For 150W, and less than $400, I'll take it.Even If I do not see half of the posts here because of ignored content I can imagine what is happening.
Anyways. First GTX 1070 review was up yesterday. It looks like it is not faster than Titan X, not faster than Fury X even, mostly on par with GTX 980 Ti and Fury. In some resolutions it is heavily bottlenecked by GDDR5 memory bandwidth.
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
I say GPUs will get x16 each, a 20 lane switch will take 3TB3 into the 8 reaming CPU lanes.
NVMe SSD (and VVMe has nothing to do with number of lanes or PCIe revision) will still be on PCH.
Or, SSD will also share the CPU lanes with a 24 lane switch instead.
AMD will not provide very much info on Polaris at Macau, they'll wait till June 29th to fully disclose the specs and leave us wondering. They'll just show off some undisclosed card and let us see some scores, of power consumption mostly as well.
NVMe SSD (and VVMe has nothing to do with number of lanes or PCIe revision) will still be on PCH.
Or, SSD will also share the CPU lanes with a 24 lane switch instead.
AMD will not provide very much info on Polaris at Macau, they'll wait till June 29th to fully disclose the specs and leave us wondering. They'll just show off some undisclosed card and let us see some scores, of power consumption mostly as well.
heh, a lot of what's said here has nothing to do with using the computers.. all of the proof and whatnot of backing these things up are gaming benchmarks..
i mean, A, it's just a benchmark.. and those are of very (Very!) low importance when it comes to gauging productivity and usability.
and B, on top of that, they're freaking gaming benchmarks.
idk, i hope apple reads these threads.. but i also hope they're reading with a grain of salt.
..and continue making computers that work well in the areas they're being designed for.
Nope. Disclosure is on 31st May. The reviews go live 29th June.
Last edited:
Oh?
Is that what that was all about?
Please, take one my GPUs. Apparently, I may not need it.
Now, back to productivity.
It pays off.
yeah.. you see all the graphics and links that keep popping up around here.. like:
...
what that's telling you is : IF you happened to be using this game on your computer AND you could have an unlimited amount of bomb-blastics (which is against the rules of the game.. you only get three of those).. And, you put your avatar in an exact place in particular scene which had a helicopter flying by in the background..
it tells you how many times you could throw your unlimited bombthing before the helicopter in the background stutters across the screen instead of moves more smoothly across the screen.
Last edited:
You're funny
Facts: PEX-swtiches allow HW desginers freedom to integrate whatever they want, sharing efficiently bandwidth (sometimes even better than w/o PEX), allows with 40 PCIE lines 2 GPU at full 16x pcie each, 3 or even more Thunderbolt 3, 10GbT, etc, the added cost its less than 100$ this debunks all your comentaries about PEX-Switches, The New MacPro is feasible with everything apple would want to include, further Apple don't need to change GPU's form factor even they can recicle for PCIe3 NVMe the current PCIe2 SSD provisions (while very likely and convenient apple to change the connector keying, to avoid cross-installing PCIe3 NVMe <=> PCIe2 SSD).
Fact: every single time we debunk your "Apple Crap Fail" theories you instead to support your arguments with new more solid arguments, you opt to disqualify, to self-promote as "tech guru", and using every common logic phalacy, very obvious sir.
You seems to ignore, this thread is followed by actual IT Related Engineers, I'll wont speak on my self, but many people here educated this thread with their knowledge, you simple use these thread as katarsis against Evil Apple or whatever is in your agenda, Please Respect, We Know Apple is Ruled By a Vacuum Salesman minded CEO, most its designers should resign and go to work at Vogue, and Apple Will do Whaever in they evil mind to anticipate user's unecessary purchase-update, all we know the evil in Apple, but WE DECIDE to Buy Apple, the nMP isn't the fastest WS possible but is an Powerful WS enough for most user need, sadly for your business model it will not feed you with dumb enthusiast looking for the latest GPU, but this machine for most us, help to run our operations, mine cost more than 7grand and it payed itself many many times, I love it to the extreme, not perfect but perfect for me (and thousands buying it again).
Sorry sir, the Little Black Trash Can, comes back with new superpowers, at time for most cMP end-of-life replacement, very likely the new GPU's wont be available for current nMP but isn't nothing technically impossible and actually apple can do it if they want (and would be and excellent move to gain User Confidence).
I've received some DM on the actual performance of the upcoming nMP and only I can say it will be socking impressive, I believe they true and the source asked me not to comment, but I can say with absolute confidence the nMP 7,1 will shut up many big mouths here.
Respect.
Last edited:
After reading the PEX-97xx capabilities (read my last link), seems a logic path to discard the 8 PCIe2 lines from the PCH and wire everything (WIFI also) to the PEX, Koyoot also have another possibility: using new AMD Fabric to interconnect both GPU w/o use PCIe lines its another option to resign the PEX (GPUs on 16x 8x, TB3 on 12x and the remaining 4 PCIE3 to the NVME, using the PCH to feed the 10GbT or the wireless (apple may opt not to include 10GbT given TB3 includes 10GbT Mac only requiring a TB3 Cable or an Tb3<->SFP+/10GbT adapter.)
Last edited:
You would have free 24 PCIe 3.0 lanes for everything on external level. Feeding the GPUs would not be a problem because without internal coherent fabric they would require 16x lanes of PCIe. When they are connected internally - they don't. AMD is working on that fabric already that will be available for ALL x86 CPUs, both Intel and AMD. When they will show it? Latest rumors point to September-October 2016 timespan. You may confuse it with the Fabric that was lately announced with other companies. That is "external" coherent fabric. We are talking about internal one.
P.S. It is quite funny that internal fabric's will by design lock your ability to upgrade your hardware in future. Think about this in the context of Mac Pro. I do not know why it is that way, however.
another question is, its cheaper new AMD's fabric or a PEX-Switch ?
I'm still on the PEX boat, one or two handling both GPUs TB3 NVMe and 10GbT.
If you would want HSAIL and the ability to see both GPUs as one big execution unit - you want coherent fabric.
If you don't need that - Switch. There are a lot of benefits of using Fabric, instead of switch, like for example increasing the efficiency if your GPUs are able to power gate itself, depending on the workload.
Dear lord. How outdated currently OS X platform looks when we think of possibilities.
i think i only champion the nmp when the discussion veers towards the design & engineering of the thing..
specs and whatnot, i generally stay out of those topics due to lack of interest so i don't think i'm hyping up mp in those regards..
a lot of my arguments around here are specifically at other people's logic, who are generally saying negative things about 6,1 since we happen to be in a mac pro forum.. but for clarity, i don't think i'm arguing in defense of 6,1.. more like arguing in offense towards other people's reasonings.
within the past 1 year, my hardware needs have been narrowed down to one specific spec.. cpu clock speed.
the faster that thing goes, the smoother i'm working.
in the past, that spec most definitely required a workstation..not so much any more.. also in the past, my workflow required, or could make use of, a whole bunch of cpu cores.. like a plethora of them.. problem is, obtaining those cores to max my workflow was way too expensive.
through advancements in software, my rendering requirements are now (like very recently) occurring on gpu.. i literally bought zero new hardware, installed a software update, and am now rendering locally at up to 10x speeds.
if i did this through past conventions, i'd be the proud new owner of a 40core computer.. how much does that cost? $25000 or something?
so, free vs $25k.. hmm
further -- options have also recently opened up for me to go to cloud with this same task and get 40x speed increases.. for around $1 each..
again.. hmm.
the problems associated with yesteryear's demanding workflows have been leapfrogged, multiple times, by software.
not only is it much (very much) cheaper for the user, the performance enhancements are much greater too.
this forum completely ignores using the computer and using software and would lead anyone reading this stuff (and most of the people who write it believe it as well) to conclude that the only way to improving productivity is via more expensive and/or newer hardware technology..
[edit] for example.. dude in the post just above this one was asked a question about software and user tasks.. i'm now being called a troll for asking such things.. it's not welcome around here[/edit]
every day that passes, this 90s view on computing becomes less and less true or relevant.
no, i don't think that.
i understand things like video contain a heck of a lot of data.. it's the nature of the medium.
but i don't work in video.. so i brag that i don't need a whole lot of cables or drives to hold my work.. in fact, i have no wires currently except display and power cables (and a usb keyboard)..
i don't know, i'm getting mixed messages regarding this.. people complain around here because of rats nests of cables and what not but then, if someone doesn't have a bunch of cables, they're not capable of saying something ala "wow.. lucky you"... instead it's "well your work is lightweight.. you're not a pro.. you need a mac mini." : /
(and hey, the only reason i'm able to connect wirelessly is because of recent hardware advancements (well, also some newer software stuff and services available) .. it's another thing, for me, that i needed in the past but no longer.. again, tech is improving people's workstations in many more ways than this forum's grails
that's not true.
i think users with applications which are highly reliant on clock speed (and this is literally about 97% of professional applications out there) will have smoother flow on an imac vs a 22core cpu.
to me, in my opinion based on all of the various computing environments i'm around, i feel the best generalized mac configuration for today's workstation users would be a 6 or 8 core 6,1 with mid gpus, 32-64 ram.. 1TB SSD.
that would be a very good general system for most, by far, users with demanding needs.
not an imac.
people don't use 44core machines.. and if they do, they've wasted their money.
44core machine should be used in scenarios where 44cores are constantly, or near constantly, being pegged.
if you're on a 44core machine and it's not being maxed out, you've wasted your money. use a much cheaper computer with the amount of cores you typically use.
if the 44cores are being pegged then you're not using the computer.. it's in use and not available for user input.
right.. not option for osx on 44core machines.
(well, i use 48,000 core machines and i'm on os x so this isn't entirely accurate.. if you want 44core machines as a number cruncher and run your application through osx then this is doable)
Last edited:
I know you are not allowed to comment on 7,1....but it would be great if GPU was upgradeable.
That would be me.
I want to know if I'm heading in the right direction with adding more Mac Pros.
It feels right so far.
Ah, yes. We shouldn't compare the 6,1 to real computers using real tests because it might look slow, or ridiculous, or overpriced. We should only compare it to other "special" computers using "special" tests where it will have a chance to look good.
Great idea. If only there were another retailer selling "new" 7970s in 2016. But neither Dell nor hp has the gall to sell such things as "new" anymore.
Speaking of which, where are those GPU upgrades you predicted 3 years ago?
Still "right around the corner"?
Last edited by a moderator:
Yawn.
Can we get back to discussing what the chances are of a nMP this year and what it might have onboard based on what is almost out there now in terms of parts?
As graphic designer with 20years of Mac use behind me I'd like a desktop computer that will drive a 30" or better color accurate monitor and run photoshop filters in as near real time as possible (files up to 1Gb). I'd like it to be thermally capable of being left switched on 24/7 without overheating (I don't ned something that can render 24/7). Pre-cheesegraters I ran a couple of iMacs but they both failed after about 3 years. I realise the 6.1 isn't designed for me but I'm hopeful that the 7.1 (if it ever materialises) although again, not a graphics machine, might be more useful than an HP workstation for my purposes. I know there are flaws in my argument but I'm not asking for a critique, just some better quality discussion on what might be upcoming.
Thank you.
Can we get back to discussing what the chances are of a nMP this year and what it might have onboard based on what is almost out there now in terms of parts?
As graphic designer with 20years of Mac use behind me I'd like a desktop computer that will drive a 30" or better color accurate monitor and run photoshop filters in as near real time as possible (files up to 1Gb). I'd like it to be thermally capable of being left switched on 24/7 without overheating (I don't ned something that can render 24/7). Pre-cheesegraters I ran a couple of iMacs but they both failed after about 3 years. I realise the 6.1 isn't designed for me but I'm hopeful that the 7.1 (if it ever materialises) although again, not a graphics machine, might be more useful than an HP workstation for my purposes. I know there are flaws in my argument but I'm not asking for a critique, just some better quality discussion on what might be upcoming.
Thank you.
Actually this part is 100% right. Whether you like the nMP or not, saying that it’s blazing fast with Final CutX or whatever it’s called is just another benchmark. Just one in the form of an application that happens to be made by Apple. We should only be looking at cross platform benchmaks where you can level the playing field as much as possible or not at all.
Last edited by a moderator:
http://www.anandtech.com/show/10375/arm-unveils-bifrost-and-mali-g71/4
Heh. Heterogenous architecture came to mobile front. Very, very powerful in mobile space, and very interesting in the context of upcoming world.
Even this GPU has Asynchronous Compute Engines.
That is another rock in the garden of ideas what we have been discussing over last few pages.
Heh. Heterogenous architecture came to mobile front. Very, very powerful in mobile space, and very interesting in the context of upcoming world.
Even this GPU has Asynchronous Compute Engines.
That is another rock in the garden of ideas what we have been discussing over last few pages.
I used to do a lot of multi-gigapixel computational photography, and similar size to you photoshop files on a 2010 mini with 8gb of ram - honestly, unless you need more than 16 (and hopefully 32gb soon) of ram you may find the sweet spot for your workflow is a mini with a eGPU, and with all hints that the upcoming (as in "released to manufacture") Retina 5k display IS an eEPU, TB3 eGPUs may well be the "non imac" less over-specced than a Mac Pro sweet spot you're looking for.
I went for a 2009 pro because I needed bulk ram for the panoramic stitching, which eats 20+ GB of ram during the render (and seeing all 16 - soon to be 24 virtual cores pegged to 100% with no change in noise levels has its certain jollies), but it I hadn't needed it at the time, for specific images I needed to make, I probably would have waited to see what happened in the next couple of months.
For 3k I would expect more than mid-range GPU...Anything with Dual R9 390X level of performance, 6 core Broadwell-E/EP CPU 32 GB of RAM and at least 512 GB's of SSD will fit my needs perfectly.
Oh, and if it would cost 3000$. That would be even more perfect
Ok, i might do some wishing too then
With Apple PCB shrinking knowhow, build a dual GPU board and put it where the current CPU is, this would give space for making a dual CPU nMP on the 2 other sides of the thermal core, 4 dimm slots per CPU would be nice too
So, configuration as follow
2 E5-2687W
256GB ram
2 Pascal GPU:s with GTX 1080 performance
2 NVMe drives
6-8 Thunderbolt 3 ports
2 10Gbit RJ45 ports
This will never happen though, i really wish, even with the current nMP there would have been a config option for doing 2 CPU:s and 1 GPU instead of 1 CPU and 2 GPU:s but this is also very unlike so after the nnMP reveal this year i will most likely just buy a real workstation from Dell or HP instead
With Apple PCB shrinking knowhow, build a dual GPU board and put it where the current CPU is, this would give space for making a dual CPU nMP on the 2 other sides of the thermal core, 4 dimm slots per CPU would be nice too
So, configuration as follow
2 E5-2687W
256GB ram
2 Pascal GPU:s with GTX 1080 performance
2 NVMe drives
6-8 Thunderbolt 3 ports
2 10Gbit RJ45 ports
This will never happen though, i really wish, even with the current nMP there would have been a config option for doing 2 CPU:s and 1 GPU instead of 1 CPU and 2 GPU:s but this is also very unlike so after the nnMP reveal this year i will most likely just buy a real workstation from Dell or HP instead
Who knows...but the chance of releasing it is higher. It may be another fcpx machine...but I wouldn't know. I'm just basing this on the last model. It was mentioned by someone that it suppose to take you from planet Yavin to Death Star in less 30 seconds with awesome power.
[doublepost=1464625328][/doublepost]
I was wondering. When Sterve Jobs was still alive, did he focus (I know he was very focused with his products) on making Mac pros to be more like z1 workstation?
- Status