MSI GT70 Dragon Edition Notebook Review: Haswell and the GTX 780M
by Dustin Sklavos on June 18, 2013 12:01 AM ESTGaming Performance
I had been hoping the MSI GT70 Dragon Edition would be an able demonstration of the performance of Intel's Haswell and NVIDIA's GeForce GTX 780M. While we can isolate the CPU performance easily enough, isolating GPU performance is much trickier. NVIDIA is using boost clocks on the GTX 780M, which means it's able to turbo up depending on thermal and power headroom, and there's actually a healthy enough variation in clocks that different chassis will be able to produce different levels of performance.
There's also the cooling system of the MSI GT70 Dragon Edition, which either doesn't have or just barely has the capacity to handle a combined 150W of heat.
Entry-level gaming results are in Bench, but suffice to say the GTX 780M is more than adequate for those settings, and so for the review I'm going to stick to Mainstream and Enthusiast level benchmarks.
In situations where the CPU is not a limiting factor, the 780M is able to boast a healthy lead on the 680M. But you'll notice that the GT70 Dragon is actually underperforming in certain cases; the CPU is getting throttled due to heat. The superior cooling system of the Alienware M17x is able to dissipate far more heat than the GT70's is.
Situations that stress the GPU more exclusively can result in healthy gains over the 680M, but overall stress on the CPU stemming from processor-intensive games like Skyrim and StarCraft II, as well as the hit from TressFX in Tomb Raider, effectively keeps the 780M from really stretching its legs.
Remember that on paper, at stock clocks, the 780M has at least 22% more shader power than the 680M and 39% more memory bandwidth. That means that, bare minimum, the 780M should be roughly 15%-20% faster than its predecessor. We're getting that in the traditionally GPU intensive Sleeping Dogs and Metro: Last Light, and most of it in BioShock: Infinite. But other games see lower gains, or are even slower on the GT70 Dragon Edition despite it having directly superior hardware.
Facebook
Twitter
RSS
115 Comments
View All Comments
huaxshin - Tuesday, June 18, 2013 - link
I used to own a GT70. It had GTX 680M and i7 3610QM. The components reached maybe max 72C after gaming for many hours. So I`m raising big question about this review since the thermal capacity have improved along with a GPU with slightly higher Core count.And the fact that you got HIGHER temperatures than Notebookcheck when they pushed the GPU and the CPU to the very limit by using artificial benchmarks like Prime95 and Furmark. Programs known to have killed a dozen of systems because of the stress they put on the components.
How do you explain that?
I absolutely believe you got a notebook with some crappy paste job. And that caused the game tests you have to show a incorrect picture of the newest GTX 780M as well as MSIs own notebook. I hope you have contacted MSI to get a new system or atleast some explanation, because what this review shows is not normal.
Nor is fair to compare Razer, very thin notebook with low end components, with a pure gaming notebook, which is very thick, and have the industries biggest fan to cool off the components. Which you failed to write about in the review. Its not just 1 fan, but a big one.
180W PSU is also more than enough, shown by internal tests by resellers I know. It even allows overclocking. The CPU throttling didn`t come from lack of power, it came from bad temperatures.
huaxshin - Tuesday, June 18, 2013 - link
Time to redo the review. For the sake of Anandtech as well as MSIs reputation.imo
ZeDestructor - Tuesday, June 18, 2013 - link
"I used to own a GT70. It had GTX 680M and i7 3610QM. The components reached maybe max 72C after gaming for many hours. So I`m raising big question about this review since the thermal capacity have improved along with a GPU with slightly higher Core count."Did you read the review? The older model CLEARLY had a different cooling design. Like, 2 fans vs 1. On opposite ends of the chassis.
huaxshin - Tuesday, June 18, 2013 - link
Eh, no. MSI have never had 2 fans.And yes, the older model had a different cooling design. A worse one.
ZeDestructor - Tuesday, June 18, 2013 - link
Actually, I take that back, here's the old model: http://cdn.goodgearguide.com.au/dimg/700x700/dimg/...as you may notice: SEPERATE heatpipes. Somehow they didn't account for the VRMs moving on chip.... Gee, maybe they should've paid attention when Intel announced quite proudly that Haswell would have integrated VRMs. and started redesigning there and then.
The 680M is a cut-down, lower power version of the 780M too, so naturally it will run cooler.
BobBobson - Tuesday, June 18, 2013 - link
Gee Whizz....Are you still pumping your gums about this?
Call the MSI secret police and hit squad, a negative review has hit the interweb...how dare they!
Darkstone - Tuesday, June 18, 2013 - link
It's unfair to compare the temperatures of notebookcheck vs anandtech, because notebookcheck measured quite serve throtteling issues as well. The CPU was basely running at 1/4th of the designed power consumption. It's clearly in the screenshots, yet the reviewer didn't even mention it.mercutiouk - Tuesday, June 18, 2013 - link
I notice our shill doesn't respond to this.ZeDestructor - Tuesday, June 18, 2013 - link
"And I absolutely believe engineers would consider 98C for a CPU core to be normal."As an engineering student who used to run an old acer at an idle of 85+°C and load of 102°C (with the bottom panel completely off), I know exactly what you mean. Sometimes constraints happen and stuff like this gets forced through QA and validation. Someone in marketing probably wanted a quieter machine, so the engineers tried something proven in a rather new way: single heatsink with a single giant fan. While the concept is great, it really needs more refinement. See the old Dell Precision M4400 (quad-core with top-end Quadro FX parts) for single-fan setups done right. ish. They probably run quite a bit hotter than my E6500 ¬_¬
Since I got my desktop, my priorities regarding laptops have changed, so I rock an awesome little X220 tablet for uni/mobile work. with a much more reasonable 40-75°C range. I can make it hit 97°C using IBT, but IBT is a pretty special little torture test, much like FurMark.
Talking of which, you guys should review convertible tablet PCs and more enterprise laptops that aren't HP Elitebooks. There's a severe lack of ThinkPad and Latitude reviews around here :(
cjl - Tuesday, June 18, 2013 - link
The fact of the matter is, the test sample (which I would tend to think would be tested pretty heavily before it was sent out to a reviewer) had a woefully inadequate cooling system. Whether it was a bad thermal paste job (possible), or simply the fact that they were trying to cool a combined ~150W of TDP with a single small fan (more likely IMHO). Look at the Alienware design - two separate thermal systems, one for the CPU with a similar fan to the one in the MSI, and one significantly thicker fan with a different design (more similar to the blowers on high-end video cards) for the GPU, along with 3 dedicated heatpipes for the GPU and 2 for the CPU. The MSI design has two dedicated heatpipes for each, and one shared heatpipe (and I'm skeptical how useful a shared heatpipe would be, regardless of what kind of fancy marketing-speak MSI uses to describe it), all cooled by a single fan that looks similar to the Alienware's CPU fan alone.Would improved TIM help? It would probably be good for a couple of degrees, but I doubt it would reduce the CPU temps to acceptable levels. The simple fact is, a modern high-end gaming notebook should have multiple fans to provide optimal cooling, and this notebook falls short in that area. As for whether the computer would have been designed for that? A surprising number of modern notebooks have overheating problems at full load (due to the competing constraints of form factor, appearance, marketing, and component placement), and it's therefore much more plausible (to me) that MSI really did botch the cooling system design than that Anandtech is incompetent (given past reviews).