Sept 2025 Apple iPhone Event

[rdrr]

I was pretty jazzed on a 17 Air - thinness being key. But until I better understand the camera compromises (testing via credible reviews and sources) over my 16PM, I'm holding off for now. And will continue to be annoyed putting my silicone-cased iPhone in my shirt pocket a few dozen times a day.

I would highly recommend the total AppleCare+ (theft and loss) if you pull the trigger. I guarantee any "repair" for a cracked screen is a total replacement.

 

[Citysnaps]

Eh... I just read that the camera in the iPhone 17 Air can't record photos in RAW format.

As much as I want a thinner iPhone, the above makes the 17 Air a pass for me.

 

[Jimmyjames]

Possible Geekerwan result!

iPhone18,2 - Geekbench

Benchmark results for an iPhone18,2 with an ARM processor.

browser.geekbench.com

 

[dada_dave]

Possible Geekerwan result!

iPhone18,2 - Geekbench

Benchmark results for an iPhone18,2 with an ARM processor.

browser.geekbench.com

View attachment 36552

What do you think? Liquid nitrogen cooled? Or just a freezer?

 

[Jimmyjames]

What do you think? Liquid nitrogen cooled? Or just a freezer?

Why not both?

 

[Alli]

Best preorder experience yet. At 7 I opened the app and got the message to stay right there. By 7:02, my phone was ordered with 9/25 delivery date and I was going back too sleep.

 

[rdrr]

Best preorder experience yet. At 7 I opened the app and got the message to stay right there. By 7:02, my phone was ordered with 9/25 delivery date and I was going back too sleep.

So yeah, my sleepy head couldn't do math this morning. Woke up at 4:30 AM, couldn't possibly doom scroll so early, so I opened the store app. Got the message they appreciated my early bird attitude, but come back at 5:00 AM PST. I didn't check back until 9:00 AM, DOH! Oct 6-13th for me. Oh well, happy birthday to me.

 

[Jimmyjames]

Geekbench AI GPU score for the 17 Pro Max.

iPhone18,2 - Geekbench

Benchmark results for an iPhone18,2 with an ARM processor.

browser.geekbench.com

 

[Alli]

And just like that my order will now be arriving on the 19th. Happy birthday to us both, @rdrr !

 

[leman]

will Metal 4 take advantage of the new AI cores in A19? Or is it just for Core ML

There are metal APIs that can target the MXU hardware. They have been present for several years, but Metal4 introduced new APIs with more features and easier programmability. We’d need GB6 to be updated to use these new APIs to see if additional uplifts are possible.

 

[leman]

Well, I we t hiking in deep Sardinian mountains early in the morning. By the end of 12 hours treck (there were times I’d though I’d die) the delivery times slipped and now I’m looking at mid-October

 

[diamond.g]

My order changed from 9/25 to 9/19.

 

[DT]

My order changed from 9/25 to 9/19.

Ours too!

 

[Roller]

I was pretty jazzed on a 17 Air - thinness being key. But until I better understand the camera compromises (testing via credible reviews and sources) over my 16PM, I'm holding off for now. And will continue to be annoyed putting my silicone-cased iPhone in my shirt pocket a few dozen times a day.

Have you tried cases that are a bit less grippy? Mt case is ribbed and doesn’t slide too easily but isn’t hard to put into or take out of my pocket.

 

[Herdfan]

I would highly recommend the total AppleCare+ (theft and loss) if you pull the trigger. I guarantee any "repair" for a cracked screen is a total replacement.

I'm sure it is. But I had a scuff right in the middle of my screen on my 15 Pro and when I went to have it replaced by Assurion (through Verizon's protection plan) they told me they couldn't guarantee it still being waterproof. So I traded it in on a 16 Pro. No way is that acceptable.

 

[Citysnaps]

Have you tried cases that are a bit less grippy? Mt case is ribbed and doesn’t slide too easily but isn’t hard to put into or take out of my pocket.

I haven't. I went with a grippy case on my 16PM as a deterrent for getting my phone jacked out of my pocket while walking in downtown San Francisco making photos.

A few times I've seen people walking with phones in shirt pockets and having a criminal quickly running by in the opposite direction, snatching their phone without stopping, and keep on running.

With the shirts I like to wear (Dickies long and short sleeve T-shirts), getting it out at an angle (from someone running by) is pretty much impossible. And it gives me a second or two to repel that - a feature made possible by the silicone case. Slowly pulling it straight up is no problem. It's the thickness (not the width) of the 16PM that I fight with as the pockets are dead flat.

I know the above sounds a little bizarre, But it is what it is. And I just don't like my phone in my jeans pocket (habit, I guess).

 

[leman]

I was tagged into this. It purports to show a 2.8x increase in FP16, 3.05x increase on int8 and a 60% uplift on FP32 on Geekbench AI between the 16 Plus and the Air. I'm a little suspicious giiven the version mismatch (1.4 vs 1.5). The notes say Windows on Arm cannot be compared between versions. There is a comparison between the 16 Pro Max and the Air which is only 2x for FP16, but the Pro Max has an extra core. Hmmm

View attachment 36542

I only see 2x increase in the benchmarks I’ve seen, consistent with the 2x FP16 pipe. The a18 Pro GPU half-precision score is around 7500 in pretty much every result I’ve seen ob GB6 ML

 

[Jimmyjames]

I only see 2x increase in the benchmarks I’ve seen, consistent with the 2x FP16 pipe. The a18 Pro GPU half-precision score is around 7500 in pretty much every result I’ve seen ob GB6 ML

True, this was a comparison between the base A18 and the binned A19 Pro, which was less than ideal.

So are you saying these scores are not due to the matmul cores but just the doubling of FP16 as stated? In other words, we can expect higher performance for software that uses the matmul cores?

 

[Altaic]

So are you saying these scores are not due to the matmul cores but just the doubling of FP16 as stated? In other words, we can expect higher performance for software that uses the matmul cores?

I suspect that’s what we’re seeing. Given that the Metal tensor ops weren’t really useful before, it’s unlikely that GB currently uses them.

 

[leman]

So are you saying these scores are not due to the matmul cores but just the doubling of FP16 as stated? In other words, we can expect higher performance for software that uses the matmul cores?

I wouldnt know. All I’m saying is that the speed up in this test is closer to 2x.

I suspect that’s what we’re seeing. Given that the Metal tensor ops weren’t really useful before, it’s unlikely that GB currently uses them.

They were quite useful as it’s impossible to achieve perfect ALU utilization for matrix multiplication without using the matmul intrinsics. So id be surprised if GB6 does not use them.