Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Status
Not open for further replies.
Battle continues.

looks like neither was right.

power consumption of fury x isn't that big (on average could be even less then 980ti), but it still has 2x8-pin, maxing at 375Watts (in reality 2x8-pin might give even more)

This graph from thg may explain it

30-100-microseconds.png



Occasional spikes to 400 watts, that's why it need 8xpin, otherise 6+8 pin would be enough.

Graphics Card Total: Min 25.53W Max 453.58W Avg 220.69W

Amd's power limiter isn't working well yet, unlike nvidia's. Hope they will fix in in 14nm cards.

It also tries not to suck pwoer from pcie slot, maxing at 31W out of 75
More http://www.tomshardware.com/reviews/amd-radeon-r9-fury-x,4196-7.html
 
Last edited:
That is correct Netkas. Its exactly in line with every other review of FuryX that shows max power draw recorded.

http://wccftech.com/amd-radeon-r9-nano-delayed-september-wip/ But you have here full Fiji chip locked in 175W of TDP and single 8 pin connector.

If we will remember that Fury X2 will have the same dual 8 pin connectors, we know pretty much why. 175W x 2 = 350 from 375W of max power.

So in the end, we argued about Asus Strix Fury, or Sapphire Tri-X Fury, or Fury X? ;)

Because those three chips all have different voltage settings and therefore different power consumption.

Which shows how flexible is that architecture and GPU itself.
 
koyoot, great news regarding Broadwell-EP. I was starting to believe that Intel wouldn't talk about it at IDF but apparently they had the good sense to break out the good news.
Guess that was what Apple was waiting on for the nMP, hope so at least.
DDR4 2400 looks sweet, but cost should be quite higher than DDR3 so we might see yet another rise in the base price.
Or maybe they get Grenada GPUs at a bargain, since they're not new tech, to compensate.
Q4'15 is right around the corner, we could see a late 2015 or early 2016 nMP 7,1 if all goes well.
Still, new mouse and keyboard BT4.2 should only come in February, and who knows about new 5K TBD.
Next year seems more likely.

Samsung will start production of HBM as well, so the cards are set.
Nano might be coming in a week, I'm curious about the actual configuration and power draw.
But next year's GPUs are massive, those will be beasts. Up to almost 20B transistors, 32GB VRAM.
 
R9 nano will have 1 8 pin, which gives it maximum of 225 watts, but in reality it could be 250-260 watts for these spikes easily (I mean 8-pin allows to draw a bit more than 150 watts). Average would stay under 175 watts, but these spikes may make it no go even for cMP on 6-pins (2x6-pin -> 8-pin adapter).

In heavy scene the cMP could just shutdown
 
Even GTX980 has high spikes but remember these spikes last micro seconds. That's something that is not immediately obvious when you look at these power graphs.
 
Time will tell Manuel what will happen, however, everything starts to fall in line so far.

Im curious if Apple would give 512 GB SSD as a base config in Mac Pro.

Netkas, I think better idea about Nano is to look how it will perform. Because it may very well be much under that 225 at Max power consumption ;).

I don't think you take into account the voltages on the GPU. Asus Strix Fury has 43 mV lover voltage in comparison to Fury X and Sapphire Tri-X Fury. And Strix Fury maxes at 270W. With 1000 MHz core clock. Maxes. Its the highest recorded power draw.
http://tpucdn.com/reviews/ASUS/R9_Fury_Strix/images/power_maximum.gif

Imagine a GPU with even lower voltages and slightly lower core clocks.

One more thing. Full Fiji chip is able to get to 1135 MHz on -48 mV voltage on core. 85 MHz higher than standard, and power consumption goes down by 40-50W.
http://tpucdn.com/reviews/AMD/R9_Fury_X_Overvoltage/images/power.gif
http://tpucdn.com/reviews/AMD/R9_Fury_X_Overvoltage/images/scaling.gif

Edit. P.S. We are talking about Fury cards in Mac Pro 5.1 and older, or potential cards in Mac Pro 7.1? Are we talking about power consumption of Full Fiji, or down clocked and downvolted Fiji chips that could go into Mac Pro 7.1?
 
Last edited:
Maybe AMD has had the time to improve their drivers and soon we'll see some improvements.
A Nano (or Pico) would in fact look great on the nMP.

16GB should be the minimum now for RAM, but 512GB SSD as base model seem unlikely to me, but that would be great of course. It would make sense if 2TB models would be out by then, but again the cost would go up certainly.
 
http://raytracey.blogspot.com
http://developer.amd.com/tools-and-sdks/graphics-development/firepro-sdk/firerays-sdk/
http://developer.amd.com/community/blog/2015/08/14/amd-firerays-library/

Not bad. ;)

RayTracey blog said:
There are already a few OpenCL based path tracers available today such as Blender's Cycles engine and LuxRays (even V-Ray RT GPU was OpenCL based at some point), but none of them have been able to challenge their CUDA based GPU rendering brethren. AMD's OpenCL dev tools have historically been lagging behind Nvidia's CUDA SDK tools which made compiling large and complex OpenCL kernels a nightmare (splitting the megakernel in smaller parts was the only option). Hopefully the OpenCL developer tools have gotten a makeover as well with the release of this SDK, but at least I'm happy to see AMD taking GPU ray tracing serious. This move could truly bring superfast GPU rendering to the masses and with the two big GPU vendors in the ray tracing race, there will hopefully be more ray tracing specific hardware improvements in future GPU architectures.

It looks like AMD started to take seriously the hardware and software tandem. In the past all what they did was hardware with completely, utterly useless software. Now tho, it turn out they may change that view a bit, which may be good for end users. Mantle, OpenCL drivers, now this API. It looks like it starts to fall in line. A lot of hidden power and capabilities from "old" GCN architecture will come out in future, that is for sure.

One small thing more.

http://www.overclock.net/t/1569897/...singularity-dx12-benchmarks/400#post_24321843
You have answer why DirectX 12 is a big game changer for both Nvidia and AMD. Absolutely phenomenal article. In Ashes of Singularity DirectX 12 there is absolutely no errors. It behave exactly how DX12 is designed to behave.
People wondering why Nvidia is doing a bit better in DX11 than DX12. That's because Nvidia optimized their DX11 path in their drivers for Ashes of the Singularity. With DX12 there are no tangible driver optimizations because the Game Engine speaks almost directly to the Graphics Hardware. So none were made. Nvidia is at the mercy of the programmers talents as well as their own Maxwell architectures thread parallelism performance under DX12. The Devellopers programmed for thread parallelism in Ashes of the Singularity in order to be able to better draw all those objects on the screen. Therefore what were seeing with the Nvidia numbers is the Nvidia draw call bottleneck showing up under DX12. Nvidia works around this with its own optimizations in DX11 by prioritizing workloads and replacing shaders. Yes, the nVIDIA driver contains a compiler which re-compiles and replaces shaders which are not fine tuned to their architecture on a per game basis. NVidia's driver is also Multi-Threaded, making use of the idling CPU cores in order to recompile/replace shaders. The work nVIDIA does in software, under DX11, is the work AMD do in Hardware, under DX12, with their Asynchronous Compute Engines.

PS. Don't count on better Direct X 12 drivers from nVIDIA. DirectX 12 is closer to Metal and it's all on the developer to make efficient use of both nVIDIA and AMDs architectures.

In simple words. Mantle just exposed hardware for software.
 
Last edited:
  • Like
Reactions: Bradleyone
Thats interesting about Broadwell-EP being released in Q4. Things are shaping up nicely for a fall update. Broadwell-EP, Fiji GPUs, Thunderbolt 3 and improved SSD. Not to mention improvements to the platform that would include DDR4 memory. This would prove to be a pretty big step up from the current Mac Pro, as both the CPU and GPU would be two generations newer. Apple could also decide to wait until next year...

Rumors have the Fury Nano coming out next week. It will be interesting to see what Fiji is capable of at a reduced power envelope. Between those Tom's hardware results and the recent Directx 12 benchmarks, Fiji could prove to be a really nice performer, especially in compute and Metal/Directx 12 type work loads.
 
  • Like
Reactions: flat five
Great post there, and makes sense.
Once we get the same performance with Metal the AMD cards might seem a wise choice forma Apple.
I'm still not convinced we'll get Fiji with 7,1 though.
 
Lets sum up a bit.

Fiji GPUs.
Broadwell-EP CPUs.
DDR4 2400 MHz
faster SSDs with up to 2GB/s transfer rates and 2 TB storage.
Thunderbolt 3.
OpenCL drivers for OSX.
RayTracing API for OpenCL in OSX.

That may be quite nice update. Only thing that makes me slightly apprehensive is the TB3. I have no idea how Apple would pack all of this on only 40 lanes of PCiEx.

What I would love to see as an addition to all this is... reliability.
 
Lets sum up a bit.

Fiji GPUs.
Broadwell-EP CPUs.
DDR4 2400 MHz
faster SSDs with up to 2GB/s transfer rates and 2 TB storage.
Thunderbolt 3.
OpenCL drivers for OSX.
RayTracing API for OpenCL in OSX.

That may be quite nice update. Only thing that makes me slightly apprehensive is the TB3. I have no idea how Apple would pack all of this on only 40 lanes of PCiEx.

What I would love to see as an addition to all this is... reliability.

What will the base config be? That's the question. Because if it's some **** graphics that will be slow by 2017 and a quad core CPU and costs £2500UK then I'm still not tempted.
 
... Intel just assured that Broadwell-EP is heading towards us. And its Q4 release date..

Intel's slides were saying Xeon E5 v4 (Broadwell-EP) in Q1-2016 not all that long ago.... sooooo where is the source for this? Whisperings of a friend of a friend in some IDF backroom or what?

Jan-Feb 2016 to Nov-Dec 2015 isn't that big of leap, but not all that long ago that wasn't the roadmap.
 
Intel's slides were saying Xeon E5 v4 (Broadwell-EP) in Q1-2016 not all that long ago.... sooooo where is the source for this? Whisperings of a friend of a friend in some IDF backroom or what?
More or less. According to the "highly reputable" wccftech, a german website was told by an intel rep at IDF that they are shipping this year.

While it is certainly far from a concrete report, it doesn't seem entirely impossible. Even if it was available in limited quantities, it wouldn't be the first time Apple got first dips at a new intel chip.
 
Newbie? Ive been here for years, sure, may have not logged on in a while.

<<they make the AUDIO CHIPS FOR APOLLO UAD, and what I do’t get is where is all the outrage?
...
Anyway, delays were due to Apple trying to make FPU faster thru GPU, never happened, thats why we have n 14//18 core, the ****** littler box can’t go over 135 watts, thus the more cores, the lower the speed, sure more cores are great, but for audio...>>
You had me at Apollo UAD! :) I do understand this. People who don't use multi-track audio recording apps (i.e., Pro Tools and Logic X) generally have a hard time understanding that having a great GPU and an okay-performing multi core CPU doesn't help us get discrete, multi-track, high-sample-rate 24-bit .BWF audio recorded into our rigs with low latency. -This is why the 64-bit single-core performance spec is important to me, even though PT11 can use "several cores" This is why things like the 2010MP 3.33 and maybe a OWC Turnkey Upgrade are looking so great to me.

FWIW, my super-noisy 2003 G5 DP 2 GHz/3GB RAM running 10.4.11 and Pro Tools 7.4 records 16 channels of 24/44.1 .WAV audio via FW400 with a buffer of 256 where most newer machines would drop to their knees. The silent 2012 Mini i7 does well with this, but get's a little warm.

BTW, would you dare buy a Thunderbolt interface these days? :eek:
 
What will the base config be? That's the question. Because if it's some **** graphics that will be slow by 2017 and a quad core CPU and costs £2500UK then I'm still not tempted.
Base config? I will not hazard a guess at this stage ;).

dec60: actually the source for this info is computerbase.de that was told by Intel representative at IDF that the CPUs are coming in Q4.
 
...
dec60: actually the source for this info is computerbase.de that was told by Intel representative at IDF that the CPUs are coming in Q4.

This article?
Somewhat butchered English that Google Translate comes up with
"... Currently the product is already in the big test phase, so that the timetable be complied with, and the latest in the first quarter of 2016, the products can be formally introduced. ... "
http://www.computerbase.de/2015-08/broadwell-ep-xeon-e5-2600-v4-mit-ddr4-2400-zum-jahreswechsel/

If testing goes very smoothly, it would not be surprising to see some early demos or maybe limited to HPC release in November ( Supercomputing '15 conference and the next round of the Top500 stats come out then ). But the nominal market introduction is still on 2016 for the complete line-up. The Core-i7 x9xx variants might go in late Q4 as the testing likely isn't as rigorous. Q4 for the E5 line-up looks like hopeful, best case thinking. If all the stars happen to line up it could. That is not necessarily saying that Intel is setting a hard deadline for Q4. It won't help Intel to just "suppress" a finished product so they could do some releases in Q4.

However, I don't see Apple keeping a production line ramped up with all of the other components waiting for E5 to maybe get green lighted. Apple could actually use some non-holiday Quarter shopping revenue.
 
More or less. According to the "highly reputable" wccftech, a german website was told by an intel rep at IDF that they are shipping this year.

reputable and wccftech in the same sentence is an oxymoron. Their articles are by and large clickbait. I almost always regret clicking on one of their links because that articles are very often flawed . ...... just like this one.
 
The setup should be more or less like that.
Fiji is a little limiting though, so I'm not sure, I'd bet more on Grenada and maybe one Fiji SKU.
However, it would be weird to see Grenada 8GB/16GB and Fiji 4GB. How would they label these cards? Clearly Fiji should be the top D7x0 card, but I don't see lower cards with more mem as a possible line up.
They can however do a normal setup like Antiqua 4GB, Grenada 8GB and Grenada 16GB (although I doubt we'll see a 16GB card) as D310/D510/D710 or whatever they'll be called, and an enthusiast card D800 with Fiji.
But I don't see Apple going enthusiast on nMP, otherwise they'd do a Core i7 CPU as well I guess.
Regarding base config, I don't think we'll see a lot of changes. Quad core should be the choice still (for cost reasons and if Intel didn't kill quads in Broadwell), 16GB mem, same 256GB SSD I reckon and TB3, which is indeed still unknown how it will be implemented (we discussed this back a few pages). Yet, for TB2 they found a solution in the current nMP. Now, the requirements are higher but at least USB3 is on the PCH now. A switch will have to be used somewhere.

Still, great news regarding raytracing. I hope the trend picks up fast.
 
Hello Manuel,:)

One thing that I believe is crucial, is that there is need (?) for a small upgrade to the PSU, I think that this should be the place to begin with, all the other parts are relying on the PSU, a little better one (more powerful) should open better possibilities for the new nMP and his options, especially GPUs.

We have probably mentioned it previously but the PSU is the base of the system.

Obviously I really don't know if it is possible to enlarge the PSU and have the absolutely same dimensions, obviously the thermals should rise.
 
Last edited:
Skylake 6700K are already being reviewed and the quad core stacks up pretty well against older CPUs with more cores. Reviewers have clocked them at 4.6 and 4.8ghz with standard cooling. At 4ghz the quad core is already achieving 28,000 geekbenches. We definitely don't need Xeons for single CPU systems unless there's some kind of scientific need for it or because Apple wants to maintain high prices.
 
  • Like
Reactions: JamesPDX
As for the GPU I personally don't have high expectations because of,
a. the thermal envelope,
b. power requirements,
c. Apple's tradition with GPUs, as long as I remember, there has never been offered in the past (officially) a high end card option, especially in desktops. MacBooks had better treatment.
 
However, I don't see Apple keeping a production line ramped up with all of the other components waiting for E5 to maybe get green lighted. Apple could actually use some non-holiday Quarter shopping revenue.

It could fit their timetable to announce the Mac Pro update in October along with other Mac focused refreshes - OS X El Capitan, iMacs big & small, the Apple Retina 5K Display, etc - and then actually make the Mac Pro available to order midnight December 19, 2015.

That would be exactly 2 years to the day that the nMP 6,1 became available. As one of those watching when it went live, and seeing the orders almost immediately balloon out to February and beyond, it would fit Apple's make-em-wait-dammit schedule (or, you'll-get-your-Mac-when-we-get-our-components).

And I of course would curse myself for not getting my order in the minute after midnight (of all things, my credit limit didn't go that high) and petulantly decide not to buy one at all and build a much better hackintosh instead that caused no end of frustration and... this time, this time, I'll be ready!
 
  • Like
Reactions: JamesPDX and filmak
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.