or not.

But I can definitely say that Vari-Mus -- once you dial in the settings right -- sound fabulous on piano tracks, especially slower ones where the Piano carries the arrangement.