Separate names with a comma.
Discussion in 'iMac' started by barefeats, Oct 18, 2013.
(includes Mac Pro with GTX 680 for perspective)
Very nice, I've been looking for benchmarks like this everywhere.
Thanks for your work on this!
Great link thanks
I got the Haswell i7 with 780M now!
Some insights regarding X-Plane 10 if you're interested.
I play X-Plane 10 heavily: high settings, lots and lost of very highly detailed scenery (pay ware and freeware) with many, many highly detail aircraft.
I went from a Mac Pro '08, 8 x 2.8 GHz, 16 GB RAM, Radeon HD 7950 3 GB VRAM (Mac edition) with SSD boot disk (but X-Plane on a 7200 rpm HDD) to a iMac 27" i7 Haswell 3.5, 32 GB RAM, GeForce MX780M 4 GB VRAM with X-Plane on the SSD boot disk (1 TB).
Both running GM of Mavericks.
Points of interest purely regarding X-Plane 10:
• As the setup with the huge amount of scenery easily exceeds 2 GB VRA, having 3 GB VRAM or more is really a necessity. Game over for the Mac Edition of the 680 (and 680 MX).
• The system uses > 12 GB of RAM total when flying over huge scenery with many AI aircraft and other plugins.
• SSD helps loading-time (a lot!), but does nothing regarding FPS.
All settings were experimented with using the Mac Pro's 1920 x 1200 settings.
My findings in the game:
- In daylight rendering the iMac with 780M is 2 x as fast the Mac Pro with 7950.
It already seemed that the CPU / BUS / RAM speed was bottleneck regarding X-Plane on the ol' Mac Pro. The iMac seems to justify that.
- In night-time rendering, where other lighting needs to be rendered the iMac is slower by about 30%!
With daytime rendering "on" I can crank up the settings much higher than the Mac Pro could, and get better FPS!
With night-time rendering.... well, I avoid that now.
So realistically, at best the 780m is about 10 fps better than the 680mx, and the 680mx is pulling already very good frame rates.
Thank you for your results. It doesn't tell us much of anything, really, as you have a COMPLETELY different system - a Mac Pro! Different CPU, different RAM etc etc. So much to be different that qualifying the benchmark results is impossible. Still fun to see, of course..
I'd say the difference is to 680MX is only slightly better than neglible. In the 2-3 cases 780M pulls ahead, the overall score doesn't quite fit with the GTX680s performance, which is in these benchmarks quite a bit lower than the 680MX. This doesn't seem right...
Anyways, the real performance of the 680MX is unlocked when overclocking. I wonder how well the 780M overclocks.
Do you think the i5 780M combo would give you much less performance in X-Plane compared to the i7?
That's a tough one...
On my maxed out iMac the GPU is the limiting factor, wheres on my old Mac Pro the CPU was.
I think the i5 will give slightly less performance then the i7. Not much at all.
Thanks , but I've seen this thread. There's no proper overclocking tests there yet (really pushing the card), but I'm sure they'll come eventually.
While the 780m is only 5-10% ahead of the 680mx, there's a couple things to remember -
1. These are mostly games that don't take advantage of the extra 2gb of VRAM. In games that do, I'd expect the 780 to have an increased advantage.
2. these tests are all done on the basic bootcamp drivers, at least I assume so. As far as I know, the new nvidia drivers for the 780m aren't working on the iMac due to some glitch (unless this has since been fixed).
With newer and better drivers the 780 will probably pull ahead a few more %.
On my comfiguration crysis 3 with everything set to maximum, and filters at 2x (smaa and anisotropic) is perfect! Off course at 2550x1440
Yep, I'm waiting for some OC'd FPS differences.
At what FPS?
I don't know how many frame, but it's perfect. I've activate the vsync too.
How could i see the frame rate in crysis 3?
Ask for specific games/benchmarks in the thread, otherwise I can only guess. I'm running 150/225 (450 effective) core/mem with good results, and posted a 3DMark bench (before/after) in there too. I might try my luck for a faster overclock, but without being able to increase the voltage it may not be stable.
"Real" games are harder to benchmark unless they have decent benchmarking tools. BioShock Infinite has one, but I can't seem to find anything for Dishonored (lest I just bench the opening cutscenes) or Tomb Raider (that doesn't require FRAPS). If you have word otherwise, link it up in the overclocking thread.
CoD: BO2 has an option to show FPS in game. I'm interested in the performance boost OC'd in BF3 and CoD: BO2.
Well, it would be interesting to see some tests in 3dmark 11, as that was a the main benchmark test program for the 680MX overclocking thread.
I game at 250/375 with 100% stability. 225/350 is a also a good overclock speed as it produce slightly less heat (still in the best performance vs. heat zone for my 680MX card).
Again, if you use the i7 CPU, you would benefit from lower temps by disabling the CPU turbo boost. The 780M draws less power than the 680MX, so you could potentially overclock quite a bit with decent temperatures (lubbo's fan control is recommended though, setting the fan at a static speed, like 2300-2500rpm, not much higher though.)
PS: You should NOT touch the voltage (I never did), and you should monitor the temperatures carefully. I did get issues with my PSU (random reboots) when gaming at 285/425. I turned off CPU turbo boost, and then my iMac became 100% stable. I still went down to 250/375 just in case..never push the computer to the very limit as I'm sure you know... Also, no reason to push the video ram much higher than what I do (vs. gpu) in my oc's, as I could not see any real performance increase - only more heat.
Way too subjective. You need a repeatable benchmark: just running around has too many variables.
I'll run some '11 tests (was just running the '13 tests).
Although the 780M has a lower TDP I'm not sure how well it will maintain a stable core at (say) 225/350. The links in the other thread indicate that some (Windows laptop 780M) users had to slightly push the voltage up to remain stable. I'm not keen on doing that whatsoever.
I'm using a decent fan control that ramps up based on temp: I have it set to start spinning up up over 70C and to max out at 85C. Works well and stops it from going crazy when not in-game.
Sounds great, thanks! Yeah, one should never change voltage settings in a closed system like the iMac in my opinion. Of course, my wish was to see how high you can overclock the 780M being 100% stable and not getting too hot. That would be the real comparison between 780M and 680MX, as I always overclock at 250/375 when playing demanding games, and the difference in performance from stock speed is quite big.
Download FRAPS and run it while you play.
Is there any benchmark / test comparing different configuration iMacs 2012 vs 2013 in both sizes?
I'd like to know how does a 2012 21.5" w/650MX compare with a 2013 one in several games at native 1920x1080, as well as how much better is a 27" with 680MX/780 running at native and 1920x1080....
3D Mark (vanilla, 2013 version) Fire Strike scores:
Fan spins to max (2400RPM) at 81C thanks to my fan speed curve. Temp never gets past 81C even at the faster clocks. Haven't noticed any artefacts so far. I don't see much of a point increasing the memory clock based on reports from others: the gain is minimal and often introduces artefacts.
Might try a run at 250/450 and see how it goes. Still running the stock Bootcamp 311.71 drivers while I wait for the 331.58 WHQL drivers to go up (the desktop ones were just released).
3D Mark '11
200/225 (450): 7988 (max 86C)
Judging by the results others have posted it looks like I could do better: this guy hit 8500 with a 130/450 overclock using 327.23 WHQL. I'm probably going to keep it hovering around 180/450 for now until I get the 331.58 drivers on this successfully and will re-run it then.
For reference, back in February I benchmarked my 2012 iMac, i7 3.4Ghz, Geforce GTX 680MX, and got 8646 in 3D Mark 2011 with an overclock of 250/375.
Thanks for the data. I ran some more tests just then on the new 331.58 drivers and had much better results when I pushed the memory down a little:
180/450: 7079 (max 81C using 331.58 - this seems like an outlier)
200/400: 8364 (max 73C using 331.58)
Will run a test at 230/370 and see how we go. That extra 50MHz on the RAM seems to have made a difference on the core temp (bus speeds, perhaps?) for some reason. Notably it's a slightly cooler morning here as well, so that might be helping (it's hard to control the variables!)
230/370: 8518 (max 76C using 331.58) - that's 1005/5376 effective clocks.