You need the DIARrhea Dongle. Things will flow much better at that point, but you gotta monitor it closely.
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.
2016 nMP
- Thread starter ManuelGomes
- Start date
-
- Tags
- nmp
- Sort by reaction score
- Status
You need the DIARrhea Dongle. Things will flow much better at that point, but you gotta monitor it closely.
http://www.amazon.com/Intel-BX80660...F8&qid=1459437348&sr=8-17&keywords=xeon+e5+v4
Also, it looks like Broadwell-EP officially launched.
P.S. http://forums.anandtech.com/showpost.php?p=38135009&postcount=3473
Also, it looks like Broadwell-EP officially launched.
P.S. http://forums.anandtech.com/showpost.php?p=38135009&postcount=3473
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Well, that's good.
WWDC is 10 weeks away.
We'll see.
Five WWDC possibilities.
1. A fabulous new MP.
2. A minor cpu/gpu/port speed bump with the same $10, 450 watt psp and mid-range re-branded Radeons.
3. An EOL announcement.
4. The silence of deep space.
5. Glow in the dark watch bands.
6. Dre is going to be the next CEO of Apple
My order of likelihood.
2
4
5
3
6
1
WWDC is 10 weeks away.
We'll see.
Five WWDC possibilities.
1. A fabulous new MP.
2. A minor cpu/gpu/port speed bump with the same $10, 450 watt psp and mid-range re-branded Radeons.
3. An EOL announcement.
4. The silence of deep space.
5. Glow in the dark watch bands.
6. Dre is going to be the next CEO of Apple
My order of likelihood.
2
4
5
3
6
1
I'd go:
4
3
5
2
6
1
[doublepost=1459441150][/doublepost]Got an email from Thinkmate, they already have workstations ready to have the E5-2600 v4 chips dropped in them.
http://www.thinkmate.com/system/hpx-xs4-2460v3
Its says v3, but the v4 CPU options are in there. And just, damn, up to 44 cores....
Last edited:
http://www.anandtech.com/show/10158/the-intel-xeon-e5-v4-review/6
http://www.anandtech.com/show/10212...luded-free-with-visual-studio-and-open-source
Couple it with HLSL being Open Source from now on and we see that whole business, not only Apple, is migrating, and changing.
http://www.anandtech.com/show/10212...luded-free-with-visual-studio-and-open-source
Couple it with HLSL being Open Source from now on and we see that whole business, not only Apple, is migrating, and changing.
The only two viable CPUs for MP that so far launched are 14 and 18 core model with 135W TDP, and considerable pricing levels.
quad core model prices start at 996 USD, and by comparing it to previous quad core model and its price(293 USD) we will not see that CPU in Mac Pro.
P.S. I think that the two CPU that will appear from this in MP are that 14 and 18 core models: E5-2697v4, E5-2690v4.
P.S.2. almost complete pricing list for CPUs: http://www.nextplatform.com/wp-content/uploads/2016/03/intel-xeon-e5-v4-table.jpg
P.S.3 http://images.anandtech.com/doci/10158/Xeon_678x452.jpg Is it only me, or does this CPU look... "pretty"?
quad core model prices start at 996 USD, and by comparing it to previous quad core model and its price(293 USD) we will not see that CPU in Mac Pro.
P.S. I think that the two CPU that will appear from this in MP are that 14 and 18 core models: E5-2697v4, E5-2690v4.
P.S.2. almost complete pricing list for CPUs: http://www.nextplatform.com/wp-content/uploads/2016/03/intel-xeon-e5-v4-table.jpg
P.S.3 http://images.anandtech.com/doci/10158/Xeon_678x452.jpg Is it only me, or does this CPU look... "pretty"?
Should be great!The only two viable CPUs for MP that so far launched are 14 and 18 core model with 135W TDP, and considerable pricing levels.
quad core model prices start at 996 USD, and by comparing it to previous quad core model and its price(293 USD) we will not see that CPU in Mac Pro.
P.S. I think that the two CPU that will appear from this in MP are that 14 and 18 core models: E5-2697v4, E5-2690v4.
P.S.2. almost complete pricing list for CPUs: http://www.nextplatform.com/wp-content/uploads/2016/03/intel-xeon-e5-v4-table.jpg
P.S.3 http://images.anandtech.com/doci/10158/Xeon_678x452.jpg Is it only me, or does this CPU look... "pretty"?
So Broadwell-EP is out. Remember, the caveat here is that these are the processors suitable for dual processor workstations/servers, of which the mac pro is not. The lower core count mac pro (4-10 cores for Broadwell) will use Xeon 16XXv4) and tend to be clocked slightly higher than their 26XX counterparts. Rumor has these chips coming out in early June with the enthusiast Broadwell-E release. Of course Apple uses 26XX processors for the higher core count configurations but they come at extreme cost.
However, given all that, Broadwell looks to be unimpressive. Little IPC improvements and only slightly higher core count than Haswell-EP (18 cores to 22). Shrinking from 22 nm to 14 nm has reduced clock speeds, not increased them. This is unsurprising, given the same trend has existed on the consumer CPUs released on 14 nm but it is still disappointing.
Its probably safe to say this is not the tech that has been holding up releasing a new mac pro. At this point Apple is probably waiting on GPUs from AMD that will come over the summer.
However, given all that, Broadwell looks to be unimpressive. Little IPC improvements and only slightly higher core count than Haswell-EP (18 cores to 22). Shrinking from 22 nm to 14 nm has reduced clock speeds, not increased them. This is unsurprising, given the same trend has existed on the consumer CPUs released on 14 nm but it is still disappointing.
Its probably safe to say this is not the tech that has been holding up releasing a new mac pro. At this point Apple is probably waiting on GPUs from AMD that will come over the summer.
Two possibilities for GPUs. Polaris 10 as D310, declocked Fury as D510 and the same situation for Fury X as D710(as we have seen, on S9300x2 FirePro - Dual Fiji locked at 850 MHz can run at 300W - 150W for each GPU).
Or sneak peak, and paper launch of new MP with Polaris 10 as D310, and waiting for Vega GPUs...
None of those options are ideal. Pascal is a no-go for Apple.
Or sneak peak, and paper launch of new MP with Polaris 10 as D310, and waiting for Vega GPUs...
None of those options are ideal. Pascal is a no-go for Apple.
What if Pascal comes in with better thermal profiles?
Various Maxwells used 50-100 fewer watts than AMD's offerings.
Not an insignificant difference when they're powered by 450 watt psp
from a $700 i5 PC.
I tend to think Apple is going to ignore Fury. Its a graphics/gaming oriented chip with limited amounts of VRAM. Add to that it is a huge die which makes it expensive. If we take a look at the D700, it is a compute focused chip with lots of VRAM (for its time). If Apple follows this methodology it will likely pick a more compute focused Polaris chip that has graphics performance in the range of the 390X to Fury and better compute performance than both.
If Apple was going to use a Fury based chip, they could have released a new mac pro 9 months ago. Instead we are still here waiting. I tend to think Apple will go all in on Polaris across the product line. Dual Polaris 10 on the mac pro, Polaris 10 on the 27" iMac and Polaris 11 on the 15" macbook pro. Makes a nice standard platform to target for metal developers and Apple can introduce a Skylake-EP/Vega system in a year or two.
Apple has been ignoring Nvidia for roughly 3 years now. Given how efficient maxwell is as a graphics (not compute) chip and that it didn't end up in either the 27" iMac or the 15" Macbook Pro I think there are some bridges that have been burned between the two. Another theory is that Apple likes some features AMD brings to the table such as asynchronous compute and less driver tuning required to get good performance. However, what Apple probably likes best is the good discounts it gets on AMD's GPUs.
No chances. Pascal is what Maxwell was supposed to be on 20 nm, but TSMC failed to deliver that node. Thats why Nvidia castrated Pascal from FP64 and sold it as Maxwell. Now we have new node, but they bring back the FP64. So do not expect 2 times better scaling. I think we can expect Titan X performance in GTX980 thermal envelope(or slightly higher like just a mere under 200W) and price level.
And yet Maxwell was around 20% more efficient even though everybody was stuck on .28 process for four years.
But I think Apple will stick with AMD because they can get them cheaper.
They know how desperate AMD is and that they will price at low levels to keep
the silicon moving at a high enough rate to stay in business for a few more years.
But I think Apple will stick with AMD because they can get them cheaper.
They know how desperate AMD is and that they will price at low levels to keep
the silicon moving at a high enough rate to stay in business for a few more years.
I would not call it that is the case. 250W GPU from AMD has 6 TFLOPs of compute power. 250W GPU from Nvidia has 6.1 TFLOPs of compute power.
275W GPU from AMD has 8.2 TFLOPs of compute power, and there is nothing from Nvidia that can be compared to this. Games are different thing. DX12 shows that compute power will reflect gaming performance in future. Other thing is that, with current drivers, and reference models of GPUs AMD get performance lead in every price/performance bracket. It leads in 4K, it leads in 1440p it leads in 1080p. I have written this over, and over, and yet people tend to still say the same not actual things about the GPUs that are on the market.
Only benefit for Nvidia at this moment is CUDA locked software. But it is possible to run CUDA on AMD GPUs without any penalty... So guess what will happen?
P.S. There is one thing that will be shining in Pascal. Properly fed ROPs. 96 of them to be precise, coupled with HBM2. It will be VERY interesting to see the numbers but I think it will be first time when we will see that type of front end on any GPU.
275W GPU from AMD has 8.2 TFLOPs of compute power, and there is nothing from Nvidia that can be compared to this. Games are different thing. DX12 shows that compute power will reflect gaming performance in future. Other thing is that, with current drivers, and reference models of GPUs AMD get performance lead in every price/performance bracket. It leads in 4K, it leads in 1440p it leads in 1080p. I have written this over, and over, and yet people tend to still say the same not actual things about the GPUs that are on the market.
Only benefit for Nvidia at this moment is CUDA locked software. But it is possible to run CUDA on AMD GPUs without any penalty... So guess what will happen?
P.S. There is one thing that will be shining in Pascal. Properly fed ROPs. 96 of them to be precise, coupled with HBM2. It will be VERY interesting to see the numbers but I think it will be first time when we will see that type of front end on any GPU.
There are new rumors out on pascal. May be announced next week with availability towards June. Guesses are performance will be more compute oriented and graphics performance will be around the Titan X/980 Ti. It may use GDDR5X, which would deliver lots of bandwidth and high VRAM capacities.
The one thing to keep in mind is that Apple likely won't ignore Nvidia forever. So then the question becomes when will they choose Nvidia again. I am not saying its going to be soon, but if there is one thing Apple likes, its having suppliers compete for their contracts. Despite how one sided Apple's GPU offerings are, they could switch back at any time.
The one thing to keep in mind is that Apple likely won't ignore Nvidia forever. So then the question becomes when will they choose Nvidia again. I am not saying its going to be soon, but if there is one thing Apple likes, its having suppliers compete for their contracts. Despite how one sided Apple's GPU offerings are, they could switch back at any time.
Like I have said, If we will look at GP104 and the fact that Nvidia coupled ROP ratio to memory at 16:1 we will see most likely 64 ROP, 3072 CUDA core GPU that will have similar performance to Titan X, despite the fact that it has lower amount of ROPs, and narrower memory bus, but they will be properly fed, because of GDDR5X. 6000 MHz will bring 384 GB/s of bandwidth from 256 Bit memory bus. Very efficient design on the front end. What it will lack however is Asynchronous Compute Engines, and Hardware Schedulers. Scheduling will be the job of CPU.
P.S. Read the colored part on the bottom. http://www.anandtech.com/show/2549/7 Oh, the irony...
P.S. Read the colored part on the bottom. http://www.anandtech.com/show/2549/7 Oh, the irony...
The 2699v3 to 2699v4 saw something like 10-20% gains in performance (depending on the benchmark) by that anandtech review. I don't know how you want to define "impressive", but that's certainly a noticeable increase. Especially, when you release most people are jumping 2-3 generations. The nMP is on v2, and if you compare the 2697v2 (2.7GHz 12 core, ~$2600) to the 2695v4 (2.1GHz 18 core, ~$2500) you're looking at around 50% performance gains.
If this isn't what Apple is waiting for, then they are living in fantasy land.
One more thing. I think people should read this post about Pascal and Polaris. http://forums.anandtech.com/showpost.php?p=38123824&postcount=201
Maybe and maybe not. The initial set of GDDR5X out of Micron won't do very high capacity.
http://anandtech.com/show/10193/micron-begins-to-sample-gddr5x-memory
[ Higher than HBM v1 but will max out at 8GB (using more physical board space.. which may/may not be available). Simpler (less wide) and better Perf/W, but for the huge capacity gap not so much far in advance of HBM v2's arrival. ]
It is a two way street. Apple likes multiple suppliers but they also like paying the best price and/or customizability. If Nvidia won't compete on cost then they will continue to be invited to bake-offs, but will loose. Is Nvidia going to let Apple do custom "Pro" cards?
Apple has three GPU suppliers for the Mac products. Only two of them can do integrated with x86 solutions.
Last edited:
That's something Conan the barbarian would if he wants slay his rivals. Broadwell... Not broadsword.
- Status