Not that I disagree, but the iMac Pro was announced at WWDC 2017 and went on sale as a late 2017 model in December right?
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.
Apple Registers Several New Mac and iPad Models in Eurasia
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not that I disagree, but the iMac Pro was announced at WWDC 2017 and went on sale as a late 2017 model in December right?
Sure, who cares when their business excel calculation has an error? Who cares when their scientific calculations have an error? Who cares when their taxes are off? Who cares about their data being corrupted?
Correctness should be king, and it's an abysmal state of affairs that it isn't. Just like it's abysmal that APFS checksums the meta data, but not the actual data. It cares about itself more than your it does about your data.
Well Mac sales keep climbing and PC sales (as an aggregate) keep falling, so maybe it's all just people new to computers migrating from smartphones and tablets (most likely running iOS).
Which of course is why XCode will never be ported to Windows.
Just introduce a MBP 13" with a new keyboard and I'll be the first one in line. Some extra ports next to the thunderbolt 3 ports would be grate but not a must. I'm waiting for over a year to upgrade my MB Air to a new Pro model, if this year introduction has no better keyboard I'll just have to go for it.
You’re correct yet didn’t ship until 2018 first week of January for those the pre ordered. Also wasn’t in stores until last week of January in te USA.
We've now had donglegate for 2 years, and it shows no signs to slowing down. How long do you think it's going to last?
The USB-C docks are about as expensive as the USB-A docks, and offer the same level of functionality, too.
The problem with TB3 is precisely that it tries to be everything, so that means it's got to be a super powerful AND numerious port. This is very expensive to make. It's awesome, but expensive, and because it's so expensive it just won't get adopted, because the vast majority of consumers just cannot afford it OR they need to go donglegate.
As a software engineer working with CAD tools and scientific computing, I absolutely could not care less about ECC memory. Please give me one non-server use-case where ECC is important. Maybe if your calculations take days I suppose? But in those circumstances I'd offload it to the cloud or a supercomputer, which does use ECC memory, and is a server.
Yes, it does depend on the workload a little, but even then...
A 8750H in a Windows laptop scores 5037 SC, 21582 MC in geekbench, while an iMac Pro scores 5009 SC and 30541 MC - clearly higher on the MC front - however a 2018 Aero 15X with a 6 core + GTX 1070 Max-Q will export 4K video ~15% faster than the baseline iMac Pro in Premiere Pro according to Dave2D. Also the fact that such a cheap laptop can even approach the iMac Pro, a super expensive desktop, really says everything that needs to be said.
I think we agree that everyone wants more cores. The core i9 exists for that purpose. It's significantly cheaper and just as fast.
Xeon is for you if you want ECC memory or more than 16 cores. There is no other reason to get it.
It is if I want an iMac release this year with a decently modern chipset and graphics that don't suck.
If the iMac had been upgraded to the 8000 series CPU's and VEGA GPU's, I'd be a lot more lenient on Apple - but they haven't done that.
[doublepost=1531173650][/doublepost]
The chances of your Excel calculation going wrong because of a bit-flip is basically 0. Not only does the data need to corrupt in the first place, which is very rare, but it then also needs to corrupt in such a way that the data is changed but the system and code is not and remains stable. The chance of this is so low I can't even tell you what it is, but we're definitely going to need scientific notation. xD
For longer running jobs like scientific calculations and tax calculations, both of which I have worked with, are offloaded to a server and done there, into which you do use ECC.
The problem with this absolutist view is that it just isn't practical. ECC memory and Xeon chips are very expensive, and they save you from an extremely, extremely small amount of pain in a workstation. Doing checksums on the actual files being written is certainly a good idea, but it NEEDS a co-processor. The main CPU should not be tasked with thiis, as it would radically slow down the computer. I'm not against the idea though - it's actually a pretty good idea. Certainly much more useful than ECC memory for the average computer user.
As long as obstinate people remain obstinate? I replaced all my USB-A to Lightning / USB-A / USB-B / MicroUSB / MiniUSB cables with USB-C versions and now everything works with USB-C just as it did with USB-A. Cost me literally peanuts via Monoprice and Amazon Basics.
Ironically, the only dongle I have is the Apple Thunderbolt 2 to 3 adapter to plug my TB2 Drobo 5D in because nobody makes a TB2 to TB3 cable.
They need to go to DongleGate with USB-C, too. My company is a mix of Dells (all with USB-C in addition to USB-A) and MacBook Pro 13s and 15s. Some of the Dells (the 7300 series) only have one USB-A and two USB-C ports (and recharge via USB-C like the MBPs) and no wired Ethernet plus only micro-HDMI for video output so they need docks or dongles just like the MBPs.
And you don't have to use expensive Thunderbolt peripherals. You can use vastly cheaper USB-C peripherals in the same port. So while you are paying more for the port, you do not have to pay more for what you plug into the port unless you want the performance / flexibility boost TB offers over USB.
Premiere Pro leverages nVidia CUDA so the GTX 1070 is why it can export 4K video so quickly, not the i7-8750H. Pair an i7-8750H with an AMD Vega 56/64 and the performance should drop a fair bit I would imagine.
Of course, the iMac Pro isn't really aimed at Premiere Pro for video editing as much as it is Final Cut Pro, which can leverage the AMD's GPU architecture as it's optimized for it.
Last edited:
So if anyone says that something is better than a Mac (or even hopes the features of that other thing come to a Mac) they have to immediately leave the forum and never mention it to other users
talking about the low level of gratitudeExample: Your a PhD student writing your dissertation. You've spent hundreds of hours editing it in Word. During this time, the text has of course been sitting in RAM. Unbeknownst to you, a single bit error flips a "<" into a ">". This dramatically changes the meaning of your dissertation. Nobody catches the error - they just assume you're arguing the opposite case. Only years later is the erroneous assertion disproven.
ECC memory shouldn't cost a lot - just 12.5% more for a "9th" RAM chip. AMD includes ECC functionality on every Ryzen CPU. It is just Intel that makes it available only as a premium feature.
Any self-respecting PhD student should be writing their thesis in LaTeX.
The chances of a random bit flip resulting in this happening is well beyond negligible. Add to this, there would be software level checks to make sure this would never happen. You can what hardware ECC does in software, its just going to be imperceptibly slower for this kind of stuff.
[doublepost=1531196364][/doublepost]
Also, I don't think you can do this with just a 9th ram chip. If you find an error, which bit in the incorrect byte is wrong? I can see it working with two extra modules, that way you can checksum both horizontally and vertically across blocks of 8 bytes. (Assuming only one error per 8 byte block)
Of course, I could just wiki how ECC works, but I'm too lazy.
Last edited:
Exactly.
And not only does Premiere Pro utilize CUDA... the April 2018 update to Premiere Pro added the capability of using the unused integrated Intel GPU that is built into the i7-8750H.
So you can use the processor's 6 cores and 12 threads... the Intel iGPU... and the NVidia GTX 1070.
Windows video editing in Premiere just got a lot more interesting!
Hell... even on my 4th-gen Haswell desktop... this Adobe update drastically improved my Premiere Pro experience. Whereas simply scrubbing the timeline used to make my computer stutter and my fans spin up... now it's buttery smooth. And H.264 MP4 exports can be up to 30% faster depending on the file. And this is an "old" processor!
Imagine what a modern processor will do!
For once... an Adobe update made a real difference!
Last edited:
IMHO as an outsider, I don't think they have thickness/weight requirements other than "not significantly worse than the previous generation". They chose parts based on power requirements (including runtime of 10h for certain tasks), then made the smallest case those parts fit in.
That said, the rumors were that they were originally shooting for a 100Wh battery in the current 15 by using the MacBook battery tech, and for some reason pulled it late in the design process. I've been curious if this refresh also boosts the total battery size (even if that increases weight).
I also have personally speculated that they want to move more functionality into the ARM chip for "Power Nap", so that they can hibernate the machine to allow them to use the more power-hungry RAM tech intel currently mandates for hitting 32 GB. However, even if that is a feature they are planning to roll out this refresh, I don't think that is why they are a little behind their annual refresh target. Instead, I think that is related to controlling messaging for the keyboard replacement program. If thats the case, these machines are probably ready to go any Tuesday now.
I think you're right, however it's not as simple as 'making the smallest case that fits'. If the case was bigger, there would be more room for better cooling, which would mean even the same internal hardware would perform better because of Turbo Boost. So even the choice to make the case 'as small as possible' is a choice to not have the computer run as fast as it could, or last as long as it could (heat degradation).
Darn, I hope so. The battery size in the modern MBPs are pretty pitiful compared to what they used to be in W.hr terms, even ignoring the weight benefits that have come from advancing technology.
Although I'm a fan of using hibernate to avoid high standby power draw of DDR4 RAM, I'm not sure if Apple would be willing to go this way because of the time taken to wake from sleep. Maybe if Apple can put a faster SSD in somehow (although its already very fast) loading 32GB of RAM backup into RAM would take like 10 seconds. Of course, that should only happen if the RAM is filled to the brim with open documents, and the MBP has been left sleeping off battery power for days, so not a common occurrence.
I hope they do this though.
Way back in my Mac 128K/512K/Plus days, I believe I saw two random memory flips. In the first, the screen should have been all black, yet one white pixel appeared. In the second, the program had zeroed out a block of memory and yet there was a bit set in the block. A random memory error, or a software bug that poked a random memory location?
To fix single bit errors and detect double bit errors in a 16 bit word requires 6 extra bits. For a 32 bit word, the same requires 7 extra bits. For a 64 bit word, 8 bits are needed. This is why one can readily buy 72 bit wide DIMMs.
(I'm wondering when ECC correction will be built into DRAM chips: The extra bits required to ECC a 1024 bit row are negligible. The only real issue is speed - the required XOR tree is very deep.)
Or you can just not do that and have all the cables for free bundled with the phone or accessory?
But you see, you're evading the point. You paid an $800 premium for those ports, and then you paid for all the cables as well. For those that don't want to pay such a colossal premium, they don't get to use TB3 or even USB-C. And the thing is - most people don't.
In fairness you can just use regular USB-3.1 in the USB-C format, but these ports are really, really rare on desktops and laptops.
Correct, you can plug cheap stuff into your expensive port - but if that's what you're going to do, why get the expensive port?
My point is TB3 needs to about the same price as USB 3.1 before it receives any adoption whatsoever. Right now it's used for things that nothing else can do, such as eGPU's, and that's it.
Intel also needs to allow AMD and ARM and others to use it. Before that happens it'll never be a standard.
And you may very well sell a lot of donglegate Dells - I wouldn't know about those, I'm not interested.
Well it'll drop by ~25% at most. If you genuinely believe that getting a desktop that costs over twice as much for a 20% improvement is worth it, you're an idiot. I know that's very harsh, but it's the reality. There are much better options for you.
If you do it because you prefer Final Cut Pro and MacOS, however, you're certainly not an idiot. I can respect that, 100%. But think about how much better Final Cut Pro and MacOS could be if Apple actually made decent computers again. (Speaking strictly about the hardware side here)
I suppose there's a reason the Hackintosh community has grown to such an enormous size.
This exactly is utterly ridiculous. The chance of a bitflip happening at all is extremely small. The chance of that hitting your dissertation is dozens of orders of magnitude smaller. The chance of you not catching it during proofreading, and then also having the logic flipped in your head when you have to defend it instead of just being able to say "wait wtf? How did that get in there?" after that is 0. Your example, or anything like it, has LITERALLY NEVER EVER HAPPENED. It's about as likely as you spontaneously teleporting outside your window and falling down because of quantum mechanics - it can happen, but the possibility is so remote it'd be stupid to consider it.
And by the way, ECC can fail too. The chance is just lower.
Well Intel has made the license open so if AMD and ARM want to adopt it, they now can.
As I understand it based on various posts in other threads, AMD's CPUs offer more PCIe lanes than their Intel equivalents so it may not be as much of a problem.
I think most bit flips are caused by cosmic rays (apart from actual hardware faults, and I'm sure there are also other ways too) Some quick googling suggested that bit flip errors are on the order of 3 per year, per memory stick (when used 24x7x365.25). So for your old computer with 1 million bits of RAM, each bit is physically large, and easy to be hit by a speeding helium nucleus. In a modern computer where we have RAM sticks with 128 billion bits, the chances of that cosmic ray hitting a bit that actually matters is orders of magnitude lower. Also, there's also a lot more spare RAM for software based checksums.
Can you explain how this works? I'm not convinced these numbers are right.
At the same time, the 128 billion are each much smaller, easier to flip.
With recent MacOS'es using every spare bit of RAM to cache files, the chance that there's something important in any particular RAM location has gone way up.
Time to go down to the dungeon, er, archives... (creak)(cough)(cough)(grrr)(ouch!)(git!)(poke) 1981 Texas Instruments TTL Data Book (slam)(whew). The 74LS630 is an Error Detection And Correction chip for 16 bit memory. It uses 6 check bits. Each is generated by computing the parity of a carefully selected subset of 8 data bits. Each of the 16 data bits is used in three check bits.
To verify memory read back data, newly computed check bits are compared with the retrieved bits. If three don't match, then one data bit has flipped (the inverted bit affects three check bits). If one doesn't match, then that one check bit was flipped. Other mismatches mean a more complex error.
For today's 64 bit memory, you'd take the parity of 8 subsets of 32 bits, with each data bit appearing in 4 parity trees.
To be convinced that 8 bits is enough to protect 64: How many bits wide is a number selecting one of the 72 bits? The answer is 7. Then add an extra bit so that two bit errors can be detected.
Aye, this is true.
This could have the quite ironic effect of substantially boosting AMD.
But I guess that's not all the filed numbers, is it? And BTW, how many filed numbers are still left? I mean, do the new releases today count as 4 models, as 3 models, or as 2 models? If the new MBPs are 2 models, then we have a real surprise here, as everybody thought that the 3 last numbers were MBPs, but in this case it would be the opposite: the 2 first numbers were MBPs, and we would have 3 new Macs awaiting their release...
Well we do have Kuo's remarks of a refreshed Mac Mini and a new MacBook Air / larger MacBook on the way.
Which I guess are also going to drop via mid-week PR like the new MBP did this morning because why have an event for them and not include the new MBPs?