Are you ready for G1 MP, /g/?
http://www.overclock.net/t/1595065/wccftech-nvidia-pascal-gtx-1080-1070-1060-benchmarks-leaked-3dmark-11-performance-entries-spotted
>not faster than a 980Ti
Who gives a shit about your poorfag GPUs
>>53596307
More efficiency tho
>>53597118
>I'm too poor to pay for electricity and my government subsidy won't let me use more
>>53597118
So what?
If you live in a first world country energy costs are cheap as fuck. If you can buy that level card you can afford the extra $5/year in energy costs.
tl;dr motherfucker
>wccftech
Ah yes, they're the most trusted news tech site
Who can forget the Phenom 4 of course!
http://wccftech.com/rumor-amd-phenom-iv-x12-170-baeca-25nm-cpu-leaked-features-12-cores-6-ghz-core-clock-am4-socket-compatbility/
Or the legendary Faraway Islands GPU series!
http://wccftech.com/amds-future-gpu-architecture-code-leaked-faraway-islands-features-20nm-coming-year/
>>53596100
Old news
>>53559720
>>53596100
hnggggggggg
Literally named their card gimp, expects us to take them seriously. I guess nVida really is the way it's meant to be gimped.
>>53596100
>wccftech
I'll leave it here too
>>53597480
Damn it Iel.
Nah, not buying Nvidia after getting memed into getting a 970 (before Nvidia's kikery was exposed).
Hope Polaris is gud. I don't know anything about it but for the first time I'm actually kind of hoping for power efficiency over maximum performance because I've got a really nice passive PSU but it's only 520W.
i'm on r9 390, bought it 3 months ago. planning to skip pascal/polaris entirely, maybe even another iteration too, depends if it'll be refresh or not. still on 1080p and i don't plan to upgrade for some time so whatever, i guess 2-3 years will be easy with 390.
>>53597481
fake
>>53597560
>>the 970 meme
the 970 is one of the best cards ever made.
>>53597214
what show is this?
>>53597611
>slower than 290x
>slower than 390
>3.5GB stutter
>>53596100
>G1 MP
>gimp
Some one is taking the piss
All 20 gallons of it
>>53597611
>one of the best cards ever made
>>53597704
guess your webm really proved him wrong
>>53597663
overclocking it puts it well ahead of those cards, even with air cooler. it almost matches the 980ti when overclocked. NO better card for overclocking. if you aren't playing at 1440p plus 4gb of ram is enough. yes, it is a four fucking gigabyte ram card. if you say or think that it is 3.5, you are a fucking retarded parrot monkey poorfag shill.its simply not 3.5. that's not fucking reality.
>>53597732
>Implying you can prove a shill wrong
>>53597704
zozzle
>>53597740
C'mon man, try at least a bit next time. If you think the 290/390/X don't get gains from OCing then you are crazy.
>>53597481
>NVIDIA Confidential
fake as fuck
anybody can put that on there in fucking msft paint
>>53597740
>it almost matches the 980ti when overclocked
>Once G1 MP is released, Maxwell owners won't be pleased
what happened to the "Nvidia's Pascal is 10x as powerful as the Titan X" meme
this is a little bump up compared to the 980 ti what the dicks
>>53598352
Considering it's the first gen of 16nm I think they're playing it safe and focusing on performance per watt, it's not like console ports are hard to run anyway
>>53598405
so they lied
>>53598460
Not the first time
>>53598352
That was for compute not graphics, do you people even pay attention.
>x80 ti will only be 25% faster than 980 ti
>980 ti overclocked gets 40% performance boost
>Polaris will only be 4gb
>still weaker than fury x
Waitfags btfo
>>53598460
No, you're just retarded. The 10x performance metric was in extremely specific compute tasks. Of course, Nvidia were happy to let retards like you misinterpret it as across the board performance and shill their cards for free.
>>53597740
>it almost matches the 980ti when overclocked
How far do you have to go to justify your purchase? Unbelievable.
>>53598714
980 ti performance is a bit far fetched, but it does beat the a stock 980 once overclocked.
All maxwell cards have quite a bit of headroom so his argument makes the most sense only when put up against the nearly capped out competing AMD products.
>>53598352
Here's the thing, if your card is performs 10x better per watt you can either deliver more performance, or you can reduce hte power draw and the heat dissipation.
Guess what the GPU companies are going to do?
>>53598768
I just hate it when people claim 'yeah well when I overclock X product I just bought it has performance close to Y, a much much more expensive and powerful product!'. That can happen, and I'm sure it could reach 980 levels, but I find it delusional to think a 970 can reach 980ti performance.
>>53597641
Musaigen no Phantom World
>>53597704
Holy fucking kek. Seriously though, anything under 980ti you're better off with amd or waiting. 390 is the sweet spot for bang/buck this gen.
>>53597791
You and I laughed at the same thing on the internetwork of world computers. May we have sex?
>>53598768
>mfw my 45% overclocked 7850 2gb performs at a bit shy of 960 levels
>mfw a 960 today costs as much as my 7850 did 3 years ago
>>53597704
what was this webm made with?
>>53599090
Think it's called beseiged or something like that.
>>53599039
960 competes with the 380x, which is a slightly more powerful (5%) 380, which itself is a rebrand of a 285 meant as a small upgrade to the 280x. The 280x is a 7970 rebrand and it takes 2x 7850 to match it.
tldr: "no"
>>53599124
Looks interesting, thanks anon.
>>53599156
>960 competes with the 380x
>>53599039
My GTX 970 is almost as twice as powerful as my reference HD7950 and it was cheaper too, people say that the HD7000 series can still keep up but I never see those results myself.
![hqdefault[1].jpg](https://i.imgur.com/BV5I3bn.jpg "hqdefault[1].jpg")
>>53599156
>960 competes with the 380x
>>53599322
Those cards are both well and beyond the next step(s) down.
>>53597740
Almost matches a 980 in a best case scenario but the 980 ti? No way.
It's a good card in general though. Great performance for the price in my country. It's cheaper than a 390 here so it's the obvious choice for new builds and upgrades.
FYI I have a 980.
>>53599322
But to add, saying a 970 is twice as powerful as a 7950 is silly. Unless you're playing Tesselated Air and Gameworks Hair: the Game.
>>53596307
Thanks. I don't have to even read.
Still planning on the evga 980ti hybrid for the oculus rift.
>>53596100
this is some shit, my 980ti scores much higher
looking at GPU score only here
>>53599215
>>53599441
Not him but most benches I've seen at 1080p there's only a 5~ fps difference advantage for either card for certain games. Most of the time theyre neck and neck. The 960 is also £20/30 cheaper so I'd say it has the better value. It also uses just under 100w less power than the 380x. I'm not trying to start any fanboy war, just putting this info out there.
>>53599661
Without context, that's not saying much. 15fps to 20 fps is a huge jump.
>>53599587
That's before Nvidia downgrades your 980 Ti's performance with a new driver update.
>>53599661
380x rests about right in the middle of the 960 and 970, which is a huge power gap. Comparing the 960 to the 380x is about equivalent to comparing the 380x to the 970.
>>53599689
I was using this article as my source
http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review/15
I think there was only one game where the 280x had a bigger than 5 fps advantage than the 960. I still think the 960 is better value because if this asus pre overclocked 380x is only beating the 960 by a few fps then a measly oc to the 960 will most likely match it or even beat it.
I quote:
> Meanwhile looking at the competition, by virtue of the GPU configurations AMD and NVIDIA went with for this generation, the R9 380X has no true competitor from NVIDIA. This doesn’t give AMD much freedom – the card is only 10% faster than the GTX 960, so they have to stay within reason on pricing – but it means that they’re the only game in town for a $200-$250 video card family. Otherwise the one tradeoff here (as has been the case with most of AMD’s cards this year) will be on power efficiency; R9 380X doesn’t improve on AMD’s efficiency at all, resulting in R9 380X drawing a lot more power for its 10% advantage over GTX 960. We will add however that a 10% gap means that the R9 380X’s performance isn’t outside the potential reach of factory overclocked GTX 960 cards, but that is very much on a case-by-case basis as opposed to today’s look at baseline performance for each video card series.
All in all I don't think the 280x is particularly good value especially compared to the 960 which is cheaper.
>>53599773
Okay, now get benches for the cards we are actually talking about.
>>53599773
Shit I said 280x I meant 380x**
My mistake.
>>53599773
From your article:
>Meanwhile looking at the competition, by virtue of the GPU configurations AMD and NVIDIA went with for this generation, the R9 380X has no true competitor from NVIDIA
There is no comparable card from nvidia, unless they come out with a 965 before pascal.
>>53599797
ignore this because this>>53599798
>>53599773
>http://www.anandtech.com/show/9784/the-amd-radeon-r9-380x-review/15
>The challenge to the R9 380X then doesn’t come from below, but from above.
There's no real contest between the 380x and the 960 as the 380x beats the 960 quite nicely. The 960 is more competitive against the 380 but even there the 380 is faster. In that case you're trading general performance for lower power consumption and gameworks support, which unfortunately doesn't last very long.
>>53599472
Nope, it it almost as twice as powerful in any game with the same settings even games that do not have GameWorks.
>>53599818
Yes you're right there isn't a true competitor since the 380x holds an advantage in most games. However the 960 does have the capability to close that gap to a stock 380x with an overclock. That's not saying the 380x can't be overclocked itself. You also never know when nvidia will release a gameworks game that will cripple the advantage the 380x currently has.
So Nvidia is using 16nm process and AMD is using 14nm
Will there be any discernible difference between a 2nm difference in die size?
>>53599923
discernible? Yes. But its really a false equivalence because to properly compare the two you'd have to use the same arch but on two different nodes.
>>53599923
depends of the arcitecture
I was going to finish my build, guess I'll just wait
GeForce GIMP? This is a joke right?
>>53599923
Price point, possibly.
>>53599964
Oh, and overclocking headroom.
>>53596100
>G1-MP
KEK
>>53599938
>>53599941
I know they're different architectures, and it's a little like apples and oranges, but will AMD be able to have slightly more powerful cards by using a smaller process node?
It would be nice for them to have the top card instead of Nvidia for once
>>53597481
And the fury completely destroys the 980ti in flops, so what's your point? You gonna be using that x80ti for protein folding?
>>53599923
I think the finfet process makes any advantage from either side negligible. I'm not sure though. Someone was explaining it in another thread and the general gist I got from it was that there won't be a different because of the finfet process.
>>53597740
>970
>4GB vram
Pick one because the jews at Nvidia say you can't have both
>>53597704
>>53599923
None of them are true 14nm and 16nm.
Only intel has true 14nm.
>>53596100
Why not be smart and wait for HBM2 they are only using GDDR5X because its a cheaper alternative before they finish testing HBM2 design.
>>53599978
>but will AMD be able to have slightly more powerful cards by using a smaller process node?
That entirely depends on how tightly the components are packed inside the GPU relative to each other. It might be a negligible difference in small-to-medium sized dies, but once those dies get bigger, then you'll start to see a difference. I think the 14nm process AMD is going with is smaller than the 16nm TSMC process by some degree, so we might see cheaper and equally strong GPUs out of AMD than Nvidia. Titans will always be $999, but if AMD can sell a GPU with about 80% of the die size with the same overall performance, then AMD has a scaling and price advantage due to the tighter fitting process.
Of course, other factors will come to play that will ultimately determine how much AMD charges for a smaller die size, like process maturity, yields, and global demands.
>>53600046
Because the wait is becoming a meme at this point.
I'm ready for Maki....
>>53599923
From a quick google search I found this article which was clearly written up by some spastic but you can take bits of info from it. They compare the iPhone 14nm chip vs the 16nm chip which are both produced by the same companies amd and nvidia are going to be using.
http://www.techgrapple.com/apple-a9-14nm-vs-16nm-performance-test-with-iphone-6s/
>>53600046
>Waits for HBM2
>"Why don't you just wait for the ti :^)"
>>53600087
Interesting thanks
>>53600087
>http://www.techgrapple.com/apple-a9-14nm-vs-16nm-performance-test-with-iphone-6s/
There are a few problems here. First off, the results are within the margin for error, secondly the sample size is too small and we know there can be a pretty big difference between chips even from the same wafer. Lastly it's hard to know how relevant this comparison is to larger GPUs as they're not really the same.
>>53597740
delusional
>>53600123
Regarding the sample size, the 14nm FinFET process has one of the most consistent average performance potential out of any process that came before it, similar to the 20nm FinFET it was based directly on. They have a much smaller disparity in performance between batches, about 20% smaller compared to the previous 28nm process.
>>53600175
Do you have a source for that? Even then, there's also the variance on the 16nm chip to take into consideration. I'd be much happier accepting these results if there were at least five samples of both chips.
>>53600123
True.
Here's another article about it. It seems it wasn't only this guy who did the benchmarks that was getting these weird results. I'm genuinely interested in why the 14nm chip, which in theory should perform and cool better, is performing worse across the board. Hopefully this has no relation to gpu performance as you said.
http://bgr.com/2015/10/08/iphone-6s-a9-processor-samsung-tsmc-batterygate/
>>53600045
Explain
>>53596100
Time for a new reference cooler no?
>>53600053
It should also be considered that AMD is likely using HBM again for its cards, as there was that report about the passively cooled AMD card running VR... The report said that the card was the size of Fury Nano, which means it was using HBM. That being the case, AMD can make their dies a little smaller because they won't need as wide a memory bus to get use able bandwidth out of the memory. Nvidia is using GDDR5 again, which will require a larger memory bus, and force a larger die size.
That's all assuming that the guy saying the AMD card was small was legit. But if it is legit, oh boy I might buy a new card for my ITX build this summer.
>>53600205
It was based on a slide from AMD back in May 2015 and a slide from GloFo around the same time, but I think it was in regards to Zen, not Polaris.
I need to find that slide, though.
>>53600207
Someone in the comments:
>Called it. All the usual idiots on here who know nothing about fab processes automatically assumed 14nm is better than 16nm, well, because their grade school math tells them 14<16, and smaller is better.
I posted a long time ago that Samsung 14nm isn't actually 100% 14nm because portions of their devices use a larger process (like interconnects). I also posted about leakage current, and if not kept in check a smaller process could actually consume more power. This is widely believed to be the reason Intel took longer to release 14nm chips because they were dealing with the same issues.
Regardless, it appears TSMC's 16nm process is actually mature (if these results hold up).
If true this could potentially be a small advantage for Nvidia. As mentioned before though you never know how this'll affect a gpu.
>>53598650
Glad my 980 ti's will be in my rig until 2017 or beyond. I've grown found of them.
>>53600233
>The report said that the card was the size of Fury Nano, which means it was using HBM
I doubt it, because none of the Polaris cards are going to use HBM (that's for Vega only) and HBM runs HOT compared to GDDR5 due to how close it is to the GPU and how dense the NANDs for the HBM are packed.
Fiji is supposed to be a much cooler-running GPU design (due to its better efficiency), but the card itself runs as hot as the overclocked 390X. HBM is the reason why.
>>53600278
Vega gets HBM2. I don't think AMD has said one way or the other if they want to use HBM again. Many have said that HBM is too expensive for anything but high end cards, but if AMD thinks they can produce cards that outperform Nadia's offerings significantly, they may include HBM and sell the cards at a premium, just as Nvidia has done in the past.
This is an interesting generation. Much more so than past generations, anyway.
>>53600325
>I don't think AMD has said one way or the other if they want to use HBM again
SK Hynix announced the winding down in HBM production to make way for HBM2. Fiji will be the only GPU to use regular HBM.
>>53596100
>Nvidia GIMP
Ok, so it's just some gimped gpu?
>>53600410
Technically speaking everything but the highest end versions of every chip are gimped. The 7950 was a gimped 7970. The 7850 was a gimped 7870. The 970 was a gimped 980 and so on.
>>53600456
>The 970 was a gimped 980 and so on.
That's why you can overclock a 970 to 980 oc levels right?
>>53600485
Not really. You might be able to get close but the 980 always has the hardware advantage. Instead of a 970 being a gimped 980 it's more of a cut down 980 with lesser specs.
>>53600485
I... what? That's some of the dumbest logic I've seen. The reason you can OC a 970 to 980 levels is because the original difference between them isn't that great and the headroom for oveclocking covers the gap. Keep in mind that you can OC the 980 as well.
>>53596100
what a shitty name
>>53596100
>GIMP
>Geforce GIMP
Oh come on.
>>53597740
>makes claims like "it is a four fucking gigabyte ram card" and asserts if you think otherwise you're a thoughtless parrot
>doesn't provide proof of contrary claims
>>53596100
>G1MP
>>53600852
How is Gigabyte not suing? G1 is their model name.
>G1MP
is this real life?
>>53602037
Because it's fake. It's an edited pic of a gtx 780 reference.
>The first entry we have is of an unidentified Nvidia GeForce graphics card with 7,680 MB of graphics memory, 512MB short of 8GB
>>53597214
Kirby a cute.
