ITT: overhyped meme technologies
daily reminder that ai is a meme
>>53500238
Two years ago everyone was shitting himself because of Google Glass, but nowadays it's as dead as Tutanchamun
inb4 the guy assblasted guy who got touched by a Noctua cooler as a kid
aio liquid cooling.
>>53500366
Wat bruh?
More like vaporware. See you next year faggots.
>>53500238
Drake is technology
>>53500238
How will the cooler make contact with the GPU when the HBM dies are taller than the GPU die?
If they made the cooler not flat, won't some retard accidently crush the HBM die with the part of the cooler that goes on the GPU die?
Is this why no one buys AMD GPUs?
>>53501039
there's heatspreaders that bring the entire package level with eachother
HBM1 is pretty much a meme that AMD is stuck with for Polaris while Nvidia delivers the real deal with being first to HBM2
TOP KEK
>>53501085
Too bad we won't see anything with hbm2 until at least late q4 anyways and by then hopefully amd will also have it
>>53500803
Its the Adobe Flash of the API world.
>>53501650
except flash is actually good.
>>53501039
You know people get paid a lot of money to do this shit, right?
>>53500238
Spot the buttblasted NVidia shill.
>GDDR5
Turns out despite all the shilling, x86 is still more powerful and energy efficient than it.
OP is a paid nvidiot who likes to shill.
>>53501934
This
>>53501934
It's pretty sad that it's inferior though. Intel x86 licensing eliminates almost all competition
>>53500238
nVidiots shilling this hard.
>>53501727
Yes. When it doesn't crash, slows a website down by 1000%, uploads all information about your computer to NSA or gives your computer AIDS.
>>53500238
Meme = HBM1
Godtier = HBM2
>>53500803
Yes.
>>53501039
No, its because the Fury is a 4K card with only 4 GB. If they had made it 8 GB the 980ti would be a shit card.
Will possibly buy the 1080Ti, but open for anything with 6000 shaders and 8 GB HBM2 (or more).
>>53500238
>>53502031
you mean AMD x86-64 licencing?
>>53503505
x86-64 is an extension of x86. Intel licences x86 to AMD, AMD liscences the x64 extension to Intel. And Via is in there somewhere.
>>53501844
>>53501961
>>53502048
Why are AMDead fags so insecure? HBM2 cards won't be out until spring-summer of 2017. This is not about AMD. And it's true that HBM1 is a meme.
>>53503505
Can't license the 64 bit extension if you don't have the 32 bit base.
>>53500238
HBM would have been awesome at 390x pricing. It's a shame that it was launched at $650, at that price point you'd expect a little more than 4GB. If they could have squeezed in an extra two stacks then it would have been godlike but that probably would have increased the cost dramatically.
>>53500238
"Smart" objects that don't need to be smart. Somehow a bunch of retards think adding batteries and chips and complications to mundane everyday objects will make the experience better.
99% of those "smart" objects are shit. I don't want a wi-fi connected toothbrush, I don't want a "smart" trashcan that tells me when I should take out the trash. All this shit does is complicate things. Even if the products sound good on paper, I've never heard of any of those "smart" or "connected" everyday items being any good.
Just like smartwatches. A fucking watch that only lasts you a day, that is ugly, and that has a screen too small to do anything with. Worst of all, some of those smartwatches don't even have their screen on all the time.
>>53503584
At this point, I'm convinced HBM2 cards in 2017 will cost $550-650 and the 970/980 tier cards will have GDDR5X.
AMD Zen
Samsung technology licensed and rebranded by AMD because they are to poor/shit to actually do R&D on their own.
Lucky if it has Haswell performance and will probably be bought in masses because "MUH UNDERDOG".
>>53501934
Has anyone done a like for like comparison? Same clock speeds and as similar ram as possible?
>>53500238
>GDDR5^4
What's gonna be next? GDDR6^{(7*5)+(5*4)}?
>>53503613
Yeah, that's looking like it'll be the case. It's a shame though, the Fury really has shown what's possible with extra memory bandwidth.
>>53503616
>will probably be bought in masses because "MUH UNDERDOG".
Heh, no it won't. Those who buy a product to support a company on principle are few and far between. If AMD don't deliver now then they probably won't last as an independent company.
>>53503631
Why would you normalise clock speed when you're talking about performance per watt and raw performance? That makes absolutely no sense as different chips scale differently with clock speed and power consumption. Some are just made for high clock speed at low power consumption while others are made for low clock speed at low power consumption. If you look at AMD carrizo, an extra 133% power results in a ~20% performance boost.
>>53500684
>Vaporware
You do realize the fury x uses 4GB of hbm right?
>>53504178
>GDDR5
>HBM
KEK
>>53504447
Are you an idiot?
>>53500238
>AMD talking about power savings
Almost chuckled a bit.
>>53504447
https://en.m.wikipedia.org/wiki/High_Bandwidth_Memory
"The first chip utilizing HBM is AMD Fiji which was released on June 24, 2015 powering the AMD Radeon R9 Fury X."
>>53503616
>>53504505
>>53503703
So as to identify which architecture is better.
But quite an arbitrary test I'll admit.
>>53503631
x86 silvermont cores already match if not exceed krait 400 cores (which in turn have better performance than Cortex A53 cores) in terms of performance.
The new 14nm goldmont atoms are said to have significantly better IPC over silvermont atoms and be more energy efficient as well. ARM is doomed.
![418093-solid-state-drive[1].jpg](https://i.imgur.com/FZ9eFyk.jpg "418093-solid-state-drive[1].jpg")
>>53504612
Additionally x86 silvermont has beaten both krait 400 and Cortex A53 cores in terms of energy efficiency.
>>53504612
Geek bench explicitly says not to attempt to compare results cross operating system and cross architecture
Why is everything about power consumption now? I thought it was all about FPS
>>53504674
I don't understand this either.
>builds $2000 gaymen rig
>worried about a $30 annual difference on electric bill
>>53504667
Look at the integer and integer multi-core, that's what really matters here. Yeah the overall score is virtually meaningless since SoCs can cheat the score by having uber data encryption performance or some shit. What I care about is the integer performance.
>>53504674
No, just complaints
When a card gets better benchmarks it tends to be a brand new architecture, and first iterations of any architect tend to have greater power consumption
So rivals yell "HOUSE FIRES AND POWER BILL!" To save face
>>53504724
Ok, I thought maybe FPS/graphics improvements had come to such a standstill that the only thing new GPUs can offer is less electricity usage. That's honestly the last thing I think about when it comes to desktops
>>53500819
drakeposters are cancer
>>53503542
>its a meme because I don't have it
Average NVidia user not able to accept their company doesn't innovate.
>>53504791
Na, same thing happened when nvidia released the 400 line with its 400W 110°C cards
Then again for amd with their 290x line
Right now they are pretty even in power use but since amd was the more recent high temp, high power it's them that are labeled inefficient
>>53501085
>Nvidia will be the first with HBM2
Bullshit. AMD worked with the developers of HBM, they've been working on HBM2 for a while. Nvidia only recently jumped on that bus after the new memory technology they were working on failed miserably.
>>53500238
> watson
>>53503542
>half the power draw of GDDR5
>twice the bandwidth at the same clocks
>4x the density
Gosh, I'm sure glad you'll never procreate.
>>53504874
>save only 15w..........
>expensive
>only 4GB max
HBM1 is a marketing meme
>>53503616
I do not see it being bought in mass. Even when AMD was the top dog performance wise for a couple years they never beat Intel's volume since Intel has a tight grip on the business and datacenter market. AMD just can't afford to offer the incentives to desktop and server builders. There is a lot more fine print in AMD's warranty compared with Intel's.
I have a Faildozer right now, not an Intel Fanboi but ill be perfectly honest I do not see AMD Zen doing anything revolutionary.
>>53502092
battlefield 4, paragon, witcher all work fine on 1 card @ 4k
I also have freesync.
>>53504674
because it sucks to have a blow dryer going off right next to you for extended periods.
>>53505024
new tech always starts off expensive.
>>53504600
Again, you can't decide which architecture is better by completely ignoring clock speed. How an architecture handles higher clock speeds varies wildly and this is a huge part of chip design. Normalising for clock speed to compare raw performance or performance per watt is stupid to the point of absurdity. Seriously, I'm struggling to find words for how idiotic that would be.
>>53503606
Worse, they relate their product to "internet of things" or another buzzword crap for this kind of garbage.
>>53504701
More like $3.00
