Yes, the more recent implementation doesn't use the physical link anymore. And in the case of dual gpu card, that point is moot anyway since both processor are on the same pcb.
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.
Will there be a new Mac Pro in 2015?
- Thread starter Anto38x
- Start date
- Sort by reaction score
- Status
Yes, the more recent implementation doesn't use the physical link anymore. And in the case of dual gpu card, that point is moot anyway since both processor are on the same pcb.
Metal isn't not a cure all that solves world hunger. The point of Crossfire is to present a virtual GPU which transparently divides the work and merges the results together to output one set of ports. Metal does little to nothing to remove that since that should be happening at the next level down. Metal isn't the lowest level graphics driver present.
Metal does have a shared memory model (concurrent, coherent access is handled) on some implementations. All the data for a computation or task doesn't have to be present on the GPU. You don't have to make a local copy just to access the data (besides copies via cache mechanism hierarchy). But that goes to why statically overstuffing the GPU in the first place. If can easily get data from memory in another GPU card, you can get it from the CPU also. If the interconnect is PCIe v3 to both it really isn't all that different.
And what It will give? Im not an expert but I think it was good for situations where your game was "thinking" in streamlined way. And I think that Asynchronous Shaders was here to get rid of it. Because you just simply get the workload for what resources you have. Not synch them.
You are mental. Seriously, flat-out mental. I see my previous comment directed towards you was deleted but please be advised that it still stands.
Crossfire is used so that your GPU will do parallel task between them... How can you do something in parallel if you don't know the status of the other task being executed?
Let it be... He's a pro at getting people in time out around here. Just put him on ignore and be done with it.
Yeah, noted. I gathered that about him already but still... can you imagine what it'd be like to work with him? Dude must've sent more people Postal than My Lai.
Both at the same nominal design power draw and clock rates? Probably none. The water cooled unit is over resourced for overclocking, but that probably isn't the default, "out of the box" mode. If run within the parameters I'm pretty sure you are not going to find any substantive differences. Those who want to run a normal stock card, at stock speeds will buy the cheaper air cooled unit. It will be cheaper and simpler to just plug/swap into the computer and just go.
The folks chasing max possible performance at any cost will get the water cooled unit. It is rated to dissipate up 400-500W. That is about the whole Mac Pro power supply. They'll need internal room for the hose and a home for the case edge fan the dumps the heat.
Don't really need one if not trying to maximize power leakage and inching forward on a diminishing performance/Watt curve. The central thermal core basically serves the same purpose ( centralize and exchange heat through a larger diameter fan at the edge of enclosure. ).
The current Mac Pro design is diametrically opposed to overclocking though. It is design where the elements are thermal balanced by Apple. The range of dynamic shifting is more down than up over specs. The power supply is scoped to be just enough to get the job done on a quite broad wide range of workloads ( but not any possible workload). .
Well we don't know yet if the air cooled fury will be spec the same as the wc one. Time will tell especially since the water cooled one is released first. Both AMD and NVidia release their reference model first historically. It's a first, to my knowledge, for a GPU maker to release the oc model first. If the wc fury x can be tweaked to higher performance with less noise (1 fan vs 2 or 3), who will buy it?
Yeah, he frustrates me as well, so much so I just bought the new DLC today for the game Postal 2! Thanks for putting 2+2 together for me. Wondered why I got reinterested in that game! LOL!
I gave up and now ignore him in case you hadn't noticed.
Yes, that is a problem. I forgot completely about Parallel Computing. Thanks for that.
Project Quantum seems to have an i7 processor, even AMD thinks it's a better proc for a gaming machine. I'd imagine they'd use their own proc, even with lower performance, but hey...
Fury X2 inside, what a monster...
Only Nano's specs are not officially known, Fury's are.
Nano looks really good, 175W and faster than Grenada.
The fact that it is closer to Antiqua is a good sign, massive bandwidth with extra compression, is it all needed?
Benchmarks for Fury X are also out, provided by AMD but seem legit.
Fury X2 inside, what a monster...
Only Nano's specs are not officially known, Fury's are.
Nano looks really good, 175W and faster than Grenada.
The fact that it is closer to Antiqua is a good sign, massive bandwidth with extra compression, is it all needed?
Benchmarks for Fury X are also out, provided by AMD but seem legit.
don't you think people get themselves in timeouts?
i challenge people's opinions with logic.. they argue back with emotions.. and say even stupider crap than they began with.. (like blegg did when he started talking about roosters.. or smthg).
my very main argument in this whole forum is many of you take this stuff waayy too seriously.. place way too much importance on the computer itself for creative work when really, the computer ranks very low on the totem pole as to what is needed..
you want to do great work? you need a decent computer.. that's that.
have fun talking about specs and new tech and what not.. it's a sweet little thing to be interested in. but keep it in perspective.
Last edited:
i don't get it. (not much of a gamer).. what's DLC?
---
---
(edit- sorry.. not used to the new forum software.. i thought this post would of be added to the previous one.. guess not)
I think there is a difference between AMD's implementations and what Fury X ( 4096 SP ) and Fury ( fewer ) are. I don't think there is a requirement that the other card implementers have to use water cooling on Fury X (and vice versa on Fury). That is just what AMD's cards do. I think there are going to be some card vendors jockeying for position on volume with various feature mixes and "defacto" differences in street prices (as opposed to the suggested one).
Tactically AMD is in a hole. Nvidia has had its 8B transistor offering on the market for 3 months. If production yields are slowly ramping then the lower volume of the higher priced card are easier to do. It is a bit less risky for something no one has done in volume yet (HBM+GPU+interposer packages ) . The prices are higher so the margin should be a bit higher in magnitude ( if not percentage ). AMD has a red ink problem too. Cash flow from sales needs to trend up sooner rather than later.
Also tactically I think AMD wants to draw out some risk takers ( not just plug in and go ). AMD suffers a bit from primarily just being the "cheaper" option. The lowest stuff is toward the end because (presuming don't have yield problems) it probably will go over well in their core market. The dual doesn't make sense until can get enough volume to fulfill the other three.
Looping back to the Mac Pro, if Apple could consume a signficant chunk after the initial wave starts to cool off that would help AMD drive more volume when get to the higher yield zone. Competitively priced that could be a win-win for both sides.
Even if slightly behind the Nvidia option if it represents a higher performance/$ .... probably enough buyers to make it worthwhile.
Some say, that in drivers of 10.11 are "traces" of ... Fury X.
Well. I think you can see where this is going
.
If they were able to squeeze 3584 GCN core GPU to 175W of TDP, im wondering how flixible is that GPU in temrs of performance... We already know that Hawaii, really badly binned GPU was able to go at 85-90% of regular performance at roughly 145W of power. Im wondering, how good will be Fury, at this...
Or they will give more power to the power supply, and will make it water cooled. That would be huge for MP...
Well. I think you can see where this is going
.If they were able to squeeze 3584 GCN core GPU to 175W of TDP, im wondering how flixible is that GPU in temrs of performance... We already know that Hawaii, really badly binned GPU was able to go at 85-90% of regular performance at roughly 145W of power. Im wondering, how good will be Fury, at this...
Or they will give more power to the power supply, and will make it water cooled. That would be huge for MP...
Some say, that in drivers of 10.11 are "traces" of ... Fury X.
Well. I think you can see where this is going
If they were able to squeeze 3584 GCN core GPU to 175W of TDP, im wondering how flixible is that GPU in temrs of performance... We already know that Hawaii, really badly binned GPU was able to go at 85-90% of regular performance at roughly 145W of power. Im wondering, how good will be Fury, at this...
Or they will give more power to the power supply, and will make it water cooled. That would be huge for MP...
Can't water cool the nMP without a whole new design of the case...
Same for the PSU. The thermal enveloppe of the nMP is already pretty much to its limit.
No space for it. Liquid colling take quite a bit of space. It could be done but it would mean a new design.
Yeah, I don't see Apple doing liquid cooling. They tried it once with the Power Mac G5. There are a few problems with it.
1. Reliability. The Power Mac G5 was infamous with issues regarding the liquid cooling. It seems like the all in one coolers have come a ways since then, but as far as I know, you don't see them a lot on workstation type machines.
2. Noise. There is additional noise due to the flowing coolant, which is audible at idle and louder than a large diameter, slowly rotating fan. The gain of course is better cooling and lower noise at full load. You can probably mask this somewhat in a giant tower, but less so in a small form factor.
3. Radiator size. You still don't get around the problem of blowing air over a large surface area to cool it.
My guess is:
Q1-Q2 2016
Intel Xeon E5 v4 (Broadwell EP)
DDR4-2400 RAM (16GB min)
New PCI-e 3.0 SSD (256GB min, but cheaper 512GB-1TB options)
Double internal slot for SSD (1 left empty in the 256GB version)
6 Thunderbolt 3/USB 3.1 ports
4 USB 3.0 ports
AMD FirePro D320-520-720 (based on Titan or Fury)
Has High Density 3D Memory been implemented by AMD?
My secret hope is to see some "Jobs style" anticipated purchase of Skylake EP processors, like they did with the first Xeon (first Intel Mac Pro). But I think we need to see whether Intel wants to skip the Broadwell EP generation.
Q1-Q2 2016
Intel Xeon E5 v4 (Broadwell EP)
DDR4-2400 RAM (16GB min)
New PCI-e 3.0 SSD (256GB min, but cheaper 512GB-1TB options)
Double internal slot for SSD (1 left empty in the 256GB version)
6 Thunderbolt 3/USB 3.1 ports
4 USB 3.0 ports
AMD FirePro D320-520-720 (based on Titan or Fury)
Has High Density 3D Memory been implemented by AMD?
My secret hope is to see some "Jobs style" anticipated purchase of Skylake EP processors, like they did with the first Xeon (first Intel Mac Pro). But I think we need to see whether Intel wants to skip the Broadwell EP generation.
- Status