Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Nobody owns a 680MX yet so the benchmarks haven't been run. However, based on available information, notebookcheck.net rates it approximately equal to the 660ti.

http://www.notebookcheck.net/NVIDIA-GeForce-GTX-680MX.83519.0.html

If you had to decide between 660M and 675MX what would you go for mate?

You probably know what sort of user I am from our board interactions in the run-up to the iMac release. Generally do some light video editing (but not too heavy as the longest rendering run I have had to do was c8 hours on a 2007 MacBook Pro), use multiple applications at simulaneously (some of them being large excel spreadsheets), don't need bootcamp, play some games in a small way (but I am not an avid gamer per se) like Civilization V etc.
 
Last edited:

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
If you had to decide between 660M and 675MX what would you go for mate?

You probably know what sort of user I am from our board interactions in the run-up to the iMac release. Generally do some light video editing (but not too heavey as the longest rendering run I have had to do was c8 hours on a 2007 MacBook Pro), use multiple applications at simulaneously (some of them being large excel spreadsheets), don't need bootcamp, play some games in a small way (but I am not an avid gamer per se) like Civilization V etc.

Unfortunately this isn't a straightforward question. The 660M isn't sufficiently better than the 650M to account for the difference in native resolution between the 27" and 21". On balance, in live usage the 660M model will perform somewhere between the two 21" models. For rendering the 660M would be more than sufficient, and better than the 21", but if you plan to run any games at all (and Civ V likes to eat GPUs for breakfast for no apparent reason) then I'd have to advise for the 675M.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Unfortunately this isn't a straightforward question. The 660M isn't sufficiently better than the 650M to account for the difference in native resolution between the 27" and 21". On balance, in live usage the 660M model will perform somewhere between the two 21" models. For rendering the 660M would be more than sufficient, and better than the 21", but if you plan to run any games at all (and Civ V likes to eat GPUs for breakfast for no apparent reason) then I'd have to advise for the 675M.

Thanks for that. I hear you re. increased GPU burden to drive the larger display. On a clarification, you mean 675MX though, right? Or is it one and the same 675M and 675MX
 
Last edited:

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
Thanks for that. I hear you re. increased GPU burden to drive the larger display. On a clarification, you mean 675MX though, right? Or is it one and the same 675M and 675MX

My finger missed the X key :)

The 675M is a Fermi chip (last year's technology) and isn't getting anywhere near Apple's computers.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
My finger missed the X key :)

The 675M is a Fermi chip (last year's technology) and isn't getting anywhere near Apple's computers.

Thank goodness for that! One final question (you should be charging for this but I haven't found anyone on these boards as knowledgable on GPUs as you are), can you clarify something for me. I have been told that the i7 processor (with its hyperthreading) also interects more efficiently with the 1GB of the 675MX compared to the 512MB of the 660M. Or is that just sales pitch for the supremely gallible? Just don't know how to go about checking this out...

PS. I do realise hyperthreading is less than desirable for the serious gamer so you probably haven't given it much thought...
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
Thank goodness for that! One final question (you should be charging for this but I haven't found anyone on these boards as knowledgable on GPUs as you are), can you clarify something for me. I have been told that the i7 processor (with its hyperthreading) also interects more efficiently with the 1GB of the 675MX compared to the 512MB of the 660M. Or is that just sales pitch for the supremely gallible? Just don't know how to go about checking this out...

PS. I do realise hyperthreading is less than desirable for the serious gamer so you probably haven't given it much thought...

That sounds like someone speaking out of their rear. The CPU never touches the VRAM at all; that's the GPU's responsibility. Unless you do high-intensity sound processing or crazy serious photo editing, avoid the i7. I'm getting the 2GB 680MX and the i5.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
That sounds like someone speaking out of their rear. The CPU never touches the VRAM at all; that's the GPU's responsibility. Unless you do high-intensity sound processing or crazy serious photo editing, avoid the i7. I'm getting the 2GB 680MX and the i5.

Thank you
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Unfortunately this isn't a straightforward question. The 660M isn't sufficiently better than the 650M to account for the difference in native resolution between the 27" and 21". On balance, in live usage the 660M model will perform somewhere between the two 21" models. For rendering the 660M would be more than sufficient, and better than the 21", but if you plan to run any games at all (and Civ V likes to eat GPUs for breakfast for no apparent reason) then I'd have to advise for the 675M.

Seen this thread on 680MX reviews? It's even got me thinking about changing my order...

https://forums.macrumors.com/threads/1506807/
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
Seen this thread on 680MX reviews? It's even got me thinking about changing my order...

https://forums.macrumors.com/threads/1506807/

Yeah, I was just nerdgasming in there. And these aren't even the controlled environment / paid versions of the benchmarks yet.

The 675MX is a fine card and will serve the occasional 3D gamer pretty well at least for the next 18 months or so. The 680MX is a work of art.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Yeah, I was just nerdgasming in there. And these aren't even the controlled environment / paid versions of the benchmarks yet.

The 675MX is a fine card and will serve the occasional 3D gamer pretty well at least for the next 18 months or so. The 680MX is a work of art.

When you put it like that (I intend to keep this machine for 4-5 years if it lasts), I should really change my order to what I want and 'wear' the delay in shipping...it's holidays anyway and unlikely to get mine before 2 Jan anyway, so may as well receive it on 9th or 10th and get what I want.

Question I am still stalling on is HEAT generation differential between 75 and 80 and how is that going to affect the life of the machine.

Or do you think the offset in stepping down to the i5 would compensate?
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
When you put it like that (I intend to keep this machine for 4-5 years if it lasts), I should really change my order to what I want and 'wear' the delay in shipping...it's holidays anyway and unlikely to get mine before 2 Jan anyway, so may as well receive it on 9th or 10th and get what I want.

Question I am still stalling on is HEAT generation differential between 75 and 80 and how is that going to affect the life of the machine.

Or do you think the offset in stepping down to the i5 would compensate?

Preliminary reports coming out of the 21" is that it has a much cooler heat profile. And really, the primary benefit of Ivy Bridge and a key benefit of Kepler is smaller die size, and therefore lower heat output. I wouldn't worry about it too much. AppleCare should definitely be on the menu, and by the time it expires you'll be past the point where today's cards can play then-modern games at high settings anyway.
 

Lasweetlife

macrumors newbie
Dec 13, 2012
15
0
A better point of comparison, at least going by the benchmarks, is the regular GTX 660 (not the ti edition).
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Unashamedly ripping from a post that 42 posted elsewhere which gives some benchmarking against the 660Ti which is the title of this thread.

Some benchmarks appear here

http://www.notebookcheck.net/Mobile-Graphics-Cards-Benchmark-List.844.0.html


Would be interested in people's thoughts please...


>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
Official benchmarks for the 680MX are starting to be posted on notebookcheck.net. From what they have so far:
~60% faster than the 675MX
~10% faster than the 680M
Between ~10% faster and ~25% slower than the desktop GTX 660 Ti, but this varies wildly by benchmark -- there are cases where the 680MX outperforms the 660 Ti.
 

Mac32

Suspended
Nov 20, 2010
1,263
454
Since 680MX is a downclocked desktop GTX680, the performance is a bit disappointing all thing considered. Hopefully newer drivers will help, but then again maybe not, since it's not a new chip, just downclocked...?
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
Since 680MX is a downclocked desktop GTX680, the performance is a bit disappointing all thing considered. Hopefully newer drivers will help, but then again maybe not, since it's not a new chip, just downclocked...?

Drivers do make a difference.. some folks have been reporting ~7% improvement from switching from the Bootcamp driver to Nvidia's beta driver.

We can't expect the 680MX to catch the 680, but catching the 660 Ti is within reach. Notably, all 3 cards are the same basic chip (GK104), just with varying #'s of cores and clock speeds. The 680MX has about 10% more cores and about a 15% slower clock than the 660 Ti. Makes sense, then, that the chips would be comparable but not equal in every task.

Things I learned today: 670's are made from chips rejected for 680's, and 660 Ti's are made from chips rejected for 670's.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Things I learned today: 670's are made from chips rejected for 680's, and 660 Ti's are made from chips rejected for 670's.

Mind boggles. Presumably marginal decisions and not necessarily significantly inferior? But don't know enough about the manufacturing process to make a comment
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
Mind boggles. Presumably marginal decisions and not necessarily significantly inferior? But don't know enough about the manufacturing process to make a comment

Basically, they make the chip once, and load it up with decreasingly stressful firmware until it passes a test, and that determined whether the chip sells for $500, $400, or $300. It helps them increase yield rates by reusing otherwise defective chips.
 

Confuzzzed

macrumors 68000
Aug 7, 2011
1,630
0
Liverpool, UK
Basically, they make the chip once, and load it up with decreasingly stressful firmware until it passes a test, and that determined whether the chip sells for $500, $400, or $300. It helps them increase yield rates by reusing otherwise defective chips.

So I theory as drivers get more optimised for overclocking and software developers make these units work harder we can perhaps expect the 680s to outlive the 670s and 660ti units? All else being equal of course (like miles on the clock)
 

forty2j

macrumors 68030
Jul 11, 2008
2,585
2
NJ
So I theory as drivers get more optimised for overclocking and software developers make these units work harder we can perhaps expect the 680s to outlive the 670s and 660ti units? All else being equal of course (like miles on the clock)

Not necessarily. One of the shader cores may have been misprinted to make the chip unusable as a 680 but qualify as a 670. That doesn't mean the problem will cascade to other sections of the chip. The firmware will disable the defective section, and the rest will continue as normal.

Depending on the degree of defects, sometimes you can squeeze a broader range of power out of the lower chips. If one "just missed" being a 680, it will be sold as a 670 but you can probably rev it up close to 680 performance. On the other hand, it may have just barely passed the 670 test and crap out if you try to rev it at all. You have no way of knowing when you buy it - all you know is that it is promised to perform within stated specifications. And again, this doesn't suggest anything about the future failure rate when used to specification. (The drivers won't overlock the chip; you need to do that yourself with special utilities.)
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.