NVIDIA announces the GeForce GTX 10-series mobile Pascal family


I’ve been patiently waiting for NVIDIA to announce their mobile Pascal GeForce lineup, and this year it came much later than expected. Normally the mobile side of the family launches before the desktop cards make it to the shipping crate, but this time NVIDIA’s launch was much more cautious, and it looks like they were playing it by ear to see how far they could spread their chip production to satisfy all their customers. Now that the desktop market is settling in and stocks are at good levels, they’re launching more Pascal graphics cards, this time for the mobile market. These GPUs will replace all existing solutions in NVIDIA’s mobile lineup, and all are much faster than the GPUs they replace at the same level.

NVIDIA GeForce Mobile Pascal Lineup

GTX 1080 GTX 1080 GTX 1070 GTX 1070 GXT 1060 GTX 1060
GPU Variant Mobile Desktop Mobile Desktop Mobile Desktop
CUDA cores 2560 2560 2048 1920 1280 1280
Texture units 160 160 128 120 80 80
ROPs 64 64 64 64 48 48
Core Clock 1556MHz 1607MHz 1442MHz 1506MHz 1404MHz 1506MHz
Boost Clock 1733MHz 1733MHz 1645MHz 1683MHz 1670MHz 1709MHz
Memory clock 10.0GHz 10.0GHz 8.0GHz 8.0GHz 8.0GHz 8.0GHz
Memory Bandwidth 320GB/s 256GB/s 256GB/s 256GB/s 192GB/s 192GB/s
Memory bus width 256-bit 256-bit 256-bit 256-bit 192-bit 192-bit

What’s interesting here is that NVIDIA hasn’t skimped much at all on the specifications. The cards are all mostly similar with the exception of the mobile GTX 1070, and they all seem to lie very close to each other in performance as well. This honestly shouldn’t surprise anyone, not after AMD announced that it was slipping the Radeon RX 460 into notebooks launching soon, but it’s nice to see that we’re finally approaching parity with the desktop and mobile graphics cards.

Right away, we see that the bang-for-buck option will be the middle of the road, the GeForce GTX 1070. While the GTX 1060 will have a better price-performance ratio, the GTX 1070 may find itself in some notebooks that allow overclocking, which means that with a few tweaks here and there, you might be posting faster scores on your laptop than your dedicated gaming rig. Thermal throttling will be an issue though. Despite the fact that these GPUs all share the same silicon, the mobile variants are going to have tighter power budgets to work with, which means that on average they’ll still be slower than their desktop counterparts.

Of course, thanks to the GTX 1080, we now also have the first examples of GDDR5X memory shipping in a mobile product. GDDR5X is generally faster and more power-efficient than GDDR5, and delivers a heck of a lot more memory bandwidth for the money. If you’re serious about VR experiences, the GTX 1080 is going to be the best choice for you. I wish AMD had some kind of counter-offer to make here, but we’ll have to wait for the rumored Vega launch in September to see if they’ll do anything. Along with the single-chip setups, NVIDIA is also qualifying the GTX 1070 and GTX 1080 for SLI, just as they currently do on the desktop. This makes sense, since the GTX 1060 usually never sees SLI setups anyway, but I can’t help but wonder if the lack of SLI in the mobile space was the catalyst for NVIDIA to begin only supporting two-way SLI.


While the mobile Pascal lineup includes the same features and capabilites as the desktop lineup, it benefits from one additional feature that is derived through work on the desktop – BatteryBoost 2.0. This is related to NVIDIA’s work on making Pascal hold a certain clock speed or performance level in a tighter fashion, as the old power management technology, Boost 2.0, would vary load levels according to temperature and voltage, which wasn’t aways ideal (Boost 1.0, which came with Kepler, was TDP-based, and flawed for a lot of workloads that would max out the TDP (thermal design point)), but receive poor performance in return). The new version of Boost is much better, and able to hold either a particular performance level or temperature for as long as necessary.

Using that new tech, NVIDIA now can target a dead-on 30fps framerate from notebooks using the battery for its power source instead of a wall socket. This means that variations in the GPU’s performance won’t be as high or low as they were in the past, and battery life predictions can be a little more accurate as a result. NVIDIA says it also makes for a smoother experience, and I’m eager to see how this bears out on high-end machines in the future. Acer’s Predator 17 ran for just over two hours on the battery playing Borderlands: The Pre-Sequel, and perhaps this helps eke out maybe another thirty minutes of runtime.

According to Anandtech, there’s also a drastic change in terms of how NVIDIA qualifies Pascal notebooks and their G-Sync or 120Hz variations. Vendors that choose to put Pascal graphics in their designs need to stick to the base clock speeds so that there’s a minimum baseline for performance across the board. If it isn’t able to keep up at least the base clock speed, then NVIDIA won’t approve the design and won’t sell chips to the vendors for that SKU. It’s a bit heavy-handed, to be honest, but it’ll probably work. I wish this was done for discrete GPU options that use DDR3 instead of GDDR5 memory.

Factory overclocks are also supported, so that’s new. So long as there’s enough thermal headroom, a brand like MSI or ASUS can easily sell a GTX 1080 notebook that boosts to 2.1GHz and call it the “OC Edition” of that variant. It opens up the gates for vendors to start offering slightly more powerful designs and charge more for the privilege, but I don’t think NVIDIA would be too happy with anyone selling an OC Edition variant with a 25% price hike, or something silly like that.

Multiple vendors will be announcing and shipping their refreshed lineups with Pascal graphics in the coming week. Keep an eye out for those models if you’re interested in the extra performance, or play it safe and pick up an older model that has had it’s price reduced to clear out stock.

That Destiny 2 Shadowkeep expansion came from the Moon