Thinkpad T410s Real Battery Life Tested.

I thought the NVS is not running by default....sigh....

 

Sounds like you're enjoying your T410s! I just ordered my T420s and I hope I like it as much as I like my T400.

By the way, I did actually own an 11.6" MacBook Air which I ended up returning because I wasn't happy with the performance or the ergonomics. I hate buttonless touchpads, which is a big part of why I selected the T420s over the X220 (the other reason was the higher resolution screen on the T420s).

I chose the T420s over the T420 for three main reasons: the bay battery, USB 3.0, and thinness. I already have a bay battery for my T400 which I will probably re-use for the T420s. I don't use my optical drive very often so it's a pretty big win to use that space for something useful. The T420 doesn't support a bay battery and I really don't want to make my machine larger with a slice battery. With the bay battery the T420s has 76Wh of battery capacity which should mean 8+ hours of light usage and 4+ hours of medium-heavy usage. All of that in a machine that's about pound lighter than my T400, is thinner, has a higher resolution screen, and is considerably faster.

When I get my T420s I'll probably post an un-boxing video if someone else hasn't done so already.

 

Congrats!!!! I found out NVS Optimus is not that good...>"<....I think your T420s integrated graphic card will out perform my Optimus....pls post performance of your T420s after you received it....So looking forward to T420s..

 

In 3DMark06 the Sandy Bridge HD Graphics 3000 (~3800) does do better than the NVS3100M in the T410s (~3200). Keep in mind that the HD Graphics 3000 is available in different power versions so performance can vary between the 35W/45W/desktop versions of Sandy Bridge. The 3800 score was based on a review of a X220 which has the same CPU/GPU (i5-2520M) as my T420s.

In real games the HD 3000 is going to lose vs. the NVS3100M as much as it wins. A lot of this is because of performance issues in the Intel drivers, and thankfully the situation is getting better. But at this point I would still chose the NVS3100M over the HD 3000 purely from a performance standpoint. That said, considering that the performance is as close as it is, I would absolutely take the HD 3000 over the NVS3100M because it uses less power (cooler/quieter) and you don't have to deal with the Optimus mess. Intel GPUs are also the way to go for Linux because the open-source drivers are supported by Intel (unlike with NVIDIA/ATI where the official drivers are a binary-only mess and the open-source drivers are hit or miss). There were some compatibility issues with Sandy Bridge and Linux but they have either been fixed already or are going to be fixed in the next releases of major distributions.

The NVS4100M option in the T420/T420s is a bigger question. Frankly the performance is underwhelming. It's much better than the NVS3100M but it's still only about 1.4x the performance of the HD 3000. Then you have to deal with the driver/compatibility mess of Optimus. If you don't game, obviously the HD 3000 is the obvious choice - it's cheaper and plenty for desktop/video tasks. The thing is though, even if you do game (as I do) the HD 3000 might still be the right choice. It runs popular games like StarCraft II, WoW, TF2, Portal, CS 1.6, CS:S, CoD: Black Ops, Civ5, and many others well. No, you're not going to play Crysis 2 on ultra, but then again the NVS4100M isn't going to run that well either. In fact, while it's easy to find games that run well on mid-range mobile GPU (like a GTX 460M) and don't run well on the HD 3000, it's really hard to find games that run OK on the NVS4100M but not on the HD 3000.

The case for low-end laptop GPUs (like the NVS4100M) is increasingly weak.

On my T400 the dedicated GPU is around 2.5x the performance of the integrated GPU. On the T410 the dedicated GPU is around 2x the performance of the integrated GPU. On the T420/T420s the dedicated GPU is now only about 1.4x the performance of the integrated GPU.

The reason that the numbers are getting closer isn't that the dedicated GPUs aren't getting better. In three generations of the T4x0 the 3DMark06 has gone from ~2500 to ~3500 to ~5500 on the dedicated GPUs. The problem is that performance has improved way more on the Intel integrated GPUs - from ~750 to ~1500 to ~3800.

From the GMA 950 to the HD 3000, Intel's graphics performance has improved around 20x. The HD 3000 is now faster than a GeForce 7800 GTX, which was the fastest card you could get five years ago. That's not bad for an integrated GPU.

 

There are a couple of reasons. The laws of physics are closing this gap. If you put a fast GPU in a thin design, guess what happens? You burn the ThinkPad logo into your thigh.

You also might remember not long ago the envelope got pushed and guess what happened? GPU meltdown city.

Now, they are trying to strike a balance between performance and battery life. With the slate pressures from the iPad and other devices, the pendulum swung the other direction so the gap on switchable has closed.

There is still a difference however. Keith reported a 2 hour difference between Optimus and discrete only on the W520. In that case, you can crank the W520 up on high while plugged into the wall and crank through something intensive, then back it all off and conserve power for the trip home.

Ultrathins or moderate thins like the T420s or T410s don't have the upper end cycles to spread the gap.

 

Well TDP constraints are certainly a big reason why we haven't seen notebook GPUs improve at the same rate as desktop GPUs. A few generations ago the highest-end GPU was 75W and now we are seeing single-GPU cards that push 250W (GTX 580) and multi-GPU cards that are pushing 350W (GTX 590). We're pretty much at the limits of practicality on the desktop side too.

I wouldn't say that the pendulum has swung back on TDP. The TDP for GPUs in the T4xx series has actually increased over time. The NVS 140M (T61) was 10W, The 3470 (T400) was 12W, the NVS3100M (T410) was 14W, and the NVS4200M (T420) is 17W.

The reason T61 GPUs melted down (I owned one) was due to a design defect, not because the GPU was super-high-power.

Intel's GPU has the advantage of sharing TDP with the CPU. Increasingly system TDP is the limiting factor in terms of cooling and power. If you want a higher-power GPU you need a lower-power CPU, and vice-versa. GPU Turbo Boost is cool because for a given total TDP you can have better CPU performance when the GPU is idle and better GPU performance when the CPU is not 100% loaded. If you tried to do this without turbo you could end up with a situation where concurrent high GPU and CPU load overwhelm the system.

Apple had this very problem in the new MacBook Pro (which only has an 85W power adapter but also has a 55W GPU and a 45W CPU) - many people reported that the system would overheat and fail if you tried to hammer both the CPU and GPU at the same time. Apple "fixed" the problem by detecting high GPU current and significantly under-clocking the CPU to keep power consumption in check.

GPU Turbo is fundamentally a better solution than the Apple "fix" because it has better defined thermal margins and because it reacts faster and is more precise.

Then you add the fact that Intel has other advantages like better process technology (32nm vs 40nm) the shared L3 cache, and it's clear why they can do better on power usage than a discrete solution.

Ivy Bridge (the successor to Sandy Bridge) is going to provide another big boost in integrated GPU performance. We're close to the point where discrete notebook GPUs are going to be found only in gaming, workstation, and desktop replacement laptops. I would not be surprised if the T430/T530 is only available with integrated graphics.

Still, at least with Optimus having a dedicated GPU no longer destroys your battery life.