Dell Readies New XPS 15: 4K, Quad-Core Kaby Lake, GeForce GTX 1050

102 Comments

View All Comments

• ddriver - Monday, December 26, 2016 - link

  Actually a small correction, scarlett 2i2 doesn't have the midi io and the digital in. So if we compare it to say the 6i6, which IMO has the edge in terms of functionality (you can't use the clarett extra digital in without additional hardware) it is still half the price.
  
  The 18i20 is 500$, so the highes end USB interface is the same price as the entry level TB interface. Not a good deal that TB...

• Nexing - Tuesday, December 27, 2016 - link

  @ddriver
  Regarding your afirmation; "interface makers have already been able to implement a walk-around for USB audio latency by using proprietary protocols supported at driver levels".
  Until a bare couple of years ago, the only DAC manufacturer with a successful USB implementation was RME, still with as low as 5 to 6ms all round latency with very streamlined hardware/software configurations.
  Technicaly at 2016, we are witnessing DACs pivotal evolution in terms of latency; yes from 5 to 10 or more, going now into 2 to 3ms together with a very noticeable improvement of sound to noise ratios, usually from 95 dBs (on DACs like the ones you mention above) right up to 110 to 120 dBs in nowadays DACs like marvelous Focusrite Clarett line, costing from $500 to $799 for specs that only last year was attainable for no less than $1300 (with luck).
  
  That said USB Pro Audio related problems are inherent to the system processing interrupts proper of the low level access to memory that USB has. Even the phased out Firewire connector has DMA (Direct Memory Access), same as Thunderbolt, something that USB doesn't.
  Even an old connector as RJ-45 (via Dante) is way up better than USB in terms of latency, though DACs implementing are still very expensive.
  For me a mobile/center piece like DELL XPS 15, with (hopefully) two TB3s would be a deciding piece of gear to invest on.

• ddriver - Tuesday, December 27, 2016 - link

  The lack of DMA is not all that detrimental because audio is fairly low bandwidth. This is evident from the fact that firewire interfaces show to tangible latency improvements to usb. Finally, usb 3.1 does have DMA and it is highly likely that we will see audio interfaces utilizing that in 2017.
  
  "RME only" is a rather snobbish attitude of a small subset of people who are into audio production. There has been a plethora of very high quality USB audio interface for many years. While RME was definitely top notch in terms of quality, when it comes to price performance ratio it was pretty much at the bottom due to inflated prices.
  
  DAC latency is unavoidable, and there isn't really any room for improvement, because of how DACs work, that is converting continuous to discrete simply has to do some buffering, therefore latency is unavoidable, and also fairly static.
  
  The DAC is not a pivoting point, as modern interfaces have numerous DACs, it is the microcontroller that receives the audio and interfaces it to the computer that matter. And the driver on the computer side. Using the default USB audio specification alone adds like 4-5 msec of latency, which dwarfs the 1 msec polling latency of USB. By working around that through implementing a proprietary driver to replace the latent default audio implementation you can put usb interfaces back on the table.
  
  I have a live setup around a second gen scarlett 18i8 with live DSP on like 20 channels, and I don't mean a plugin or two total, but a plugin or two on each of those 20 channels, and I get rock-solid operation at 10 msec total latency, which is good enough, below the threshold of perception, and identical to listening to your guitar amp from 3 meters. And this is a live, concert setup. For studio work and offline processing latency is even less of an issue.
  
  So I am not overly keen on spending 100% price premium on TB on account of its hype. Plus very few PCs come with TB, especially workstation mobos, because lets be real, nobody uses ultrabooks for serious audio production, what people use is high performance rackmount workstations.

• Nexing - Wednesday, December 28, 2016 - link

  @ddriver,
  Unlike you, I need computers for live performance, something hundreds of thousands of people do most everyday around the planet. In these cases latency matters. Add then it is not just the musicans, the FOH technicians plus the lights and video projection people, so at least three teams per event will need such powerful and portable "ultrabook". And we don't want/risk carrying Precisions around, but light, upgradeable, well constructed, thin bezel 15 XPS are almost perfect.
  About latency perception; 5ms is common for percussionist and something not far from this number gets troubling for DJs, who compare the attack of beats. The 10 ms figure you cite is assumed for live musicians (using monitoring systems and for their ensemble work).
  Again, you may utilize USB (2.0 or 3.0) with very streamlined hardware/software sets -as your scarlett based one-, but 2016's live software requires the usage of plugins for the incredible sound effects common nowadays... and there is where the problem starts!! You need powerful CPUs when you are running 6 to 10 simultaneous Ableton live tracks each with several demanding plugins. And contrary to the usual numbers/arguments the hardware is barely there. For such scenarios 2017 DELL XPS with Thunderbolt 3, weighting less than 2 kilos and with a 15" screen theoretically will achieve smooth operation, audio glitch free performances up to a point, because -as a musician/performer- you want to add more effects live (try Zynaptic Pitchmap) and you reach a point when SUCH setup chokes very soon, and I am not even considering any Spectral audio software tool, because gear is still many years far from ready for live performance with that whole category of software.
  As you accept, musicians at studio work still have to wait seconds and even minutes that top notch computers require to complete normal audio tasks, this is a fact not many people at these tech sites seem to know.

• Nexing - Wednesday, December 28, 2016 - link

  @ddriver
  Plus you have to state that most probably you had to disable your wifi, bluetooth, many windows services and what not? limiting configuration in order to achieve "second gen scarlett 18i8 with live DSP on like 20 channels,(...) rock-solid operation at 10 msec total latency.".
  Then you install most any common software for video, printer... a windows update and your "live set up" starts to produce glitches or worst. Thunderbolt is a way to spare that out from the live equation, and also from a problem free general usage. Most people is far from the care you had to apply to you system, for sure, and want to use the tools and effects at hand. Something that in the last decades, I have been helping/servicing people systems applying from minor to extreme measures in order to get to a functional point for Audio set ups, live or in studio. At 2017 this is finally ending. Musicians won't need to specialize in configuring a PC, These are good news. And the mature XPS line seems for such recommendation.

• ddriver - Wednesday, December 28, 2016 - link

  Nope, there is no need to do that. Well, all those things are disabled, because that particular machine serves one specific purpose only, it sits in a rack and it doesn't even have a display connected, so everything that is not essential to its job is disabled.
  
  But I also have a smaller scarlett on my desktop workstation and latency is just as good with wifi and bluetooth and all services and whatnot enabled, and it is a system with a lot of hardware on it. It is only bad drivers that cause glitches, and such drivers will cause glitches with TB as well, just like they cause glitches with internal PCIe audio interfaces. Performance is rock solid, even considering some extreme cases, when I'd be tracking audio to video in real time, with the video processing happening at the same time, and video processing is immensely more taxing than audio processing. Still rock solid 10 msec round-trip latency, it could possibly go even lower than that, but there is no need to, I prefer to be a little extra generous with the buffering which is also softer on the CPU.
  
  And 10 msec is pretty darn good for any case, even if you are say a technical death metal drummer who blasts like crazy at 250 BPM. I've produced a few of those, no complains on latency. What really kills it is the flanging effect you get from phase fighting, which is evident even at as low as 1 msec latency, but the remedy for that is simple - just cut out the dry audio, use isolation headphones or whatever. You should not trust too much what arbitrary people say, I've heard people who say that vocals are even more sensitive to latency than drums, which is laughable, considering that voice doesn't have nowhere near the attack of percussion, and if you put those snobs to a blind test you'd realize they will epically fail to tell 5 msec latency from 15, much like every form of audiophile out there.
  
  As for people who buy cheap gear and expect it will work for everything - that's their own problem. Whether you or someone else - things always have to be tested. Do you know that the XPS will have good drivers? Do you know what its DPC latency will be in advance of any retail units? I've seen people buy crazy expensive workstation laptops, hook up an audio interface and get crazy dropouts. Because an "audio workstation" is not just about any workstation, there are specific requirements. I personally don't think it is in people's interest if they know nothing about nothing and get things just handed to them. The less you know the less you can do. And it is not like it is saving your mental capacity to use for other things, people are just lazy dummies who hate learning new things to their own detriment.

• ddriver - Wednesday, December 28, 2016 - link

  Nice try, however, nobody in the world uses software plugins for high end setups. They don't use PC audio interfaces either, they use huge ass digital mixing consoles which have everything done in hardware and has pretty much zero latency. Their only latency is the ADCs and the DACs, resulting in about 1 msec of latency total, including all the processing. Concerts use truckloads of gear, entire racks with dedicated analog and digital sound processors, which are used as channel inserts. Nobody cares about lightweight ultrabooks whose performance is garbage compared to a decent workstation.
  
  Nobody who matters in the live sound industry uses computers or plugins or any of that nonsense. Also, live sound is about something different than quality entirely, even the world's best live sound guys' setups sound like crap compared to a modestly produced studio recording, because live sound is shit, it is not about sound quality, it is about the show, at those arbitrary venues with acoustics that would vary throughout the show, at those levels of power "quality" is the least concern. What people care the most is make instruments as audible as possible and make everything as loud as possible without producing feedback loops. A lot of the big ass concert grade digital mixing consoles are still 16bit/48khz because that's just cheaper and easier and nobody on the show will ever be able to tell the difference between that and 24/192 audio.

• drajitshnew - Monday, December 26, 2016 - link

  I think almost any eGPU implementation will be at least 1070-- anything else does not make sense.
  Further, as I have already said a lack of PCIe lanes is a DESIGN DECISION rather than a CPU restriction. Probably to protect alien ware

• qap - Saturday, December 24, 2016 - link

  Difference between x4 and x16 of PCIe 2.0 is NOT negligible. For GTX 1080 it is around 15% on average in FullHD (i wount be giving link for competing site, but you can find the test). That would be acceptable. Unfortunately it is extremely variable between games and for example in Far Cry, Just Cause or Warhammer it is around 30%. And most of the difference is between x4 and x8.

• Rampart - Sunday, December 25, 2016 - link

  The main problem is less about the hit to the GPU and more about any kind of CPU bottleneck from a mobile chip.