|
|
Nvidia announces five 40nm GeForce 200M GPUs
Author: Tim SmalleyPublished: 15th June 2009
Nvidia has released five new 40nm GeForce 200M GPUs this afternoon, which add support for DirectX 10.1 and GDDR5 to Nvidia's products for the first time.
The company's previous GeForce 200M products were, in fact, based on the older 55nm G92 GPU which has found its way into many of Nvidia's desktop products in the past 18 months and are derived from the two and a half year-old G80 architecture.
These latest mobile products have the underpinnings of the company's GT200 GPU which is featured in the GeForce GTX 200 series desktop cards, but they're targeted at more mainstream audiences.
The GeForce G210M is the lowest end part, which features just 16 stream processors running at 1,500MHz and a 64-bit memory interface. The G210M has a thermal design power of just 14W and its core operates at 625MHz, while the 512MB of GDDR3 memory hums along at 1,600MHz (effective).
Next up are two fairly similar parts, the GeForce GT 230M and GT 240M, which both feature 48 stream processors running at 1,100MHz and 1,210MHz respectively. Core clocks are 500MHz and 550MHz respectively and both have a 23W TDP. They also both feature 1GB of GDDR3 memory clocked at 1,600MHz (effective) on a 128-bit memory interface.
The GeForce GTS 250M and GTS 260M are also fairly similar parts, as both have 96 stream processors; they're clocked at 1,250MHz and 1,375MHz respectively. Unlike the GT 230M and GT 240M parts, there's a 10W difference in thermal design power - the GTS 250M is rated at 28W, while the GTS 260M's TDP is 38W.
Nvidia has set the GTS 250M's core speed at 500MHz, just like the GT 230M, while the GTS 260M's core clock replicates the GT 240M's at 550MHz. These parts, however, are the first Nvidia GPUs to support GDDR5 memory - there's 1GB of it on board connected via a 128-bit memory interface. It runs at 3,200MHz on the GTS 250M and at 3,600MHz on the GTS 260M.
The company said that it has made "adjustments in the micro-architecture to improve battery life and overall graphics performance compared to the previous generation." It claims the architectural improvements actually make the GTS 260M almost as fast as the GeForce GTX 260M, despite having fewer stream processors that run at the same frequency and a smaller memory bus width. Of course, the introduction of GDDR5 is undoubtedly going to make up for a lot of the deficit associated with the smaller memory bus width, but it's not going to make up for the shader unit deficit.
Nvidia said that we can expect to see these new GPUs in "over 100" notebook designs by the end of the year and when asked about the reported problems with TSMC's 40nm process, the company said it had encountered no such issues. "We're using around 80 per cent of TSMC's 40nm capacity," a spokesperson claimed.
Discuss in the forums.
Related
Comments (21)- Site Links:
- About
- Email Editor
- Newsletter
- Privacy
- RSS
- Score Guide
- Our Other Websites:
- Auto Express
- Channel Pro
- Computer Buyer
- Computer Shopper
- Den of Geek
All content © bit-tech.net 2000-2010
© Copyright Dennis Publishing Limited licensed by Felden
Why pair 512MB of 1600Mhz DDR3 with 16 shaders on a 64-bit interface? Surely 128 or 256MB of DDR2 would have been more suitable for such a low end part..... or has DDR3 really fallen that much in price?
You forget the "MOAR IS BETTAR!!1!" mentality of the average consumer.
Now that's interesting.
Cheesecake on plate on table, please.
More likely "we are using 80% of TSMC's 'available' 40nm capacity..."....
64-bit requires fewer memory chips than 128-bit, so I assume that'll help with power savings.
power and memory capacity does NOT go hand and hand, remember memory capacity is used for storing textures, the more memory the larger the texture that can be stored or the amount, so it comes in handy for people with higher res monitors or those doing graphical rendering/work.
for gaming and real time rendering yes, not so for drafting or graphical work like on photoshop
In that case, you have main system memory; there's no need for VRAM.
Actually it makes a lot of sense. Think about elements like the Tessellator, which requires the gpu to have an actual hardware piece dedicated to tessellation. AMD has done this long ago even though it's been completely unused by the market, while Nvidia never bothered to do it, and for the new Dx11 cards they'll need to have it. I see these 2xxM cards as a "practice" version, cos the hardware requisite differences for 10.1 and 11 are basically nonexistent, whereas from 10.0 there are a few differences.
By adding 10.1 now, which is pointless, I agree, it does make sense from the perspective that it could be Nvidia's "training" to see how whatever they're implementing in Dx11 would work, so they first release it in these minor mobile cards.
Yeah, it makes me wonder too.. a 1,600 MHz of DDR3 @ 64-bit would make it on the same level as a 800 MHz of DDR2 @ 64-bit. Does the price difference of a 1,600 DDR3 over a DDR2 is really less than a 128-bit PCB over a 64-bit PCB? I think even if it's true, they would only make a very little savings here, maybe in amount of a couple dollars, 5 dollars anyone? :P Well at least they're trying :P
Yep, I'm don't know what they're thinking too.. With DX11 just around the corner, I'm not sure this kind of updates would bring any difference. Not to mention that they had trying to kill support for the DX10.1. So, what's the point? Maybe they gone nuts? LOL :P