So now that based AMD has revealed the specs of the RX 480, what do you think the specs of the fully unlocked RX 480X will be? How well will it perform? How much more power will it consume, how much will it cost etc.?
>>54852644
Performance: Still less than a 1070
Power consumption: Housefire
Cost: Your house
RX 480 is the '480X' you speak of.
There will be an R9 480 which is what you appear to think the RX 480 is.
>>54852655
Spot the nvidiot shill
>>54852661
Thanks Pajeet, was confused about that as well.
>>54852644
i m wondeeing if this card will be in new macs. i want to upgrade my hackintosh gpu
http://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility/
>In the simplest terms AMD has created a product that runs hotter and slower than its competition's new architecture by a potentially significant margin.
This HOUSEFIRE is already at 150W TDP, more shaders enabled means over 150W TDP and still much slower than GTX 1070 150W TDP
>>54852682
You are buttdevestated.
>>54852682
>website was not invited to launch event
>starts ranting without any sources
>>54852682
Does it still hurt, Kyle?
>>54852682
Why is HardOCP so anti AMD?
At least they were more subtle before they were refused the nano...
>>54852644
>>54852682
>150W TDP
TDP != power consumtion
>>54852913
Protip: TDP gets measured be meauring the power consumption because measuring the actual heat is close to impossible.
>>54852930
No it an approximate figure for cooler designers. For instance, Intel CPUs usually use more power than their TDP specifies
>>54852644
According to videocardz Benchmark leaks. Performance is on Fury/ gtx 980 Level.
Pretty great for 200$ IMHO.
>>54852644
I doubt we're going to see a card with the full Polaris die available on shelves for quite a while because those are likely all being sent directly to OEMs.
>>54852655
>When you just bought nvidia again after their 970 Lie Fiasco. And now after the AMD 480, their upper mudrange Release, you are salty as fuck and try to justify your purchase to yourself.
-The Post
>>54852795
You just answered your question. Kyle threw a bitch fit because AMD wouldn't send him a free card that he could shitpost about on his clickbait website. He's been buttblasted ever since then.
>>54852979
For reference, look at how long it took for them to release a full Tonga chip to end users.
>>54853050
Isn't it still cut, just less?
>>54853103
The r9 380x is the full Tonga, although they renamed it Antigua for the 300 series.
>>54853184
Not him, but I always assumed tonga was just a cut down fury
>>54853229
Not even close, Fury uses Fiji which utilizes HBM. Tonga uses GDDR5.
>>54853050
Yep, R9 285's launch date was 09/02/14, while R9 380X's one was 11/19/15. Hopefully this won't be the case here.
>>54853247
Can they not simply replace the memory controller?
Lots of budget cards come with either ddr3 or gddr5.
>>54853276
That would entail a completely separate chip since the memory controller is built in.
Do we have an idea of the temperature ? will it be an housefire like a lot of people says here ?
>>54852644
>>54852661
RX480X is going to be exclusive for iMacs all 2016. The performance is on par with Fury non X (as by leaked benchmarks). RX480 is slightly below 980, and almost 390X performance wise.
>>54853302
So they make multiple versions of the same gpu?
My laptop gpu has 2 models, 2gb ddr3 or 1gb gddr5.
>>54852839
source?
>>54853184
but full tonga has broader memory controller than 380x
>>54854831
http://www.pcper.com/news/Graphics-Cards/AMD-Confirms-Tonga-384-bit-Memory-Bus-Not-Enabled-Any-Products
>>54854831
>>54854843
Makes sense that they did that, though.
Enabling the additional 128 bit would have increased power draw by about 20W (11%), but performance would have been boosted by less than 5%. Also they would have had to add another 2 GiB of memory, which would have increased the price.
In the end, they would have increased power draw and price of the card by over 10%, for at best 5% more performance. Even spending that 20W on the core would have been more efficient, since it actually is limited by the default power target.
>>54852644
Pretty sure that 36CU is the full die.
>>54855714
Power consumption for the 5K MacMonitor isn't bad though. It looks like the GPUs they sourced are an extremely high binning, or they have some power saving IP not enabled in consumer Tonga, closer to what is seen in Fury. It makes sense as they're both 3rd gen GCN but consumer R9 285 cards were surprisingly lacking.
>>54852644
rumored before 2560 CUs and probably 100-150mghz higher clock
>>54852720
Because Kyle got in argument with AMD over something stupid and now wont top shitting over because he didn't get his free samples in time.
What an autist
>>54856083
This would make the most sense.
>>54856083
>MGHz
That's one beefy ass VRM
>>54856083
I sincerely doubt GCN is pushing clocks that high, its not worth the increase in power.
>>54852682
Thanks for the info buddy, now neck yourself
How does the 480 do compared to a 970 overclocked to 1500 MHz? I already have the 970 but I want something with more than 3.5 GB and 1070 will most likely be 500€ or more.
>>54856600
It's a worthwhile upgrade if you've got the extra cash imo, and you can always get another later
>>54856600
Well a 480 in cf beats a 1080 for cheaper.
I'd stick with the 970 though unless it's not cutting it for whatever it is you do with it.
>>54856600
Who knows, there aren't many benchmarks available yet. It's most likely not quite as powerful as marketed except in some handpicked situations. They are just doing what 90% of companies do before launching a big title, hyping it up as the second coming of Christ. If the actual performance is anywhere near what's currently being advertised, I'm buying it though.
>>54852644
480 has 36 CU
480x will probably have 40 as opposed to 44, because otherwise the performance gap between it and Vega will be too small
>>54856600
Stock 480 should be at least 15-20% faster than a 970. (390x and 980 are about 17% faster)
A heavily overclocked 970 would come fairly close to a stock 480, but then you can still overclock the 480.
what's the the 59hz on hdmi post
What's the difference between 60 and 59?
Does that mean I only get 59 fps?
>>54853496
That's because whatever chip is in your card has a memory controller that is able to utilize ddr3 or gddr5. That's not that difficult to do since gddr5 is just an evolution of ddr3, which is an evolution of ddr2 which is an evolution of... and so on and so forth. Look at NV's 1080/1070. The 1070 is just a cut down 1080 that uses gddr5 instead of gddr5x. The memory controller is able to do that because gddr5 and gddr5x are almost identical from a structural standpoint. That's not the case for HBM. HBM is actually made up of stacks sitting on top of each other, which is totally different from all of the various iterations of ddr that are placed on a plane. It might be possible to make a mem controller that could address HBM as well as traditional ddr, but it would make absolutely no sense when you factor in the cost to develop and deploy it. Also, HBM sits in a completely different performance bracket, so it makes no sense to waste money planning out a die that would be used in such disparate market segments.
(yeah I know that I've oversimplified it to the point of being almost factually wrong)
>>54857548
gddr and ddr are totally different types of memory tough. Please don't mistake gddr3 with ddr3, they are different.
>>54857680
GDDR5 and DDR3 are pin compatible, though they handle signaling differently.
GDDR5 memory controllers can in fact use the same DDR3 DRAM used in desktop DIMMS.
>>54857680
Of course they're different, just like how a pajeet is different from a man. HBM, however, would be more like a silverback gorilla.
>>54852644
>tfw just bought a Geforce 960 for 200€
>together with a new monitor that requires a better GPU to even work
My timing is always glorious.
>>54852839
AMD wins again
>>54856701
Not in CF. Using the multi GPU technologies that DX12 has.
NVIDIA WON
>>54852661
That doesn't sound like a very smart way to name your GPU, since many people refer to card solely by the numbering scheme, which AMD is definitely aware of.
https://www.youtube.com/watch?v=04ITA1_XoqM
are we just going to overlook amd's marketing tactics?
>>54857933
Considering everyone has been aware that this was coming since last year, you're pretty fuckin' dumb bro.
>>54857933
What's it like to not do research before buying things?
>>54858088
being so buttblasted you post a video on youtube about it
holy kek
And they accuse him of being an Nvidia shill
>>54856701
beauty of mantle and every fork of it, is that it plays for AMD strengths and weaknesses.
This is why Async is such a big thing, because AMD always had it, while nVidia saw no benefits in it prior.
>>54858126
you can be sponsored by Intel, AMD and Microsoft, yet still be an nvidia shill.
>>54858094
>>54858115
I didn't know there were leaks of this GPU beforehand.
>>54858150
I think it's safe to say he shills for both Nvidia and AMD
>>54858126
I thought he was a corsair shill
>>54858182
He shills for anyone willing to pay
>>54858182
He's a Noctua fetishist.
>>54858173
doesnt that just make him a positive person
>>54852682
And its also significantly cheaper ;^)
>>54854831
>>54854843
>>54855714
> tonga and no 384b bus
ah yes, the time honored Radeon tradition of wasting die space on shit that never got used...
>>54858277
Neutral you mean
>>54852930
>Protip: TDP gets measured be meauring the power consumption
What? No! You knuckle dragging mouth breather, why do you post when you have no idea what you're talking about?
All the nivida shills are so butthurt, this shit is golden
>>54852644
>So now that based AMD has revealed the specs of the RX 480, what do you think the specs of the fully unlocked RX 480X will be?
RX 480 is already the uncut die. It's only 230 mm^2, so it won't need as much binning as the 312 mm^2 GP104/1070/1080.
>>54856979
>480x will probably have 40 as opposed to 44, because otherwise the performance gap between it and Vega will be too small
Do you honestly think AMD's going to build a HBM2/interposer using GPU that's not at least 50% larger than Polaris?
Each HMB2 module provides 250 GB/s of bandwidth, so you'd need at least two to make a chip worthwhile.
A 2x/3xHBM small and a 3x/4xHBM large Vega would both be well beyond what Polaris 10 could ever hope to offer.
>>54852839
This.
Also Nvidia shills on suicide watch:^)))))))
>>54858508
If it was the full chip they would call it 480x. That's almost certain
>>54858126
He looks like an east european gay porn star
>>54852644
> RX 480X
Fuck that. GIve me 490X.
>>54858555
It's new branding, genius.
> R9 3??[X]
> RX 480
Just like ATI went from the 9000 series to the X000 series ca. 2005, they probably wanted to bump R9, kept the X=10 thing, and decided that too many Xs in a name is confusing.
>>54858473
Tumblr memes belong on >>>/tumblr/
Very well priced given the performance it will offer.
Glad amd is putting some value on the line in GPUs. Hopefully it will force Nvidia to reduce their as well.
My 680 isn't dead yet but with vr becoming more common place I am looking at upgrade options.
>>54858707
>yfw RX 480 Pro
YOUR QUEEN.
>>54858928
>RX 480X
>Two X
>Twice as fast than GTX
>OPPA
>>54853335
Pajeet says it will run cool.
>>54858126
He shills for anyone who will pay him. Nvidia does it so amd is doing it too now.
>RX vs 1060 FUCKING WHEN
>>54858473
>Via 9gag.com
KYS
>>54852682
https://hardforum.com/threads/from-ati-to-amd-back-to-ati-a-journey-in-futility-h.1900681/
does everyone on that forums sucks the op's dick?
>>54858928
9700 Pro was such a fucking godly card, I'm just not sure the "Pro" moniker belongs on anything in the Polaris lineup given Vega's near future existence.
but yeah, ATI/AMD naming has been a shitshow at various times in the past
> Rage ('95-'99): Pro, XL, VR, GL, Ultra, MAXX
> Radeon 8000 ('01): LE, SE, Pro
> 9000 ('02): Pro, SE, XT, XXL
> X000 ('04): LE, SE, XT, Pro, GT, GTO, XL, XT PE, XT Platinum, XT Crossfire Master
> X1000 ('05): Pro, XT, GT, GTO, XTX,
> HD 2000 ('07): Pro, XT
> HD 3000 ('07): X2
> HD 4000 ('08): X2
> HD 5000 ('09): Eyefinity Edition
> HD 6000 ('11): [just numbers]
> HD 7000 ('12): GHz Edition, Boost
> HD 8000 ('13): [just numbers]
> Rx 200 ('13): E, X, X2
> Rx 300 ('15): X, Fury(X), Nano, Pro Duo
> X700/X800 (')
>>54853496
gddr5 is just ddr3.
Anyone integrated graphics here? Feels good to be master race.
> tfw Iris Pro
I almost came.
>>54852971
I'm waiting on the final reviews from independent sources, but if this is true for 200$ I might get one. or I might wait for vega
>>54859193
You get banned if you don't.
>>54859218
GDDR5 has a much higher bandwidth, 2GB of GDDR3 bottleneck any card since the GT 8000 series/ ATi 1000 series
>>54859286
$200 for 4GB, $230 for 8GB widely confirmed at this point.
I'll probably get the 8GB model ASAP then upgrade to Vega/GP102 next Spring or get a 2nd RX 480 depending on how things go with DX12 multi-GPU.
>>54859370
Might as well just buy the 4gb card if you are going to upgrade so soon.
>>54859387
hedging bet for CrossFire.
>>54859195
>tfw building my first computer, going through product names for CPUs and giving up in frustration on AMD's side, thus choosing Intel
>>54859339
It's still the same type of memory, just sits behind a different bus.
A good comparison is when car companies use the same basic big block engine in multiple types of cars, just sitting behind a different gearbox. A heavy duty vehicle like a pickup gets a box for high torque, but a low RPM while a sports car gets a box for low torque, but at a high RPM.
>>54858707
>too many X's
This is the company that released the X1950 XTX CrossFireX Edition from XFX
>>54852655
Very creative...
>>54859195
There was also the 7870 XT
Was a cut down 7970, bretty good value buy actually
>>54852644
>How well will it perform?
Below Fury X. I don't care about anything else, because it's too slow for what I want anyway.
>>54852795
Because Kyle is a butthurt faggot.
>>54859218
>>54859487
GDDR5 is not DDR3 in any way shape or form
Example: GDDR5 supports a read and write in the same cycle, something DDR cannot do, and that requires a vastly different memory cell architecture.
>>54859750
GDDR5 is modification of DDR3 for graphical needs.
>>54859387
>>54859408
DX12 and Vulkan will allow you to span the use of CF/SLI vRAM. Get the 4gb versions.
>>54860710
>dx12
this won't be a thing till late 2018. this meme needs to end. dx11 is still VERY much the norm.
>>54860710
Vulkan/DX12 "allow" pretty much anything in theory but are reliant on game devs (some of the most retarded SW developers on the planet) to actually implement it, and do so correctly.
In any case, pushing textures between cards via PCIe is no faster than just pulling them from main RAM, and I'm not a poorfag with <16GB anyway.
>>54860710
> try to save $60 from a $400+ purchase in the hopes that every dev will implement some semi-theoretical feature
shiggy. diggy.
>>54860806
it will be the norm in situations where performance matters - e.g. new games with graphics demanding enough to push devs to learn DX12 or Vulkan.
>>54860809
The first part of your statement is true - devs need to implement it, though it's one of the easier things to implement (they still might not do it just because of how unpopular CF/SLI is).
The second bit about PCI-E being a bottleneck is not true, however. http://www.guru3d.com/news-story/both-mantle-and-dx12-can-combine-video-memory.html
The way these APIs work, there's no swapping of texture data from one card to another; they both have independent resources and tasks to work on in order to get the whole picture.
>>54860843
Well yeah, saving money, and also just not fucking needing 8gb unless you're gaming in 4K or VR. At which point you're already dropping big bucks, why the fuck are you considering any type of RX 480 in the first place?
>>54852655
isn't the power consumption theoretically about 150-200 watts?
>>54861075
Please, you're making AMD users look even more like poorfags than normal.
A 15% price premium as insurance against bloated assets in future games is pretty damn minimal. When consoles keep something like 5.5-6 GB of their 8 GB of RAM reserved for graphics assets, it's at minimum a safeguard against lazy ports.
>>54861447
1070/1080 is billed as 150/180W.
RX 480 is supposedly 150W as well.
In the last few years though AMD's power estimates seem to be slightly more conservative, so make of it what you will until you see actual review of both products side by side.
>the (reference) RX 480 can draw a maximum of 150W (75W PCI-E + 75W PCI-E 6 pin connector)
>the (reference) GTX 1070 can draw a maximum of 225W (75W PCI-E + 150W PCI-E 8 pin connector)
>both have a TDP of 150W
Guess which card is more likely to go over its TDP.
>>54857415
it means you have like 59.97Hz but the application truncates the number to only show integers.
>>54861517
iirc the 150W was draw. It's rated <150W as per koduri from the linus video. He says less than 150W
>>54852839
I knew somebody would make a cropped image of this as soon as I saw it.
>>54854245
http://exhentai.org/g/932413/f84f5afc97/
>>54861471
well, being totally honest it's more like I just don't see a need. I'm still getting all 60+fps on my 1080p setup with GTX 760 2gb SLI... so even 4gb seems rather high for me.
Then again, I tend not to play lazy console ports, so there's that...
>less faster than my 390x
NO THANKS
>>54861842
>less faster
Pajeet please.
Also that's on paper.
>>54861669
>>54861767
1070 seems to draw 135W-140W in normal gaming, not sure what it would do under FurMark, etc.
RX 480 benchmarks will be very interesting, though for a <150W card it seems kind of academic anyway.
> http://www.pcper.com/reviews/Graphics-Cards/GeForce-GTX-1070-8GB-Founders-Edition-Review/PC-Perspective-Advanced-Power-Te
>>54861884
https://youtu.be/rkiZdSBux9E?t=6m7s
Pardon my french but it can hit 217W draw. It has a theoretical of 225W it wouldn't be surprising to see it go even past that at peak. And that's without OC/custom PCBs.
https://youtu.be/eqCo0LTUhsE?t=4m55s
Less than 150W Raja says.
>>54862009
>comparing real world testing to canned indian marketing statements
nice try raja, AMD made no improvement in mm2/w this generation (438mm2/275w is roughly the same as 232/150w). this means that AMD is masking their design flaws (no real improvements in power efficiency) by putting out physically smaller and cheaper GPUs.
the only real offering AMD has with the 480 is price/perf, so we'll have to wait and see how the $200 variant performs before making any judgement now that we know the GPU is simply another rehash of GCN.
>>54861884
Well, 150w is the max limit on the power lane. If they used 150w completely, then there's no OC room at all.
Chances are, it will use about ~110-130w and depending on games.
>>54863274
which is pretty terrible for a card that's supposed to be 2/3rds the size and 2/3rds the performance of GP104. so much for '''power efficiency'''.
>>54863245
This post right here is pretty cringeworthy.
>>54863293
What's 16 divided by 14?
>>54863318
>facts are cringeworthy
maybe once trump deports AMD's indian staff we'll get real GPUs from them, then people won't have to be cucked by 3.5 and 'premium' reference cards
>>54863342
Ok kid.
