This is obviously a reference to Apples use of AMD GPUs in the MacBook pro. Ryzen is not even a laptop chip.I am not an expert in alternative factsI explicitly wrote: "at the presentation".
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.
2017 iMac refresh wish list [merged]
- Thread starter Sirmausalot
- Start date
- Sort by reaction score
Was actually thinking about this during the presentation yesterday. Ryzen is giving AMD an edge on much more affordable multicore chips. It will be the perfect candidate for any sort of video work. Could definitely see apple utilizing the ryzen + vega combo, which from what we have seen so far, is pretty competitive to nvidia offerings as well.
Welp, you are taking it out of context. She was talking about achievements of AMD as a whole in the past 12 months at that point, just like the biz talks of Tim Cook at the beginning of events.I am not an expert in alternative facts
So while you might not be an expert, you might accidentally make a piece.
The iMac is far too thin to host a Vega card without melting
And that's the real crime... the word "thin" being used to describe a desktop computer!
The iMac is far too thin to host a Vega card without melting
We'll have to wait and see. I consider the current design to be slightly over the edge for its maxed out GPU (~125-150 TDP GPU), it can throttle in certain circumstances therefore its needs better cooling. The 1080m (150-180 TDP GPU) will push it well over the limit.
Admittedly a more throttled 1080m > a less throttled m395x when it comes to performance. Not even a contest in my opinion.
These new GPUs are more efficient PER WATT but they are exponentially more powerful. Current rumors of the desktop variant of Vega would make it seem its a blast furnace however again its also rumored to be powerhouse. And we have no clue what the mobile variant will provide.
The obvious solution to me. Redesign the iMac to be compatible for the parts within. If you are building a PC and it throttles, then you are built it wrong. Space constraints and physics aside Apple has the capital to make it work.
To be honest the TDP of the 6700k and 7700k push the cooling system to the brink.
Here is an idea for you Apple, free of charge. Offer the maxed out GPU option ONLY with a blade type SSD, use the HDD space for additional cooling components (fan, heatsink, etc, whatever). Delid the CPU and GPU and use a liquid metal thermal compound. Or just make the damn thing bigger.
[doublepost=1488052818][/doublepost]
It would also help with fan noise. The two lower model 8 core CPU's have much lower TDP's.
However its unlikely to happen. A lot of those same professionals that would benefit from more cores and threads also utilize Intel proprietary tech like Thunderbolt and QuickSync amongst a range of other things.
Also there is no iGPU (APU) option yet with Ryzen so the lower end machines would need dGPU's or Intel chips. The other problem is without it being an APU its unlikely Ryzen will currently support h265/VP9 10bit encoding and decoding like Kaby Lake has native support for. This will be beneficial for professionals (and even average users) now and even more in the future.
If I was to build a new PC even now with how tempting Ryzen is and how much I would benefit (I do a lot of video encoding) I'm not sure I would go with it or not. Its hard for me to justify more cores thus faster encode times now at the expense of slower encoding in the future using a different encoder. Then again I could be wrong I'm just theorizing whether Ryzen will have native support for h265/VP9 support.
It is exciting though (if you are a geek like me). I feel Intel is hurting pretty badly. They are having tons of trouble with their 10nm process, tons of lay offs, horrible business accusations, just tons of lack luster. And the Intel tick (process), tock (architecture) plan went to process > architecture > optimization and directly to process > architecture > optimization > optimization. With rumors of ANOTHER optimization step to help battle Ryzen. Do you see what COULD happen? A Intel CPU that is running near 6ghz and is hotter then the surface of the sun. Does that sound familiar? Hmm..... AMD much?
And if you are interested in this subject there is somewhat of a credible rumors about Intel working with AMD for integrated graphics designs. Its all rumors from the best I can tell however if you step back and think about it its not out of the realm of possibility. Compare gaming benchmarks of Intels iGPU vs AMD APU. I'm finding +10 FPS on AMD's A8-7670K vs Intels 6700k using integrated graphics in Grand Theft Auto which is very much CPU intensive also. Think about that, 10 FPS more on a A8-7670K vs Intels 6700K.
I made the mistake of becoming an Intel fanboy. I considered AMD stock when it was 2 dollars but refused to give them the benefit of the doubt. Idiot!
i think for sure 2 things will happen, the shrink in bezels, the trend already started with the new MBP,and new improved speakers, this trend also happen with the first ipad pro (12.9). Apple already put "retina" display from iphone - imacs, now they have to shrink the bezels and improved the speakers also in all their products...so expect this bezel thing, to happen also in next ipad and iphone 8
As much as I do agree that nVidia sells really good GPUs, I am pretty sure Apple will stick with AMD. iMac architecture is very close to the MacBook and it would be more convenient to follow the same choices. It has been quite a long time since Apple adopted an nVidia card.
About the TDP, maybe you're right. I was just comparing it to the past iMac CPUs. No one went over 63W.
Already is floating around 45W TDP 8Core/16T Eng Sample of this CPU
.For mobile most interesting parts, are for sure the APUs. I promised in Zen thread in Mac Pro subforum, that I will write a bit about Ryzen APUs, and its interesting feature, but I think this is good opportunity to reveal this.
With Ryzen APUs you get three things combined. Infinity Fabric, Ryzen CPU CCX, with 4 cores, and 8 Threads, 12 CU IPv9(Vega architecture) GPU. This is mobile. Desktop gets two APUs: Snowy Owl, 16C/32T monster with 4096 GCN core GPU, 64 PCie lanes, and 16 GB of HBM2, and 4C/8T CPU with 16 CU+4 GB HBM2 APU.
The APUs will be named NPU from this moment on. What is important is how they work in tandem with other GPUs.
Mobile APU, without HBM2 still has High Bandwidth Cache Controller. Effect? If your computer has acces to a GPU with HBM2 memory, regardless if it is internally, or externally connected - it will automatically increase the capabilities of the GPU.
Imagine situation like this.
You have MacBook Pro with 4C/8T+12 CU, and you connect to this through Thunderbolt 3 Vega 11 GPU, which will have 48 CU's and 8 GB's of HBM2. Total package available for your application will be 60 CU's, regardless of the API you will use. DX11, DX12, Vulkan, OpenCL. Does not matter. Drivers will automatically balance the load available for your GPUs.
Imagine a Trash Can Mac Pro, with 3 Snowy Owl APUs, and 192 PCie lanes available, and that you are connecting a rack made from 10-20 GPUs. You software automatically balances the load available for you hardware.
It does not matter if your hardware has HBM2 on package or doesn't. It balances the load accordingly.
If you would be aware of me previous posts in RX 480 thread in Mac Pro forum, you would know How I worded it then:
"Vega has next generation schedulers"
About the load balancing, and the expandability of compute capabilities, well... I have been writing this 2 years ago guys, on this very forum in at least 2 dead Mac Pro threads
.I will link this post to the Zen thread, in question.
The iMac is far too thin to host a Vega card without melting
Vega is being designed to run at the same TDP as the 390-395x.
There are two Vega chips. Vega 10 is the larger one with a TDP of 225-250 W (like the 390X) and a smaller Vega 11 that will likely be faster than the RX 480 but slower than the bigger Vega 10. We will probably know a lot more about both after tomorrows press conference by AMD.
Also, keep in mind that Apple has been known to downrate the settings while keeping the shaders the same. Thanks for the heads up tomorrow.
My guess it that happens either next year or even later. Apple has a pattern of doing these sort of things late in the game.
Given the last major update to the iMac was 2015, it's already behind. That's really the supply delays as Apple would only waist money releasing an iMac update with the older video card. The CPU isn't going to be much different from what we see now, but the SSD may be significantly larger given the newer tech released. Also, the video cards is the big push for an update. If they wait until next year, Apple will then either release with the Vega (TDP tuned for the iMac) or newer cards. If they release next year with the current line, they'd be selling a previous gen card by then. I expect them to release this year, but we may not see it until Fall.
Well, looks like Vega will be out in time for the iMac as it is officially on the shelves on 4 Apr. According to sources, AMD all but skipped the high end gaming video cards by releasing to the low to mid-grade systems (Polaris 11). Polaris 10 has yet to release with the highest being 480. Because of this, I highly doubt that AMD will release a line that will be discontinued in less than 30 days.
In the past, AMD branded the Vega card as being the optimized FINFET card (the graphics card version of a tock). AMD has announced that they're launching their higher clocked, higher TDP cards with that optimization package so we are essentially seeing them skipping right to the higher end Vega. The question I have now is will the Polaris 11 see the Vega upgrade with this release?
How will this impact us? TDP has been brought up time & again & people keep saying that the Polaris 10 (an now Vega) cannot be used in the iMac. This is not true for a couple reasons:
1) Apple has time & again requested certain clocks & voltages to support the systems so they can get away with the highest end card they can cram in these compact systems. Reference the 295x & the temp issues with them being topped out.
2) Finfet by nature is a reduction in chipset size that allows AMD & nVidia to produce double the clock speeds while using THE SAME AMOUNT OF VOLTAGE as the previous generation chips. The currently supplied AMD 395x was not replaced with an equivalent die shrunk chip. This would fall within the Polaris 10 architecture, which means we may see the iMac jump to Vega. Given that the 395x pushes the temp limit of the generational design, I hope that AMD doesn't push to create a chip higher than the the FINFET upgrade could support, or else we'll see a repeat of nVidia's 7000-9000 & AMD's 6000 debacle.
To understand what happened then, ATI was purchased by AMD & the graphics division was shut down for a while. nVidia took advantage of this by maximizing the chip clocks & violating the rules engineers set to keep cards from failing:
80C=allows GPU to survive for the expected life=10 years
for every 5C over that, divide the life of the chip by half.
Example, I had a nVidia 8400 DESIGNED to run at 100C & it required replacing every 8 months for 3 years until the warranty finally ran out. Let's do the math:
80C=10 years,
85C=5 years,
90C=2.5 years
95C=1.25 years
100C=.625 years = 7.5 months. Being off by 15 days is not bad.
I would suspect we will see either Polaris 11 chips on the low end systems (with or without the Vega upgrade), & we will likely see Polaris 10 Vega chips on the high end iMacs (specifically the 27"). I also think that AMD will likely try to max out the Vega over what it can handle this year, but don't think the iMac will see those chips.
http://www.techspot.com/news/68361-amd-radeon-rx-500-gpus-reportedly-polaris-rebrand.html
http://www.notebookcheck.net/AMD-Radeon-RX-480M.164291.0.html
In the past, AMD branded the Vega card as being the optimized FINFET card (the graphics card version of a tock). AMD has announced that they're launching their higher clocked, higher TDP cards with that optimization package so we are essentially seeing them skipping right to the higher end Vega. The question I have now is will the Polaris 11 see the Vega upgrade with this release?
How will this impact us? TDP has been brought up time & again & people keep saying that the Polaris 10 (an now Vega) cannot be used in the iMac. This is not true for a couple reasons:
1) Apple has time & again requested certain clocks & voltages to support the systems so they can get away with the highest end card they can cram in these compact systems. Reference the 295x & the temp issues with them being topped out.
2) Finfet by nature is a reduction in chipset size that allows AMD & nVidia to produce double the clock speeds while using THE SAME AMOUNT OF VOLTAGE as the previous generation chips. The currently supplied AMD 395x was not replaced with an equivalent die shrunk chip. This would fall within the Polaris 10 architecture, which means we may see the iMac jump to Vega. Given that the 395x pushes the temp limit of the generational design, I hope that AMD doesn't push to create a chip higher than the the FINFET upgrade could support, or else we'll see a repeat of nVidia's 7000-9000 & AMD's 6000 debacle.
To understand what happened then, ATI was purchased by AMD & the graphics division was shut down for a while. nVidia took advantage of this by maximizing the chip clocks & violating the rules engineers set to keep cards from failing:
80C=allows GPU to survive for the expected life=10 years
for every 5C over that, divide the life of the chip by half.
Example, I had a nVidia 8400 DESIGNED to run at 100C & it required replacing every 8 months for 3 years until the warranty finally ran out. Let's do the math:
80C=10 years,
85C=5 years,
90C=2.5 years
95C=1.25 years
100C=.625 years = 7.5 months. Being off by 15 days is not bad.
I would suspect we will see either Polaris 11 chips on the low end systems (with or without the Vega upgrade), & we will likely see Polaris 10 Vega chips on the high end iMacs (specifically the 27"). I also think that AMD will likely try to max out the Vega over what it can handle this year, but don't think the iMac will see those chips.
http://www.techspot.com/news/68361-amd-radeon-rx-500-gpus-reportedly-polaris-rebrand.html
http://www.notebookcheck.net/AMD-Radeon-RX-480M.164291.0.html
Keep in my the only Vega card that AMD has really talked about is Vega 10, which is supposed to be a 12 TFLOP monster card. You are right in that you can lower the clocks of some cards so that the thermals fit into the iMac but this tends not to make sense for the highest end cards. There is a point of diminishing in returns in that if you have to lower the clock speeds a significant amount its not worth paying the premium for the bigger GPU. For instance it made sense to fit the ~380 mm2 Tonga (M395X) from a 180 W desktop card into the 125 W limit of the iMac. However it didn't make sense to try and fit in the 250 W 600 mm2 Fiji (Fury) into the iMac.
I think the smaller Vega 11 chip would be a good choice for the iMac. If we assume performance and thermal wise it will land between the 150 W Polaris 10 and 225-250 W Vega 10 then it could easily fit. Its very unlikely Vega 10 makes it into an iMac. I'm skeptical it would even make sense in the current form of the Mac Pro.
The Surface Studio display can be as thin as it is because it's just a display. Microsoft moved all the guts to the base, but even there they sacrificed performance in areas compared to the iMac due to lack of thermal headroom in the smaller volume compared to the iMac's back.
As CWallace said above... Microsoft made other tradeoffs due to cramming an entire computer into a tiny box.
And that was my point. There's no need for "desktop" computers to have the words "thin" or "tiny" describing it.
Processors need decent heatsinks. They need decent airflow. GPUs need the same. But you can't have that when you cram them into tiny boxes or behind thin monitors.
BTW... the Surface Studio is not immune to complaints.
It starts at $3,000 for a mobile i5 processor and a mobile GPU.
I repeat... mobile parts in a "desktop" that starts at $3,000
And that goes back to the problem I described earlier. Why did Microsoft put a laptop CPU and GPU in something that is supposed to be a desktop?
Because they're trying to cram stuff into a tiny box!
Uh, you're right, I mis-read the start of that first article.
@Stacc,
Keep in mind that Apple goes with the Polaris 11, we're only going to see about a 25-50% due to the fact the card will be using about 50% less power than the current Tongas. As you pointed out, the Polaris 10s would be extremely ideal running 150w vs the current 180w (down clocked to 125w). The Vega 11s & 10s are both supposed to be architecture tweaked, not overclocked. Due to that, they'll only preform a small percentage better than their Polaris counterparts.