Pundits have been speculating on when the Apple Pencil would be adapted to the iPhone since the day it was announced. Heck, folks on this forum probably beat most of them.
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
Pundits have been speculating on when the Apple Pencil would be adapted to the iPhone since the day it was announced. Heck, folks on this forum probably beat most of them.
It's been discussed here, but the chipset needs to have knowledge of Thunderbolt, which is why there are no Thunderbolt options for even older Intel chips. If it were that easy there would be Thunderbolt on cMPs already.
ASUS was reportedly going to sell a Thunderbolt + AMD mobo in 2012 but never seems to have gone anywhere.
[doublepost=1507915603][/doublepost]
Pundits have been speculating on when the Apple Pencil would be adapted to the iPhone since the day it was announced. Heck, folks on this forum probably beat most of them.
Yeah. Lot's of people have speculated on this.
DNG's prediction just seems really unrealistic after only six months. In six months Apple has gone from nothing to a nearly complete AMD Mac Pro? Yeah don't buy it.
PhD (I think's) rumor sounded a lot more realistic for where Apple would be six months in. Two concepts that they're still arguing over. Not vouching for that either though. I haven't heard anything to back that one up either.
That is the one thing I heard is that Apple has nothing in April. Zero. Zilch. So keep that in mind for these rumors.
Seems the iMacPro will use special -W Xeons (2140B and 2150B) with lower base freq than the regular SKUs.
It also seems it has an iGPU but that is disabled by Apple. Odd thing having an iGPU, I'd assume a new socket would be needed for that. Maybe only the relevant blocks?!
Pike found it in the firmware:
https://pikeralpha.wordpress.com/20...apples-new-imac-pro-with-special-xeon-w-skus/
To lower power and heat output for sure.
The mMP shouldn't have that problem I hope.
It also seems it has an iGPU but that is disabled by Apple. Odd thing having an iGPU, I'd assume a new socket would be needed for that. Maybe only the relevant blocks?!
Pike found it in the firmware:
https://pikeralpha.wordpress.com/20...apples-new-imac-pro-with-special-xeon-w-skus/
To lower power and heat output for sure.
The mMP shouldn't have that problem I hope.
Last edited:
From: https://pikeralpha.wordpress.com/20...apples-new-imac-pro-with-special-xeon-w-skus/
Intel Xeon W-2140B @ 3.2GHz ( 8 cores / 16 threads, CPUID 0x050654)
Intel Xeon W-2150B @ 3.0GHz (10 cores / 20 threads, CPUID 0x050654)
The W-2145 runs at 3.7GHz and the W-2155 at 3.3GHz so not much of a clock-speed hit. They also appear to have Speed Shift, which allows the CPU to control the turbo modes which provides faster switching to and from those modes than depending on OS calls.
These CPUs may also have an iGPU. Current iMac Pro firmware appears to disable the iGPU, but not sure it needs to be on for Quick Sync Video to actually work or just it's presence in the CPU core is sufficient.
The first base config iMac Pro GeekBench score is now available, though note the clock-speed for the Vega 56 is probably lower than what will ship so the final score should be higher.
Intel Xeon W-2140B @ 3.2GHz ( 8 cores / 16 threads, CPUID 0x050654)
Intel Xeon W-2150B @ 3.0GHz (10 cores / 20 threads, CPUID 0x050654)
The W-2145 runs at 3.7GHz and the W-2155 at 3.3GHz so not much of a clock-speed hit. They also appear to have Speed Shift, which allows the CPU to control the turbo modes which provides faster switching to and from those modes than depending on OS calls.
These CPUs may also have an iGPU. Current iMac Pro firmware appears to disable the iGPU, but not sure it needs to be on for Quick Sync Video to actually work or just it's presence in the CPU core is sufficient.
The first base config iMac Pro GeekBench score is now available, though note the clock-speed for the Vega 56 is probably lower than what will ship so the final score should be higher.
Last edited:
More word salad.
If you're inlining assembly code - you're not leaving it to the compiler. Write simple scalar high-level code, and leave it to the compiler to vectorize it on AVX.
A simple generalization is that "AVX" is "loop unrolling in hardware". Instead of doing 8 million loops, you do 1 million loops processing 8 elements per instruction.
Compilers have been unrolling loops since before most MacRumours readers were born. Believe that they understand how to unroll with AVX*.
Luke - trust the compiler.
Last edited:
The part about multithreading and vector instructions being somehow opposed is odd too.
Not True, and FYI Gygabyte already integrated TB3 into one of its X399 threadripper motherboard, will enable it later, maybe on jan 1 with an bios update.
There is no TB2 AMD mobo coz there is almost no marker for PC with TB2, as very small it for Tb3, plus Intel wages, so from jan 1 more likely you'll see more TB3.
Only restriction for TB3 is having 2 or ideally 4 PCIe lines available, of course the bios need to be aware of the TB3 controller which its seem by the cpu as another PCIe Muxer.
Dear, I'm your cheff
You missed that Gigabyte scrubbed that because of.... : drumroll : CPU support.
https://www.anandtech.com/show/11847/gigabyte-announces-x399-designare-ex
If it's being worked through on the CPU side, could be an option eventually. Maybe AMD is working on it to try to push Apple more. Who knows, maybe Apple is even helping out. But I doubt it.
no, you can have both and usually have a much more efficient application, the point is im my case, it dont work, unless I ran it on a Xeon Phi, much cheaper and easy to maintain is to go full on SMT on a GPU, I may keep the current AVX optimized code, and offload on the GPU with non AVX tricked code, this creates an problem, I have to maintain 2 very complex libraries for an practically worthless performance gain, I'm working now on Cuda, maybe later I'll migrate it to OpenCL 2, or even Metal (unlikely since it is deployed on linux in production environment, but will be an good exercise).
I hope nVidia/Apple soon allow Pascal on TB3, it will easy my work.
[doublepost=1507931978][/doublepost]
What it means, they do not need to add nothing to the MOBO, just work on the GPU signal since Threadripper do not have iGPU and they need to work on the video feed for TB3 Header (raven ridge its AMD' Vega/Zen APU).
Losing Multiplatform is a big downside to GPU code. Especially these days with OpenCL not universal. (OpenCL can be kind of a bear of a technology to work with anyway...)
A GPU is basically doing vectorization on a single CPU. Except a GPU does it's vectorization as actual cores.
Vectorization vs. GPU is pretty much the same thing concept, just depends which device is going to be more efficient for your specific case. Mixing them unnecessarily could be trouble though. Going to introduce slowdowns in the pipeline while the data is funneled somewhere else.
It means exactly what I said. Threadripper does not support Thunderbolt. And Gigabytes statement implies iGPU is not the only problem.
Intel and Apple made a huge mistake in bundling video (display port) along with PCIe on T-Bolt.
T-Bolt might have been successful if it were only an external PCIe link.
video feed for TB3 Header??? the HP-Z pro workstations with server cpus with no igpu had an loop back DP cable from the pci-e video card.
I dont work at Gigabyte but I think they may have some noise or level related issue with their peculiar TB3 implementation, Apple have much better experience integrating TB3 with dGPU.
There are N reasons they are experimenting Issues with TB3, but unlikely is the CPU or the Chipset, as they state their TB3 has issues with the video feed, but all other TB3 modes works, they may enable it as a TB3 w/o video as they seem suggest in case they dont solve the issue feeding the video, the problems related to the video input as I write above are more likely noise or signal's voltage level related, I ve seen this before.
Meanwhile, the Full AMD Macs may implement an in-house baked TB3 Header instead purchasing expensive intel Alpine Ridge controllers, it would solve not only such 'compatibility' issues, but also improve the USB-C Alt modes enabling DP 1.4 instead DP1,2 in USB-C alt mode (in TB3 mode DP1.2 is mandatory),
And here's the proof! This is Wolfgang Puck's Spago in Las Vegas (circled in yellow). As you can see everything's being set up to be "MAGO!" The "S" and "P" have been removed from the neon signs! If any "M's" from AMD come up missing we know were they will end up!
I got tired of hearing about "DNG" (never did buy he/she/its "info" for a second). But in a vacuum, lots of people will take anything seriously. For those so inclined, here are the secrets to a good rumor:
- Give people something that they almost universally want, e.g. an OLED screen; the bait.
- Come up with a juicy Apple-like advance to focus on, e.g. "phased array speakers for VR-like immersion" that that sounds futuristic yet credible. This stokes imagination and sets the hook.
- Pick something old that goes away yet is arguably better the new way, e.g. an "M.2 RAID" with no 2.5 or 3.5" drive bays. That's the Apple way.
- Pick something that people don't want yet would have to live with, e.g. a not-quite-standard video card. Again, that kind of thing is so familiar to Apple folk (yet annoying) that it will add credibility.
- Focus on the sizzle and not the steak. In other words, discuss how sexy something will be and not on the details. E.g. "the Monolith" - a space gray case on a polished black solid aluminum base, with just enough access for most people's needs. Sounds good!
- Sprinkle in just enough realistic specs to give rumors the sense of having a credible technical basis, e.g. 8 memory slots (more than the iMac Pro), etc., but:
- Don't give too many precise specs so that you risk getting caught out by something highly improbable.
- Always offer an alternative to prevent a backlash, e.g. the "second, larger case that's not as far along in development ..."
- Leave an opening for more details to come.
neither me, but its funny, but sometimes somehow scores a hit.
- PCIe Slots
Chessegrater NG
Easy: TB3, NVMe...
No CD/DVD, AMD CPUs
Side Glass Panels
4TB SSD (easy)
Three GPU?
Not DNG
Not DNG
Likely, but if this unicorn we name MP7,1 comes with 3 GPUs you know where you read it first.
This are my own personal (not DNG's) predictions on the MP7,1 based on what everyone knows:
There is real chance to switch from CPU vendor to AMD, either Threadripper or Epyc.
Form factor: no idea but one thing will be sure, GPUs will use a custom connector and the mMP will be modular only while on production line, or flexible enough for a wide range of GPU/CPU TDPs.
Also there is real chance for first time no matter if the CPU is AMD or Intel that nVidia GPUs again will be an option along AMD GPUs offerings, but unlikely you'll be able to flash any gpu to put on the mac.
Last edited:
Except of course the reason that your rumor would be so believable is this is exactly how Apple's marketing and engineering operates internally early in the design process.
That 10 core gets a heck of geek bench. But I'm curious how the 8-core and its 5K price tag will be received. It geeked only 23,536 on multicore. That's more or less what the 8700Ks have been hitting.
Last edited:
GeekBench is a synthetic benchmark so real-world results on software optimized for as many cores as it can get could show real difference between a 4-core (8700K), 8-core (W-2140B) and 10-core (W2150B).