Pretty good info, but note that some CPU hungry plugins like core speed as well, especially when stacked in the same channel (master). In the latter case, a single track may overload a processing thread, which causes dropouts while all other cores may be idling.

Note that VEP is an extremely efficient host, so using plugins in VEP where possible certainly offloads CPU demand of most other hosts.

At least on Windows, low latencies are pretty easy to reach with current systems. We test our VSL systems with a 64 samples buffer at 44.1kHz. Especially the faster systems (hexacore 3.5GHz and dual Xeon systems) handle large/complex mockups.

Well, as being the guy who configured and tested this system I can confirm the workstation is an amazing system.

It is the fastest workstation CPU available in that generation, 8 x 3.1GHz, and it still is able to crunch most demanding projects.

The OS drive in the system contains a hidden partition with an OS backup without installed applications, so you can restore the system to our original Windows installation with audio tweaks if preferred. Unless this space was used to enlarge the drive capacity.

Having communicated a lot with OP, I also am convinced it indeed is in excellent state. So will be the Angelbird Master SSDs.

Cheers,

Leon

XI-MACHINES

If the system is not connected to a DHCP server device it might interfere if he hasn't got a static IP.
Also, an SSD and especially RAID will not fix his issues.

OP, do you use the VST3 version of a VEP audio input by any chance (and have it active in said project)?

Here a vote for Amsterdam as well. I studied there for 7 years and it is fantastic, but hard to avoid tourist traps. Try to stay in the "Jordaan" area.

Vierwaldstädtersee is nice as well.
But in the end, I prefer Thailand over anything else.

Hi Jeff,
this is weird behavior, I can only imagine that mixing 2 and 4GB modules might be the culprit here. Doe the same happen if you try it with 2 x 2 GB (identical modules) and 2 x 4 GB (identical modules)?
Also, can you use a buffer of 512 OR 1024 samples? 768 is pretty unusual (MIR wouldn't accept it), worth a try.

Well, don't let those figures fool you. This tool mainly measures the SSD cache which is flushed every 0.5-1 second. Also, it is more a theoretical value, real world use is different.



RAID0 for libraries is ok if you have a backup, but I would not use it for anything else, especially not for writing recordings or projects where data loss easily becomes a small desaster.


RAID 1 is the most elegant way, but many SSD/controller combinations are pretty tricky with RAID 1. RAID 5 is a fast and safe solution, but also requires some homework to make sure your configuration is reliable.

BTW here is a test result of 4 x Angelbird 240GB in RAID 0. Note that this tool is tricked by the cache of the Samsung. The lower write speed is due to the Sandforce controller, which handles uncompressable data not as fast as compressable data. For libraries a non-topic. So please take the screenshots with the Blackmagic tool with a huge bucket of salt...

Installing the software also installs the drivers, some of them being pretty bad (Intensity series). Also, the tool is more of a gadget IMO, other tools show more consequent results.

I have attached a screenshot of 4 Samsung 840 Pro SSDs in RAID 0 on a single Hexacore system. They are only 256 GB instead of the faster (more NANDs) 512 GB models in your article. Note the weird negative results of the software, which seems to be the maximum the software will calculate. We don't have Angelbirds in stock at the moment to do such a test. From my memory, they have similar test results, with the big difference that Angelbird does cherry picking of all components and the Master series having an onboard UPS for added reliability. The Samsungs have a great price/performance ratio though.


In my opinion, RAID0 is a bit risky in a professional environment. If one drive fails, everything is lost. For liraries not a big drama, but you will have longer downtime to get the library up and running again. VSL claims separate drives is faster, I have no reasons to doubt that since they have tested it.

The discussion around these benchmarks have evolved to a point that there has not been a significant change in either OS performance to claim a major difference from these results.


Cubase indeed performs worse on Mac, so does every other cross-platform DAW software. That is some kind of indication if you ask me.


Logic has a fantastic hybrid engine which uses a high playback buffer as long as no tracks are record enabled and actually performing, which means that on playback, you get an amazing performance since even though the buffer is set at 32samples, it actually uses a buffer of 1024 samples - a value I haven't used in 6 years. A great workflow improvement, but this also confirms my point on why benchmarking is so hard to do.


Feel free to send me an email with some files, screenshots and buffer settings so I can show you what various of our systems perform, but note that this is time consuming and not very high on our priority list, so don't expect a quick reply.


Note that Í still don't see this as a competition. Many people love Mac/OSX and will never change, which is fine. But please respect that fact that many people (especially professionals) are not interested in a competition regarding voice count or MB/s of their SSDs. Either a system does the job, or it doesn't.

Thank you Miklos, for your review. The team is really happy it has exceeded your expectations. Thank you as well, Roy!


But why anybody would be "afraid to give feedback" to a file might have to do with the fact that it is not a competition.

Most users here are to make music and make a living with that. I know Miklos has tons of other things to do, and as long
as our workstation is doing its job, he has no need to waste his time with benchmarking.



Regarding the benchmarking: other than synthetic tests, it is hard and time intensive to do honest comparable benchmarking, especially between different operating systems.

There are lots of parameters to set, the host has to be the same with the same settings, and the plugins of course as
well.



Probably the most fair benchmark would be using ProTools with VEPro on the same system, since it is the only cross-

platform DAW which shows similar performance on Mac and PC, but only if the buffer is not lower than 256 samples, which is
too high for most composers in the writing stage.





Regarding the fastest system: the dual E5-2687W is currently the fastest "normal" system which is available for composers,

and frankly, it is way ahead of the dual X5670 which is in the MacPro 12 core 2010 (or even the X5690 for that matter) because
those chips are 2 generations newer - and Intel made a huge performance jump with it.


Also, you might have heard of www.dawbench.com, which shows that OSX does usually not perform better. In fact, on the exact same MacPro, every cross-platform application shows (much) better low latency results on Windows than on Mac. The only situation where an application runs better on Mac is Pro Tools, as long as the buffer is set to 256 samples or higher.



But that is not the point. OSX and Apple are very popular, a lot of things are working really well, lots of audio people are enjoying it for lots of years. Because of this I am not trying to push people into one direction. However, when it comes to low latency performance, Apple is far behind at this moment.