Power And Final Words
Before we wrap up, we need to clarify a few points. We made this clear before, but we need to go over it again. When people travel, they travel with computers smaller than desktop replacements (by that I mean screens smaller than 15.6"). The whole tablet, netbook, and CULV notebook trend makes the footprint "norm" even smaller. We should also consider that a battery for a DTR functions more like a UPS and less like a power plant away from a wall socket. When a professional or gamer carries his mobile workstation around, he isn't going to use that system between power plugs.
That is part of what makes talking about a desktop replacement somewhat unique. We still care more about the performance aspect of the hardware. The new 55 W and 45 W Core i7s bring the latest in desktop-class performance at a fraction of the power consumption. In many cases, this will still translate to bulky enclosures to lug around. But lower TDPs will be much better suited to sliding into svelte chassis.
If there is a point that Intel is going to make, it's that mobile workstations no longer necessarily need a discrete graphics in order to get adequate performance. The HD Graphics 3000 engine delivers everything a mobile workstation needs. Professionals who use desktop replacements for productivity work, this is a huge benefit, as discrete graphics can account for a large portion of a notebook's power profile. For gamers, this means that systems bordering on the DTR classification can get by more efficiently in tasks like Flash playback without having to resort to switching graphics (provided you have a notebook capable of switchable graphics).
But the big story here actually isn't the incremental bump in 3D performance Intel is pushing with its Sandy Bridge architecture. We naturally assume that a next-generation product will deliver better performance, so it's understandable that we'd be looking for Sandy Bridge to beat the Arrandale, Clarksfield, and Clarkdale designs we tested.
What we wouldn't necessarily expect, though, would be better performance and lower power consumption at the same rated TDP levels.
We were naturally curious about power, so on a whim we ran the IE8 workload of our own RLUMark 2010 benchmark. We only recently have had time to update the test's code to account for the latest IE8 and Office 2010 patches. If you read our Fall 2010 netbook roundup, then you know this test is like MobileMark, except that it is scripted in what I call "user time."
We want to stress real-world usage, which is perhaps one of the biggest reasons we decided to have an in-house-developed benchmark. This benchmark mimics what you should expect from everyday life. Right now, the battery life metric is programmed to simulate a user typing at ~45 WPM and reading at ~200 WPM. So, this is a “Real Life Use” benchmark, hence the name--RLUMark (at least until I think of a better name or the Tom's community recommends one).
Remember that IE8 is threaded (one tab, one thread). So, this multi-threaded benchmark includes navigation of Web sites, including scrolling calculated to a reading speed of ~200 WPM. On Amazon, we shopped for GPS units and common Biology textbooks.
This session consists of the following run sequentially in the order listed:
- 24 minutes of Wikipedia Reading (4 Entries): one tab per entry
- 4 minutes of Amazon.com: two tabs
- 3 minutes of CNN.com: two tabs
- 2 minutes of Google Finance: one tab
- 2 minutes of Accuweather: one tab
- 25 minutes of Flash 10.1 (YouTube 360p, H.264, hardware acceleration enabled): one tab
We should point out that we are isolating platform power consumption here. Both notebooks had their video output via HDMI to a desktop monitor at 1680x1050, factoring out each system's LCD. The notebook's AC power adapter is only powering the SSD, motherboard, processor, and wireless card.
We initially ran the battery analysis measuring DC numbers, but the results totally caught us off-guard: the clear winner was Intel's Core i7-2820QM. How could the 45 W high-end mobile Sandy Bridge-based chip use less power than the 35 W mid-range mobile Arrandale? We reran the benchmark by measuring AC power use without the internal battery. These figures confirmed the 7 W margin (average) that we were seeing on the DC side.
With an average power consumption of 17.3 W, the Arrandale-based Core i5 almost doubles the power consumption of the new Core i7's power consumption (10.1 W) in our real-world workload. Even when both processors see a power use spike mid-way through the benchmark due to the start of YouTube video playback, the Core i7 still sees a comparatively small increase. We can't even imagine how lopsided this benchmark would have looked with a Core i7-920XM (Clarksfield).
These findings go a long way to explain the momentum behind Intel's mobile Sandy Bridge launch. Remember that TDP is Thermal Design Power. Per Wikipedia: this represents the maximum amount of power the cooling system in a computer is required to dissipate. Of course, when we're trying to deliberately tax a system in the benchmarks, power use is high. But in the real-world usage that RLUMark represents, Sandy Bridge doesn't come anywhere near its TDP.
As a result of power saving technologies and lithography advances, we see that, despite the Core i7-2820QM's 45 W TDP, it turns out to be a faster and more power-conscious processor during general-use scenarios. Why is there a different power consumption increase when Flash video playback kicks in?
This is where the architecture comes into play. Remember that the Arrandale design employs a 32 nm dual-core processor and a separate 45 nm die containing the graphics core and memory controller. That's not ideal, either from a performance or an efficiency standpoint. Sandy Bridge was purpose-built to enable the best of both worlds using a single die manufactured at 32 nm. Talk about a nearly unprecedented improvement, considering this processor is Intel's second-fastest Core i7, while the Core i5-540M is the third-fastest mobile Core i5 from the Arrandale-based family.
If this reference build that Intel sent us is representative of what we can expect to see from the entire mobile Sandy Bridge lineup, then we're super stoked about what those low voltage and ultra-low voltage Core i7s should be able to deliver. If you're in the market for a full-sized notebook, hold off until you can get your hands on a Sandy Bridge-based platform. You are going to get better performance, along with a substantial increase in battery life compared to Clarksfield/Arrandale-class processors. Once you add the benefits of Quick Sync into the mix (covered in our earlier Sandy Bridge review), the platform gets even more impressive (so long as you aren't a gamer needing discrete graphics).
If AMD is paying attention, it needs to get its act in order. Brazos is one step up from being a pawn in the AMD Fusion chess set. It's not that pawns can be tossed away willy-nilly, and we don't mean to take away from the platform's potential. However, taking over the netbook space, if that even happens, isn't going to be what puts AMD in the position it wants to assume. Let's put things into perspective. The 2010 Danube platform has been out since last May. If you look at AMD's official 2010 platform lineup, the most powerful mobile CPU available seems to be the 35 W Phenom II quad-core N950 (2.1 GHz) on the Acer AS7552G-6061. This is the fifth most powerful mobile CPU listed, and to be fair was part of the Danube refresh rolled out in October (there was another batch announced earlier this month, including the X940). Notebooks that ship with mobile quad-core AMD processors are extremely rare. Its mobile Phenom II Black Edition chips are virtually nonexistant. So, the mobile space is decidedly dominated by Intel, and AMD has a lot of work to do in order to change that.
