@Psycogeek and that's not how "walking on water" was described
first you would need to make the computer sentient, then convince it that there some god and then convince it that, if the computer trusts in said god, it will not sink
Anyone know the agency responsible for appearing on MTV's show "Jackass" :-). I recently found out I am perfect for that show. NMS is a bad port because it's a console application that only later got a PC version, and they made no PC modifications, to the interface. Additionally they used OpenGL instead of Vulkan/DX12 on PC
What can I say, nothing could upset me, tonight.
Except hearing NMS is a horrible port considering I almost got that horrible port on Tuesday :$
@Dog Didn't they say the "mobile" 1080 would just be the desktop 1080 in MXM with perhaps some software/firmware thermal throttling as-needed (depending on how good the laptop form factor heat dissipation can be)?
@Dog I can see the argument for size (more a concern about the size of the cooling solution, not the size of the chip - chip is tiny), but battery life concerns should easily be mitigated by turning off some cores or something
I've been wishing for a system with a very high-end desktop GPU that will turn off like 50-75% of the cores on battery power and basically give you a GPU like we have in full-fat laptops today; but when you plug in the A/C, it operates as fast as the thermal solution allows, approaching actual desktop performance
CPU should just do CPU stuff, then have a full-fat Nvidia GPU that just uses as much power as the current power solution (battery or A/C) can reasonably pull without killing your battery in 3 minutes
@allquicatic Yeah, but for the past few years, NVidia/AMD dGPUs have still been fabbed on 28nm and were considerably less efficient than Intel's 14nm iGPUs
Case in point: a 28w i7 with Iris graphics performed the same as an NVidia 35-45w dGPU
@Dog I don't think the fab size really had much to do with it; it was more that Intel's GPU architecture is designed from the ground up to be low power and power efficient, whereas Nvidia and AMD's chips are designed to run hotter and hungrier for performance
although now that they are nearly exactly the same fab size, and Intel won't be shrinking for several more years, that argument is moot
@allquicatic If an Intel iGPU can perform the same using 15w as a dGPU that uses 35w+ then it doesn't really matter why, its still a league of difference.
Thankfully Pascal seems to have caught up in efficiency, now I'm just waiting to be able to buy it in a <14" machine
Unfortunately AMD's Polaris doesn't seem to be quite as efficient... though some leaked benches suggest it might be on par, actual products seem to be still only as efficient as NVidia's Maxwell 2
@allquicatic I'm still dreaming of the day I can lug a lightweight 12-14" machine around that can do everything - tablet, laptop, gaming desktop (when docked)
@Dog I don't think Polaris is really, in actuality, competitive with Pascal in any real material way, so all AMD can do is try to finagle the value proposition of price:performance using marketing tactics to try and carve out a small share of the market somewhere where they're "better" for the price and maybe not go bankrupt
if you held a gun to AMD's head and made them release the full-scale Polaris chip at the fastest practical performance level this year, I don't think it would match the 1080
@Dog Polaris and Vega are still fundamentally based on the GCN architecture, just with iterative enhancements and different design goals, but it's fundamentally the same thing; they haven't completely reinvented the wheel with either architecture, and they probably have more similarities than differences, even comparing Polaris/Vega to, say, Vega and GCN 1.0
if they can't compete with the 1080 with Polaris, that's only because they chose not to scale it up to the transistor count / shader count / VRAM size / TDP comparable to the 1080
@Dog just seems to me like AMD is admitting defeat -- I mean, why can't they release a >= 150W chip with >= 8 GB GDDR 5X that goes toe to toe with the 1080? Why wait a minimum of 3-4 months (and possibly longer if it gets delayed) to release your counter offer, long after they've sold 1080s like hotcakes for months?
I mean NVidia's not in a position to release a HBM2 flagship yet either. And if AMD designed Vega from the ground up to use HBM2, it's going to be more work to re-adapt it to GDDR5(X).
Which makes me wonder if the rumoured/leaked RX 490 is actually going to be a Vega part, I'd be thinking more along the lines of Fury X2 or summit.
Then they're fools for not having a "bridge" chip between GDDR5 and HBM2. I'm sure Nvidia will get around to HBM2 eventually too, but they put in the extra work to sell a "good enough for now" chip while HBM2 is in development.
@Dog Polaris is targeting only a small subset of the market, though! Not the entire market. Pascal, as it stands, right now, is targeting the entire dGPU market
@Dog Gross revenue or unit volume? Very important distinction. Me paying $700 to Nvidia (eventually, after it filters down through the profit of the vendors, tax etc.) pays for more than a couple entire units of GTX 1060 sales.
@allquicatic And IMO NVidia wasted a lot of resources to do that... building a "bridge chip" between GDDR5 and HBM2? A dead-end architecture that'll be replaced in 3-6 months? Meh. That said, it does look like Pascal as a whole is adapted to use both, which I guess was a good play.
@allquicatic Both. If 90% of sales are non-flagships, you'd have to price all their non-flagship cards at an average price of $60 before they get beat in revenue.
@Dog Is Nvidia really going to release a new flagship based on HBM2 in 3 - 6 months? I doubt it. I think the only reason they'd do that is if, somehow, just having HBM2 enables AMD to absolutely smoke Nvidia on price/performance and take the top-end performance crown from them.
Pascal's "full implementation" chip already runs on HBM2 - so it's just scaling it to the point they can sell it to consumers instead of the Tesla P100 at $5000 a pop or something
@Dog Are you sure? If they do to the 1080 Ti what they did with the 980 Ti, it will just be the (new) Titan X with less of an emphasis on double-precision FP64, higher clocks, and about 1% better performance while being priced more than the 1080 and less than the Titan.
@Dog Ah, I thought they tweaked the Titans to focus more on FP64 than the standard consumer GeForces, but apparently they don't
if you want double precision you still either buy an AMD card (any card, really, even the consumer ones are great) or Tesla...
@Dog I think Nvidia will release HBM2 consumer cards after AMD does. But I don't think that the mere fact they're late on HBM2 will have an effect on market share, or necessarily even price/performance or top-end performance, at least in gaming, because memory bandwidth is not universally recognized as the gaming bottleneck anyway.
@allquicatic HBM2 isn't the be all and end all, but Vega isn't just Polaris + HBM2. Again, looking back at the Fury X it'll be a scaled up architecture and craptons of memory bandwidth
(And if you're lucky, a swanky watercooler slapped on too)
But from what I can tell, HBM2 is an integral part of how Vega was designed, hence not being released with GDDR5X
I would be happy if Vega could compete with or even beat the 1080, but something tells me that the release schedule will go something like this: 1080 -> Vega -> 1080 Ti, and the Ti will reclaim the Nvidia performance crown after a ridiculously short Vega kingship or tie, and the 1080 Ti won't even need HBM2 to do it.
Or it'll go 1080 -> 1080 Ti -> Vega, in which case Vega may never take a performance crown, not even for one second.
@allquicatic Maybe. But again, few people actually buy the halo/flagship products. Even I rarely touch the top card. Last time I bought a flagship card was 2009
@allquicatic Quite possible. After all NVidia did only bring out the 980 Ti to prevent AMD's Fury X taking the performance crown.
> Tech site PCPer got its hands on the latest GPU market report from Mercury Research, and relays that AMD increased its market share, quarter-to-quarter by 3.2 per cent, to gain a total 29.4 per cent of the discrete GPU market in Q1 2016. It is noted that while AMD has achieved a decent gain, the Radeon range has, in recent history, enjoyed a much larger slice of the GPU market share pie.
Seems like every other dude has a 1070 or 1080. Or two.
Unless you want to talk about people who aren't even gamers at all, who might buy whatever random crap is at Best Buy for $100, but then those people almost never even need a dGPU to begin with.
@Dog Still, the profit margins on flagship GPUs are extremely high, and they still sell millions of them (maybe not many million 1080s, but certainly when you rope in the 1070s, it's going to be many many millions). That's billions of dollars.
@Dog Pascal "just" came out, though. You can't expect everyone to upgrade within the first 6 months of it coming out. But the 1070 will replace a large chunk of those 970s by this time next year.
@Dog I consider an x70 card to be a flagship. It basically is. It's damn close.
A 970/1070 is still an extremely expensive card by most peoples' standards. It's still a ridiculously fast card even by Star Citizen / Crysis standards. It's a damn flagship.
@Dog My point is more that, whether you're AMD or Nvidia, if you are going to release a card like the 1080, then releasing a card like the 1070 is an extremely simple and low-cost engineering effort that involves part binning, disabling a few cores, docking some VRAM and calling it a day. It's effectively the same chip, just mildly cut-down. It's not a whole different animal. It's the same family chip and it's in the same market bracket (high-end).
You wouldn't release just a 1070 and be like "Welp, that's all for this year, folks!" -- and neither would you just drop a 1080 and call it done. It's a one-two punch. "Can't afford the 1080? NP, spend a few hundred less on this thing that delivers 80% the perf."
But AMD skipped out entirely on releasing anything flagship-caliber in any time frame that could be considered (in this fast-moving economy) to be contemporaneous with the 1080.
Heck the fact they manage to still be in the game, while fighting companies with four times the revenue or more in both CPU and GPU markets is impressive
I was convinced five years ago they'd lost all hope and would die a painful death. Yet they still keep going...
@allquicatic Part of AMD's "excuse" for Polaris is they'd rather concentrate on improving their drivers than releasing "top end" hardware with crappy software support
And now AMD have been releasing better drivers more often.
@JourneymanGeek And Nvidia could literally send AMD to Chapter 11 bankruptcy court by docking their prices on the GTX 1060 or 1070 by a few percentage points. Bam; goodbye AMD. Like squishing a bug.
You can't have it both ways... say it's a bad decision to not release a high-end chip, then also say working on their shitty drivers is a bad idea too.
@Dog Yes, you can have it both ways. It's easy. You just have a company that has enough money to hire enough engineers to do both. Nvidia does it. I have yet to hit a single bug with the 1080 despite starting out using it with the initial release driver. The initial release! AMD's initial release drivers are always absolute garbage.
@Dog Back when AMD was riding high on extremely competitive and impressive chips like Evergreen (HD5870), they could afford to do shit like that. Life was good for AMD's graphics division. Quite good.
They were in a position of strength to potentially even cower Nvidia in the market (like Nvidia is now doing to them) and they slipped. Multiple times.
@Dog It's called VC and/or debt. Unfortunately, I doubt many VCs want to bet on AMD, and they have too much debt already. If they had a clean bill of financial health to start from, they could probably rustle up some VC or debt, put in the engineering hours to get it done, and compete properly.
@Dog It wasn't an argument. Not every single thing I say is pointed at trying to say you're wrong. If you're right, I'm not trying to disagree. And in that case you are right. They're in a tough spot financially.
They need some way to play even tougher, though, and just see it through. A monopoly market is bad for us all.
@JourneymanGeek Intel has been at that point for a while; did you see how long they sustained Itanium on life support, continuing to iterate new processors that sold in extremely low volumes (and ended in severe amounts of write-offs for Intel, spending far more on R&D than they could recoup in sales), just because a few vendors were interested in using it and a few governments bought in?
@JourneymanGeek I thought that just before Phenom bombed. And then Bulldozer bombed. Yet they're still alive... being kept on life support by console sales it seems
@Dog keep in mind that their desktop/laptop graphics division also helped to prop up the company through the 2000s and early 2010s, because up until the past few years they've been extremely competitive with Nvidia if you're willing to overlook the occasional driver bug
also they have to be kicking themselves for selling the mobile Radeon IP -- Adreno, an anagram for Radeon -- to Imagination Technologies. How much could they be making from that if they kept it?
@Dog Adreno was built off of ATI's technology. Obviously Imagination Tech has taken it a long way since then as they continue to make it better (without AMD's help), but ATI gave them a big jump start as they had been doing internal R&D on smartphone form factor Radeon for a while.
They bought it in 2009, so they probably got some IP of a level of sophistication comparable to Evergreen (HD5870 series). Not the literal desktop core IP, but a lot of concepts and specifications from that era.
> "By supporting OpenGL ES 1.2 along with high-resolution (640 x 480) display support, the Imageon 2380 opens up a whole new world of advanced, detailed rendering that makes it possible to bring high-end immersive experiences on handsets. As game developers, we can now deliver an end user experience that can hold its own or even augment our console, portable and PC projects."
LOL. 640 x 480 is a high-resolution smartphone display :D
But doesnt that also mean your relativly expendable? Now if instead you made your self indispensable somehow, they might even like miss you when your gone :-)
Course business will certannly pay a "receptionist" to spend the whole day putting on her make-up and doing her nails, as long as she is on-call to handle the spam calls and spam e-mail, and that one true customer who calls.
they will fork out huge sums for a security guard to be the first one shot , if anything goes wrong at the location, without any big expectation that they do real work :-)
@Bob True. I'm actually wondering now why they decided to drop Itanium, unless the patent / IP worries over AMD64 have completely gone away now and they no longer have a reason to build a contingency.
@allquicatic I think they've gotten themselves in enough of a cross licencing mess, and AMD can't really get acquired by anyone without losing the x86 licence
Emulation to run existing x86 applications and operating systems was particularly poor, with one benchmark in 2001 reporting that it was equivalent at best to a 100 MHz Pentium in this mode (1.1 GHz Pentiums were on the market at that time).[96]
There were none. AMD licensed x86 from Intel, I am pretty sure Intel forced them to license the x64 extension they created, or I hope Intel had lawyers smart enough to think to put that clause in their contract :$
I wish I could work from home one weekend. At my own desk, i totally would rewrite, this application of ours. I just need internet handy in order to do it, but yeah ..., that isn't an option.
@Ramhound from what I understand amd and intel are basically locked into a situation where they can't really do a modern x86 system without patents from the other
I have no doubt if Intel and AMD wanted to make a new architecture they could. Honestly I don't think the consumers care for another platform that isn't compatible with their current system
This new architecture would have need 100% support x86 without any performance decrease
but if you do that, just tack it onto x86, why bother doing the other 2 billion in R&D
I hate Amazon sometimes, so there was a bug on their website that if you allowed an Amazon Day deal expired ( lighting deal ) that you could run into a bug where you wouldn't be allowed to add it again to your cart
So I missed out on the $40 off a 16GB Kindle Fire 10"