NVIDIA GeForce GT 640M: Kepler Goes Mobile
Introduction and Specifications
The fact that NVIDIA is ready with both desktop and mobile variants, in and of itself, is fairly impressive but their mobile platform demonstration vehicle may also surprise you.
NVIDIA presented us with a 15-inch Acer Ultrabook to show off their midrange mobile chip based on their new GeForce 600M series architecture. It's just 20mm thin, weighs about 5lbs and claims to offer up to 8 hrs of battery life, though you can bet that figure is not under a gaming workload.
Regardless, NVIDIA obviously decided to step out in current fashion and is looking to up the "ultra" factor in Ultrabooks. Although it seems like an unlikely place for a discrete GPU, NVIDIA's 28nm Kepler design trimmed down to "only" 384 processor cores and a 128-bit memory interface, sips a lot less power than its higher-end desktop counterpart. Of course, the Acer Timeline Ultra M3 is also a bit larger than most of the 13 and 12-inch Ultrabooks we've looked at thus far as well. But it's still a pretty svelte and light machine.
Let's look at the specifics of the NVIDIA's new GeForce GT 640M mobile graphics engine...
|
Here's a quick look at NVIDIA's new notebook graphics product stack:
Mainstream GeForce 610M |
Performance GeForce GT 620M GeForce GT 630M GeForce GT 635M GeForce GT 640M LE GeForce GT 640M GeForce GT 650M |
Enthusiast GeForce GTX 660M GeForce GTX 670M GeForce GTX 675M |
As you can see, there's a GeForce 600M series now in all segments, but not all are based on the new Kepler architecture. Here's what a few of the GeForce GT 600M series GPUs look like, beyond the GeForce GT 640M we'll be stepping through today...
GeForce GT 650M | GeForce GT 640M |
GeForce GT 640M LE |
GeForce GT 635M | ||
Process | 28nm | 28nm | 28nm | 40nm | 40nm |
Architecture | Kepler | Kepler | Kepler | Fermi | Fermi |
Processor Cores | Up to 384 | Up to 384 | Up to 384 | Up to 96 | Up to 144 |
Processor Clock | Up to 850MHz | Up to 625MHz | Up to 500MHz | Up to 762MHz | Up to 675MHz |
Memory
Interface |
Up to 2GB
GDDR3 or GDDR5 |
Up to 2GB
GDDR3 or GDDR5 |
Up to 2GB
GDDR3 |
Up to 2GB
GDDR3 or GDDR5 |
Up to 2GB GDDR5 |
Memory Width | Up to 128-bit | Up to 128-bit | Up to 128-bit | Up to 128-bit | Up to 192-bit |
Memory Bandwidth | Up to 64.0 | Up to 64.0 | Up to 28.8 | Up to 50.2 | Up to 43.2 |
NVIDIA is blending some Fermi designs into the lineup, but re-branding them so as to limit confusion in product placement and performance. However, there is a fair bit of Kepler technology through the top end of the stack. NVIDIA's formal press release can be found here. Also, if you'd like a deep dive look into NVIDIA's new Kepler GPU architecture, we'd suggest heading over to our Geforce GTX 680 article for the full low-down. Marco goes through all the heavy lifting and details there.