backtop


Print 15 comment(s) - last by EasyC.. on Mar 20 at 10:37 AM

Demoed unit has four times the number of CUDA cores

At 5 lb and 20 mm thick, Acer Inc.'s (TPE:2353) TimelineU M3 is a bit of a beefy 15.6-inch pseudo-Ultrabook.  But what it lacks in slender figure, it makes up for in pure power.

It comes with a Intel Corp. (INTC) Core i7-2637M, a 1.7 GHz dual-core Sandy Bridge part that can Turbo up to 2.8 GHz, a 1366x768 screen, 4 GB of 1333 MHz DDR3 RAM, a 256 GB SSD from Lite-On, and a DVD+RW drive.  But its real treasure is its GeForce GT 640M GPU.

Yes, that's a Kepler GPU.  

NVIDIA Corp. (NVDA) has been noticeably absent from the graphics race this spring while rival Advanced Micro Devices, Inc. (AMD) rounded out its lineup [1][2][3].  However, NVIDIA is charging back, leveraging its strength of GPU computing.  The GT 640M comes with 384 CUDA cores (8 streaming multi-processors), putting it on par with a GTX 460 or a GTX 570M.  

Like AMD, NVIDIA is targeting lower clock speeds for its first 28 nm parts.  The GT 640M is clocked at 625 MHz for the core (versus 672 MHz for the GT 540M).  The part is somewhat limited by a 128-bit bus, and DDR3, instead of dedicated GDDR5.  Still gaming performance in Anandtech's benchmarks falls just short of the GTX 560M, but is overall solid.

TimelineU M3
The Acer Aspire TimelineU M3 is a slick piece of hardware, thanks to its Kepler GPU.
[Image Source: Anandtech]

The laptop showed an impressive 417 min. of battery life on its 55 Wh battery -- rather impressive.  The laptop is also cooler than previous powerhouses.  In both metrics the advantage of NVIDIA's new 28 nm and Intel's 32 nm processes are showing.

Sadly, details about Kepler architecture-wise are scarce as there's an embargo on them.  In fact Acer broke the embargo (accidentally) by airing the laptop, hence giving the world a sneak peek at Kepler.  If the GT 640M is any indication, expect CUDA cores to receive around a 4x bump, putting NVIDIA likely ahead of AMD in computing power.  At the same time expect lower clock speeds as NVIDIA -- like AMD -- adjusts to the perils of 28 nm part-making with its silicon partner Taiwan Semiconductor Manufacturing Comp., Ltd. (TPE:2330).

Source: Anandtech



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Should be...
By mandrews on 3/19/2012 4:16:41 PM , Rating: 2
quote:
Don't you mean 640m?
yup, fixed, thanks!
quote:
I have to be semantic... But nVidia doesn't manufacture it's chips, TSCM does. So that would be TSCM's process node that is doing well.
Yup, and I say that in the article. It's "NVIDIA's" process to the extent that it's purchasing the fab contract... but I agree you could make that argument semantically. I think when you read that I mention that TSMC is the fab contractor it becomes clear though.

Regardless, I would be hesitant to say whether it is "doing well". Both NVIDIA and AMD have showed off 28 nm parts that cram more compute cores onboard -- that's more or less a given, due to the die shrink.

However, their cores are featuring LOWER clock speeds in both cases. This is likely a sign of quality issues on TSMC's 28 nm node. If all were well you'd see a nice clockspeed bump, given the lower power consumption/lower heat dissipation. Clearly something not-so-good is going on, it seems.


"A politician stumbles over himself... Then they pick it out. They edit it. He runs the clip, and then he makes a funny face, and the whole audience has a Pavlovian response." -- Joe Scarborough on John Stewart over Jim Cramer














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki