Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
WilliamG,
I don't know why are you complaining and what did you expect from Apple for this update. I knew it'll get 780m (with 4gb vRAM, hallelujah!), Haswell CPUs, 802.11ac and pcie-based flash (FD included) for several months now, simply because there's no other hardware Apple could use for it.
 
I would love to see proof of this on the 4GB GPU 2013 iMac vs last year's 2GB GPU iMac. I await benchmarks!

Simply stated:

The AMD HD 7950, GeForce GTX 680MX and GeForce GTX 780M seem quite similar in performance overall.

The 7950 has 3 GB VRAM, the 680MX 2 GB and the 780M has 4 GB.

In most games (if not, all, except X-Plane 10) the performance figures of the 3 grfx cards will probably be quite similar (give or take 10% here and there).

It's just that X-Plane 10, with extra highly detailed scenery add-ons installed, uses more VRAM than the 680MX can offer.
This really is true.

Attached are 2 sreenshots of X-Plane 10 running on my current system (Mac Pro '08, 8 x 2.8, 16 GB RAM, Radeon 7950, SSD boot)

One screenshot is the scenery of New York City loaded, the other are my grfx settings including the amount of VRAM consumed (it is now 2587 MB, i.e 2.5 GB) once NYC is loaded.
 

Attachments

  • NY-2.jpg
    NY-2.jpg
    238.3 KB · Views: 375
  • NY-1.jpg
    NY-1.jpg
    724 KB · Views: 476
Simply stated:

The AMD HD 7950, GeForce GTX 680MX and GeForce GTX 780M seem quite similar in performance overall.

The 7950 has 3 GB VRAM, the 680MX 2 GB and the 780M has 4 GB.

In most games (if not, all, except X-Plane 10) the performance figures of the 3 grfx cards will probably be quite similar (give or take 10% here and there).

It's just that X-Plane 10, with extra highly detailed scenery add-ons installed, uses more VRAM than the 680MX can offer.
This really is true.

Attached are 2 sreenshots of X-Plane 10 running on my current system (Mac Pro '08, 8 x 2.8, 16 GB RAM, Radeon 7950, SSD boot)

One screenshot is the scenery of New York City loaded, the other are my grfx settings including the amount of VRAM consumed (it is now 2587 MB, i.e 2.5 GB) once NYC is loaded.

I don't doubt those screenshots. What I doubt is any tangible frame-rate difference between that and the same/similar card with more RAM. Really, in the case of the iMac.

We'll just have to wait and see, though, what people report.
 
I don't doubt those screenshots. What I doubt is any tangible frame-rate difference between that and the same/similar card with more RAM. Really, in the case of the iMac.

We'll just have to wait and see, though, what people report.

If a game using the selected settings wants more VRAM than the card has on board, then, of course, the FPS will become a slide show, i.e. around 1 - 2 FPS.

So, in my example above, the iMac with GTX 680MX (with 2 GB VRAM) cannot handle it.
If the same GTX 680MX had 3 or more GB of VRAM then it wouldn't be a problem. I'm sure of that!

IMHO, the 780M is 7 - 10% faster than the 680MX (due to the slightly higher clock). The 4 GB just gives X-Plane 10 breathing space the 680MX cannot give.

All other games, AFAIK, don't consume so much VRAM, so the two Macs will score pretty identical (just 7 - 10% advantage for the 2013 iMac)
 
If a game using the selected settings wants more VRAM than the card has on board, then, of course, the FPS will become a slide show, i.e. around 1 - 2 FPS.

So, in my example above, the iMac with GTX 680MX (with 2 GB VRAM) cannot handle it.
If the same GTX 680MX had 3 or more GB of VRAM then it wouldn't be a problem. I'm sure of that!

IMHO, the 780M is 7 - 10% faster than the 680MX (due to the slightly higher clock). The 4 GB just gives X-Plane 10 breathing space the 680MX cannot give.

All other games, AFAIK, don't consume so much VRAM, so the two Macs will score pretty identical (just 7 - 10% advantage for the 2013 iMac)

I really do understand the theory behind more VRAM. I'm still not convinced it's anything more than marketing in these iMacs. Again, only benchmarks will prove it either way.
 
I really do understand the theory behind more VRAM. I'm still not convinced it's anything more than marketing in these iMacs. Again, only benchmarks will prove it either way.

It is not just marketing if the game actually consumes the VRAM... and clearly X-Plane 10 does that!

I agree that other games probably don't. So, if you're a non-X-Plane 10 gamer on a Mac, then there is no benefit in 4 GB VRAM.
 
It is not just marketing if the game actually consumes the VRAM... and clearly X-Plane 10 does that!

I agree that other games probably don't. So, if you're a non-X-Plane 10 gamer on a Mac, then there is no benefit in 4 GB VRAM.

Bioshock Infinite also consumes more VRAM than 1GB, but as I recall the 2GB cards didn't perform much/any better.

Again, we're going in circles. I wanna see benches! :D
 
Some fact regarding BF4 beta..with i5 3.2GHZ 680MX i get on 1440p+high settings around 44fps and under i5 3.4Ghz 780M i get at the same settings and same map around 52-54fps (same nvidia driver beta)
 
Some fact regarding BF4 beta..with i5 3.2GHZ 680MX i get on 1440p+high settings around 44fps and under i5 3.4Ghz 780M i get at the same settings and same map around 52-54fps (same nvidia driver beta)

Game runs pretty badly on the iMacs, sadly. I can't get 60fps at 1080p with all the settings on LOW (newest BETA). Not really sure what's going on. (I have the i7 680MX)
 
BF4 recommended specs are 3gb vram, so 780m should perform better in that game than 680MX. Also it's optimized for i7 so I can't wait to find out how much difference the i5 vs i7 makes.
 
Some fact regarding BF4 beta..with i5 3.2GHZ 680MX i get on 1440p+high settings around 44fps and under i5 3.4Ghz 780M i get at the same settings and same map around 52-54fps (same nvidia driver beta)

Honestly that has nothing to do with the iMacs or their GPU. Its become well documented in only 3 days (if you work for EA, Dice or Origin you probably wished you'd booked you holiday for October) that BF4 has some serious issues with certain hardware at present. That will be rectified. They can't afford not to.

The fact is its a beta and cannot be used to represent the difference or similarity in GPU performance for any machine, mac or pc. We need to see solid benchmarks on established released games and for me on hi-res triple screen gaming as the 4gb of the 780m should hopefully give a nice boost.

Its the 1st time I've thought of upgrading my 2011 6970m 1gb & my bank manager looks worried :)
 
thanks very much guys

..for all the answers and help I got from you guys. I'll stay with all I got. Very satisfied with this iMac model.
 
There isn't a difference between each gfx. If you want a 780M then flash the 680MX to one.
 
It is not just marketing if the game actually consumes the VRAM... and clearly X-Plane 10 does that!

This is more likely inefficient texture management by the engine though then anything else. There is no reason for a flight simulator to consume so much memory. A 2048x2048 texture with a full mipmap chain is 32Mb (less when compressed - and you SHOULD compress textures), a GB of VRAM could easily fit 20 of those (which should be MORE then enough for caching) in addition to the framebuffer and all the other stuff.
 
This is more likely inefficient texture management by the engine though then anything else. There is no reason for a flight simulator to consume so much memory. A 2048x2048 texture with a full mipmap chain is 32Mb (less when compressed - and you SHOULD compress textures), a GB of VRAM could easily fit 20 of those (which should be MORE then enough for caching) in addition to the framebuffer and all the other stuff.

I'm not an expert on this but I do read the X-plane.org forums a lot.

I trust that Laminar Research know what they're doing.

I understand that because of the amount of scenery that has to be drawn is so huge in a flight sim (you can see for miles and miles around) that VRAM gets consumed so easily. Lots of scenery gets cached.
Flying over Manhattan when you have detailed scenery installed at high resolution surely gives your grfx card a lot to draw.
Take a look at my screen shots.
 
I'm not an expert on this but I do read the X-plane.org forums a lot.

I trust that Laminar Research know what they're doing.

I understand that because of the amount of scenery that has to be drawn is so huge in a flight sim (you can see for miles and miles around) that VRAM gets consumed so easily. Lots of scenery gets cached.
Flying over Manhattan when you have detailed scenery installed at high resolution surely gives your grfx card a lot to draw.
Take a look at my screen shots.

I wouldn't call myself an expert, but I do have some experience with graphics programming. The screenshot you show illustrates how the engine manages the LOD (look how the upper fifth of the image becomes blurry). The high detailed parts of the image is middle and bottom. For a 1920x1200 display, this level of quality is easily supplied by 512x512 textures (which are all 1MB before compression). Assuming that the rendered scene utilises 36 (6x6) tiles, its well under 40MB of VRAM (under 60 if you include bump maps). What indeed would use lots of VRAM is the highest-level of detail, but you only need to load it if you are really close to the ground, basically, if all other tiles are not or just barely visible. Let's, for the sake of the argument, assume that you need the highest level of detail when your plane is below 5km. Assuming cruising altitude of 8000 km, even if you direct your plane straight down to the ground at mach 1, your engine has 8 seconds to stream the texture data to the card. This is actually enough to open the file, load the texture, decompress it on the CPU and upload it to the GPU. This way or another, there is no reason why the scene you post should use up over 2GB of VRAM, because no texture level over 512x512 is even being utilised there! The only explanation I can find is that engine programmers are 'lazy' and upload all texture levels simultaneously, instead of streaming the highest level of detail dynamically as required. I mean, ID software managed to utilise HUGE textures for their games, without using up too much VRAM and maintaining very good performance levels - because of smart engine programming.
 

Attachments

  • FPS.jpg
    FPS.jpg
    39.2 KB · Views: 224
Last edited:
*cough*

He didn't use grfx that uses more than 2 GB VRAM... trust me...

See: https://forums.macrumors.com/posts/18166620/

EDIT:
This test does not help us find out what happens if VRAM exceeds 2 GB. That really, really, really is important.

I don't trust anything except benchmarks between two VERY similar cards (VRAM aside). Comparing an AMD card and an Nvidia card isn't telling me much. Heck, everything is different in that link. Different CPU, RAM, GPU etc. Couldn't get more different if it tried!

I'll go with there being basically no tangible difference unless proven otherwise. I have no idea what settings were being used on the actual benchmark, but it seems to me to be a realistic result, no?
 
wow I'm surprised the 780m is faster than the 680 mx nvidia did a really nice job this time.

even so no worth upgrading
 
Settings comparison 7950 - 780M

In my earlier post I made a few screenshots on my Mac Pro with 7950.

Now attached similar screenshots made with my maxed out Haswell iMac.
Checkout the differences.
 

Attachments

  • 3.jpg
    3.jpg
    520 KB · Views: 230
  • 4.jpg
    4.jpg
    782.6 KB · Views: 247
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.