Why does nvidia always fuck up when they shrink die but AMD doesn't?
HD5xxx/6xxx vs thermi for example.
And now polaris vs pascal.
>>54605459
>why does a card that's much faster need more power and produces more heat
>>54605527
the 1080 supposedly needed less, using only 1 8-pin power. it might be that they are having trouble dissipating heat from the shrunken components.
>>54605573
Probably this
>>54605527
Gtx 970 vs 670?
Hd5870 vs 6870?
Gtx 650 vs 950?
Hd5770 vs r7 260?
>>54605573
It has a TDP of 180 vs 165 for the 980
im an AMDshill but the Fury X is watercooled, the Nvidia card is a 1 fan reference design
>>54606088
It's still 24 degrees over the 980, and 4 degrees over boiling point of water.
>>54606088
And they're selling it at $700 DOLAROOS
100+C is literally housefire tier.
AMD R9 290X reference was going 90+C.
>>54605459
Well, it's an overclocked older architecture with new VRAM, this was to be expected
>>54606168
AMD didn't have it this bad though. Pretty sure only the 290x uber got close to 100 degrees and even then, it still stayed under.
>>54605459
sauce on that graph?
>>54606257
>>54606234
And the 290X was the big card of that generation. If the 1080 is 100+C tier hot, what will the 1080 Ti be? 120+ C?
>>54606353
Imagine the Titan
>>54606353
1080 is like ~85 full load OC'd
>>54607024
Its max temp threshold has been set to 85c.
Tomshardware tried to overclock it and the gpu clock keep dropping lower and lower till it reach base clock.
>>54605459
>reference card
>>54607070
>$699 reference card
Still hotter than 970, 980 and 980ti when it use the same cooler minus the fancy cut.
>>54605459
What clockrate do they have the 1080 at? I don't see why it's temp would be so much higher under full load on furmark when it's cysis 3 temp is about the same as the 980.
If that 1607 is their clockrate I don't see why it gets so hot for them, tomshardware shows furmark temp the same as game load temp.
>>54607100
pretty sure the GPU drops boost until it hits base at 85C, and if it doesn't work it throttles near 100C
Pascal is a filler to fill the void before Volta.
I seriously hope that AMD did pull the same bullshit with Polaris.
>>54607161
Isn't Polaris just midrange until vega?
>>54607070
>still hotter than previous gen reference cards
>>54605459
still no async compute, its shit
>>54607267
midrange of current gen (1070)
it was estimated to be around flagship of last gen cards
>>54607389
Don't worry, they'll just bribe developers to not use async.
>>54607267
>>54607410
Polaris will be equivalent to a 390X
>>54607592
Polaris 11 will be. Polaris 10 will be a 980 Ti
>>54607743
Either one should be a decent upgrade from my HD5870.
>2500 cuda cores (vs 2000 on the 980) clocked 1ghz higher than previous gen on a massively smaller die
>y is it hot lol
>>54607864
>have to rely on clock rate to make some ground
>against a fucking 28nm gpu
>>54607892
>tfw mtk6752 performs better and runs cooler than a SD615.
It's not all about the die size you fucking retard.
>buying nvidia ever
https://www.youtube.com/watch?v=G7IbbqX6wxA
>>54605459
>>54609672
>Things get hot with the backplate off
WOWEE.
>>54609727
Nvidiot confirmed
>>54605459
Because their running the card at 2GHZ+ with no fucking watercooling or proper heat exchange
>>54609904
Look at the clocks.....
>>54606088
The Nano is a 1 fan reference design as well. It's still not a house fire.
>not placing a mug with water on top of a 1080 in the winter so it boils the water so you can have a nice cup of hot chocolate when you're done gaming
This sounds perfect.
>>54607267
and isnt the 10x0 nvidia series a midrange series too? Its just like GTX6x0 series... midrange chips being sold as highend until GTX7x0 came.
>>54610327
>water
>hot chocolate
>>54610408
Gtx 680 wasn't high end?
Pretty sure gtx x70 or x80 and above would be considered high end.
>>54605459
the 5000 and 6000 cards were much worse temperature wise. reference versions of those cards needed fans speeds at 100% during gaming, whereas my 1080 sample doesn't need to go above 50% fan speed to maintain safe temperature.
>>54607743
No, Polaris 11 is aimed at low power portables (budget "gamer" laptops) so it's shooting for r370 speeds with ~50w tdp. Polaris 10 is targeted at mid-high tier laptops and mid tier desktops, again, with low power requirements; probably around 150w tdp. The full Polaris 10 should be close to the performance level of the gtx1070 though, when you consider that the 1070 has had ~25% of it's resources neutered.
>>54610475
Not who you're replying to but most people enjoy hot chocolate by heating water, then adding chocolate mix into it. Swiss Miss is a brand that sells them in packets or containers.
>>54609904
It must be hard to see when you're down there sucking that chink dick, but when you come up for air, take a look at the clockspeeds. Hint: it's less than 2ghz.
>>54605459
The fact a Fury out benches a Fury X leads me to believe this is a BS benchmark. And the fact that a regular 980 is close to a 1080 is impossible unless shillvidia is lying again about their performance gains. Wouldn't shock me.
>>54610520
>amdrone doesn't understand what heat output is
>>54610494
Oh.
I usually mix the milk chocolate into a tiny bit of boiling water in the cup, add the milk and then heat it in the microwave.
>>54607998
1080 doesn't run cooler.
1080 uses 10% less cores, but 60% higher clocks and barely comes out ahead of the 980Ti.
1080 has 25% more cores, and 60% higher clocks but doesn't even come close to 40% higher performance compared to 980.
>>54610730
I would laugh so hard if it couldn't even overclock well.
>>54610802
based on the heat this thing is dangerously overvolted and im worried about a legit housefire if someone tried to push a 2ghz overclock
even at 1600mhz its over 100C thats fucking crazy this is gtx480/fermi all over again
>>54610514
you mean 1600-1700mhz? Still fucking WAY to hot for those low clockspeeds at 16nm
This card is a housefire
Still waiting for 140°c 1080 ti
>>54611369
Not really, the Reference 980Ti hit 82ºC.
Lots of people seem to want to comment on this "issues" without having knowledge about Nvidia cooling approach on reference cards.
What you see here is the fan profile on the bios for a reference GTX970, at 78.16ºC the fans are only hitting 1861rpm, this card has a temperature throttle set to 79ºC(Temp Tager, Reference 980Ti is set to 82ºC), meaning that at 79ºC the core clock will start to throttle down to keep temperatures under 79ºC, if for some reason the GPU can't keep temperatures under 79ºC even with heavy throttling, the card will ramp up the fan RPM until it hits 91ºC which then the system will shutdown as it hit the Max Temp Target.
On aftermarket cards the TPM12 target is usually set around 60ºC, which is why you see aftermarket cards cooling better, because they ramp up the fans early, which is not the case on the reference models as they ramp up the fans as last ditch effort.
>>54605459
>trusting some random fucking soviet benchmarks
you amdorks will cling to anything, won't you?
>>54611479
Still a retarded card though. Has a higher tdp and power consumption than previous gen.
>>54611526
Power usage is just slightly higher than a GTX970 while performing stronger than a Titan X. The Titan X alone is twice as strong as a GTX970, just let that sink for a moment.
1080 is the new king no matter what kind fault people try to find and aftermarket cards are going to be even better.
>>54611558
That's all good, but the fact that the 1080 runs hotter and consumes more power than the 980 will give AMD an advantage if they can offer highly efficient GPUs.
>>54611567
>but the fact that the 1080 runs hotter
You don't seem to understand how reference cards work, the temp target will be hit at stock fan profile on any reference card, even the 970, in fact as you can see in here >>54611479 the fans don't even ramp up before hitting temperature throttling, if you want cooler temps use a custom fan profile, this is how ALL Nvidia reference cards work.
>and consumes more power than the 980 will
Yes, while being almost as twice as powerful.
You are retarded.
AMD can't compete at all, even if they wanted to get the cheap $150-$250 market, Nvidia could just release a card with GTX980 performance and half the power usage of a reference GTX970 without problems at that price range.
Unless AMD has something stronger than the 1080 they are done for, even so the 1080 is a cut in half Pascal chip, the full chip set should be near twice as fast as the 1080, unless they can blow away the 1080 by 20% or more they are doomed even with cheap prices.
>>54611606
>shilling this hard
>>54611631
Not shilling, just schooling you.
I'm not even going to buy Pascal, I'm going to wait for Volta.
>>54611479
Which begs the question why the hell dont they ship cards with better stock coolers like the fury nano and x
Its like back when intel still had stock coolers on their highend CPU's they where woefull
>>54611606
>even so the 1080 is a cut in half Pascal chip, the full chip set should be near twice as fast as the 1080
kek, look at this retard. Pretending to know what he is talking about when he's pulling everything he's got from his arse to shill nvidia.
so, from all reviews I read they made next gen r9 Nano in power draw department
but big, more loud, with slightly less throttling
>>54605459
>>54611558
>>54611558
Size of the die, heat and clock speed... Breaking a sweat. With the upcoming new gen cards, this gpu will go down as the first cuck marketing and the price will be slashed in half in six months
9 series was an end game, this is a new platform
Comparing both is kinda lame... You say low power became a priority but the clock for clock perf needed to go thru the roof justifying an extra cooler so third parties get a blueprint to control temps
HBM2 is where it's at
>>54605459
>>54611558
>>54611869
Where's the problem?
>>54611874
Not having the technology to transmit data at the speed in which the memory could allow.
Pissing in the wind.
>>54607410
>>54607592
>>54607743
>>54610488
it's actually incredible how much conflicting misinformation is being thrown around, half the shit i read here i don't think is even reported as a rumor on any tech site it's just assumed by retards from /v/ or something.
>>54611949
every freaking tech site reported those rumors, don't you read newspapers everyday?
>>54607161
and volta is a filler to fill the void before x hurr
>>54605459
It's literaly has the power of 2 GTX 980ti or 10 Fury Xs in there while using less power. At least Nvidia doesn't need to make a dedicated water cooler for it just to keep it cool like the Fury X.
I'm wondering whether to hold out for Polaris 10 or upgrade now.
How shitty are AMD drivers really? I don't want to gamble on getting a Geforce 900 series card without coil whine.
>>54612746
They're about the same now. Back a few years ago they really did suck though.
>>54612746
Higher price of 980 ti classified or lightning - bought titan x sc, retains value
Bought maxwell last year - not interested w/ first gen middle of the road
Buying full pascal - flagship cycle
It's relative to your usage and resolution, the upgrades are plotted out and you simply need to cycle your purchases, it isnt really definite unless your usage becomes definite
I'm not a fan of waiting also leads to higher hysteria and isn't always healthy, balance of impulse and reason
>>54612746
much better now, and getting better with each game release
hell they even had drivers for doom before it launched and a hotfix just today
>>54612860
Say that to my hd6870
>>54611479
>MFW my 780Ti runs at 62° on 1300 MHz
>>54612856
I wouldn't do that, I had a Titan and will go with improved phases next year. I'm aiming for the 1070 but will refuse it if its shitty, I just need to survive with a healthy wallet until the 1080Ti comes, then I can buy 2 of them for 4K.
>>54614332
Yeah we know everything not GCN was abandoned. So are Nvidia cards in the 300's series.
>>54610730
It's on average 36% faster than a 980 at 1080p. Stop lying just because their business practices suck.
Do you guys have any idea how much effort it takes to shrink your manufacturing process? It's not "just Maxwell with higher clocks." It's a whole fucking new architecture. I guess I can't expect /v/ leak over to understand anything about hardware engineering.
>>54614927
There was no 300 series?
It went 200 series to 400?
I went from a gtx 260 to a hd 6870 for dx11.
Looks like I'll have to upgrade again for dx12, Just my fucking luck.
>>54605573
>measuring power with pins
>my pc works on 2 pin 230 volt connector
>>54605459
>>54605459
>>54605459
>>54605459
>>54605459
spending 1 month for a market research to get a potentially 20 dollar of value
welcome to /g/
>>54614332
This. I might consider switiching to AMD if they bring something 4K worthy cheaper and I see they give driver support for older GPUs. But I don't count on it. Isn't the 5000 and 6000 series already considered as 'legacy'?
Even if Nvidia doesn't gimp directly, they brought more new cards (NEW architecture not new name) in the same time where AMD only rebranded its highend Hawaii.
Otherwise, I wish to know why my 780Ti hits less FPS than a 290X or 970 in Witcher 3. Some people said 9xx have more tesselation performance, but my 780Ti already has more than a 290X. But I don't care, this is just a shit game I got for free.
>>54615108
It doesn't matter to most people since they upgrade every 2-3 years
>>54605459
AMD isnt retarded is all.
When you die shrink your supposed to also rework the architecture (Vega)
Nvidia just took maxwell, slapped it on 16nm and gave it some gddr5x with a heavy overclock and thats why you have a 100C house fire tier 1080.
>>54609727
>backplate meme
>>54607389
>still no async compute, its shit
https://www.youtube.com/watch?v=Bh7ECiXfMWQ
>>54607389
But it does:
http://wccftech.com/async-compute-detailed/
>>54615004
>2 pin 230 volt connector
>not 3-pin
>>54619438
you are literally dumber than a curry nigger.
GP104 like Maxwell can only do one thing at a time per compute block.
The only things that changed are that it can do a context switch at any arbitrary point and that the switch only takes ~0.1 ms (150k-200k cycles) instead of 1+ ms like it did on Maxwell.
GCN's async juggles potentially many dozens of instruction streams across the ALU array with no single global context and no penalty for switching between instruction streams.
Nvidia's slides have something called "asynchronous time warp" which is a VR technique not related to async compute shader processing.
>>54611949
Everything that you read about Polaris is just rumors
>>54618398
Its not actually built into the hardware
>>54611683
the ti is 20% bigger and titan is 40%, there is no 100% bigger chip.
Does Vulkan use async or is it DX12 only?
>>54610484
so they are that afraid of the old 5000 blower sound meme that they refuse to allow the cooler do its damn job?
>>54612717
its not even 10% faster then a 980ti oc.
>104C
That's no housefire, that's a fucking inferno.
Have some OC.
>>54612717
Fury x doesn't need a water cooler. It doesn't output much more heat than a fury you retard
>>54623597
LOL, 10/10 anon.
>>54623597
Thanks for the kek
>>54622916
>Does Vulkan use async or is it DX12 only?
It's exposed in DX12, Vulkan, and Mantle.
All three APIs have the concept of separate work queues and inter-task dependency ordering, and it's just a matter of hardware support for being able to truly concurrently process different tasks on a GPU vs. just preempting/context switching back and forth rapidly.
>>54623597
>gaming processing unit
Are you really this uninformed?
>>54624806
It's targeted at gaymers,so why not?
>>54624892
>I'll buy anything labelled "gaming edition"
good goy
>>54624935
Where did I say that?
>>54611676
Because the blower fan design works pretty much independant of case airflow since it exhausts out of the case. Multi-fan, open heatsinks can actually perform worse than a blower depending on case capacity and fans.
>>54625055
Why not a blower with a larger fan and better exhaust port?
Maybe add a couple of heatpipes? My reference 570 had what appears to be a large copper block though, but the heatsink itself was rather small. Should extend it further down the card with well spaced fins to allow better cooling.
>>54605459
Well 5xx was an improvement over 4xx, so this contradicts what you are saying.
Also you have 180W on a really small die, that is alot.
>>54609672
Well what did you expect.
Those are the fucking VRMs that are that hot.
This is nothing unusual.
>>54610730
Rumor has it that some first OC tests with better cooling got it to roughly 2.5GHz
>>54611479
I also bet you that nVidia gimped on the stock cooler.
Most aftermarket companies like ASUS , Gigabyte, MSI and so were pissed at the Titan cooler.
It caused them trouble making something better without vamping up the price too much.
>>54611869
Since when does any kind of RAM eat so much power.
>>54625755
Vram because much higher bandwidth and frequency.
>>54625755
>muh 7000mhz gddr5
>>54615914
>Nvidia just took maxwell, slapped it on 16nm and gave it some gddr5x with a heavy overclock and thats why you have a 100C house fire tier 1080.
The way it's meant to be played.
>>54605459
with thermi, they went full retard with the compute cores.
with pascal they went full bulldozer.
>>54611402
>air brushed to death
she must be ugly in real life.
>>54627332
>implying you wouldn't fuck that
Here, have another.
>>54627267
The way we're meant to be played?
>>54620592
>grounding your pc
>not living life on the edge
feg.
>>54606059
TDP doesn't man the card is more powerful. Less pins means that you saturate the one connector instead of using half of two, it's the same thing and my point still stands.
>>54628880
Remember the 295x2?
Imagine enthusiasts overclocking it on a cheap psu with poor insulation and cheap wires
>>54629014
Don't you realize that tdp actually measures wattage per hour. Its not gonna go past 77 C when you are overclocking within tdp max ranges.
>>54629230
You don't remember the 295x2 anon.
How unfortunate
>>54611606
>being this retarded
>>54621130
thank you
>>54621130
Very well written and detailed post anon!
What exactly is an AIB partner and what do they do? Nvidia and AMD sell them base GPUs and then AIBs design cooling and provide warranty support?
Also today I learned that PC partner owns both Sapphire and Zotac, so Zotacs are basically Nvidia Sahppires.
>>54631486
nvm wiki was wrong about the Zotacs
>>54610582
fucken degenerates, not just putting milk in a pot and heating that
>>54631750
On the stove?
>>54606088
and the fury isn't watercooled whats your point?
>>54609727
>hurr 101c is normal
>>54607096
>when it use the same cooler
Eh no. More heat = More heat disippation. Learn physics you knob. Besides, 1080 uses less watts so there's really no telling how hot non reference cards will be.
>>54632038
>More heat
>Smaller die area
That's the physics at work here. Pascal is inherently hotter than Maxwell or Kepler.
>104C
JUST
>>54632143
Yes
