Dell XPS 15 2016 (9550) InfinityEdge Notebook Review

Our review unit is equipped with the matte FHD version of the IPS panel. A resolution of 1920x1080 pixels results in 141 PPI. If you are not satisfied with that, you can also opt for the 4K Ultra HD display with 3840x2160 pixels (282.4 PPI, review coming soon). While the Full HD display has a matte surface, the 4K display comes with a glossy coating. The latter also supports touch inputs, which is not the case for the FHD version. Our FHD configuration is exclusively available in combination with the i5-6300HQ and 8 GB RAM (price: 1479 Euros, ~$1608). This means you cannot combine a matte panel and the Core i7 / 16 GB, which will annoy some customers.

The luminance of up to 400 cd/m² and the contrast of more than 1,000:1 are a good start for the display. We can still see some backlight bleeding, even though it is only visible with a black picture and when you look for them. We used a long exposure to show it in the picture. Those areas are not visible in practice, independent of the brightness level.

The performance is somewhat of a mixed bag. The processor runs at 2.8 GHz in the Multi-Test when the device is warmed-up, and the Single-Test is executed with 2.9 GHz. 3.0 GHz is very rare and can only be maintained for seconds, for example, when the XPS 15 is still cool. The system can therefore not utilize the full performance potential of the i5-6300HQ CPU (Turbo: 3.2 GHz).

Running on battery power will result in a throttling to just 1.6 GHz. This basically halves the possible scores; Cinebench R15 drops to 246/68 points. The system, unfortunately, does not recognize the mains power when you plug the PSU in; when the 1.6 GHz is set, it is not going to change, and it obviously reduces the performance. An application that is already running will therefore not immediately run faster when you attach the PSU. We had to put the system in standby for a short while before the XPS 15 lifted the clock reduction. Otherwise, the clock would stay at 1.6 GHz for multiple CPU benchmarks, so we had to repeat them. This limitation is independent of the power profile; we also used the "High Performance" setting on battery power. We hope that this issue will be fixed in an upcoming update.

The bar chart includes the XPS 15 from 2014 with its quad-core i7-4712HQ. Our i5-6300HQ can almost keep up with this chip, but only in the single-core benchmarks. Extensive multithread load like heavily parallelized software will not bring the system to its limits, but the Haswell rivals like the Apple MacBook Pro Retina 15 perform much better. The Aspire V Nitro VN7-792G shows: The 6300HQ is a bit faster in a thicker chassis with a more powerful cooling solution. +15% (Single R15) is the biggest difference; the other results are more or less on par. The small construction also seems to affect the performance on mains power.

The Asus Zenbook Pro UX501JW – actually an important rival because of its high-quality and thin construction – can manage an advantage of 11 (Single) and 30% (Multi). The XPS 13 with the only 15-Watt SoC in this comparison falls behind by 34% (Multi) but is almost on par in the Single-Test.

The WD Blue (WD10SPCX) is Western Digital's slim edition with a height of just 7 millimeters. It is a SATA 6 GB/s hard drive that runs at a rather slow 5400 rpm. We know that Dell also uses other hard drives from the Golem-review, where the device was equipped with the Toshiba MQ02ABF100 (Ultra Thin 7.5 mm). This is already the largest capacity you can currently get; more than 1 TB is currently not possible for a 7-mm drive. By the way, the WD model retails for almost 100 Euros (~$108) in Germany.

The WD Blue is not alone though, it is accompanied by a 32 GB SATA-SSD in the M.2 slot, which is used as a cache. The Lite-On chip uses two flash packages from Hynix, 256 MB DDR3-1333 (DRAM cache) and a Silicon Motions SM2246EN controller. The BIOS lists a RAID array, more precisely a RAID 0: Striping – acceleration without redundancy. The user can use one single drive and does not "see" the 32 GB drive in Windows. There is actually no BIOS option to deactivate the cache partner, so we would call it a hybrid system. 

Slow hard drive?

Not at all, at least CrystalDiskMark and AS SDD determine high sequential and 4K / QD 32 read transfer rates. 238 MB/s (seq. read) is not managed by any other conventional hard drive in our database. Read 4K / QD32 can even compete with real SSDs. The hybrid drive manages 31 MB/s, while the HDD in the ProBook 470 only reaches 0.29 MB/s. We did, however, notice heavy fluctuations when we repeated the tests. Read 4K: 31 – 27 – 1.9 – 26 – 1.5 – 1.3 – 26 etc. We get the impression that data is proportionally written on the SSD and HDD, so some read accesses are fast and some are slow. The results in the HDD benchmarks are therefore erratic (read results).

All writing tests show the typical HDD picture. The advantage of this array is limited to the reading performance. So what happens if you only open new files? Can you benefit from this array or will you just be annoyed by the slow hard drive?

This cannot really be the case, because AS SDD and CrystalDiskMark don't give the algorithm a lot of time to learn. However, the HDD benchmarks as well as the PCMarks show reading speeds and storage scores, respectively, that can compete with the SSD-based rivals. Dell's hybrid solution manages to overcome the slowness of conventional hard drives by providing a high reading performance for applications. You also get a lot of storage, so the solution is a good compromise between performance and storage capacity.

Deactivate the SSD cache?

Can the 32 GB drive be used separately when we deactivate the array? We switched the SATA configuration to AHCI in the BIOS, but the operating system could not be booted any more. So it's not that easy. We would have liked to run benchmarks for the Lite-ON CS1-SP32-11 separately. One question that comes to mind is the reliability; would a failure of the Lite-On chip disable the whole WD drive? Users should definitely create backups, but this is also the case if the system just used one hard drive.

The two SATA controllers (SATA-0, SATA-1) and the SSD (M.2 PCIe SSD-0) can be activated individually in the BIOS – Dell also offers single-SSD configurations – so it is obviously also possible to deactivate the RAID array when you make a clean install of the OS. Windows could, for example, run on a bigger M.2 SSD, while the HDD is used for data.

Dual-heatpipes for the processor and the graphics card transport the heat to two fans. We could not find out if those two fans act individually, the system noise was always similar. There are two components that create noise while idling: The fans and the hard drive. The lowest curve in the graphics in cyan shows the hard drive, the light green curve is Idle Max, where the fans are active in addition. The two components are on one acoustic level, neither the cooling solution nor the hard drive leaves a negative impression. The latter is also very quiet when it reads and writes; movements of the head are not louder than the rotation.

How is the situation under load (yellow = 3DMark06)? The fans start to spin up to 37 dB(A), which is audible as a quiet murmur, but sonorous and not annoying. The fan does not pulsate, and this does not change under maximum load during our stress test (red = Load Max.). The red curve is identical to the purple curve (only GPU load). The fans will hit their maximum speed when you play games, high GPU load or in the stress test (CPU + GPU). We also had this impression during the gaming benchmarks: There is hardly a louder or quieter differential, the cooling solution works at the maximum level and reaches 47.5 dB(A) in very demanding titles. Less complex titles result in 37 dB(A), just like our 3DMark06 test.

To ensure that the two fans get sufficient fresh air via the openings in the bottom cover, you should not place the XPS 15 on a blanket, carpet or the like. The warm air is dissipated to the outside via vents underneath the display cover.

In short, the situation is similar to the past: Powerful components still need a powerful cooling solution and these components usually make the most noise in a thin chassis. The MacBook Pro Retina 15, IdeaPad Y50-70 and MSI PE60 are amongst the loudest competitors. The Zenbook Pro UX501JW appears to be a bit quieter, but it already reaches 42 dB(A) while idling (median).

Can the loud dual-fan solution keep the temperatures under control during load? This is certainly the most interesting question considering the 45-Watt processor and GTX 960M (TDP around 60 Watts). The 130-Watt PSU is powerful enough for both components, but what about the heat?

Let's have a look at the idle operation first, where the temperatures are surprisingly low. The average value at the top is just 32 °C and 34 °C at the bottom, independent of the power profile. The rivals can do that as well, the thin aluminum chassis of the Zenbook Pro UX501JW actually stays cooler, even though it is based on a Haswell processor. The MacBook Pro Retina 15 is also cooler.

Load scenarios are a real challenge for the XPS 15, the cooling can hardly manage the heat in the stress test, despite the massive throttling. A look at the temperatures shows the situation. We can measure up to 62 °C at the top and 55 °C at the bottom. This is still reasonable considering the powerful components; hardly any user will really stress the notebook for longer periods while it is on the lap. 

We can see a recurring cycle of the processor and graphics card during our stress test: The i5-6300HQ quickly throttles from its 2.8 GHz (@Cinebench tests) in steps to 1.8, then 1.4 and finally 800 MHz. The temperature of the 6300HQ drops from 95 °C (limit where the throttling sets in) to 70 °C at the same time, just to rise back to 90 °C as soon as the CPU continues to run at 1.8 GHz. This recurring cycle (800 MHz – 1.8 GHz) takes place roughly every two minutes and we can see a correspondingly fluctuating power consumption between 66 and 129 Watts.

The GPU is also affected by these temperatures. Maybe the linking of the CPU and GPU heat with dual-heatpipes was not a good idea after all. A warm processor will also warm up the graphics card. The GTX 960M therefore also throttles its clock. We can see 650 MHz in the stress test with fluctuations to 1,200 MHz, but it runs at 650 MHz most of the time. The GPU temperature does not fluctuate as much as the CPU: It can reach 90 °C, where the throttling sets in and the temperature of the GPU will level off at around 70 °C.

Throttling everywhere then? No, as we described in the section "Performance," the i5-6300HQ can maintain 2.8 GHz, but only if you don't stress the GTX 960M at the same time. This is 15% under the maximum potential of this CPU (single-thread). Only stressing the GPU (screenshot FurMark only) does not seem to be an issue and the card can run at the full 1,200 MHz. We will check whether the gaming performance is affected by throttling due to the simultaneous load in an in-depth gaming review.

The power consumption at the socket fluctuates between 10 and 17 Watts while idling, depending on the power profile, the wireless modules, the brightness and finally the background processes. The XPS 15 is therefore more frugal than a similarly equipped Haswell model (Zenbook Pro UX501JW). That Haswell does not automatically consume more power is proven by the Lenovo IdeaPad Y50-70 with 4 up to 11 Watts while idling. Our review unit also consumes more power while idling than the Acer V Nitro VN7-792G (945M), despite the same Skylake platform, and the latter is a 17-inch device.

Normal workloads simulated by 3DMark06 (without charging of the battery) pull 70 Watts from the socket, which is comparatively low. An indicator for the throttled GPU performance? We will check it. 

The stress test provokes throttling all the way down to 800 MHz, which lets the power consumption fluctuate in a circle: 129 Watts – 99 – 66 – 129 – 99 and so on. We noted the average value of 98 Watts. The reason for the fluctuations was described in the section "Emissions/Stress Test." Notebooks without throttling and comparable hardware consume between 112 up to 120 Watts (IdeaPad Y50-70, MSI PE60) in this maximum load scenario. The XPS 15 is similar to the Zenbook Pro UX501JW (73 Watts), which throttles in the stress test as well.

Loading Time / Energy Options

In terms of charging times, we cannot confirm the review from our Golem colleagues. The compact 130-Watt PSU cannot completely charge the XPS in about 1.5 hours. The charging prediction indicates two hours, but the device was actually charged to 100% after 2:47 hours. 95% is reached after around two hours, the last 5% takes the most time. You can adjust the charging behavior very comprehensively in the BIOS, even dependent of the weekday and the time.

But there are even more interesting possibilities, like the Peak Shift. The user can set the time (from … to …) of the peak load of its power consumption. The XPS 15 will not use any mains power during this time, as long as the battery capacity does not drop below 15%. You can adjust if the device will be charged or not during these peak times.

The standard setting in the Primary Battery Charge Configuration is "Adaptive". Our colleagues from Golem probably used the "Express Charge", where the battery is recharged to 90 % within 1.5 hours and is fully loaded after another hour.

The details are sophisticated; the power connector is illuminated and easy to find in the dark. The battery button on the right side of the case shows the charging status visually via four LEDs. Also, at the front of the XPS is a wide LED that indicates a critical battery state. It glows orange and white when the power adaptor is attached.

Dell equips its high-performance notebook with a 56-Wh battery, only the more expensive Core i7 version with 4K touch display and 512 GB SSD gets the big 84-Wh battery. The advantage of the higher capacity is compensated for by the higher power consumption of the 4K panel (more LEDs), which is shown by the additional runtimes of the 84-Wh model: The idle runtime is marginally shorter, Wi-Fi and video runtime even more so (-7 / -12%).

No HDD = 84-Wh battery. This equation works, because the bigger battery just occupies the space of the 7-mm hard drive Western Digital WD10SPCX.

As usual, we determine the Wi-Fi and the video (Big Buck Bunny H.264 1080p) runtimes with a reduced brightness. We could not adjust the panel to the standard of 150 cd/m², so we decided in favor of the XPS 15 and went one step below at 123 cd/m² (center, dimmed by 6 steps). The Wi-Fi test uses the “Balanced” power profile, while the energy-saving mode is active for the video playback test. The Intel processor C-states were active in the BIOS.

The two runtimes are close to each other, the 1080p video runs half-an-hour longer, probably because the flight mode is active. This means you can expect 6 hours in practice at best, which means without gaming, picture editing or video rendering. This is usable but pretty short, the Apple MacBook Pro Retina 15 2015 lasts 30% longer in the same test. It also carries a 99.5-Wh battery in return. This means the XPS 15 makes more of its battery capacity and is more efficient. But it should be, because we compare Skylake with Haswell.

The table shows: Notebooks with a similar performance have a hard time to manage long runtimes – but the manufacturers don't help with the integration of small batteries.

The runtimes are much shorter under sustained load, even though the CPU throttles to just 1.6 GHz. 140 minutes is actually not bad in comparison, but it is also a result of the throttling, which is not the case for the other rivals in the table. We also ran a loop of Cinebench R15 (Multi-Core) after Battery Eater Classic. The result was 2:38 hours (158 minutes).

We determine the idle runtime at the lowest brightness level. The notebook benefits from the active PWM (below 10% luminance). This value is not practical though and can only be used when the XPS 15 is executing tasks without necessary inputs. The Apple MacBook Pro Retina 15 2015 manages almost twice of that with its big battery. The other devices with the same form factor fall behind. It is interesting that the big 17-inch device Aspire Nitro Black Edition VN7-792G (945M) lasts 13% longer while idling.

You can order the 84-Wh battery separately at the beginning of 2016 according to Dell. This should help the runtimes, but the user will have to remove the SATA hard drive and remove the RAID array by a new installation of the OS. We might be able to test that with the 84-Wh i7 top model next week.