*gets his popcorn*

ORIGINAL: AndyW

Interesting read: http://www.uaudio.com/webzine/2008/september/doctors.html

" So … the hard part was finding the right processor at the right price, one that could fulfill all the following needs.

• Run UA’s floating-point algorithms with good average instance multipliers
• Scalable DSP power at audio industry sales volumes (MFLOPS per $)
• Interface to PCIe and high-speed memory (to run Reverb/Delay/IR-type plug-ins)
• Quality development and debugging tools (fast plug-in development)
• Processor price availability and roadmap (future-proof expansion options)

There are many exotic chips with a great deal of processing power, but when you have to consider all the requirements above, you're quickly reduced to just a few practical options. Since first samples of the 21369 SHARC processor were within our development time window, we selected that as the target processor because it offered the greatest raw computational performance for our needs for the price."



This requirement is the first statement that makes the SHARC choice make business sense. Since audio sales volumes are much lower, there must be more profit to be derived per card to pay for the R&D, therefore a cheaper, less powerful DSP is needed.

I'm seeing the exact same results as the original poster with my UAD2 Duo. Once I get the latency up to approximately 17ms (in either ASIO or WDM) all the clicking/distortion/popping disappears entirely and the CPU usage drops drastically.

It seems like you just can't push the UAD2 to high DSP usage levels at low latencies right now. This is completely fine for mixing, but not the best if you want to track a live vocal directly into a mix with UAD2 plugins. I know I can bounce as a workaround, and I'll do so, but it would be nice to see that improved.

So basically, right now with the UAD2, the more plugins you run the higher latency you'll need. Still, when it comes down to mixing time, this thing has been a God send so far.

Im still getting clicks and pops when adding various UAD plugs in a project.
Neither the UAD meter or the CPU meter is maxed out.. But when I try and add an extra plaugin I get clicks and pops.
To remedy this I take my latency up to around 20ms and it seems to fix??
NOT happy with the new software!!!!

ORIGINAL: Funkybot

ORIGINAL: plainfaced

Im still getting clicks and pops when adding various UAD plugs in a project.
Neither the UAD meter or the CPU meter is maxed out.. But when I try and add an extra plaugin I get clicks and pops.
To remedy this I take my latency up to around 20ms and it seems to fix??
NOT happy with the new software!!!!

Perhaps it's time to contact UA tech support directly.

Sorry, by "still getting" I mean only with the newest drivers/software V5.. It was running tops on the earlier version..

Does anyone know if I can go back to 4.5 (or whatever it was)??

hahaahahah - buy the "top notch dsp powerhorse" with the best sounding plugs on the planet to be forced to set latrency at 20 ms ?????
hahahah i used to run my poco pci-e (which i sold cos it was a silly dongle for average plugs) at 6 or 3 ms with no problems on a PENT IUM 4 single with 3 ghz...

and still the uad-priestst praise their tin god....

as i posted earlier - we tested how many ssl compressors we could run on a 8core mac and arrived at 250 at 3ms we stoped cos it was getting boring...

you really gotta love uad to love uad... and what a great legacy btw. : the 40 bit sharc dsps have been used by behringger since stone age - muahahahahaha !!!!!

ORIGINAL: Creator

I think that many of us can agree that we would be just as likely (if not more so) to buy UAD plug-ins if they were native as we are now that they are DSP based. I agree that DSP cards end up being an extremely effective dongle with a lot of unfortunate side effects. But...

... the plug-ins make it worth it. There are native plug-ins of the same caliber but often not for the same colors. Many comparisons of the newer Waves JJP collection plug-ins to the existing UAD ones of the same gear found that Waves were anywhere from equivalent to inferior. I have yet to hear any reviews that called them superior, despite the higher price.

However, there are several excellent native plug-ins. The Abbey Road Plugins, for instance, offer vintage emulations that no other vendor does (save for very "neutral" stylistic emulation from the Liquid Mix that don't model gain stage, distortion, etc.).

Good points.

I was a skeptic before and still think the UAD1 is underpowered enough to be called a dongle, however, a simple test I did with the UAD Nigel plug was revealing to me.

http://forum.cakewalk.com/tm.asp?m=1499055#

http://forum.cakewalk.com/tm.asp?m=1499055#

I *don't* view the UAD2 QUAD as a dongle any more that I would view my video card as a dongle because it is equivalent to 3 cores of a 3.0Ghz native CPU power, at least for this one plug. I don't see anything else being different from that by more than a factor of 2 so I think it is probably fair to extrapolate that test to say that a UAD2 Quad is between 1.5 and 3 cores worth of Intel CPU native power at 3.0Ghz.

I decided to try a test with the 33609 compressor. It takes 67% of the UAD1 DSP. It takes 5% of the UAD2Q(22% of one DSP). So it appears that a single UAD2 DSP is about 3 times more powerful than the UAD1 with this compressor. With the UAD2Q viewed as a single DSP, it is 12 times more powerful than a single UAD1. This tracks close to the claimed 10x more powerful so lets conservatively use that number and use the Nigel data to extrapolate how a 33609 would work if run natively. That means if the 33609 was native AND we assume there is no inherent efficiency advantage to DSP over CPU(VERY conservative assumption since we know special purpose DSP is more efficient) AND we extrapolate from the Nigel data that the UAD1 is about 30% of one 3.0Ghz native core, it would take 20% of one core of a native 3.0Ghz quad core to run the 33609. This is about 5% overall native CPU, meaning that for this test, the UAD2 Quad comes out as *equivalent* to the native quad core. Ergo, each core on the Quad is equivalent to a core on the native CPU.

So the Nigel test said the UAD2Q was 75% of a native Q6600@3.0Ghz and the 33609 test says the UAD2Q was 100% of a native Q6600@3.0Ghz. Conservatively(since we are discounting DSP efficiency), the average power of a UAD2Quad has got to be somewhere in that range of 1.5-4x the power of a native quad core. That, to me, is way above "dongle" status. These two comparisons demostrate to me that the UAD2 *Quad* is actually a pretty powerful DSP. Not very impressed with the Solo or Duo.

Now, one could argue that cost is a factor here...$1500 for the UAD2 versus $250 for a quad core native processor that you could use for other stuff. True...you can also get a 4-door sedan from Hyundai or BMW with about the same price spread. If you really like how the BMW drives and you have the cash, you are going to get the BMW.