Wednesday, June 15th 2016

AMD Confirms "Ellesmere" and "Baffin" GPU Specs

In its post-E3 press-deck, AMD confirmed the specifications of its Polaris10 "Ellesmere" and Polaris11 "Baffin" GPUs. The two chips will drive at least three desktop discrete SKUs between them, the Radeon RX 480, the RX 470, and the RX 460. Of these, the RX 480 and RX 470 appear to be based on the "Ellesmere" silicon. This chip features 2,304 stream processors spread across 36 compute units (CUs), and a 256-bit wide GDDR5 memory interface. The "Baffin" silicon, on the other hand, features 1,024 stream processors spread across 16 CUs, and a 128-bit GDDR5 memory interface, likely ticking at 128 GB/s.

Source: ComputerBase.de
Add your own comment

47 Comments on AMD Confirms "Ellesmere" and "Baffin" GPU Specs

#1
Chaitanya
Interesting strategy which might pay off for AMD in recovering lost market.
Looks like this Alienware VR backpack uses Polaris.
Posted on Reply
#3
medi01
Some much more impressive comment (chart is from wccftech) by yet another random internet dude (although on GAF spreading FUD is bannable):

Originally Posted by slapnuts

I been itching to say this and i can now but the Rx 480 does overclock like a beast and can reach 1070 level of perf in "games" ...these benches are not ideal for actual game performance, its even better in-game. believe it!! Im so proud of AMD for doing this, to put an end to the price gouging and giving gamers a fair deal for once!

This is one of the best bang for bucks cards in over a decade!! In the coming days more and more truth will be shown. For what ever reasons behind this...I am so embarrassed at some websites making AMD's Polaris cards far worse than what they really are. Finally some of us can speak up for AMD for once..geesh


Posted on Reply
#4
Dj-ElectriC
Judging by those results^, the RX 470 will be 58% faster than the R9 270X.

Applying it to w1z's summery here:



Will put the RX470 at about 45%-47%
If true, that's actually quite ground-breaking for about 150$.
All we can do is wait now
Posted on Reply
#5
john_
medi01 said:
I am so embarrassed at some websites making AMD's Polaris cards far worse than what they really are.
I am expecting reviewers to try to drive reader's attention away from the perf/$ figures, and to concentrate on the fact that Polaris can't offer GP104's performance out of the box. I am also expecting them to make these cards look like the poor guy/looser guy choice somewhere in their conclusions. And of course all hell will break lose if the cards are not perfect. If the card's cooling system produces 0.00001 dB more noise than the competition, I am expecting lengthy analysis about how this product failed.
Posted on Reply
#6
happita
I'm hoping that both 4GB and 8GB versions will be released at the same time so that I can get what I want on launch day. I got rid of my R9 290 for $220 over a month ago.
Seems like a perfect fit to me ;)
Posted on Reply
#7
medi01
Anand still hasn't got full 1070/1080 review, has he?
Might have got pwnd by nVidia's marketing dept once more.

PS
Oh boy, hardcop's Kyle salty predictions vs what wccftech just leaked... :D
100w
60C
At stock, above Fury ^^ and OCed 980 (1350Mhz) in Frestrike Ultra 1.1
Oh, and 299$ "mighty OC" card looming.

Promises to post OCed benches soon. (in comments section, Khalid Moammer)
http://wccftech.com/amd-rx-480-faster-than-nano-980/
Posted on Reply
#8
HumanSmoke
john_ said:
I am expecting reviewers to try to drive reader's attention away from the perf/$ figures, and to concentrate on the fact that Polaris can't offer GP104's performance out of the box. I am also expecting them to make these cards look like the poor guy/looser guy choice somewhere in their conclusions. And of course all hell will break lose if the cards are not perfect. If the card's cooling system produces 0.00001 dB more noise than the competition, I am expecting lengthy analysis about how this product failed.
Getting your preemptive whining review post in a full two weeks before the review drops. A new record I think.
Posted on Reply
#9
medi01
HumanSmoke said:
Getting your preemptive whining review post in a full two weeks before the review drops. A new record I think.
You may want to check what hardocp's Kyle has "leaked" already... :D
Preemptive, my pants.
Posted on Reply
#10
the54thvoid
medi01 said:
Anand still hasn't got full 1070/1080 review, has he?
Might have got pwnd by nVidia's marketing dept once more.

PS
Oh boy, hardcop's Kyle salty predictions vs what wccftech just leaked... :D
100w
60C
At stock, above Fury ^^ and OCed 980 (1350Mhz) in Frestrike Ultra 1.1
Oh, and 299$ "mighty OC" card looming.

Promises to post OCed benches soon. (in comments section, Khalid Moammer)
http://wccftech.com/amd-rx-480-faster-than-nano-980/
They did a preview but they always seem to be too busy covering Apple 'anything' releases. Where did Anand go to again.... oh yeah...Apple.

As for the WCCFTECH stuff, pinch of salt but we know AMD can make good hardware so if they cut the profit margins on that pricing, it's very feasible. Nvidia on the other hand are using 'flagship' material to raise profit margins on BOM.

As for the pre-emptive whining from John.... to be expected. C'mon now Mr John, let's not be too depressive. TPU will judge fairly.
Posted on Reply
#11
john_
HumanSmoke said:
Getting your preemptive whining review post in a full two weeks before the review drops. A new record I think.
Not really. It just happens to read reviews the last few years and it also just happens to not being a hardcore Nvidia fanboy, to enjoy double standards and biased reviews.

Anyway it is funny how you rush to do a little damage control after reading my post.

the54thvoid said:
As for the WCCFTECH stuff, pinch of salt but we know AMD can make good hardware so if they cut the profit margins on that pricing, it's very feasible. Nvidia on the other hand are using 'flagship' material to raise profit margins on BOM.
I don't think they have to cut any profit margins. Other than doubling their memory on the cards, I don't think an RX480 costs more than a R9 380X to be made. 14nm might be more expensive, but the chips are also smaller.
As for the pre-emptive whining from John.... to be expected. C'mon now Mr John, let's not be too depressive. TPU will judge fairly.
Believe me, I would love to have NO reason to start winning. But after a few years of reading reviews, editorials and looking at how the tech press covers the market, it stopped being just winning. It is experience. It is as easy to guess that the tech press will try to find or create negatives for Polaris, as to guess that in a football match between Germany and Malta, Germany will win.
As for TPU, other than that beautiful editorial, I don't have something negative to write. Reviews here do try to stay objective.
Posted on Reply
#12
medi01
the54thvoid said:
Where did Anand go to again.... oh yeah...Apple
Oh, did he?
That doesn't explain still not having full review though.

the54thvoid said:
TPU will judge fairly.
Like listing 8Gb as cons. Yeah, thanks... Thank god there is 4GB card available this time... :D
Although I admit TPU bias is mostly minor and often close to zero, especially compared to some other, cough, review sites.
Posted on Reply
#13
ZoneDymo
HumanSmoke said:
Getting your preemptive whining review post in a full two weeks before the review drops. A new record I think.
Seems to me its more of a clever move then anything.
Its painting a picture of what biased reviewers will come up with.
This means that if the prediction comes true, its the reviewers that come off as terrible, it discredits them.

Otherwise its a reaction, now its a prediction and a prediction in this context is much stronger.
Posted on Reply
#14
the54thvoid
john_ said:
Not really. It just happens to read reviews the last few years and it also just happens to not being a hardcore Nvidia fanboy, to enjoy double standards and biased reviews.

Anyway it is funny how you rush to do a little damage control after reading my post.



I don't think they have to cut any profit margins. Other than doubling their memory on the cards, I don't think an RX480 costs more than a R9 380X to be made. 14nm might be more expensive, but the chips are also smaller.


Believe me, I would love to have NO reason to start winning. But after a few years of reading reviews, editorials and looking at how the tech press covers the market, it stopped being just winning. It is experience. It is as easy to guess that the tech press will try to find or create negatives for Polaris, as to guess that in a football much between Germany and Malta, Germany will win.
As for TPU, other than that beautiful editorial, I don't have something negative to write. Reviews here do try to stay objective.
Bias is something we ascribe to. You are in AMD's camp - not in a fan boy way but you prefer AMD. A comment I would see as a slight critique of an AMD card, you would see as a terrible injustice. People use all forms of cognitive behaviour when we evaluate things and unfortunately an inherent bias resides in all but the best of us. This bias leads you to make the comments you have already made.
Ironically, in making those comments you have set your self up to find a negative review and use it to justify your inherent bias. When people expect to find something (be it poor reviews, ghosts or god - you will find it) it is simple human nature.

I'm in Nvidia's camp but only so much as they give me the fastest card. I get pissed off at the Async nonsense but then maybe that is my bias that Nvidia surely cant make such a mistake?

Regardless, the only way to read a review is from a purely subjective point of view - is the article stating a fact of the condition or is it conjecture about the results? And if the conclusion is justified based on evidence it is a fair conclusion. Though as always, read many reviews to smooth over a reviewers inherent bias.

The Germany versus Malta match is an unfair comparison as Germany has a far more advanced team with a massive financially invested football league. Malta are minnows. The best comparison is if you read the West's accounts of a news article and then read Russia Times account - the difference is hilarious.

medi01 said:
Like listing 8Gb as cons. Yeah, thanks... Thank god there is 4GB card available this time... :D
By your own WCCFTech linked graph, look at the 'awesome' benefits 8Gb yield over 4Gb. That's what the negative press on 8Gb was. It has very little benefit, except on cards that have the gpu grunt to actually run very high resolutions. Most folk agree that adding double the RAM to a mid tier gfx card will only serve as a marketing gimmick.

EDIT: quick maths - your graph from WCCFTech = double memory gives 1.5% improvement.
Posted on Reply
#15
john_
the54thvoid said:
Bias is something we ascribe to. You are in AMD's camp - not in a fan boy way but you prefer AMD. A comment I would see as a slight critique of an AMD card, you would see as a terrible injustice. People use all forms of cognitive behaviour when we evaluate things and unfortunately an inherent bias resides in all but the best of us. This bias leads you to make the comments you have already made.
Ironically, in making those comments you have set your self up to find a negative review and use it to justify your inherent bias. When people expect to find something (be it poor reviews, ghosts or god - you will find it) it is simple human nature.

I'm in Nvidia's camp but only so much as they give me the fastest card. I get pissed off at the Async nonsense but then maybe that is my bias that Nvidia surely cant make such a mistake?

Regardless, the only way to read a review is from a purely subjective point of view - is the article stating a fact of the condition or is it conjecture about the results? And if the conclusion is justified based on evidence it is a fair conclusion. Though as always, read many reviews to smooth over a reviewers inherent bias.

The Germany versus Malta match is an unfair comparison as Germany has a far more advanced team with a massive financially invested football league. Malta are minnows. The best comparison is if you read the West's accounts of a news article and then read Russia Times account - the difference is hilarious.



By your own WCCFTech linked graph, look at the 'awesome' benefits 8Gb yield over 4Gb. That's what the negative press on 8Gb was. It has very little benefit, except on cards that have the gpu grunt to actually run very high resolutions. Most folk agree that adding double the RAM to a mid tier gfx card will only serve as a marketing gimmick.

EDIT: quick maths - your graph from WCCFTech = double memory gives 1.5% improvement.
Of course we will see something from a -sometimes completely- different perspective. But I am not talking about me, you, or HumanSmoke. I am talking about reviews and tech news. Those should be 100% objective, and while Utopia is a dream, at least we can criticize anything that it is not objective, or when we believe we see double standards. And I try to be very specific when I point to something that looks wrong. And I only point to something when there are indications that something is wrong. I always base my... whining on facts. And the fact is that there will always be something wrong with AMD, as far as hardware reviews.
Anyway. In 15 days we will know what these new Polaris cards will offer and how the press will present them to their readers.

PS I translate and post news in a Greek forum. My articles about Nvidia and Intel are usually more than those about AMD(more products = more news articles). I always keep my opinions far far away from those articles. I am as enthusiastic about Intel and Nvidia new hardware in those articles, as I am about AMD. I keep my personal opinions for all the other posts. Never express them in the news articles. That's what I expect from tech reviews and journalists. To by objective in their reviews and editorials. They can be the crazy fanboys elsewhere.]
Posted on Reply
#16
ZoneDymo
john_ said:
Of course we will see something from a -sometimes completely- different perspective. But I am not talking about me, you, or HumanSmoke. I am talking about reviews and tech news. Those should be 100% objective, and while Utopia is a dream, at least we can criticize anything that it is not objective, or when we believe we see double standards. And I try to be very specific when I point to something that looks wrong. And I only point to something when there are indications that something is wrong. I always base my... whining on facts. And the fact is that there will always be something wrong with AMD, as far as hardware reviews.
Anyway. In 15 days we will know what these new Polaris cards will offer and how the press will present them to their readers.

PS I translate and post news in a Greek forum. My articles about Nvidia and Intel are usually more than those about AMD(more products = more news articles). I always keep my opinions far far away from those articles. I am as enthusiastic about Intel and Nvidia new hardware in those articles, as I am about AMD. I keep my personal opinions for all the other posts. Never express them in the news articles. That's what I expect from tech reviews and journalists. To by objective in their reviews and editorials. They can be the crazy fanboys elsewhere.]

Its impossible for a review to be 100% objective.
What is good?
What is loud?
What is high?
What is low?
What are good functions to have, what are not and why?

Im sure for every one of these you can make up some baseline but in an ever changing climate that will have to change soon as well.
Posted on Reply
#17
john_
I did said that Utopia is a dream :p and that the only thing we can do is criticize. As for what is good, loud, high or low, as you said you can have a baseline. And while this baseline can change with time, it should not change based on the name of the brand.


------
I am going to stop it here. If 10 people quote me, I will have to post 10 posts. In that case I would ban me myself from this thread for totally derailing it :laugh:

Next post from me I hope with more info about Polaris cards :)
Posted on Reply
#18
medi01
wccftech promises OCed results:



the54thvoid said:
By your own WCCFTech linked graph, look at the 'awesome' benefits 8Gb yield over 4Gb.
If you follow this logic 4GB version of 480 will be better than 480 version.... :D

One could continue. Say, "async" is a cons, since "not many games use it".
DirectX 12 support is const too, with the same argument.
Etc.

In my humble opinion extra options that is not of much use (or no use at all), while arguably being of no advantage, is definitely not a disadvantage. (or "cons", at least in my non- native-English speaking world)
Posted on Reply
#19
Aquinus
Resident Wat-man
the54thvoid said:
It has very little benefit, except on cards that have the gpu grunt to actually run very high resolutions.
Come on, you should know better than that. More VRAM only helps when you're already running out of it. Listing 8GB as a con is stupid because it's like saying, "Yeah, I know no games use 8GB right now but, we're fail to acknowledge than maybe, just maybe, newer games might use more than 4GB and when that happens, cards with 8GB are all set in that respect." Simple fact is that we don't know if a lot of these GPUs can handle large memory loads because we have no games to really compare it to. However, if we're looking at numbers, one could assume that the capability of a GPU to utilize memory is based on memory bandwidth and the parts of the GPU that use the most VRAM, which are textures. So now you ask yourself, well, does a GPU have enough texturing capacity if you were to fill up VRAM? We don't know but, what we do know is that AMD cards go much heavier on the TMUs than the ROPs. Take the 390 for example, it has texturing capability somewhere between a 980 and a 980 Ti (on paper,) so it does make you wonder if AMD would really fall over when games do use more than 4GB. So far, some of the very rare cases where I've seen 4-5GB of usage, I haven't noticed a negative impact.

I'm only ranting about this because listing 8GB as a con because it doesn't offer performance benefits for increased capacity is freaking insane. Since when did you upgrade your system memory and gain performance from memory capacity if you weren't already running out of system memory? The point of 8GB isn't to gain performance, it's to not lose it when VRAM usages do start going over 4GB.

Lastly, going from 4GB to 8GB really doesn't increase costs that much, so who cares if you pay an extra 20 dollars for it?
Posted on Reply
#21
john_
I'll better sell my HD 7850. RX460 for 79 is amazing. Yes yes it is not a beast, but considering that the best card you can buy for that price is probably a 250X, this is more than a huge jump.

Posted on Reply
#22
PP Mguire
What's all this? Am I the only one that ignores everything but hard performance numbers and price? I don't care what the reviewer has to say unless there was an issue with faulty hardware or something. People look too far into things to fuel their agenda, geeze. If they release a card that can overclock to comparable 1070 speeds for $200 then more power to them. That's fucking fantastic, and I'll for sure buy one for an mITX rig.
Posted on Reply
#23
P4-630
The Way It's Meant to be Played
Dj-ElectriC said:
Judging by those results^, the RX 470 will be 58% faster than the R9 270X.

Applying it to w1z's summery here:



Will put the RX470 at about 45%-47%
If true, that's actually quite ground-breaking for about 150$.
All we can do is wait now
So around GTX980 performance then.
Not bad for a cheap card.

I still don't regret my MSI GTX1070 Gaming X order though :D
Posted on Reply
#24
trog100
until nvidia come out with a 1060 amd are gonna have a clear run.. they have to win somewhere.. he he

trog
Posted on Reply
#25
bug
john_ said:
I am expecting reviewers to try to drive reader's attention away from the perf/$ figures, and to concentrate on the fact that Polaris can't offer GP104's performance out of the box. I am also expecting them to make these cards look like the poor guy/looser guy choice somewhere in their conclusions. And of course all hell will break lose if the cards are not perfect. If the card's cooling system produces 0.00001 dB more noise than the competition, I am expecting lengthy analysis about how this product failed.
I like how HardOCP does their reviews: they highlight which card play each game at which settings. Because in the end, it doesn't matter if a card is 20% than another if that doesn't translate into higher playable settings. The cost is important, too and sometimes the power draw. But fist and foremost it's what a card can do.
Posted on Reply
Add your own comment