Dell XPS 15 Gaming Performance (with ThrottleStop)

Dell XPS 15 Gaming Performance (with ThrottleStop)

This is where things get a bit messy and confusing. The CPU throttling on BIOS version A01 through A04 is a severe enough problem that out-of-box gaming performance on the XPS 15 is substantially lower than expected. In most games that we tested, CPU throttling kicks in after anywhere from 30 to 90 seconds, at which point the CPU sits at 1.2GHz until you exit the game. This makes benchmark results highly variable, with the first run always being the fastest, and frankly it’s unacceptable. How this slipped through the cracks on a shipping laptop is something of a mystery, as it took all of five minutes with DiRT 3 and Batman: Arkham City for me to determine something was wrong with performance, and to then track down the root cause. OEMs take note: both games have built-in benchmarks and your performance labs should be using these and other high-profile games to check performance before you start shipping product. End of discussion.

Now, with that out of the way, we wanted to see what the XPS 15 could do with a bit of fine tuning on CPU clocks. Under the default settings on the A04 BIOS, CPU temperatures topped out at 80C, which is lower than we’ve seen on quite a few laptops. Furthermore, CPU clocks were almost always 2.8GHz (maximum Turbo Boost for four threads or more on the i7-3612QM) before throttling down to 1.2GHz, which seems rather myopic. The whole point of Intel’s Turbo Boost is to let the CPU run at the maximum sustainable clock speed without overheating, and right now that’s not happening.

A bit of tinkering with ThrottleStop 5.0 Beta 3 yielded CPU temperatures in the 80-90C range with a 23X multiplier, with some games capable of running at 2.5 or even 2.6GHz without overheating. Again, this isn’t something an end user should be required to use in our opinion, but thankfully it does allow substantially improved performance and it’s easy enough to enable/disable as needed. Depending on the game in question, we saw performance improve by anywhere from 10% to over 100% by using a 23X multiplier with the CPU PROC HOT signal disabled—Batman and DiRT 3 showed the greatest improvements. Note also that the difference at higher quality settings isn’t as large, as the GPU becomes more of a bottleneck. We have 1080p “Enthusiast” results for the XPS 15 as well as our 768p “Value” and 900p “Mainstream” scores, but in most games 1080p Enthusiast wasn’t playable so we won’t report those scores for now. Let’s start with our Value and Mainstream results, and then we’ll list “recommended” settings and frame rates for the XPS 15 running at 1080p.

Value Gaming Performance

Our Value settings don’t pose any problem for the XPS 15, with all games easily besting 30FPS and most surpassing the 60FPS mark. The XPS 15 basically sweeps the table against the Acer V3, showing the value of using GDDR5 memory instead of DDR3 memory on the GPU. Also worth note is that Civilization V performance continues to be a bit quirky with certain driver versions; the original 296.01 drivers couldn’t even break 30FPS regardless of setting on the XPS 15, while the newer 304.79 beta drivers at least let you get 35FPS. One quick look at the W110ER results however shows that we ought to be able to get performance closer to 50FPS. I ran FRAPS while performing the Civ5 “lategameview” benchmark at one point to see what was going on, and for some reason performance appeared to fluctuate between 50FPS and 35FPS, where normally the frame rate is pretty steady. I’d guess a future NVIDIA driver will address the issue.

Mainstream Gaming Performance

Mainstream performance generally follows the same pattern, only with lower frame rates. The GT 640M with GDDR5 is close to handling all of our test suite at 900p “High” details, with Battlefield 3 and Civilization V being the only two games where we’d recommend dropping to lower detail settings (and again, the drivers are likely a factor in Civ5 performance). It’s also interesting that the Acer V3 manages to eke out a win in Batman, indicating our lowered CPU clocks might be coming into play. Otherwise, the use of GDDR5 is typically able to improve GT 640M performance by an average of 8%, with up to 25% higher performance in at least one game (Batman using our Value settings).

Most of the games’ settings are straightforward, but a couple warrant further explanation. In Batman, enabling Very High with PhysX is borderline playable, with an average frame rate of 27 FPS; DX11 without 4xAA is a similar 26 FPS, and enabling 4xAA in either mode drops a few FPS whereas FXAA is basically free. Civ5 performance is still a bit chaotic (see below), so take those scores with a grain of salt for now. Finally, our test scene in Skyrim happens to be one of the lowest performing areas we could find, so while we report 30 FPS there are many areas in the game that will run in excess of 60 FPS.

As mentioned earlier, Civilization V and Total War: Shogun 2 ran better with the most recent 304.79 beta drivers, and that’s what we used for the above settings. The remaining games all ran best with the official 296.01 Dell drivers, though at some point we expect to see a new WHQL driver that will provide the best of both worlds (and hopefully further improve Civ5 performance). Not shown in our charts is another popular game, Diablo III; given that the Acer V3 could handle maxed out settings, the same holds for the XPS 15. With GDDR5 memory, the GT 640M ends up being around 10% faster on average compared to DDR3; it can certainly handle gaming at 1080p, but in many games you’ll need to lower the detail settings somewhat to be fully playable.

ncG1vNJzZmivp6x7orrAp5utnZOde6S7zGiqoaenZINyfJJom56knGLFsb%2BMamxmpGVnfrl5w56rmqGcmrFussirqq1lnKS8rHuT