Don’t really have an opinion here but you said ’Amiga’ AND referenced the bridgeboard so you should get some positive feedback.
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.
Aspects where Intel Macs still outperform the M1 models?
- Thread starter jent
- Start date
-
- Tags
- apple silicon intel macs
- Sort by reaction score
Don’t really have an opinion here but you said ’Amiga’ AND referenced the bridgeboard so you should get some positive feedback.
Apple is getting closer to that ideal with the M1's
As for Performance/watt, I couldn't possibly care less about that. (just my opinion!)
You might not, but most users do. The market has shifted toward laptops, which are Apple's best selling machines. Even in a desktop, many users like to have machines that can run quietly, especially those doing AV work.
It becomes a thing anywhere you're considering density of packed components. Prior generations of Apple hardware, including imacs, macbooks, and the cylindrical mac pro would encounter throttling due to high internal temperatures, in addition to significant fan noise at times. That in itself is enough to make performance per watt a significant factor across basically all Macs sold today.
Are you positive they care about that, or is it only your opinion? To me, people buy laptops because you can move them around a lot easier and they have a built in battery backup. And that is just my opinion.
Anyway, I'm not totally watt unconscious,
, my car is a Prius Prime which is the most power efficient car you can buy for most uses, and I got it because I can actually see the difference in fuel costs. With a PC, I don't -- at all. My power bill doesn't go down when I replace a desktop with a modern laptop, nor does it go up when I put in a new desktop. The power differences really aren't that great. Where I work only has about 30 PCs, and one machine in the mill uses far more power than those PC's, and we have a lot of mill machines!I agree though for a server farm, that would make a difference to them!
And fwiw, I'm the guy that buys those PC's for where I work and I've backed off a little on laptops for now, Lithium ion battery failure rates have gone up, and they're a fire hazard when they fail. I can't wait for better batteries. (which should be soon)
So put in better cooling -- it's easy on the desktop side. (not so much for laptops)
Apple has constructed these things to be as thin and compact as possible for a large number of years. Do you really think they're going to change their design strategy to favor more active cooling?
Are you positive they care about that, or is it only your opinion? To me, people buy laptops because you can move them around a lot easier and they have a built in battery backup. And that is just my opinion.
Anyway, I'm not totally watt unconscious,
I agree though for a server farm, that would make a difference to them!
And fwiw, I'm the guy that buys those PC's for where I work and I've backed off a little on laptops for now, Lithium ion battery failure rates have gone up, and they're a fire hazard when they fail. I can't wait for better batteries. (which should be soon)
There have been a ton of heat threads in the MacBook Pro forum.
You have cooling costs as well, especially during the summer or in areas where there is no air conditioning and ambient temperatures are in the 80s and 90s. Or if you're working in your car on a really humid day.
We replaced 95% of our light bulbs around 2011 with LEDs and the I did see the impact in our electricity bill. I was a bit surprised at that as I took lighting for granted.
Higher performance per watt = higher performance in the same form factor. Notice how Apple didn’t downgrade the thermals on their entry level machines even though they could have, given M1 energy efficiency. M1 laptops at 15W currently deliver performance levels more closely associated with 40-50W laptops in the x86 world. Now think what this means for larger Macs.
I care about performance a lot, which is why I think perf per watt of these chips is a huge deal. Looking forward to an iMac outperforming large workstations or a 16” MBP rivaling gaming PC towers. Already the meager M1 can trade blows with fastest x86 CPUs in peak single core workloads... at 5watts!
I was speaking of desktops mainly, but YES, I want better cooling, that's where you'll get the best performance and machine lifetime! Whether they do or not is their problem, though I bet the current mac pros have pretty good cooling!
Single core loads??? Not interested with a modern OS. I have 10 cores on my workstation -- plus hyperthreading for 20 logical cores. And it cost only a little more than my M1 MBA (16G, 1TB).
We wont know until we see them. There's a big bottleneck with the M1 (the single die basically) I'm very interested in how Apple solves that problem, but I bet it wont be as power efficient.
There's no doubt it matters to some, especially as you scale up the number of machines, but not with just 25 or 30 regular business machines.
Those bulbs you replaced saved more power than the difference between PC stuff. Those old incandescents were terribly inefficient.
Single-core performance is a good predictor for multi-core performance with these chips, they can scale almost linearly exactly thanks to their low power consumption (where Intel or AMD have to cut the clock in half to thermally manage multiple cores).
All large workstation CPUs are essentially single dies... you can buy chips with 80 CPU cores on one. There is a reason why multi-socket systems are dying out in workstation space: single socket is faster, more affordable and more efficient. And when you look at large workstation CPUs from AMD and Intel it’s easy to see that the size of the CPU cluster does not diminish efficiency. Apple currently leads in this area by a factor of 3-4x, their multicore designs will benefit from this advantage the most. A cluster of 16 Firestorm cores will draw around 60-80 watts at full load with a performance rivaling that of a 200+W 24+ core Xeon.
The single biggest bottleneck of M1 is memory bandwidth, but than again this is pretty much a solved problem.
Anyway, you are right that we will only know for sure when we see the final product. But based on what we know now - and there are certainly some big unknowns - I just don’t see any reason to be skeptical about the scalability of Apple CPU cores. The picture is really very clear here.
Multiple threads running on different cores do not act together synchronously. They maintain their own stacks and do not typically write to the same locations in memory (if they do, you run the risk of false sharing). As a result, you care about how well each core performs. Aside from that, very recent versions of Clang and gcc have decent auto-vectorization capabilities, which can provide some level of single threaded synchronous parallelism via the use of vector/simd instructions and loop unrolling to hide instruction latency.
I'm not going to tell you that multi-threading is bad, just that it requires more intervention to use effectively on the part of developers, effort which they may not cram in to every part of their applications.
Totally disagree, single core perf means nothing to me, we have multithreaded everything these days. Even if the M1 had a single core performance increase over my i9 (I don't think it does), it only has 4 high performance cores, I have 10+. They just really don't compare in any way. As for the scale linearly -- that doesn't make any sense at all -- we don't know squat about scaling on an M1 because there's only 1 version right now, and it only has 4 performance cores and 4 low power cores. (now having 4 low power cores -- that's VERY innovative, but it can't compare with bigger processors as is.) The M1 cuts the clock on the 4 high performance cores in half when pushed to the limit heat wise, just like other processors. I've seen it every time I try to run more than one VM on it.
True, but the RAM isn't there, the disks aren't there, the GPU isn't there if you have a discrete GPU. (They, of course still have the built in GPU, but if it's not being used...)
No.
The trick is we don't know how they'll connect them. Having all the RAM on a massively parallel Mx processor has it's own problems. Anyway, we'll just have to wait and see given the published information. Anyway, I'm just as interested to see what comes next as you are!
All I see are problems, but we wont know until we know.
I'm really not trying to say intel, or AMD is better, heavens no, they have their own set of problems, big ones. I just got in this discussion because of the performance/watt part, like I said watts are a small time concern for me and the M1 using less just doesn't ring my bell any.
Of course, that's the benefit of them! How many PC's or Macs do you know that has thrown out the OS and just has a single threaded program running on the hardware?
Then there's the other apps, and the OS itself. Mulithreading is not only not bad, it's the only way to get decent performance. (except for some very specialized tasks running on hardware we're not talking about. Like ASIC's doing mining...)
Before this, I was referring to multi-threaded workflows. In such workflows, your cpu cores are generally saturated by independent tasks, associated with a small number of problems. Running an OS and various networks connections typically doesn't saturate cpu throughput. Those tasks are more frequently bound by various forms of IO and do not necessarily scale well with a high number of physical cores, assuming a single OS rather than multiple VMs.
It depends on what you're doing.
My workload spreads across my 8 cores/16 threads nicely. One program is display 88 real-time charts. The other is displaying 15 real-time charts with several studies per chart. I assume that it just runs each chart in a thread. I also have about ten tabs in Firefox and you can specify the number of processes/subprocesses that Firefox will use - up to 8. Most of those tabs are on auto-reload. All of these threads aren't CPU intensive so they don't get anywhere near stressing a process or thread but they do spread out nicely.
I don't understand your point here. That's the way computers run these days and good for that to me. Computers aren't a game to get the best theoretical usage, it's to get the job done, and they're pretty good at that -- at least ever since we went from single threaded, interrupt driven, state machines. And yes, that means intel processors do it pretty well, they get the job done at a reasonable cost. Saturating cpu throughput is not something that's a good thing, it's a hard limit that one should strive not to do.