I know that /g/ was quick to call HardOCP a shill for this article
http://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility#.V1FL-b4f30Q
What he claimed that his inside sources had told him that: 'Polaris was significantly slower and less efficient than Pascal'
Now that both parties have shown their 150w tdp cards, the 1070 and the 480, the first outputting 980ti performance and the second outputting 390-390x performance, I can only conclude he was right.
At its price point it's still a competitive card, but it's definately slower and less efficient than the 1070
>>54887139
>definately
>>54887139
We don't know yet.
We only know that peak consumption of the 480 is 150w.
We don't know idle power consumption (most important for electric bill) and we don't know load power consumption (for how hot the card runs).
Now you can stop speculating, and stop shitting on AMD. At least until the first benchmarks come out.
Tdp is not power consumption. 1070 and 1080 condone about 160-180w on average.
480's max output limit on PCI Lane and 6pin is 150w. The average will be around 120w
>>54887163
>We only know that peak consumption of the 480 is 150w.
eeuh no we don't?
>Tdp is not power consumption
It isn't but it is heavy correlated, look at the tdp on the 300~ series cards versus the TDP on Maxwell cards. I'd say TDP is a good proxy for power consumption.
>>54887139
He made a negative spin story based on a positive one. That's a mark of an angry shill.
>>54887201
A story that turned out to be true
>shill apologetics
We don't actually know anything for sure right now.
>>54887203
No it didn't you dumb cuck.
>>54887192
>We only know that peak consumption of the 480 is 150w.
Bad phrasing, we know the absolute max the card can get is 150w.
>>54887206
>We don't actually know anything for sure right now.
We do, we know the TDP of the 1070 and the 480, and tdp is a good proxy for power consumption
>>54887192
>Maxwell TDP's were real
nicely memed
protip: we know it's peak power because that's how much a PCI-E slot and 6 pin connector can supply. That's all the ref 480 has.
>>54887192
PCI Lane provides 75w 6pin provides 75w. That's the standard mobo specification.
Pls Kyle your stale shilling is stale.
>>54887139
>$199
>$379
>>54887219
>and tdp is a good proxy for power consumption
No.
>>54887222
That chart still shows Maxwell is well ahead in effiency
>we know it's peak power because that's how much a PCI-E slot and 6 pin connector can supply. That's all the ref 480 has.
The reference yeah, board partner cards might have two connectors.
>>54887235
>No.
Yes.
On the subject of polaris, i hope they are available in half-height/low profile formats. It'd be a pretty sweet card for a HTPC setup.
>>54887222
What kind of load were they testing here, is it synthetic stuff like furmark or was it actualy gaming loads? I'd like to see the source
>>54887139
Not a argument
>>54887252
No and read the above posts you mouthbreathing retard.
>>54887265
How was he wrong?
>r9 290x beats gtx 980ti in dx12 games
heh...what a return on my investment...truly the 7970 ghz of its generation...
>>54887243
So you're saying nvidia 1070 and 1080 might not be 150w but could be 300w because board partners will have two connectors instead of one?
Thanks for the heads up nvidiot.
>>54887274
>read all the comments that agree with me but ignore the others
Yeah, fuck off
>>54887262
it fucking says on the pictures
Metro Last Light 1080p for the AMD cards
Thief 1080p for the Nvidia
Different reviews, but both used the same voltage and amperage measuring equipment
>>54887280
>So you're saying nvidia 1070 and 1080 might not be 150w but could be 300w because board partners will have two connectors instead of one?
No, because we have already seen board partner 1080's tested and they weren't 300w
>>54887298
Cool, looks like Maxwell is still alot more efficient
>>54887275
How is he right when we dont even have the benchmarks? He may be proven right but its a little bit premature senpai
>>54887311
You can read in the article he got his info from inside sources, which turned out to be right.
The RX 480 has about the same speed as a R9 Nano and it's 2.8x as power efficient. This would put it's power consumption to
175W / 2.8 = 62.5W
>>54887293
>can't read two posts precisely explaining why his nonsense is just that
mouthbreathing retard confirmed
>>54887309
and they exceeded their TDP's massively
>>54887327
>The RX 480 has about the same speed as a R9 Nano
Lol what?
>>54887331
I read them, and none of them countered the fact that TDP is a valid proxy for power consumption
>>54887327
uh
2.8x more power efficient only applies to the GPU core
4gb of GDDR5 take up around 50 watts alone.
>>54887346
>countered the fact
>posted a baseless assertion without anything to back it up
mouthbreathing
retard
>>54887327
Its probably basing off off 390X, which is 275 TDP. A bit more OC'd version would be around 300w
300/2.8 = 107w
Thats probably their target.
>>54887339
>>54887222
Why are these results so different from the typical power draw from the wall analysis that most reviewers post?
You always see very big differences between the 200 series and Maxwell there.
If one source deviates so much from what everyone else is showing I tend to get suspicious.
I know /g/ likes their tinfoilhat conspiracy theories, but I think it's unlikely that everyone is wrong while only these guys are right
>>54887356
>posted a baseless assertion without anything to back it up
>baseless
>780ti had a much higher tdp than a 970, and also had a much higher power draw
>300~ series cards all have a much higher tdp than Maxwell cards and use up much more power
Baseless huh?
>>54887358
That's a synthetic benchmark though, AMD card tend to do well there since they have alot of raw horsepower, it doesn't always translate to extra gaming performance
>>54887369
>still thinks a card can draw more power than a rail can provide
Yes baseless. Now we're getting somewhere.
>>54887369
>much more
>29w difference
>>54887360
70 watt delta between a 290X and 970 isn't small dude.
The point was they went way over their TDP
>>54887390
That's literally irrelevant to the question whether TDP is a good proxy for power consumption, and like I already said, aftermarket cards will likely have more power connectors
>>54887396
>furmark
>290x is a lower performing card than the 980
>>54887408
You nvidiots really need to stop moving the goal posts around. It makes you look stupid when you can't stick to your own statements.
>>54887401
I guess that is rather shifty
>>54887360
>I know /g/ likes their tinfoilhat conspiracy theories, but I think it's unlikely that everyone is wrong while only these guys are right
I'm not an expert on video cards by any means, but in my experience this actually tends to be the case in practice.
Sturgeon's law applies to test websites. 90% of benchmarks are run by clueless idiots who don't know how to set up a proper test environment.
The well-executed tests are usually the outliers when you look on the internet.
>>54887422
You aren't comparing cards with similar performance, and you aren't using it in a situation where that card would be used by somebody who buys it
Basicaly, you are being disingenious
>>54887401
Reminder: 280, 280X, 290, and 290X all had the same TDP rating; you're delusional if you think all four of them drew the same amount of power.
>>54887434
> but in my experience this actually tends to be the case in practice.
An analysis of power draw is literally plug in a device and average the numbers over your gaming session. It's not rocket science. No I don't believe 90% of them are doing it wrong is very likely. It it was something more complicated I might agree
>>54887422
>A Ferrari uses more petrol than a Honda,
>Honda wins
>Not comparing it with a Lambo
>The city of being retarded like you
>>54887436
The difference between 290X and 980 are diminishing every day.
Literally, with every new game, the average performance/watt on 290X card gets closer and closer to 980. This isn't even counting DX12, simply counting DX11's mature drivers on AMD.
>>54887441
>Reminder: 280, 280X, 290, and 290X all had the same TDP rating
No they didn't, stop lying
the 280x had a 250 TDP while the 290x had a 290 TDP
https://www.techpowerup.com/gpudb/2398/radeon-r9-280x
https://www.techpowerup.com/gpudb/2460/radeon-r9-290x
>>54887451
Thank god the 980 isn't being actively sold anymore and nvidia has replaced it with a new model.
While AMD still hawks a 3 year old card as new.
>>54887451
>The difference between 290X and 980 are diminishing every day.
Resorting to memes again, top kek
>>54887436
>You aren't comparing cards with similar performance
Not even that guy, but, ????
The 290X may even inch ahead in DX12 games.
>>54887480
>DX12 games
>Not a meme
funny guy
>>54887445
>JUS PLUG A KILLAWATT IN DA WALL AN IT BE DON
People like you are why shit power supplies get sold.
See this graph? This is the 12v rail on a $300 Gold rated 1500w power supply. Most review sites didn't fucking catch how they cut as many corners as possible in the design, resulting in this. Because most review sites use shit ass setups where they stick a killawatt in the wall and call it done.
>>54887473
>Thank god the 980 isn't being actively sold anymore
LuL
>>54887473
Again with moving goal posts.
>tdp of architecture vs architecture -> performance tier of architecture -> age of retail significance of architecture
kek
JUST to humor you, 290X isn't being sold either. So we can't make ANY comparison right? fucking hilarious.
>>54887489
How is that relevant to the average power draw of a card over a gaming session?
If those spikes would happen frequently you would simply see a higher average power draw
>>54887499
>290x isn't being sold either
A 390x is a 290x though moron.
>>54887488
>He posted a graph that proved me factually wrong
>Better use the word meme to nullify his argument
Top cuck.
>>54887445
dude I have no fucking clue about benchmarks and I can think of at least 5 ways to fuck that up right off the top of my head
>not isolating power draw between the GPU and the rest of the system
>mismatched / non-equivalent driver/game settings
>not compensating for differences in cooler design or airflow
>using a different fan profile or thermal target curve
>not normalizing for / providing performance delivered
>>54887512
And 980 is still sold.
This is just too hilarious. Are you really that retarded?
>>54887513
Literally 24/7 the AMD shills crawl out to point out the only benchmark they win. Forgetting to mention that nobody actually plays those games.
>>54887513
What DX12 exclusive games (so no DX11 mode) are really out right now?
That's right, no major titles, so yeah it is a meme
Vulkan is obviously the superior option since its cross platform
>>54887522
>980 is still being sold
It's just the remaining stock, Nvidia aren't manufacturing any more.
>>54887523
I wonder if any of them actualy plays Asots?
>>54887522
http://www.techpowerup.com/221783/nvidia-reportedly-stops-production-of-certain-maxwell-gpus
They've stopped making maxwells you stupid shill
>>54887463
http://www.anandtech.com/show/7457/the-radeon-r9-290x-review/
>AMD doesn’t list the power consumption of the 290X in any of their documentation or specifications, and after asking them directly we’re only being told that the “average gaming scenario power” is 250W.
Techpowerup DB has been wrong on more than one occasion mate.
>>54887531
And you think AMD is still manufacturing 390s and 290s?
I'm beginning to worry about you nvidiafags. Do you live in a world where only nvidia exist in your imagination and your reality? Just apply the same fucking logic and reasoning to AMD. This type of thinking is just too backwards for me.
>>54887503
Because killawatts are a shit way to measure power consumption, and the example was to show how outliers can reveal shit.
>>54887559
>You think AMD is still manufacturing 390s?
Yes, they're still actively selling them.
>>54887558
Stop moving the fucking goalpost, you made a retarded statement before checking the facts and got exposed.
>>54887570
ok then, Nvidia is manufacturing Maxwell chips too then
http://www.newegg.com/Product/Product.aspx?Item=N82E16814127838
>>54887570
Again, 980 are still actively sold as well.
Logic? Please /v/, go back to your games
>>54887566
That doesn't answer what I said, why wouldn't spikes be included in their calculcation of the average power consumption?
And why should you focus on spikes instead of average consumption what actualy matters for your power bill
>>54887575
290X, 290, 280X, and 280 all have the same TDP
250 watts.
Do you think a 280 uses 250 watts? Do you think back in 2012 when the 7950 launched, that it used 250 watts? It had the same TDP back then you know, right?
>>54887374
actually AMD cards do less well in 3Dmark 11 compared to real-world. Nvidia does better in this particular synthetic.
>BTFO
>>54887570
This is just embarrassing to watch. A shill can't even see what they're typing.
>>54887523
>>54887527
Sigh, shills out in force today. Where did I even once mention one game?
I posted a graph to disprove a shill when he said the 980 and 290X don't have similar performance, they do, end of.
I couldn't give two shits about aots. I just mentioned that future DX12 games will likely favour the 290X. And DX12 games are coming, stop spouting this only 1 game meme.
>>54887590
>290X, 290, 280X, and 280 all have the same TDP
No they don't, I already showed you they don't
why are people so autistic about muh gpu? I had a r9290 and I can still run all games on max with a slight overclock.
>>54887591
Lets see some evidence for that
>>54887601
You posted a link to the Techpowerup GPU database.
I posted a link to a review of the 290X, with a statement that AMD considers the TDP to be 250 watts.
Smith, the writer of the review, thinks the TDP is a bit too low, but the official AMD statistic is that the 290X has a 250 watt TDP.
>>54887604
me2, @2560x1440...
people with 770/780/780 Ti already getting that insane gimp with new games, losing out to midrange AMD cards
>>54887639
>http://www.anandtech.com/show/7457/the-radeon-r9-290x-review/
From your own article
>>54887645
>people with 770/780/780 Ti already getting that insane gimp with new games, losing out to midrange AMD cards
Which midrange AMD cards are you talking about
>>54887655
(unofficial)
Again, as I said, the author of the review thinks the TDP is 300w, and thus put it down on there as such.
It should be noted that Anandtech measures powerdraw at the wall, not at the GPU however.
I posted the link to get the direct statement that AMD themselves consider the TDP to be 250 watts. Not for the statement by a reviewer who doesn't even have access to proper power measurement equipment.
Your picture even has "300W (Unofficial)" on it. 250w is the official.
Quit dicking about.
>>54887674
And they put it there because they think it is closer to the truth.
The 290x does not have the same typical power draw as a 280x, which should have been obvious
>>54887685
>I posted the link to get the direct statement that AMD themselves consider the TDP to be 250 watts
That just means AMD is retarded with their TDP ratings, and was probably to ashamed to put actual TDP of the 290x on it 300w is much close to the truth, and thus the 280 doesn't have the same tdp as a 290x
>>54887669
>380x, 970, 290, 390, RX 480 ,btfo
>2013 800$ card, while the 600$ at the time 290x is now at near-980 levels and rising.
You got fucking memed kid
>>54887688
>>54887700
And it's already been shown their opinion was categorically false when actual power measurement equipment was used. The 290X draws 250 watts nearly exactly. >>54887222
So yeah, 250 watt TDP for 290X, 290, 280X, and 280.
Also, TDP means Thermal Design Power. Who the fuck told you it meant Typical Power Draw?
>>54887701
>380x beats 970's in new games
Oh really, heres DOOM, you see where the 380x is ? And do you see where the 970 is?
Oh let me guess your response will be
>b-but that doesn't count! here have a benchmark of Asots!
![index[1].png](https://i.imgur.com/UBdrKhT.png "index[1].png")
>>54887718
Yeah 380x is the only card he listed that is behind the 780 Ti though.
>>54887729
Bro, the difference between the 380x and the 970 is over 20% even in your cherrypicked post
Your meme of the 380x beating the 970 in new games is just that, a meme
>>54887718
2013 $800 card means 780 Ti
Honestly, it aged like milk.
>>54887685
>>54887700
TDP is not power consumption. They are related but not same.
TDP is rated by different vendors using different tools/benchmarks done internally.
So AMD's 290X's TDP is 250w. The reviewer got confused by TDP and Power Consumption.
>>54887741
>new cards beat older cards
Stop the presses
>>54887741
>780ti
>shows 4k results
Nobody games on 4k with a 780ti, it has 3GB of Vram ffs stop being an idiot
>>54887754
780 Ti and 290 launched at the same time in late 2013, 290 was $400 on launch, 780 Ti was $750
280X, which is getting damn close to it, was released in early 2012.
>>54887760
280X has 3gb of VRAM mang.
>>54887754
>mfw 290 was released at 350$ a month before 780 Ti came out.
fucking shill btfo
>>54887771
>280X has 3gb of VRAM mang.
Older AMD cards getting closer to Keppler cards only makes sense.
Nvidia doesn't optimize for Keppler anymore, while AMD is still stuck on their third rebrand of the same architecture, so optimizations of their 'newer' cards still benefit their older cards.
This is only because AMD is still stuck on their old shitty architecture though which is not a good thing obviously
>>54887754
>$350 AMD OLDER card beat $750 nvidia newer card
>$350 AMD older card still competes competitively and have much longer life ahead due to DX12 in the future
>>54887791
So you're saying, if you buy nVidia expect your $700 GTX 1080 to last only 2-3 years
Thanks for pointing out a good point.
>>54887791
>third rebrand
>1.1 they entirely tossed out the frontend with and redid from scratch
>1.2 they entirely tossed out the memory controller and media codecs and redid from scratch
>Polaris they implemented this: http://www.freepatentsonline.com/20160085551.pdf (ie massive shader and scheduler changes)
You're just pissed off they kept the name around.
>>54887815
Yes, Nvidia actualy creates new improved architectures every 2-3 years in contrast to AMD.
The kind of people that shell out the money for a new 1080 will buy a new graphics card every year anyway so what do they care.
>>54887826
Those are just optimizations, not actualy new architectures, you fell for marketing terms, good goy
>>54887827
>Maxwell isn't just Kepler with an adjusted control/compute ratio, the crossbar removed, and redesigned ROP's
The last major Nvidia architecture change was Fermi to Kepler, the last major one before that was G70 to Tesla
>>54887841
>Maxwell wasn't a brand new architecture over Keppler
I'm not wasting any breath on you anymore, you are literally tech illiterate
>>54887827
nvidiafags = applefags of GPU
Homosexual and retards
>>54887882
You sound salty
>>54887139
TDP != power consumption.
Nvidia always advertises significantly lower TDP than AMD, yet final power consumption numbers are wildly different.
Wait for actual benchmarks you silly sausage.
>>54887882
this
>>54887922
>Nvidia always advertises significantly lower TDP than AMD
You're right, but actual power draw was still miles ahead of AMD's comparable cards
>>54887853
http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/4
>When NVIDIA introduced the Maxwell 1 architecture and the GM107 based GTX 750 series, one of the unexpected aspects of their decision was to release these parts as members of the existing 700 series rather than a newer series to communicate a difference in features. However as it turned out there really wasn’t a feature difference between it and Kepler; other than a newer NVENC block, Maxwell 1 was for all intents and purposes an optimized Kepler architecture. It was the same features built upon the efficiency improvements of the Maxwell architecture.
>With that in mind, along with the hardware/architectural changes we’ve listed earlier, the other big factor that sets Maxwell 2 apart from Maxwell 1 is its feature set. In that respect Maxwell 2 is almost a half-generational update on its own, as it implements a number of new features that were not present in Maxwell 1. This means Maxwell 2 is bringing some new features that we need to cover, but it also means that the GM204 based GTX 900 series is feature differentiated from the GTX 600/700 series in a way that the earlier GTX 750 series was not.
if you go line by line on the features, all major ones were just ROP functions
Sorry to say, but Maxwell was iterative atop Kepler, not a break like Fermi to Kepler.
>>54887948
That makes the jump in performance per watt they made from Keppler to Maxwell without a die shrink even more impressive.
Where is AMD's leap in performance per watt that isn't based solely on the die shrink?
>>54887971
GCN 1.2 was bretty good actually
Fury X uses less power than the 290X usually.
>>54887997
That is because of HBM, come on dude, don't be like that
>>54887997
also compare that to 1.0, the 280X
not as big an increase of course, but 1.0 compared to 1.2 is nearly as much as Kepler to Maxwell
>>54888009
4gb of GDDR5 at the frequency the 290X ran them at only used around 30 watts of power mang.
iirc HBM1 uses 25 watts
>>54888027
Source:your ass
>>54887580
>>54887579
They just rebranded the 290x as the 390x you stupid fucks. Nvidia has stopped making 980s. AMD still actively manufactures and sells 290x/390xs
>inb4 they're not the same card because they have a different sticker
First of all, because we're talking about AMD, I doubt the power efficiency of Polaris will be as good as Pascal. However, at mainstream level, it doesn't matter. You won't be getting into housefire territory until large dies appear
>>54887139
>Now that both parties have shown their 150w tdp cards, the 1070 and the 480
As other posters have correctly stated, TDP does not equal power consumption.
>it's definately slower than the 1070
No shit. Some more powerful card, 480x or 490, probably also with a 150W TDP if it's a 2500SP part, will be the 1070 competitor
>>54887192
Ur getting rekt m8
>>54887346
But they did you fucking moron, 6-pin + PCI-e vs 8-pin + PCI-e
>>54887441
From experience, the 280x/7970 was hot as fuck. Quite likely it put nearly as much heat into my room as a 290 would have
>>54887512
>A 390x is a 290x though moron
No, it's not. Same performance, but different arch. 280x is a straight rebrand
>>54888034
https://www.amd.com/Documents/High-Bandwidth-Memory-HBM.pdf
using stated bandwidth per watt numbers:
290X has 320gb/s / 10.66 = 30 watts
Fury X has 512gb/s / 35 = 15 watts
bretty good off the top of my head then.
>>54887139
It is not meant to compete the 1070 you retards. It's still decent for a 200 $ 980 with similar performance. The 490 will compete the 1070.
>>54888099
>a 390x is not a 290x
Lmao, a 390x is a straight 290x. The 380x is the one with the new arch. And even then it's worse than tonga.
>>54888099
>300 series not rerbrands
http://vrworld.com/2015/06/24/amd-moves-to-justify-radeon-300-series-rebrand/
Kek, it''s the 380x that's not a 'straight rebrand' the 280x actually outperforms it.
>>54888109
Nobody is arguing it is idiot.
The point of discussion is whether HardOCP was full of shit or right
>>54888099
7970/280X and 680/770 were pretty close in perf/w
390X is pretty much a 290X; it's a B1 stepping vs an A2 stepping though.
>>54888120
380X is Tonga, 280X is Tahiti; but yeah stock 380X was clocked a bit lower and the 256 bit bus vs 384 bit bus.
Hilariously Tonga does have a 384 bit bus on die, but they only enabled it in Apple products using the chip.
>>54888158
Yeah sorry I meant tahiti.
>>54887243
>That chart still shows Maxwell is well ahead in effiency
It also shows that it uses way more power than its TDP, while the 290x uses way less.
>>54888184
No shit
>>54887257
>On the subject of polaris, i hope they are available in half-height/low profile formats. It'd be a pretty sweet card for a HTPC setup.
I just want a version that isn't fucking 30cm long. The reference design they showed this week was much shorter than that, but padded to 30cm by the cooler itself. So here's hoping there will be a shorter non-reference design by someone.
>>54888184
So the moral of the story is that Maxwell is more efficient.
Thanks.
>>54888257
Moral of the story, "more efficient" is under 50w difference.
>>54888281
You mean 60w
>>54888281
Yeah you're forgetting to mention that efficiency also takes into account extra performance per watt.
Using my own 3rd party testing of Polaris, disregarding die shrink and clock speeds, Polaris (gcn1.3) HAS 29% more IPC than Fiji (gcn1.2) AMD lied about 40% improvement over Fiji, but we'll have to wait to see how much power RX480 actually uses
