Print 80 comment(s) - last by .. on Aug 30 at 7:46 PM

Low cost and integrated GPUs gain the most in Q2

GPU sales are looked at as an indicator of how well the overall computer market is doing since all computers today ship with a GPU of some sort be it of the discrete or integrated variety. So far, 2009 has been a rough year for GPU makers.

Typically, GPU makers would be disappointed with no significant growth, but with the poor economy, it's a welcome change from the significant declines other quarters have seen. Jon Peddie Research (JPR) has unveiled its latest numbers for the add-in GPU industry for Q2 2009. The numbers show that 16.81 million add-in units were shipped, up 3% from the previous quarter and down 15% from the same quarter in 2008.

Inventories had to be replenished over Q2 and JPR estimates that the replenishment at least shadowed consumption or was a bit higher. With the poor economy still hurting many consumers, sales in the GPU market moved downstream. Most of the growth in the GPU market was in the integrated segment with a 4% year-over-year increase in shipments for the quarter. Lower cost discrete cards also saw modest improvements according to JPR with higher cost discrete cards taking the worst the quarter had to offer.

The quarter was the first where AMD finally started to gain back some share from NVIDIA. JPR reports that AMD's unit share rose from 31% in Q1 to 35% in Q2 with NVIDIA seeing their share decline the same 4% to 64% overall. That means that AMD took every bit of its growth from NVIDIA's marketshare. NVIDIA and Intel GPU shipments did rebound slightly in Q1 2009.

Comments     Threshold

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

RE: Why Would You Buy NVidia?
By stmok on 8/27/2009 4:02:22 AM , Rating: 2
OpenCL will kill it. Stone dead.

CUDA is already becoming a dinosaur in the world of accelerated computing.

Why develop for one GPU segment (Nvidia), when you could develop for all three* (+ATi & Intel)? Makes no sense whatsoever. Watch CUDA die a slow death over the next 2 years.

*neglecting the smaller brands

Could you be any more ignorant?

(1) CUDA is a framework that allows you directly access Nvidia GPUs; You can also call OpenCL or DirectCompute APIs with it. Seriously dude, download the CUDA SDK and you'll see they teach you how to use OpenCL with CUDA...OpenCL sits on top of CUDA and provides an alternative way to access the GPU.

(2) If you know CUDA, you have a bit of a head start when using OpenCL. Hell, you can port CUDA code to OpenCL.

(3) CUDA calls on GPUs. OpenCL calls for all sorts of processors. This could be CPU, GPU, Cell processor, DSPs, etc.

"When an individual makes a copy of a song for himself, I suppose we can say he stole a song." -- Sony BMG attorney Jennifer Pariser
Related Articles

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki