The better graphics performance by the air is probably due to it running a much lower display resolution than the retina MBP.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Iris vs. HD 5000
- Thread starter ShaneBunting
- Start date
- Sort by reaction score
The better graphics performance by the air is probably due to it running a much lower display resolution than the retina MBP.
In this particular benchmark, the 5000 does come out on top every time. I have run it several times to be sure. This is just one small benchmark, though.
The MBA is definitely showing higher FPS, perhaps it is running at the native resolution of the screen (1440x900 on the MBA vs. 2560x1600 on the rMBP)? Interesting that the 2.6 rMBP shows 2 FPS lower than the 2.4 as well...
That's only 10% increase in GPU, which just does not explain the ridiculous increase as stated above. Right?
Max FPS spiked in that test - probably a glitch. If you look at the average FPS:
rMBP 2.4 = 14.9
rMBP 2.6 = 15.2
That's only a 2% increase for the 2.6 over the 2.4.
rMBP 2.4 = 14.9
rMBP 2.6 = 15.2
That's only a 2% increase for the 2.6 over the 2.4.
Would you be willing to try the "preheated" thermal throttling test I proposed? Launch Unigine Heaven on all 3 machines, let it run for maybe 15-20 minutes, and then start the actual benchmark at that point?
I realize that's kind of a whole procedure, let the machines all benchmark for nearly half an hour, but I think it would produce some interesting data (either the performance will show significant variance or the difference between the Air and rMBP chips isn't as significant as expected, aside from the i5 Air being massively CPU under-clocked).
I might try that later tonight. I can even throw my Mid 2012 rMBP 15" in there just for comparison's sake.
Here is the result of a 'preheated' Heaven benchmark. I ran Heaven for 15 minutes on both the Intel 5000 Macbook Air and the Intel Iris MacBook Pro before kicking off the benchmark. Temperatures registered a pretty constant 78 degrees Celsius on both machines for the duration of the test.
Attachments
The 6 fps increase on min fps seems pretty significant.
Also, would be interested in some real world results... even on windows
And thanks so much for all the tests, they are very helpful.
Also, would be interested in some real world results... even on windows
And thanks so much for all the tests, they are very helpful.
Remember, the rmbp is drawing 2x the pixels as the air. That explains why the results are so similar. It's too bad they couldn't get an iris pro into the 13" rmbp. If you want to get a comparison of the graphics cards without taking screen resolution into account, you'll need to run an off-screen benchmark.
Thanks Raibyn! I think we are getting fairly close to truth here. Basically, for graphically intensive games one could expect around 10% better performance from the rMBP (with CPU stuff taken into account), unless the rendering is bandwidth-limited.
Well, not really. Its still rendering to the same size render target, which then gets upscaled. The upscaling itself is fairly cheap - the framebuffer texture is just around 4MB. But it certainly takes something away. Furthermore, its possible that OS X optimises the whole HiDPI stuff once the application runs in the fullscreen mode - I have no idea. I didn't have the time to do consistent tests yet.
Two issues:
-A chip will exceed TDP if its thermal conditions allow it. So unless the system heats up into the 90's and is unable to bring the temperature down with its fans, the chip will continue to draw more than 15W as long as the software asks for that much power (which games and benchmarks basically always do unless they're frame capped).
What we may actually be seeing is that the cooling on the Air and rMBP 13" isn't as different as the designs would suggest. Personally I wasn't aware until recently that older MBPs (pre-underside vents) actually already drew air through the ports instead of just exchanging at the hinge only, which should also be what the Air does.
-Benchmarks like this basically only tax the GPU, unlike real games where the CPU may also have lots of work to do (handling networking and interpolation/predicton, running AI routines, physics, etc.).
There are some interesting benchmarks out there which show the i7 Air as being in between the 2.6 and 2.8 options on the rMBP, but this only applies to single core performance. When the system tries to load both cores the rMBP CPUs pull ahead, though even the 2.8 only has like 10% on the Air's i7. In full dual core load it basically matches the 13" rMBP's 2.4.
In general it seems like the ULV chips are really impressively efficient. I think Intel's entire CPU lineup this year is being marketed a little deceptively; The Air's i7 "1.7" chip is really very close to the rMBP's i7 "2.8" chip in reality. Of course if we want to talk reality they are physically the same chip with different binning and TDP regulation. We might discover that the ULVs are actually better chips with a major under clock and more aggressive power limits, which is how they produce such amazing battery life (higher binned chips hit the same clocks at lower voltages).
Last edited:
Wanted to bring this up from another thread:
According to notebook check,
the 1.3 i5 MBA (HD 5000) scores a 1,035 on 3DMark11
the 1.7 i7 MBA (HD 5000) scores a 1,070 (Some results on Futuremark's website go up to 1,200)
and the 2.4 i5 rMBP (Iris 5100) scores a 1,164 (Some results on Futuremark's website go up to 1,300)
Even though they're synthetic benchmarks, that's not even a 10% gain in performance. And all of the tests were run at the same res (3dMark11 must have looked awful on the rMBP...)
According to notebook check,
the 1.3 i5 MBA (HD 5000) scores a 1,035 on 3DMark11
the 1.7 i7 MBA (HD 5000) scores a 1,070 (Some results on Futuremark's website go up to 1,200)
and the 2.4 i5 rMBP (Iris 5100) scores a 1,164 (Some results on Futuremark's website go up to 1,300)
Even though they're synthetic benchmarks, that's not even a 10% gain in performance. And all of the tests were run at the same res (3dMark11 must have looked awful on the rMBP...)
Just a thought: A test that would really strain TDP would be any version of Counter-Strike (probably GO since it's a little more aggressive graphically) with bots. This gives the CPU plenty of work to do on top of the actual rendering. Just keep adding bots (bot_add in console) until things start to slow down.
Thank you for clearing this up. I'm still wondering though: if TDP seems to matter so little due to cooling mechanisms, why didn't they ship the rMBP's with ULV chips? They cost the same, but draw less idle power.
Edit: I guess we need a benchmark that stresses both CPU and GPU at the same time, to see the true difference the TDP can make.
Honestly I think they basically did. That's why I said the marketing on Intel's end is deceptive; I'd say those chips in the rMBP 13" idle much lower than their base clock names. Even in Ivy at the desktop level (so double the TDP of 15" rMBP parts and no iGPU in use) my 3770K, which is marketed as a base 3.5 Ghz chip, actually idles well below 3Ghz (I could go check, but I'm not at home). I think it actually hangs out around 2Ghz at idle.
IMO only look at the boost clocks and TDP when looking at Intel chips. So in that context, we have:
Air i5: 15W 2.6
Air i7: 15W 3.3
13" rMBP: 28W 2.9
13" rMBP: 28W 3.1
13" rMBP: 28W 3.3
I think when you look at it like this the performance numbers now basically line up with the stats. The Air's i7 falls just short of the rMBP's i7 on single core (and I mean within a couple percentage points), but loses out by 10% on dual core (the TDP limitation appears).
Problem is if you advertise with these stats only people will be very confused about what's better. Then again, according to the results maybe they should be.
If you think about it, the fairly significant energy savings might actually be worth the slight loss in performance ... if you don't care for the Retina display. Laptopmag did a mini review on the Haswell i7 MBA and said that it actually has a better battery life than the Haswell i5. They said that the higher powered CPU w/ the same TDP was actually more power efficient.
I think you have a point.
I was checking out some reviews on notebookcheck and they basically prove it to some extent (if my interpretation is correct).
Zenbook Infinity (Intel Core i7-4558U 2.8 GHz):
"While idle, the Zenbook UX301 requires approximately 4.5 to 7.5 Watts"
http://www.notebookcheck.net/Review-Asus-Zenbook-Infinity-UX301LA-Ultrabook.103027.0.html
Samsung ATIV Book 9 (Intel Core i5-4200U 1.6 GHz):
"When idling, the 940X3G is satisfied with a low 4.7 to 9.0 watts"
http://www.notebookcheck.net/Review-Samsung-ATIV-Book-9-Plus-940X3G-Ultrabook.103299.0.html
Those laptops are kinda similair in specs (ultrabooks with high-res 13" displays).
Addition from Anandtech, comparing the 1,3 i5 vs the 1,7 i7:
"At idle large parts of the silicon are clock gated if not fully power gated. Idle voltages are extremely low (even compared to what you find in modern smartphones) and both parts run at the same 800MHz frequency at idle, so power consumption is comparable between the two at idle."
http://www.anandtech.com/show/7113/2013-macbook-air-core-i5-4250u-vs-core-i7-4650u/3
So yeah, I guess the "normal" frequencies are just marketing terms.
Last edited:
Shh, don't let the Air i7 zealots hear you say this blasphemy...
----------
Yeah... this was completely disputed by every other reviewer - it doesn't actually make sense.
Shh, don't let the Air i7 zealots hear you say this blasphemy...
Personally - I've got no idea why anyone feels the need for the high-end i7 upgrades for the Haswell offerings... none of them seem to provide enough zing, and the great big benefit of Haswell is battery life. That 15-20% of battery life decrease with non-idle workloads (using a machine means it shouldn't be idle!) that Anand measured on the 4250/4650 comparison is still significant.
Especially if one were to pick the "13 Air - you'd want to max out your battery life if you're not going retina.
Uh-oh, I guess I'm going to get raked with hot coals now.
Thanks for the benchmarks! At the time of purchase of my MBA 11" i5 (which I had to buy from an apple store while visiting the US), I was really debating on whether to go for the 13" MBP for the retina screen and faster graphics. And despite these benchmarks showing that the MBP would indeed be faster, I'm happy that the speed/performance increase isn't so huge such as when comparing against the 15". I'm not a huge PC/Mac gamer but I'm happy that my MBA can handle (to a limited degree) some games such as SC2, D3 etc. even when output to a 1080p monitor.
OK,
so I have a 2.0MHz rMBP15" with Iris Pro and 16GB of RAM. Screen is fabulous on sRGB and AdobeRGB and it handles my Photoshop actions and stuff bazingly fast. In those test there's no comparing to 11's or 13's that can stand so close.
So... what is it with this threat? Actually, the screen of the 11 is not only smaller, it is non-retina and image quality of the 13 (or 15) rMBP is better, period.
Furthermore I 'm a litle behind on updating about multicore use in gaming software engines, but I suppose a decent game would run better in a faster quadcore over a dualcore, or even a slower dualcore, Haswell or not.
So, bottomline: I think the 11" Air is a lightweight longlasting tool to get some things done, and also a basic 'getalong' item. If you have a need (and can pay) for performance: do it, get a 15" rMBP. If you really want longlasting batterylife and low weight (like for backpackers), get an 11" for weight or 13" for longevity). But don't just drop some close benchmarks to justify buying lower end, because that's what you buy: lower end, period again. What is a fact, is that an 11" with pci-e SSD against a basic 13" non retina with HD makes people hesitate (since they both lack the retina screen, and also because the 11" performs very nice but then again the 13" lasts longer). Or did I miss something?
so I have a 2.0MHz rMBP15" with Iris Pro and 16GB of RAM. Screen is fabulous on sRGB and AdobeRGB and it handles my Photoshop actions and stuff bazingly fast. In those test there's no comparing to 11's or 13's that can stand so close.
So... what is it with this threat? Actually, the screen of the 11 is not only smaller, it is non-retina and image quality of the 13 (or 15) rMBP is better, period.
Furthermore I 'm a litle behind on updating about multicore use in gaming software engines, but I suppose a decent game would run better in a faster quadcore over a dualcore, or even a slower dualcore, Haswell or not.
So, bottomline: I think the 11" Air is a lightweight longlasting tool to get some things done, and also a basic 'getalong' item. If you have a need (and can pay) for performance: do it, get a 15" rMBP. If you really want longlasting batterylife and low weight (like for backpackers), get an 11" for weight or 13" for longevity). But don't just drop some close benchmarks to justify buying lower end, because that's what you buy: lower end, period again. What is a fact, is that an 11" with pci-e SSD against a basic 13" non retina with HD makes people hesitate (since they both lack the retina screen, and also because the 11" performs very nice but then again the 13" lasts longer). Or did I miss something?