NEW: Waves Abbey Road Chambers

[Mathematics]

On my old Intel i5 3450, Chambers takes over 30% CPU. That's more than U-He Diva, which is known for being a power hungry beast.

I guess it's time to upgrade my PC

That percentage seems a bit more like an average of what I'd expect from an i5 or even an earlier i7. That is, it should take up some CPU. You don't need to upgrade because you probably wouldn't run too many AR plugins together in one project.

I'm actually curious what cprompt is running. I was able to get just under 10% but I have to run a buffer size of 2048. For running under 10%, you need a whole lot of clock cycles. I'm talking up in the order of close to 5GHz, or pretty close. I'm not in the field of software engineering but I recall that the floating point math being processed for Waves plugins can only be done by a single thread. This, I believe is true for all plugins unless the developer, like UAD, designs a series or chain of plugins to utilize a multi-thread system, then more cores could be helpful. This is Waves...I don't think so but I'm not sure.

Any software engineers in the house that can confirm this? I'd actually like to know for myself. I have a feeling I'm not too far off though, but I would like someone in the field to check this over and validate this.

Just for fun, I do recall to better emulate analog gear, like ARC plugins, Waves would need require people to run something like 24-48 core computers. That is a whole other level of production.

[Scotty]

No formal test but as far as asio load goes it looks like Chambers uses between 30% to 80 % more ASIO bandwidth as compared most of my other reverb plugins. This is on an i7 hexicore 3930k overclocked to 4.2 ghz. The proportions seem consistent across my audio card buffer settings.

At 32 sample buffer setting with RME Raydat Chambers is using about 25% of the available ASIO when pushing audio through it...

about 20% @ 512
about 15% @ 2048

[J4R1O]

On my old Intel i5 3450, Chambers takes over 30% CPU. That's more than U-He Diva, which is known for being a power hungry beast.

I guess it's time to upgrade my PC

Intel Core i5-2320 @ 3.00GHz here... with UAD-2 Octo and 16 GB Ram, no big issues with power hungry VST instruments (ok, some issues with Diva when using its Divine mode ) or plugins until last nite when I replaced Cubase stock reverb with AR Chambers on full mix; an instant snap, crackle and pop.

Edit: btw, my 666th post here.

[Funkybot's Evil Twin]

Used it last night on a singer-songwriter 2 acoustic guitar, 1 vocal mix that had a very old school vibe. I started with the wide vocal preset, and things quickly fell into place from there. I was very happy with what I was hearing. I then threw it up on a piano, and with some work, found settings I was happy with. This may end up being a niche plugin, but we need more stuff like this that's off the more beaten path, so nice job Waves!

Using Windows 10, VST3 x64 in Studio One 3.5.6, CPU usage is about 15% of one core of my i7 4790k and the plugin reports 2.3ms of latency to the host. So I could (in theory) run about 6 per core times 8 cores, so over 50 per project (with nothing else). That's definitely on the high side for a single effect plugin, especially a reverb, but doesn't approach the CPU usage of something like Diva. If your CPU usage is that bad, I'd report it because it may be a bug.

[perpetual3]

Just in case one installs it and it doesn't show up. In Windows, go to:

C:\Program Files (x86)\Waves\WaveShells V10

and the respective location for Mac and you'll have a new WaveShell2-VST 10.0_x64.dll file. If you're using custom folders, just copy that new *.dll.

Sounds awesome. The STEED, for me, is the best part of this plugin. I like the sound of this reverb. It's definitely got a different flavor compared to AR Plates, and that one sounds good too. Here's an interesting setting I ended up with after tinkering around. Sounds like a megaphone with slight distortion in a hall. It likes its clock cycles, that's for sure.

Another lo-fi winner?

[GearNostalgia]

On my old Intel i5 3450, Chambers takes over 30% CPU. That's more than U-He Diva, which is known for being a power hungry beast.

I guess it's time to upgrade my PC

Intel Core i5-2320 @ 3.00GHz here... with UAD-2 Octo and 16 GB Ram, no big issues with power hungry VST instruments (ok, some issues with Diva when using its Divine mode ) or plugins until last nite when I replaced Cubase stock reverb with AR Chambers on full mix; an instant snap, crackle and pop.

Edit: btw, my 666th post here.

Ok, does UAD help with all kinds of VST? I thought it only worked for the custom coded that they list on their site.

The number of the beast - Iron Maiden fan?

[Mathematics]

@perpetual3, so far...that was actually a first thought of mine. In fact, I just used it for such a genre this morning. That's how i accidentally discovered the plugin functions like a de-Harsher.

[Mathematics]

Used it last night on a singer-songwriter 2 acoustic guitar, 1 vocal mix that had a very old school vibe. I started with the wide vocal preset, and things quickly fell into place from there. I was very happy with what I was hearing. I then threw it up on a piano, and with some work, found settings I was happy with. This may end up being a niche plugin, but we need more stuff like this that's off the more beaten path, so nice job Waves!

Using Windows 10, VST3 x64 in Studio One 3.5.6, CPU usage is about 15% of one core of my i7 4790k and the plugin reports 2.3ms of latency to the host. So I could (in theory) run about 6 per core times 8 cores, so over 50 per project (with nothing else). That's definitely on the high side for a single effect plugin, especially a reverb, but doesn't approach the CPU usage of something like Diva. If your CPU usage is that bad, I'd report it because it may be a bug.

I love it too. BTW, I wouldn't be surprised to see a CPU efficiency update sometime soon. To be honest...I'm taking a stab and guessing what's mostly causing the CPU consumption is the internal oversampling. If that is, I don't mind but maybe they can put a variable switch for this particular plugin and perhaps consider implementing it in the ARC. I don't think I've seen any Waves plugin with that feature, here's to wishful thinking...

[J4R1O]

Ok, does UAD help with all kinds of VST? I thought it only worked for the custom coded that they list on their site.

No, but at least I can use UAD's power hungry plugins instead of other developers power hungry native ones.

The number of the beast - Iron Maiden fan?

Not a huge fan but Powerslave and Somewhere in Time are great heavy albums.

[jens]

I'm actually curious what cprompt is running. I was able to get just under 10% but I have to run a buffer size of 2048. For running under 10%, you need a whole lot of clock cycles.

You should also wonder what host he is running, as some report % of the total system (e.g. or rather at least Reaper) while most others report per core. That of course will cause a tremendous difference. So without mentioning the host (or at least whether it's per system or per core) these numbers are pretty much meaningless, as you will have no trouble understanding (especially with a user-name such as yours ).

I assume cprompt is using Reaper, so multiply his numbers by a factor of four (I think - or is it perhaps even eight due to multi-threading?) and they should be comparable to your own measurements.

[jens]

I then threw up on a piano

*ouch*

that sounds like one expensive hangover...

[Mathematics]

I'm actually curious what cprompt is running. I was able to get just under 10% but I have to run a buffer size of 2048. For running under 10%, you need a whole lot of clock cycles.

You should also wonder what host he is running, as some report % of the total system (e.g. or rather at least Reaper) while most others report per core. That of course will cause a tremendous difference. So without mentioning the host (or at least whether it's per system or per core) these numbers are pretty much meaningless, as you will have no trouble understanding (especially with a user-name such as yours ).

I assume cprompt is using Reaper, so multiply his numbers by a factor of four (I think - or is it perhaps even eight due to multi-threading?) and they should be comparable to your own measurements.

Ah. Thank you. I didn't realize Reaper's efficiency was so good. Thinking about it, you're probably also right on that he could have reported total CPU usage. I get numbers like his checking Task Manager. Thanks again for bringing that up

[mixtur.se]

That percentage seems a bit more like an average of what I'd expect from an i5 or even an earlier i7. That is, it should take up some CPU. You don't need to upgrade because you probably wouldn't run too many AR plugins together in one project.

I'm actually curious what cprompt is running. I was able to get just under 10% but I have to run a buffer size of 2048. For running under 10%, you need a whole lot of clock cycles. I'm talking up in the order of close to 5GHz, or pretty close. I'm not in the field of software engineering but I recall that the floating point math being processed for Waves plugins can only be done by a single thread. This, I believe is true for all plugins unless the developer, like UAD, designs a series or chain of plugins to utilize a multi-thread system, then more cores could be helpful. This is Waves...I don't think so but I'm not sure.

Any software engineers in the house that can confirm this? I'd actually like to know for myself. I have a feeling I'm not too far off though, but I would like someone in the field to check this over and validate this.

Just for fun, I do recall to better emulate analog gear, like ARC plugins, Waves would need require people to run something like 24-48 core computers. That is a whole other level of production.

No ned to be a developer, you can test it yourself. The majority of plugins use one thread, and som DAW:s even have one thread per channel. Ex live uses one thread for a whole audio path https://help.ableton.com/hc/en-us/artic ... ndling-FAQ

The built in performance meter n the DAW isn’t really giving you the full story. Check at the OS level

[Scotty]

In Cubase it is all about the ASIO load. It doesn't matter how much CPU headroom you have if the ASIO is peaking and you are getting audio breakup you have to increase your buffer size or reduce the load somehow.

For the reports to be meaningful we should report what audio card we are using and at what buffers - if the reports also show a relative comparison to another reverb plugin then at least we can extrapolate from there. . Just as some hosts are more efficient (hello Reaper) some audio card drivers are as well (hello RME).

[Mathematics]

Interesting that you referenced an article from the developer. Lol.

Thanks for the article. The issue I was trying to resolve was CPU consumption with this plugin. As jens pointed out, the DAW will influence performance and cprompt was probably reporting overall CPU performance.

If it's overall CPU performance, my CPU take a 3.9% hit with this plugin while running a buffer size of 256 with an MOTU PCIe-424 audio card. This is a 7th Gen i7 at 4.x GHz.

The answer is simple, just budget for an instance of this plugin with whatever you have. Any desktop based i5 should have no problem running it.

I believe cprompt was reading overall cpu performance because 5.2% doesn't seem accurate. I'm not saying it's not possible but considering this isn't quantum computing, I think it's safe to assume the we're all hitting between 15-30% CPU usage. I think the whole reason why we're caught up on this is because we all want to make sure we're not seeing a bug. Again...we would have to ask the developer.

By the way, the article explains how Live 9 and 10 utilize multiple cores and threads to balance CPU usage as the number of tracks increases. So, what was interesting in the article is that Live is designed to use a single thread for an entire signal chain. So, in fact, the article was helpful because it proved my whole argument about needing clock cycles.

Essentially, you would need something over 4Ghz if you wanted to run multiple instances of ARC plugins. Furthermore, the way this plugin is being processed and according to the developers notes you provided, this would mean that this plugin, in an Ableton Live session will hit the CPU with a big hit every single time you instantiated it as an insert or in parallel processing. So, in this case, it's not the cores that will help, it's clock cycles.

In other words, if you want to eat faster, you can use 4-16 hands to shovel food down to your stomach but you can't ask a second person to help you eat the food. Similarly, you can go fast with a car but attaching a second car doesn't increase your speed.

With a modest setup consisting of a USB 2.0 interface, i5@3.5Ghz, 8-16GB DDR3 RAM, Live (64-bit DAW) running with a buffer size of 512 samples...you can expect the first instance would hit the available CPU allocated to your DAW, let's say on a good day, by 20%. The second would hit you an additional ~20% and so on. So, if most other advanced DAWs operate similarly, you should only be able to run about 5-6 instances on this before you max out your CPU.

I know you mentioned one doesn't need to be a developer, but referencing developers notes was indeed helpful.

Scotty, you're right too. Another big factor indeed, is the ASIO driver, buffer size, and hardware. So, if people are going to post performance stats, please state:

DAW, CPU architecture and Speed, Interface, buffer size, and CPU performance by your DAW (not overall performance).

At this point, I think we have what we need and then some. If you're experiencing more than 15% CPU hit as reported by your DAW, welcome to the club.

If something isn't completely right...sue me.