I hope you're not including me in that group of guys because I haven't taken a position one way or the other.Well, if you will take into account that 16 core Zen2 AM4 CPU will use the same amount of power, yes, it will run circles around 9900K.
Also, you guys are comparing Threadripper CPU vs 9900K. And you forget than IPC of Zen2 CPUs will be higher than Zen/Zen+, and clocks of Zen 2 AM4 CPUs will be on the same level as Coffee Lake CPUs. So Single Threaded performance will be in worst case scenario - the same as Coffee Lake. Scores for Cinebench in this picture will only be higher. 3400-3500 Pts, potentially for top-end SKU.
Yes, guys. This is running Circles around 9900K.
Imagine what 64 Core Rome is capable of. Imagine what Mac Pro using 64 Core AMD CPU could be capable of.
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.
MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
I'm always suspicious of "the future processor from XXX will be so much faster than today's processor from ZZZ" arguments.
Especially when XXX == AMD.
But at least the silly nonsense of talking about housefires has stopped.
Last edited:
PS = Post Script
After that, it is additional "p"s, not "s"s...
Post Script
Post Post Script, not Post Script Script...!
Etc. ...
Sure, sure, sure.
But, did you understand it? Or, you being a grammar popo?
Post Script Script or Post Post Script....
I think you can write it either way and the meaning would be the same.
Last edited:
Not to be all grammar popo-ish, but no it’s not the same at all. Just like random letters usually don’t make a work, random words don’t usually make sentences.
PPS is correct. PSS is not correct.
yeah, yeah, yeah. PPS is the correct usage and thanks to u and boil for being popo-ish about it. Now, I know and knowing is half the battle.
I still like PSS better tho than PPS, TBH with you.
Let me explain...
PS then S is not as random as you say. If, you think about it, it makes more senSss than the proper usage.
First off, the "S" already connects to another grammar rule regarding plurals. So, the "plural" of PS is PSS....
Also, for the casual reader, seeing "PS" then "PSS" would be easier to read than "PS," then "PPS" IMO because a casual reader might not know what "PS" stands for in the first place and the two "P's" might confuse them.
Whereas, visually, "PSS"---by the simple addition of "S's" just means "more."
CapiSsssh?
PS--Also, "PSS," sounds like "Pssst" as if one is literally saying "Pssst, there's more."
PSS--And, if you keep adding "S's" the sound of "Psst" such as "Pssssssssssst," then one can see that one is short and the other is a longer sssssound. Whereas, "PPS" and/or "PPPS" and then "PPPPPPPPS...." The "Ppppppppppp" sound is hard to make and pronounce in one's mind when one is reading or seeing it than the "Ssssss" sSound.
Last edited:
Will you guys hush up and get back to speculating about a possible nMP
nano Mac Pro...?!?
Noooo....!!!
AMD is in the business of making custom wares...
What if Apple contracts them for a special Threadripper-esque package...
Which would contain four 8C/16T 7nm chiplets to one side of the 14nm I/O die & a 7nm Radeon VII die to the other side...?!?
The world's most powerful APU ever...!
Last edited:
Threadripper and upcoming Zen 2, IMO, is not that great though for Apple to suddenly hop on board simply for core-count. If, Apple for some reason thinks core-count is somehow the magic formula, then they would make the move. But, it puts Apple in a position of willy-nilly syndrome, fair weather team company, the other side is greener mentality.
And, tbh, my main point about all of my recent post about AMD in Macs is that the other side is not that much greener. So, again, why?
You keep talking about 7nm, chiplets, more cores, IPC improvement, etc.... And, mind you, this is from a catching-up perspective. So, let's be optimistic and give AMD the BOD and say Zen 2 IPC is now around 200-ish in CB single core score. Do you think Apple is going to be so impressed with that that they would switch to AMD because AMD can now get Kaby Lake-like IPC?
And, do you think Apple is so impressed with the 7nm and chiplet design AMD is proposing that they would do it, simply to get a Radeon VII and an 8-core/16-core CPU on the same die?
I don't think that's that impressive nor enough to see a giant company fair weather it.
But, do you know who can fair weather it? You!!! You can actually get yourself a Zen 2 PC when it comes out.
Or, do you somehow want Apple-look-design, MacOS and AMD CPU together in an official package?
I mean, what can't a Windows 10 machine or even Linux with Zen 2 can't do that a Mac with Zen 2 can do?
Last edited:
Well, the next MP will need to be about regaining a customer base .
Price, compatibility, expandability, adequate performance is where it's at .
A for innovation and cutting edge, Apple doesn't do that, they are not and never have been in that niche .
Apart from packacking , of course .
Oh, taking a swipe at the whole "can't innovate my a**" thing, huh?
iPhone was pretty innovative. It spurred a whole generation and an industry and culture unlike anything ever seen since the invention, i mean, errrr.... innovation.... errr, someone stumbled over a carriage, some round looking things that resemble wheels and an engine sometime ago....
And, why do other company's laptops look like wannabe Macbook pro's?
And, wtf is this?
Or, this?
Or, this?
Last edited:
More like one of the biggest CPU Multichp Module (MCM) ever. The Radeon VII also needs 4 HBMv2 RAM stacks on the same interposer. They could put the I/O and CPU on another MCM/interpose and then attached the GPU interposer to the same mechanism , but in the very large module territory. Two large heat sources physically tightly coupled together with substantially different thermal output with 5 more of a different TDP magnitude also thrown in the mix.
It is probably not an accident that in the iMac Pro that the CPU is moved off the centerline ( and unfortunately away from the RAM door that can be hidden by the pedestal arm ) and into the space where the HDD was. The physical gap between CPU and GPU in an iMac Pro is substantively longer than in a mainstream iMac. That's likely present because Apple is trying to avoid an issue if they were much closer.
Squeezing all of that into a single MCM package is an even bigger issue. There are ways to unwind somewhat from that but it is a bigger problem, but those have their own complications.
A GPU scaled down so its TDP was in roughly the same range ( +/- 15% ) as the CPU chiplet would work much better. Also shrink the HBMv2 memory so that it was more a large L4 cache.
The other major problem is whether Apple is really going to contract for it or not. I can see Apple asking for something they effectively won't pay for ( volumes of Mac Pro or even MP + iMPro not being all that relatively high). Also tossing pragmatically same GPU at each Mac Pro user will run into similar issues that making everyone by two did. ( iMac Pro labors under that issue somewhat, but gets by. Having two system with very similar boat anchors isn't going to help Apple get a broader market reach. 0
Vega GPUs are not designed to be used in Chiplet based designs. Currently people already have found 7 nm APUs and they are monolithic, and ... Vega, not Navi, based.
There is however second APU coming. Based on Zen2, and Navi. Its codename is Dali. And it is replacement for 35W desktop Athlons. That design has 4C/8T, and 10 CU GPU. There is possibility that it is Chiplet based.
Every chiplet based APU will be based on Navi, and future GPU designs, not Vega. So if Apple will want Chiplet based Semi-Custom design, which is most likely, it would be based on Zen2+Navi chiplet designs.
The context of the question was how much of an outcry if it was like the Corsair One which can't take 3.5" drives. Yes the Corsair model has more than the Mac Pro 2013 , but can go back only 10-15 pages in this thread where there were folks who insisted impeding doom of they couldn't put all four of their 3.5" drives from they Mac Pro into a new. That four HDDs was a minimal amount that should be offered and that Apple should do more. ( often since had shoveled 2.5" drives into the 5.25" bay of the Mac Pro).
If pushing the notion of lowest $/GB 3.5" HDDs then already at the point Corsair One isn't enough. Hence, if Apple did it folks would complain.
Personally, I think that just getting Apple past the notion that a single drive is sufficient would be a breakthrough. Multiple SSDs would be a big step forward from where they have been at for the last 2-3 years. 2.5" maybe because Apple could just offer lower $/GB SSDs as BTO options and still be fully invested on their "solid state storage is the future internally" campaign. An empty 2.5" drive bay would give folks the options of putting a HDD in it, but skeptical that Apple is going to stock those as components or try to sell a substantial number of them (i.e., put multiple drive bays in the device primarily to sell more to a few folks. ).
Apple's Time capsule back-up with HDD? Gone. Mac Mini with HDD? Gone. Late derivation on iMac ... iMac Pro? HDD-less. No laptops. Apple foot dragging on putting Fusion Drive feature into APFS. ....
APFS wasn't optimized for HDDs. Are folks going to complain about that. Already have and probably will continue to. However, that probably will have substantive influence over the next Mac Pro design baseline that Apple sets up for the next 5-10 years.
Not particularly. Intel's iGPUs started out in the extra mostly "unused"/"extra" space on the die. Over time Apple (and other System makers) asked Intel throw higher and higher transistor budget at the iGPU. And they did.
The shift to multiple dies is in part driven by the slowing down by Moore's Law. As the 7nm , 5nm , 4nm ... processors get more complicated it gets more painful to make super large dies. Growing the transistor budget as the size slows down will shift more so to lashing multiple ( pragmatically more affordable) dies together to grow the overall package budget.
The Xeon E7 ( 'max' core and socket count) series got subsummed into with the shift to Xeon SP ( effectively got E5 2600 and E7 x8xx series merged into one group). Pragmatically, what going to see with the upcoming "super max" SP variants is Knights Corner being merged into SP class offerings. Xeon Phi 100 series started off around 57-61 cores. The current generation of SP tops out at 28. If they did a 2 x 28 mash up they'd have 56 ... which is pretty close to what they started out with on Phi. It would just be a another "black hole" consumption of another product line.
Intel already indicated back in 2017 that they were heading toward multiple chips to get to high core counts (before "chiplet" stuff).
"... On speaking with Diane Bryant, the 'data center gets new nodes first' is going to be achieved by using multiple small dies on a single package. But rather than use a multi-chip package as in previous multi-core products, Intel will be using EMIB as demonstrated at ISSCC: an MCP/2.5D interposer-like design with an Embedded Multi-Die Interconnect Bridge (EMIB). ..."
https://www.anandtech.com/show/1111...n-core-on-14nm-data-center-first-to-new-nodes
Once going to process of putting multiple CPU dies to get to higher core counts, it is relatively natural addition to substitute out one of those CPU focused dies and put in a GPU die. ( e.g., the Intel + AMD GPU mash up that Intel did. A future iteration where they completely toss that AMD GPU would be easy with an Intel GPU die if they had one. )
Pry Apple away? Apple spent probably 2005-present yearly handing Intel a wish list of what they wanted a GPU to do. The bulk of GPUs bough by Apple over that time span has been Intel GPUs. In terms of numbers they greatly outnumber the other two vendors. They have been dominate over the last 7+ years. Intel has 10 years of "we wish we had blah , blah , blah " lists from Apple. It isn't like they are a complete outsider and don't know. For example they have been a ground floor implementor of Metal on the Mac.
If Intel simply grows out their Mac support software stack to cover the new GPUs, doesn't try to piss on Metal, executes on their roadmaps , and manages to get rough parity with AMD performance at a incrementally lower cost. ..... it won't be a "pry" away. They'll just win the design 'bake off".
Apple does not want single source major suppliers. They don't.
If Intel had spent the money from MacAfee on GPUs they wouldn't be another year or so away from finishing Xe and moving to other memory busses than the mainstream CPU ones.
Apple's GPUs being completely dependent upon shared physical memory is exactly why is relatively far away from being a viable discrete GPU for a Mac Pro. It could be added but they don't even have the concept. Metal does, but the GPU doesn't.
Gen 11 is a integrated GPU. Some shared physical RAM is just part of the definition. Is it all shard flat addressable? No. As I outlined they deferred aspects of OpenCL 2+ to the 10+ generations. Did Intel have the requirements for Metal when they started several years ago. Probably ( Metal has been around since 2014 so if they started in 2016-2017 on this Gen 11 GPU they would have had the specs for Metal 2. Metal 2 appeared on Macs in 2017 and obvious Intel would have been told about that way ahead of time since most Macs in that time period come with Intel GPUs. )
Actually not joking. Apple Car seems to be on track to be one of the biggest boondoggles of funds since Apple spent $2+ billion to become a patent troll ( the Nortel patents). 100+ folks tossed because they have relatively little at this point. Apple had a window to be a electronics supplier to several car companies. Instead engaged in "Monkey see, Monkey do" with Tesla ... which is almost like the blind leading the blind. Apple has gobs of money so it won't hurt them and they can sweep it under the rug over time.
Apple strategically needing to do more ARM die variations for substantially higher volume products than the Mac Pro would be all the more indicative of what a waste of time , resources , and effort for Apple to plow into the discrete GPU space.
They can just buy Data Center level mainstream parts just like everyone else..... just as they are doing now.
Apple needs internally circuit design Data center solutions like they need a hole in the head. The closest major cloud competitor to that is Amazon and they simply just bought Annapurna Labs "off the shelf". They could just buy Ampere's or Cavium's Thunder solutions if they were desperate at being 'green' in their data center set ups.
if Apple wanted to attach their own FGPA solution to a x86 server chp then Intel or AMD would probably being able to do that for them if they had some special "web service specfic" hot spots they needed to cover. There is about zero rational reason for Apple to drift in to the highly internally designed CPU package space for the core web services they provide.
[doublepost=1549235952][/doublepost]
Away from the ears doesn't mean you aren't using them. The issue is whether they absolutely have to be inside the same physical box. That 1-3 drives can't heavily task a USB (gen 2 ) connection unless contrive some rolling down hill with hurricane winds at your back corner case "use case".
But yes for larger budgets where "time is money" and the folks time being sucked up are paid more than moderate amounts, it often pays to just go multiple SSDs. All the more so if sneaker netted around.
Exactly, that's why I said you have many options to connect them.
Do you need a huge rack and away from your room for noise reasons? 1-10 Gbps Ethernet (or fibre for even higher bandwidth)
Do you need a few HDDs and you don't mind some white noise? Spend some bucks in a good case and go for a 10 Gbps USB-C that can be hidden under your desk.
Do you need more speed than storage and don't mind hearing a little fan from time to time? Get a 1-2-4 HDDs TB3 cases for 40 Gbps transfers.
Did you spend all of your money into the mMP? Go for a cheap up to 8 bay USB3.0 boxes for a simple 5 Gbps backup solution.
And there is always Wi-Fi 6 and its new gen of wireless NAS coming soon...
There's still some value in a HDD bay like the ones in some towers that convert between a single 3.5" bay or two 2.5" bays. You wouldn't want your boot volume to be on anything but PCIe flash if you're in a professional use case, but adding scratch disks with comparatively cheaper SATAIII SSDs still has a lot of value.
(And if Apple is going to use proprietary SSDs for the boot volume then that's a way of getting around Apple's prices in a cheap, internal manner.)
(And if Apple is going to use proprietary SSDs for the boot volume then that's a way of getting around Apple's prices in a cheap, internal manner.)
Are you guys really that hearing sensitive? The local dogs bark louder than my system.
Maybe they're barking because the noise from your system annoys them.
Some folks criteria priority puts money saved over noise. The options are just in a single dimensions of noise. Some are just going to be far more bothered by money. Also 'race to the bottom' cheap enclosures can run into sleep/wake issues among other problems.
Also folks working in groups are on a different adoption rate to centralized storage than one person operations. Barney's tinker box in the basement has different priorities than a group of 2-5 ( Fred , Wilma, and Betty ) working on a project with tightly coupled components.
[doublepost=1549261929][/doublepost]
Chuckle... sources on the monolithic die ? Given AMD doesn't have a 7nm design for I/O and memory for the other chips where are they cooking it up for the APU version? If they are using the chiplet how can it be monolithic ? They could do a chiplet that had both x86 cores and GPU cores on it. However, if it is hooked to a 14nm I/O chip to get any I/O then the package won't be monolithic. A different, smaller I/O chip perhaps ( so cheaper and less power)
First, there is a difference between architecture and implementation. The reference standard implementation of the architecture that anyone with some money can get from ARM (the company) is skewed low power. However, folks can also buy an architecture license and do their own "bottom up" implementations. Apple has one of these. That doesn't mean they are trying not to go in general direction that ARM is going in.
However, folks could go a very separate way. ( Which actually hasn't worked out so well for vast majority of them. It has been a slow slog. For example getting a real, support Linux server distro up and running all the way through support validation/qualification stages. ).
https://www.nextplatform.com/2018/05/16/getting-logical-about-cavium-thunderx2-versus-intel-skylake/
[ This article also mentions Ampere ]
Some peaks at power (note: this is the system consumption ) .....
"... Our Gigabyte/ Cavium ThunderX2 Sabre development platform hit a peak of 823W at 100% load. We think that there are likely optimizations that can occur at the system’s firmware level, and by using GA power binned chips. At first, we thought that these numbers were way out of line so we discussed them with Cavium and that is when we were told that the ~800W range was correct for our system and pre-production chips. ..."
https://www.servethehome.com/cavium-thunderx2-review-benchmarks-real-arm-server-option/8/
If go to the first page of that article you'll see a change where the top end Thunder X2 has a TDP of 180W. For 64 cores you'd need two... so 360W just on CPU . That's isn't low amps. It is also way off the ARM reference implementation (e.g., 4 way symmetric multi-threading (SMT) . Apple has about zero SMT. )
Anandtech ran some benchmarks on a Thunder X2 system also. ( it is decent for some stuff. A single user workstation ..... in most case probably not. It isn't what is it built for. )
https://www.anandtech.com/show/12694/assessing-cavium-thunderx2-arm-server-reality/
Ampere's eMag system was benched over at https://www.phoronix.com/scan.php?page=article&item=ampere-emag-osprey&num=1
( TDP of that CPU is in the 120W range. )
Are you talking about stuff from this era ??? (2011 )
" ... “ARM microprocessors are designed for lower performance and unlikely to match x86 performance in the next few years,” Kanter said. ... "
https://www.macworld.com/article/1159856/macbook_arm.html
Which was and is true. At this point, we are more than several years past 2011. However, it is EXTREMELY illustrative of just how long this 'echo chamber' has been clamoring that Macs are going ARM just around the corner.
Not really. Apple is more than several years away from that point. But they aren't the only ARM implementors out there. The several implementations have mainly been designed for different workloads. They could be shift but highly debatable whether there is a reasonable sized market there to justify the expense. Just the server stuff has been problematical in terms of volume and reach with available OS base options.
On which benchmarks? The notion that ARM servers were going to wipe the x86 servers from every task in the Data Center and in HPC envrionments in a couple of years ... yeah that's mostly hooey. High users , high latency ( due to concentrated workload) tasks they do work. But that isn't a single user workstation context for vast majority of workstation users.
You bet .
Even though Job's 'the iPad is the future of computing' deserves more recognition, in my opinion .
The iPhone is only 12 years old, a bit early to consider it an earth shaking invention .
It wasn't Apple's invention either, but they where there at the right time with the right product , no doubt .
As for culture, the internet in general and social media in particular made mobile, online capable devices a success, not the other way around .
And, wtf is this?
Or, this?
Or, this?
I don't know, but I believe the tcMP is generally considered a complete and utter failure .
[doublepost=1549276565][/doublepost]
No problem , just deduct the cost of the external box from the price of an MP without drive slots .
For a fast and quiet 4 bay enclosure, including cables , about 800 bucks or so should do .
"noone will want trucks in the future..."
... followed by more or less the entire North American car industry cancelling all car production to focus on trucks, which remain the largest selling vehicles by an increasing margin.