amd r9 fury bench leaked?

[Waco]

From AMD:

 

Lower voltage combined with lower clock speeds (remember, this is CMOS, power usage scales with frequency) combined with direct active cooling makes the heat a non-issue.

[DanTheGamer11]

But how does being close to another major heat source make it easier to cool :')

It's weird that they went with a bare die considering abrasive paste and a mechanical load are being applied directly to it :/

 

edit for ^

True, but the gddr packages only have one die level(I saw 4 on a plane in a slide somwhere) vs the four stacks with pretty much the area of one die to dissipate it, I wonder if whoever made the hbm has datasheets for it

Edited June 26, 2015 by DanTheGamer11

[Waco]

GPU dies are usually bare. CPU dies were bare for a very long time until heatspreaders became popular (and people still delid for that).

 

The thermal density of air is so low that the core, even being close, likely contributes very little to the heat in the memory.

[El_Capitan]

For DDR3 RAM, 85C (1x refresh rate), or 95C (2x refresh rate) was the thermal operating limit.

 

For HBM, you're using a lot less power than DDR3, DDR4, and GDDR5. In the architecture itself, there are thermal dummy bumps that help with thermal dissipation between the stacks.

 

http://www.hotchips.org/wp-content/uploads/hc_archives/hc26/HC26-11-day1-epub/HC26.11-3-Technology-epub/HC26.11.310-HBM-Bandwidth-Kim-Hynix-Hot%20Chips%20HBM%202014%20v7.pdf

[bp9801]

Having everything on the die means the core and memory can be actively cooled instead of relying on passive means.

 

 

 

[DanTheGamer11]

Is the fury x not cooling the hbm modules too waco? pcb/die and interposer transfer heat too

 

Thanks el capitan for the pdf thing. If the leakage and losses for the memory things are the same then hbm seems a bit harder for cooling, didn't know about those dummy bumps, guess they saw a need for them

Did you see any of the reviewers showing the memory temp?

 

saw a video and apparently the back of the silicon chip is what you see, so that's why its black, not cause of the reflections

[Waco]

I'm not sure what you're trying to say, but you're not following along.

[DanTheGamer11]

repeat then, I'm feeling the same since you mentioned air

 

http://cdn.videocardz.com/1/2015/06/LR-Fury-X.jpg

 

lack of prints of hbm stacks means the hbm isn't cooled/doesn't use paste?

 

http://cdn.videocardz.com/1/2015/06/Biadu-R9-Fury-X-6.jpg

 

that looks nice 

Edited June 26, 2015 by DanTheGamer11

[bp9801]

The HBM chips and GPU core are all on the same die, so if you cool the GPU core, you're also cooling the HBM because they are right there next to it.

[DanTheGamer11]

I know that, I assumed that, don't see why air was mentioned in that case

Didn't see any reviews with memory temp yet, I guess people aren't interested in it since the heatsink covers all whether you want it to or not, that or it doesn't show temps for it? The layout doesn't help my question either, don't think it will be possible to compare to gddr5 or whatever modules

[bp9801]

The Fury and Nano are air cooled, Fury X is water cooled.

 

Why do you need the memory temperature? When has a review looked at the memory temperature? Who cares about the memory temperature?

 

If you're cooling the GPU, you're also cooling the HBM stacks and VRM. Everything sits under the water cooler on the Fury X, and probably will for the air cooled Fury, too.

 

 

[DanTheGamer11]

We were talking about the air (not sure why tho) between the gpu and the hmb stacks, not how the whole is cooled

 

Hmm thought overclockers looked at memory temps

 

yeah said that in previous comment

 

don't think I get an answer for that q, too many dissimilarities between gddr5 modules and the hbm stacking