Dell Precision M4700 and M6700 - Preliminary Info

Many of the specs for the Kepler Quadro GPUs are out. The K4000M, for instance, has almost three times as many Cuda cores as the 4000M (960 instead of 336, I think).

Apart from the ISV certified drivers and I think more VRAM (maybe double?), the K5000M looks roughly comparable to the GTX 680M (though of course each is designed and marketed for very different purposes).

 

Meaning what exactly? I'm referring to bottle necks along the line. Do we see a meaningful improvement, or is this the equivalent to faster RAM? Have you seen any benchmarks thus far?

I believer that is the K5000M you're referring to?

 

To loosely paraphrase another post from memory (possibly multiple posts - I've read but not bookmarked so many) regarding the CUDA topic - the Kepler architecture means that a direct comparison of core numbers isn't as relevant as the performance underlying it. Something to do with the complexity of fermi (last generation) meaning Kepler's "simpler" architecture isn't as much as it sounds in practice. [this comment comes with a heavy dose of needing confirmation by somebody who genuinely knows their GPUs - and specifically relates to CUDA not overall card performance]

I continue to use Computer Games on Laptop Graphic Cards - Notebookcheck.net Tech as my general point of GPU ranking reference, but their Kxxxx listings are equally preliminary with no benchmarks as yet attached.

 

Read my post last page. The simple answer is Kepler runs the shaders at a lower clockspeed (but has more). The shaders are otherwise identical.

For example, a Fermi with 384 shader @ 1500 Mhz = Kepler 768 shader @ 750 Mhz.

 

@extide

The Fermi and Kepler architectures cannot be compared 1:1 with each other. The shaders in Kepler are less efficient than those in Fermi. However, the entire Kepler package is actually more efficient than the entire Fermi package.

 

Let me put it this way - the new quadro's and firepro's are about 60-100% more powerful than the last gen. One exception is the Kepler's sheer computing power which is not that great compared even to the previous gen.

 

All things being equal?

 

The K5000M is similar to the 680M, which is ~7970M= (the top new FirePro). The lower end FirePro (M4700) = 7870M or ~6970M in terms of power.
Hopefully, it's not too confusing.

 

where did you find the info that the top new fire pro = ~7970m?

 

Are the new mobile firepro released? I can't find any information on AMD webpage.