Friday, July 29th 2016

AMD Polaris 11 "Baffin" ASIC Pictured Up Close

AMD's upcoming 14 nm Polaris 11 "Baffin" ASIC, which powers the Radeon RX 460, was pictured up close, and it's tiny! Pictured as part of a Sapphire Radeon RX 460 Dual-X disassembly by PCOnline.com.cn, the Polaris 11 chip features a tiny package substrate owing to its low pin-count, wiring out a 128-bit wide GDDR5 memory interface; a PCI-Express 3.0 x8 host interface (it fits into x16 slots but has wiring for just x8); and electrical pins to cope with its <75W TDP requirements. The card relies on the PCI-Express slot for all its power draw. Sapphire's Dual-X cooling solution looks beefy from the outside owing to its cooler shroud and pair of fans, but underneath is a fairly simple monoblock aluminium heatsink.
Sources: PCOnline.com.cn, VideoCardz
Add your own comment

46 Comments on AMD Polaris 11 "Baffin" ASIC Pictured Up Close

#1
Captain_Tom
Look how small that thing could be. Low Profile cards incoming.

Pretty excited to see something 20% stronger than a PS4 finally fit into a low-profile form factor.
Posted on Reply
#2
natr0n
I bet it will oc like a beast.
Posted on Reply
#3
AsRock
TPU addict
WOW, where's the rest of it :P.
Posted on Reply
#4
madness777
I feel like GPU cards will get back to their old form, where they had a ridiculously simple PCB and were made to save space rather than look good. Well, I hope, it's gonna cut the cost A LOT!
Posted on Reply
#5
Caring1
Why isn't this single slot?
Posted on Reply
#6
Fouquin
Caring1Why isn't this single slot?
Right?! This is the perfect candidate for a modern HD 4850, but less likely to heat itself to death.
Posted on Reply
#7
seronx
Made in Taiwan... In Taiwan.... Taaaiiiwwaaannn....

globalfoundries.com/about/manufacturing
East Fishkill, New York
Malta, New York
Burlington, Vermont
Dresden, Germany
Singapore


RX480 also states made in Taiwan. ugh.
Posted on Reply
#9
Assimilator
Here a free tip for you Sapphire: if you have to design a PCB with massive areas of unused space just to fit your unnecessarily large cooler, you are doing it so very wrong. And wasting money in the process.

I feel sorry for the people who get suckered into buying this hunk of junk. That cooler has so much aluminium it's probably worth more than the board itself...
Posted on Reply
#10
Frick
Fishfaced Nincompoop
This really shows how impressive the AMD Nano was.
Posted on Reply
#11
Caring1
AssimilatorHere a free tip for you Sapphire: if you have to design a PCB with massive areas of unused space just to fit your unnecessarily large cooler, you are doing it so very wrong. And wasting money in the process.

I feel sorry for the people who get suckered into buying this hunk of junk. That cooler has so much aluminium it's probably worth more than the board itself...
I think it comes back to consumer expectation.
If a card is small, it is seen as less powerful and not as good.
By keeping Graphics Cards a certain size, expectations are met and people feel they are getting value for money.
Posted on Reply
#12
AsRock
TPU addict
Caring1I think it comes back to consumer expectation.
If a card is small, it is seen as less powerful and not as good.
By keeping Graphics Cards a certain size, expectations are met and people feel they are getting value for money.
Yup, as a lot judge a book by it's cover.

But at the same time it looks like they could of made it low profile compatible and gave a it a interchangeable bracket.
Posted on Reply
#13
xorbe
If it's entirely slot powered, why would you waste power on a second fan
Posted on Reply
#14
HD64G
My guess is that first few models based on Polaris 11 will have this size but very soon all manufacturers will come us with very small and probably 1-slot AIB solutions. The rush to make it to the market is responsible for the lack of them now.
Posted on Reply
#15
lanlagger
I read here in comments - "passive", "singleslot", "slot powered", "next 750ti"... Those are are wishful thoughts... but the fact that chip is tinny does not indicate that it has excellent thermals or good performance/watt ratio.... it just indicate it will have good performance/squaremillimeter ratio... as far as we know from existing polaris - it has anything, but good performance/watt and thermals, even with custom triple fan oversized coolers - temeratures are not spectacular (to say the least)... so - yea - wishful thoughts, but not goanna happen
Posted on Reply
#16
Steevo
Perhaps the fans turn off at idle, and only run slowly at full throttle, and it can overclock like a mofo.... and thus the reason for a better cooler.

Perhaps its the process its on is less forgiving of heat.
Perhaps the die size interface area needs such a large heatsink.
Posted on Reply
#17
RejZoR
I hate the codename of this one (Baffin). I always read it as "buffoon"...

Btw, the problem with 16nm chips is that they are tiny. Which means, even if they are energy efficient, the heat is a lot more focused because they are tiny. Meaning it's hard to efficiently take heat away from it. That's the problem both NVIDIA and AMD are facing thee days on 16nm... Especially on mid end models which are really tiny but fast enough to produce decent amount of heat...
Posted on Reply
#18
Steevo
True, how do you afford to put a cooler on that costs more than the die itself? Meaning anything more than a simple extruded or cast cooler is cost prohibitive for the price bracket, so you make it bigger as that is cheaper than adding a heatpipe, using copper, using a small vapor chamber, or even a copper core.
Posted on Reply
#19
Casecutter
Wouldn't one think that as it is marked Taiwan it's a TSMC 16nm FinFET?

Might be a good chip as I have the impression the GloFo Polaris is suffering from some process issues at this point. Basically what we saw when TSMC started their 28nm process, and why there was then the 7970 GHz and why Nvidia was late with a GTX 680/670 as they waited until TSMC fixed their issue.
Posted on Reply
#20
Assimilator
CasecutterWouldn't one think that as it is marked Taiwan it's a TSMC 16nm FinFET?

Might be a good chip as I have the impression the GloFo Polaris is suffering from some process issues at this point. Basically what we saw when TSMC started their 28nm process, and why there was then the 7970 GHz and why Nvidia was late with a GTX 680/670 as they waited until TSMC fixed their issue.
Apparently GloFo likes sticking "Made in Taiwan" on chips that aren't.
Posted on Reply
#21
Casecutter
AssimilatorApparently GloFo likes sticking "Made in Taiwan" on chips that aren't.
:confused:
Posted on Reply
#22
Assimilator
Casecutter:confused:
Refer to seronx's post which lists the GloFo foundries... none of which are located in Taiwan.
Posted on Reply
#23
Casecutter
AssimilatorRefer to seronx's post which lists the GloFo foundries... none of which are located in Taiwan.
Exactly, so it's pretty apparent it's a TSMC part. So perhaps it won't be of the same shortcoming that the GloFlo Polaris seem to be having.
Posted on Reply
#24
rtwjunkie
PC Gaming Enthusiast
CasecutterExactly, so it's pretty apparent it's a TSMC part. So perhaps it won't be of the same shortcoming that the GloFlo Polaris seem to be having.
Very strange, since GloFlo is supposed to be the manufacturer. And only one of their facilities is set up for the right specs to do it, which is the Malta, NY plant, north of Albany. I'm guessing that facility is at maximum capacity and AMD realized they would have to bring in someone else.
Posted on Reply
#25
SonicZap
rtwjunkieVery strange, since GloFlo is supposed to be the manufacturer. And only one of their facilities is set up for the right specs to do it, which is the Malta, NY plant, north of Albany. I'm guessing that facility is at maximum capacity and AMD realized they would have to bring in someone else.
Another possibility is that GloFo's process has worse power efficiency, and since low power use will be critical for <75 W desktop and especially laptop design wins, AMD went with TSMC's better process for Baffin. It would also explain why the RX 480 gets completely slaughtered by Pascal when it comes to power efficiency.
Posted on Reply
Add your own comment
Apr 27th, 2024 10:36 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts