The Newbie: Performance of the Razer Blade 15 (2018)
Nov 10, 2018 - 10 minutes reading time
Before we get into the thick of things, first let's look at my XMG. Aside from the hefty weight it's just become a little dated. It was just barely enough to run the games I was making during my studies. There's some performance to be had.
The Old Hardware
XMG P704 PRO (2014)
- Intel Core i7 4710MQ
- Nvidia GeForce GTX 870M (6GB)
- 2x8GB Crucial Ballistix Sport DDR3-1600
- 2x SATA SSDs (1x 250GB, 1x 120GB)
- 17,3" Full-HD TN screen (60Hz)
- 230 watt power supply
Based on the Clevo P177SM-A barebones laptop, the XMG offered good performance for a reasonable price of around €1670. Lots of I/O ports and performance roughly on par with high-end hardware from 2011 (comparable to my old desktop system) with a pretty good thermal solution. Then, Nvidia released the 10-series of GPUs for laptops that turned the market on its head a little. The GPUs didn't differ at all from their desktop counterparts, the only difference became memory and clock speeds, and thermals of course. There was a sizeable increase in performance as a result.
The XMG accompanied me well through my studies, but it has been stretched to its limits with And Now and Tea Shop. The future-proof video memory, 6GB in size, saved the GTX 870M from aging too quickly. Through overclocking and later undervolting I could eek out about 90MHz faster clock speed. Nvidia's Kepler chips just didn't lend themselves to overclocking as well as newer chips with measly air cooling. I knew that from my desktop GTX 680. I suspect that these special laptop GPUs were also preselected (binned) to be of worse quality than the ones that landed on desktop cards. Since the power consumption in laptops is very limited, so the chip wouldn't need to run at its limits.
Unfortunately, about a month after I bought the P704, XMG followed up with the P705, a new device with M.2 NVMe slots and an IPS display. Looks like I was too quick on the draw there.
The New Hardware
The candidates
The potential candidates for a new device were all from the same category of laptop:
- MSI GS65 Stealth Thin - €2199 (Notebookcheck's review)
- Gigabyte Aero 15X - €2499 (Notebookcheck's review)
- Razer Blade 15 (Advanced) - €2449 (Notebookcheck's review)
- Apple MacBook Pro 15 (2018) - at least €3299 (Notebookcheck's review)
All four were released in 2018 and have, aside from the MacBook, the same main components: An Intel Core i7 8750H, 16GB of RAM and an Nvidia GeForce GTX 1070 Max-Q. This is topped off with a 144Hz Full-HD IPS display with roughly 99% coverage of the sRGB color space. The MacBook would have a Core i7 8850H, a special variant of the 8750H with a slightly higher boost clock. The MacBook also features an AMD Radeon Pro 560X, that roughly matches the RX560 desktop chip, and an IPS Retina Display with 60Hz. Apple plans to have AMD Vega GPUs in MacBooks later this year, but I suspect they will still be performing worse than Nvidia's GPUs.
The four candidates all fulfill my requirement for weight. They all weigh around 2-2.5kg. All are premium devices price-wise. But two need to go immediately. The MacBook Pro because of its high price and significantly lower graphics performance that's relevant for video games. The MSI is excluded because of numerous reports of weak build quality. I really didn't want to deal with that given the price I'm looking to pay. The Gigabyte and Razer remain then.
A friend of mine bought the Gigabyte Aero 15X based on my recommendation, and was mostly happy with it. I'm not averse to this laptop and would recommend it again when you need good performance and the maximum possible battery life. The Razer Blade 15 was just about to be released when I was browsing around. I liked it more than the Gigabyte from a looks standpoint, with its gamery Crysis keyboard font. The Razer Blade looked like a MacBook, but with some gaming DNA. Pretty neat, I thought. The first reviews also presented the device as being solid.
Razer Blade 15
My chosen configuration has the following hardware for the mentioned €2449:
- Intel Core i7 8750H
- Nvidia GeForce GTX 1070 Max-Q (8GB)
- 2x8GB Samsung M471A1K43CB1-CTD (DDR4-2666 CL19)
- 1x 256GB M.2 PCIe SSD
- 15,6" Full-HD IPS screen (144Hz)
- 230W power supply
Testing methodology
I'm not looking for the most neutral comparison here, but how much performance I gained from the new device, and how much performance I'm leaving on the table compared to my desktop PC.
The desktop PC has an Intel Core i7 7700K, 32GB DDR4-3000 CL15 RAM and an EVGA GeForce GTX 1080Ti SC2. And of course I also tested the old laptop. All three systems were running stable overclocks/undervolts that I was using daily.
I tested the three systems in 3DMark Fire Strike, 3DMark Time Spy and Unigine Superposition. Unfortunately, because the systems are so different, they don't offer the same access to sensor information, so I can't present as much data as I did previously. Still, I tried to collect as much sensible info as I could. Sadly, the laptops didn't let me log the fan speeds. The data is again logged to the second. For framerate comparisons I took the framerates output by the respective benchmarking software.
This time I also had the opportunity to log the room temperature while the tests were running.
Benchmarks
3DMark Fire Strike
Let's first look at the results (average of three runs):
We can immediately see a large difference in performance between the three systems. The Razer Blade 15 allows the desktop with its GTX 1080Ti a 1.8 times increase in performance, while being 4.5 times faster than the XMG. The two extra CPU cores allow the i7 8750H in the Blade 15 to be a little faster than the i7 7700K in my desktop, which is amazing!
Let's look at the rest of the data. You can switch between devices via the following buttons:
First of all, the relation of core temperatures to power consumption is interesting. We're comparing a GTX 1080Ti with a power limit of 300 watts (250 watts stock + 20% set in MSI Afterburner), a GTX 1070 Max-Q with 90 watts and a GTX 870M with 100 watts. Neither MSI Afterburner, nor other software could read how much power the latter was taking. All GPUs try to stay within their power limits. The undervolted GTX 1080Ti never fully utilizes it.
The desktop sees maximum CPU and GPU temperatures of 66°C and 68°C respectively at an average of 35.6 watts and 192.1 watts of power consumption throughout the four tests. The Blade 15 has 96°C and 66°C respectively at an average of 18.7 watts and 75.3 watts. The P704 reaches a maximum of 73°C and 84°C at an average 11.6 watts for the CPU. It has to be stated here that the graphics tests don't stress the CPU as much, while the GPU isn't stressed as much during the CPU test. That's reflected in the mentioned average values. Game benchmarks are made to simulate realistic loads on the game-relevant components. For the most part, that would be the GPU. The CPU usually only has to be fast enough to gather all the needed data to send it off to the GPU, to form draw calls. The GPU is usually the limiting factor. And that would be the case with the three systems here. The GPU usage is around 100% most of the time, while the CPU usage is far below that. This behavior will also be noticeable in the other tests.
3DMark Time Spy
Again, first a look at the average results of three runs:
It becomes clear that Time Spy, a DirectX 12 benchmark, runs better on modern hardware. The better results are not only attributable to the faster GPU, but also optimizations that were made specifically for this API that Nvidia started supporting with their Pascal architecture. Again, the i7 8750H sees a significant uplift compared to the i7 7700K, and it does so using less power.
In a modern benchmark such as this, the GTX 1070 Max-Q gets about 3.7 times the performance of the GTX 870M in the XMG P704. The GTX 1080Ti in the desktop again gets roughly double the performance of the Blade 15. The difference in performance compared to the desktop is pretty significant, but given the power limitations it's absolutely reasonable. The GTX 1080Ti is about 6.88 times faster than the GTX 870M, an impressive result.
The rest of the data:
If you didn't notice it in the last graph: While its GPU temperatures are under control, the Blade 15 has a problem with CPU temperatures. I'll get to that later. In the longer and more thorough Time Spy benchmark the CPU reaches 99°C in the CPU test, and there's thermal throttling, reducing of the clock speed to keep temperatures under control. Undervolting the P704 would be a possibility to squeeze a little more performance and a more stable clock speed out. I didn't have time to do that for the Blade 15 yet.
By the way: The bouncy usage and temperatures at the beginning and end of every test occur because of the 3DMark Launcher that plays animations and stresses CPU and GPU a little.
Unigine Superposition
To round out the benchmarks, here's Unigine's Superposition benchmark with the 1080p Medium preset. The average scores of three runs:
And the rest of the data:
It becomes clear that the Razer Blade 15 is thermally limited. Even a comparatively low load on the CPU makes it pretty hot. The GPU has an easier time dissipating its heat via its larger chip surface. It could also be that the contact of the CPU to the vapor chamber cooler is just not very good. Or the thermal paste used is the limiting factor.
Anyway, the Blade 15 seems to hold up quite well, despite its shortcomings. Given the power limits of the components it becomes clear that, while it's not giving ultra high-end performance, it's pretty efficient at what it's doing. In Superposition the new laptop yields about 76% the performance of the desktop at a fraction of the power consumption.
Conclusion
The Razer Blade 15 is an interesting device with pretty ample power in a small footprint. There are slight doubts about the longevity and the time of purchase, since Razer seems to have improved the production since the first devices have come from the factory, so that the displays show less backlight bleeding for example. Still, in the four months that I had the Razer Blade, and especially after changing the thermal paste, I'm content with it all things considered.