On that I agree but there is a segment (or was) of middle range iMacs (current 27 inch) that is not pro. I meant Mac mini "pro", not Mac Pro mini
.
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.
New cores (A15)
- Thread starter cmaier
- Start date
- Sort by reaction score
.so are you sure guys we will have just 3 segments for the silicon macs? i mean the Mn series for the macbook air, 24" imac, ipad pro etc, the middle of the road MnX for the 14" 16" macbook pro, bigger imac and the silicon for the mac pro?On that I agree but there is a segment (or was) of middle range iMacs (current 27 inch) that is not pro. I meant Mac mini "pro", not Mac Pro mini
So only 3 segments? i am ok with this too if it will be so but just wondering
i wonder if the mac pro will be very priced configurable with 2x Mn or 3x Mn(for the base prices of 2 possible configurations) , other step up in perf and price - config 2x MnX or 3x Mnx and so on OR will be different SoC's entirely ?!
Mac Pro Starting from $2000 - dual Mn SoC with up to 32 gb Ram, up to 4T ssd etc
$2500 - triple Mn SoC with up to 48 gb ram , 4T ssd etc
Mac Pro - $2999dual MnX SoC with up to 64 gb Ram, 8T ssd etc
- $3500 triple MnX SoC with up to 128 gb ram and so on
is this possible ? to be so config from a cheap price starting and supporting up to over $5000 based on the SoC and number of them?
Last edited:
I believe there will be only two chips/platforms: vanilla M-series for low-power consumer Macs and a scalable professional chip for everything else. Bigger Macs will use multiple scalable chips to build more powerful systems.
I am very skeptical about Apple changing their price structure this much. The Mac Pro will most likely remain at close to $5000 as it is now. But we might get a smaller new intermediate model.
So, on balance, I think you are probably right.
But... playing devil's advocate for a bit, doing multiple chips in a single system is potentially quite hard. Apple will have multiple SoCs with a lot of stuff doubled up for no obvious benefit.
So I still wonder whether it would be cheaper/easier for Apple to have a 'big M' chip that adds extra cores.
A three chip line-up could have some merits in having one model that is 'desktop' oriented and not engineered with power-constraints in mind.
I think the barrier to this is economics - I have read that it costs $1bn to tape out a 7nm chip (maybe the same for 5N?), but if it's just the 'same but different', perhaps it gets to a place that makes sense? The GPU makers seem to have halo models that can't sell in the millions, yet which seem to be profitable, so it must be at least possible.
It doesn’t cost a billion to tape out a 7nm chip. Not sure what is supposed to be included in that billion dollar figure, but, just, nope.
An inefficient team designing one of these, and taking a year to do it, might have, say, 300 people on it. If each one gets $600,000 in salary and benefits and overhead costs (they don’t), that’s only $180M. In reality, a variation of an existing SoC takes a lot fewer than 300 people a lot less than a year. (And the cost of each employee, on average, is a lot less than $600 grand)
I didn‘t mean multiple SoCs, I meant a single system on a package, stitched together from multiple „tiles“ aka. chiplets. As to redundant components (SSD controller, display engine etc.), they could be either off-loaded to a separate chip (on the same package of course), or just disabled. It’s not like they are taking much space. Or maybe, multiple display engines can be used to enable more displays etc.
The problem with big chips is that they are getting progressively more and more expensive, not just because of the size, but also because it gets very hard to make a defect-free chip. Smaller chips simply have higher yields. At the current level of complexity, it really starts to be one a problem. That’s why industry is moving towards scalable technologies, where you make small chips and connect the, together to act as one big chip. AMD is a success story in this area, Intel is introducing their scalable tech next year, and it will probably come to GPUs sooner rather than later too.
Apple is the owner of some very advanced chiplet-related patents, they have a focused product line and a lot of experience with economies of scale (just look at M1, it’s a marvelous product). Chiplets would be exactly their style as they would allow them to offer multiple configurations at various performance levels at a very reasonable amortized R&D cost. And of course, the leaks point towards this as well.
I had a meeting with those guys. A lot of people I used to know are there, or are involved in advisory or funding capacity. I wish them luck.
Curious why you think that? Hardware improvements look to be slowing down as we hit really small lithography limits and transistor counts.
The biggest changes will most likely come from more feature specific silicon and improved algorithms (ML, cryptography, etc).
A couple of things coming that do look pretty awesome though:
- Wi-fi 7 up to 30Gbps (and wi-fi 6e now even looks capable of near 2Gbps)
- AV1 decoding should hopefully keep cpu utilization low for streaming
I really just want a cool/quiet machine. Soooooo tired of laptops burning my lap and fans spinning up.
I agree with him, actually. Compare the period from the late ’70s to ’80s to what happened when the world settled on Wintel. Everything stagnated - everything became faster, better, more capable, but it was all still really the same thing over and over again.
Page 4 - Question - Apple A15 announced
Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
forums.anandtech.com
Write up from the first benchmark video (Chinese) came out.
Page 4 - Question - Apple A15 announced
Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Write up from the first benchmark video (Chinese) came out.
Hm, what’s the difference to A14 then? Weird…
They usually don’t announce any events until two weeks in advance. And last year’s event was end of October or into November iirc.
I expect interesting things to come out of advanced packaging more than lithography advances. Of course, they’ll both contribute. And there are interesting possibilities in GAA MBCFET, but from a laymans perspective the cool structures also look non-trivial to mass produce. Really non-trivial. So maybe holding ones breath is not recommended. :-D
Rumour is 12th October event… the day after the education “back to” offer ends… like last year…
the date is always coming around 7 days earlier to the event
So, keep an eye after first of October, around 4 or 5th of October
Tell that to cerebras
More like tell that to Cerebras's customers.
"... Current customers of CS-2 units include national laboratories, supercomputing centers, pharmacology, biotechnology, the military, and other intelligence services. At cost of several million each, it’s a large bite to take all at once, hence the announcement today. ...
.... however cloud rental costs at Cirrascale will run to $60k a week, or $180k a month,...."
Cerebras In The Cloud: Get Your Wafer Scale in an Instance
Their WSE chips aren't defect free. They build in extra redundant components on the wafer so can they work around the majority of the defects. But that just folds back into the chips being more expensive.
Apple would probably prefer to make $60K per day selling six $10K systems, rather than shrink the volume as low as Cerebras does.
As long as Apple can cover the whole laptop line up with monolithic dies ( that use less power) they will probably will be willing to grow as big as necessary to do that. The desktops that are "left over" at this point , ( iMac large screen , Mac Pro , current case Mini ) , could have more "give" on power consumption and Apple switch over to. "cheaper for us". If they can cover the desktop line-up with some small changes to a "big" laptop chip then can mean more saved costs for Apple.
Yeah this is basically what you would expect from upclocked firestorms like in the M1 - in some ways even worse as where did the power savings/extra performance from the new node go? They were expected to be small but where are they?
I should state I’m not against them taking a year off in core designs - neither Intel nor AMD else releases brand new cores every year (well Intel starting to because reportedly they have such a huge backlog of designs from their manufacturing woes). CPU designers deserve a good work-life balance too.
Well let’s see Andrei’s results when he gets them @anandtech.
Edit: low power mode is more efficient but why are there no savings anywhere else except when it’s downclocked?
Last edited:
1. Better energy profile when not single thread drag racing for bragging rights.
( vast majority of folks would rather have a phone with enough charge to last a busy day that is not primarily occupied with gaming )
2. Much higher synergy with cameras on most of the systems with A15's inside them.
( e.g., ProRes video content creation. )
3. Better ML/AI. ( which these benchmarks are almost completely opaque to. )
4. Fairly decent chance has higher synergy with VR goggles at the end of 2022 for ones with 5th GPU.
I think he meant the CPU cores so it should read how is avalanche different from firestorm? Which means the other points don’t really apply and 1) is only true *in low power mode* which most people don’t keep their phones in constantly. Otherwise, the default is that it goes faster for more power and for pretty much the same power increase you’d expect from an upclocked firestorm core on N5.
Edit: isn’t this GB MT score for the A15 lower than what we were seeing previously? Wasn’t it closer to a 18-21% increase compared to the A14? I mean I expect variation between models and units and even runs but even so … that’s not insignificant.
Last edited:
I meant the CPU cores, sorry. So far it looks just like Firestorm at 3.2ghz. Why claim new cores if you just ship old cores at higher clock?