Asus ROG G55 & G75 in April with gtx 670m

[octiceps]

Well the 580m is 30% faster than the 570M.

You should expect each 600M card to be 30% faster than the one it replaces.

The 670M being 30% faster than the 570M would mean it at least equals the 580M.

LOL did you make this up? Or do you somehow have inside information on stuff that doesn't exist yet? Or are you just trolling?

The last time ASUS updated their ROG notebook line, they came out with a GTX 560M that was 10% to -30% faster than the GTX 460M.

[sarge_]

-30%? 

[octiceps]

-30%? 

Yeah. Negative thirty percent. That's speaking of the nerfed 128-bit GTX 560M of course.

[Kevin]

LOL did you make this up? Or do you somehow have inside information on stuff that doesn't exist yet? Or are you just trolling?

The last time ASUS updated their ROG notebook line, they came out with a GTX 560M that was 10% to -30% faster than the GTX 460M.

You're actually helping to prove my point.

460M -> 560M was merely a transition to a slightly more efficient core, with higher clocks, and even that provided between a 10 and 15 percent increase in stock performance.

The move to 600M, is a die shrink and architectural improvement. The cards will have more cores and higher clocks to go with them.

The 570M has 336 cores. Don't be shocked if the 670M has over 400, with higher base clocks included. It will be 30% faster than the 570M, EASILY. I'm actually being conservative.

[Cloudfire]

I`m with Kevin on this one. You can`t compare going from Fermi to Kepler with 460M to 560M. Not even close

[octiceps]

You're actually helping to prove my point.

460M -> 560M was merely a transition to a slightly more efficient core, with higher clocks, and even that provided between a 10 and 15 percent increase in stock performance.

The move to 600M, is a die shrink and architectural improvement. The cards will have more cores and higher clocks to go with them.

The 570M has 336 cores. Don't be shocked if the 670M has over 400, with higher base clocks included. It will be 30% faster than the 570M, EASILY. I'm actually being conservative.

But what about the memory bus? Everything I've seen points to the GTX 670M as still being a 192-bit GPU. I don't see how it can ever be much faster if the memory bus is holding it back.

But if ASUS breaks the mold and starts going back to putting top-of-the-line mobile GPUs in their ROG laptops and keeping the price competitive, good for them; they'll have a winner on their hands.

[octiceps]

I`m with Kevin on this one. You can`t compare going from Fermi to Kepler with 460M to 560M. Not even close

This image could mean a bunch of different things. Kepler could be considerably faster than Fermi, OR it can be more efficient, OR it can be both, at least according to this chart. You could definitely make the argument, just based on this picture, that Kepler, being a die shrink, is only more efficient by being more powerful per watt and not necessarily more powerful overall. That's basically what Intel is doing in their CPUs with Ivy Bridge -- a die shrink that is provides roughly the same performance as Sandy Bridge while using less power.

[Cloudfire]

Does it really matters how they achieve the increased performance?

Faster clock for clock (like Sandy Bridge was to Nehalem). Im OK with that.
Not much faster clock for clock but due to the die shrink they have much bigger thermal envelope to push the performance forward? Im OK with that too.

But like mentioned, it is a new architecture, not just a die shrink, or small increase in core efficiency like 560M was to 460M. That is the whole point here. To further compare this with CPUs and Intel, it is kinda like Quick sync was to Nehalem architecture. So yeah, I think we are talking big increase in performance with Kepler. We will see anyways "soon"

[Kevin]

But what about the memory bus? Everything I've seen points to the GTX 670M as still being a 192-bit GPU. I don't see how it can ever be much faster if the memory bus is holding it back.

In what situation is the memory bus holding the 570M back? It never hits a point where increasing clocks doesn't increase performance.

OR it can be more efficient, OR it can be both, at least according to this chart. You could definitely make the argument, just based on this picture, that Kepler, being a die shrink, is only more efficient by being more powerful per watt and not necessarily more powerful overall.

More powerful per watt is the absolute best kind of more powerful.

It means, for instance, you can now deliver the power of a 100W GPU in a 75W GPU. Or imagine that the GTX 680M is a 100W GPU, which packs the power of what in 40nm was only possible with a 130W card.

It raises the ceiling on potential horsepower, without also raising the power consumption.

[octiceps]

In what situation is the memory bus holding the 570M back?

Uhhh...in pretty much every modern game when you crank up the resolution, AA, and AF. That's part of the reason why the performance of the GTX 570M is closer to that of the 560M than the 580M.

It never hits a point where increasing clocks doesn't increase performance.

Actually, there is a point. At the point when you can no longer hold a stable overclock and your card starts overheating, throttling, and crashing, your performance is no longer increasing. But that goes for every graphics card.