ASUS Readies Radeon R9 Fury STRIX with DirectCU III Cooler

[HumanSmoke]

No because I'm hoping ASUS doesn't take all that nonchalantly. You can add all the hyperbolizing you what but... It's not just adding "hold down bracket"

Right, because Asus aren't able to make a bracket to the tolerances required (or to use a full cold plate/support bracket), yet TSMC mass produces the same bracket for every AMD GPU. So either TSMC are geniuses, or AIB's are dumb as a short bus full of redneck zombies.

the problem I see is with HBM module are small, so for a heat-pipe to work excellent they have to be perfectly placed over the center of a heat pipe (right on top of the module), to do it right. I can't see them having a gap straddling an HBM module (in such case that would be at least two modules) at best the pipes might be reused. Now unless those modules as to heat aren't all that worrisome, but as of yet I have heard either way. Lastly the cooler fins probably won't work as they need to be stamp in a pattern that provides a larger opening for the block. So sure it a cooler perhaps something like that.

HBM is just stacked DRAM running at 1.2V using 3.3 watts per stack....

....so just how much heat are you expecting HBM to generate considering many 7Gbps GDDR5 equipped cards using 1.55V generating twice (or more) the amount of heat frequently receive no cooling whatsoever?

from what I read on several places, and even in forums of semiaccurate that this is the same cooler used on the non-x version of the fury. Usually semiaccurate isn't full of bs like some places. Also the PowerColor Devil 13 picture was on semiaccurate which will be used for the dual Fury card based on the surface area/ water cooler plate.

I would think that the cold plate is a logical choice, if only to negate the chance of breaking the silicon during assembly. The plate is little more than a copper plate with a support edge to maintain rigidity - basically the same as the Fury X, without the need for machine screws that seal the lower and upper halves of the water chamber

 

[Casecutter]

yet TSMC mass produces the same bracket for every AMD GPU.

I'm not sure what you term as the "bracket" I believe you might be talking about the metal surround or "frame" that the substrate is package on? So your picture and this fury package are the same size bracket/frame?

so just how much heat are you expecting HBM to generate considering many 7Gbps GDDR5 equipped cards using 1.55V

I'm sure it not anything like what the TitanX rolls with but, HBM being condense on such a small surface I really haven't heard if even on the lower power there's a concentration of heat they need to dissipate. But relooking at the package and imagining the surface perhaps tube placement directly over the HBM die wouldn't be that big of issue in the infancy stage. Though like you are saying a plate that provides a flat uninterrupted surface is probably a safer bet than using direct-touch pipes at least at first. But then I suppose AIB have cycle stressed enough of the interposer HBM packages on direct-touch pipe to get a real world idea if there will be problems.

 

[HumanSmoke]

I'm not sure what you term as the "bracket" I believe you might be talking about the metal surround or "frame" that the substrate is package on? So your picture and this fury package are the same size bracket/frame?

Yes and No.
Yes, the support bracket is the raised steel surround that adds rigidity to the package and is high enough to ensure that once the cooler assembly is secured to the GPU, the correct spacing between the cooler heat transfer surface and GPU allows for heat transfer without crushing the silicon.
No. All current AMD GPUs aside from Fiji are packaged by TSMC. The Fury X is packaged by Amkor (TSMC supply the GPU, Hynix supply the HBM and interposer) - it they who manufacture the support bracket.

 

[Casecutter]

No. All current AMD GPUs aside from Fiji are packaged by TSMC. The Fury X is packaged by Amkor (TSMC supply the GPU, Hynix supply the HBM and interposer) - it they who manufacture the support bracket

Not sure why that's a "No" definitely what present here has nothing to do with the support bracket/ frame being a larger part for Fury and the interposer package.

AMD had to find a group that could take on the mounting of the Die and HBM to the interposer. I wonder if there's more profit for AMD, as once packaged and even possibly mounted to the PCB, AIB aren't doing much. AMD is taking looking at most all the risk at that point.

 

[HumanSmoke]

Not sure why that's a "No" definitely what present here has nothing to do with the support bracket/ frame being a larger part for Fury and the interposer package.

FFS, I was answering your question.

So your picture and this fury package are the same size bracket/frame?

You seem to be indulging in some deliberate trolling. You have already supplied pictures in this thread - did you manage to cull pictures yet completely fail to see the difference in shape and size between this and this? Unless you're totally blind it should be readily apparent from scaling the fixed points of reference like the PCI-E slot finger, that the package framing is different.

 

[Casecutter]

I'm guessing you don't disassemble many graphics cards.

You trolled me with that cut, I didn't say anything to you, I just used the picture and pointed out the why Asus couldn't utilize that same cooler. You seem to be the one with thin-skin.

You seem to be indulging in some deliberate trolling. You have already supplied pictures in this thread - did you manage to cull pictures yet completely fail to see the difference in shape and size between this and this? Unless you're totally blind it should be readily apparent from scaling the fixed points of reference like the PCI-E slot finger, that the package framing is different.

I'm not trolling... nice spin and redirect (smoke and mirrors). I'm not the one failing to see!
I've been presenting the case that Asus can't just "probably" utilize that same cooler... And if the members re-read the sequence of this it obvious who's "indulging".

It'll probably utilize same DirectCU cooler as found on Strix GTX 980Ti.

Probably. The GPU+HBM interposer looks tailor-made for the large heatpipe base footprint of the cooler

J'ai fini

 

[HumanSmoke]

You trolled me with that cut, I didn't say anything to you, I just used the picture and pointed out the why Asus couldn't utilize that same cooler. You seem to be the one with thin-skin.

Well, so much for your assertion. 100% wrong.

...so a string of posts arguing why Asus couldn't use the same Strix DC3 as the 980Ti, and lo and behold....the same cooler!

I'm not trolling... nice spin and redirect (smoke and mirrors). I'm not the one failing to see! I've been presenting the case that Asus can't just "probably" utilize that same cooler... And if the members re-read the sequence of this it obvious who's "indulging".

For some of us that follow the tech, the fact that Asus would use the same cooler was obvious.

 

[newtekie1]

Not to say they couldn't make some changes or adapter, but it doesn’t appear to be a simple bolt on.

Well, so much for your assertion. 100% wrong.

I'm just saying, technically @Casecutter was right from the beginning. They did, in fact, make an adapter to fit the 980Ti cooler on the Fury.

 

[HumanSmoke]

I'm just saying, technically @Casecutter was right from the beginning. They did, in fact, make an adapter to fit the 980Ti cooler on the Fury.

That was my original point back in post #6

The spacing of the mounting holes between cards using the same cooler on both vendors cards has differed for generations. Currently the spacing for Hawaii/Pitcairn/Tonga is 53.2mm, while GM 200/204/GK 110/104/GF 100/110 is 58.4mm. It is a relatively simple matter to machine a hold-down adapter (and placement of heatsink plates for power phases) for using the same cooler with both vendors.

...and then he segued into a whole new diatribe about reworking the design...

No because I'm hoping ASUS doesn't take all that nonchalantly. You can add all the hyperbolizing you what but... It's not just adding "hold down bracket", it means a new block (larger bolt spacing) that the pipes then need to be swedged into. Sure I'm not saying it can't be done, the problem I see is with HBM module are small, so for a heat-pipe to work excellent they have to be perfectly placed over the center of a heat pipe (right on top of the module), to do it right. I can't see them having a gap straddling an HBM module (in such case that would be at least two modules) at best the pipes might be reused. Now unless those modules as to heat aren't all that worrisome, but as of yet I have heard either way. Lastly the cooler fins probably won't work as they need to be stamp in a pattern that provides a larger opening for the block. So sure it a cooler perhaps something like that.

...when all that was done was to reuse the exact same cooler with a thermal pad. Hold down brackets are particular to every set of GPU packages for coolers depending upon the package size.