R9 390X Cooling

[DiogoDX]

Please AMD return the backplate for reference cards. I loved the backplate in my old 5970. I'm totally in for a 390X with AIO and backplate and at least 50% faster than 290X.:thumbsup:

 

[HurleyBird]

Please AMD return the backplate for reference cards. I loved the backplate in my old 5970. I'm totally in for a 390X with AIO and backplate and at least 50% faster than 290X.:thumbsup:

+1

Backplates are awesome.

 

[tollingalong]

Please AMD return the backplate for reference cards. I loved the backplate in my old 5970. I'm totally in for a 390X with AIO and backplate and at least 50% faster than 290X.:thumbsup:

Last rumor I saw was the 390X might only be 40% faster than the 290X so you're out of luck there (nuclear power plant needed to power the thing wasn't included).

 

[JDG1980]

Last rumor I saw was the 390X might only be 40% faster than the 290X so you're out of luck there (nuclear power plant needed to power the thing wasn't included).

I wonder if the R9 395 X2 will need a double-fan radiator (240mm x 120mm). It's also interesting to speculate how many PCIe power connectors it's going to take.

 

[el etro]

Last rumor I saw was the 390X might only be 40% faster than the 290X so you're out of luck there (nuclear power plant needed to power the thing wasn't included).

Source? You really know what you're talking or you're trolling?



----------

Let's wait now more info for the upcoming 295x/390x then. We know barely nothing about the card at this time.


----------

A bet of mine: in order to battle the earlier forms of Maxwell that is still at 28nm, they needed to make a bigger GPU in order to remain competitives at high end, and they know that sooner or later the Big Maxwell will come to market. And, due to 16/20nm not coming soon as they expected, they needed to built an architecture to work as a bridge for the Volcanic Islands to the Next-Gen at current node, and this bridge is PI GPUs.

One thing we know is that, opposing to Nvidia's decision, AMD is building bigger GPUs as they can using smaller die sizes as they can. It results on hotter and more power hungry chips(due to higher transistor density), as we saw on 290x.
Build a GPU bigger than Hawaii with so few architectural improvements and virtually no reduction on power consumption per SP could result on a Video Card with a inappropriate total power consumption. IMO that's why AMD claimed for Asetek help in order to cool this coming up Video Card.

Would be good to see the impact of cooling solution in the thermal and power performance of the card. If the water cooling cuts out 50W for the power consumption of this big card and lower 40 degrees of temperature, it will be a good decision. Summed with HBM, it can bring a outstanding reduction on power used by this GPU.

About the GPU itself: No more than 512bit is needed(and 512 bit is already an exaggeration that only serves for 4k purposes), and i wish they build a not so complex PCB in order to make the 390x/295x easier to be made. If the GPU is exactly twiced Tonga with no more architectural improvements, then there is no much besides more performance to wait from it. I personally want to see what new things Fiji brings to the table. I want to see a 4,5GB Vram 384bit Video Card, if HBM clocking permits the GPU be fed with the proper bandwidth.

 

[tollingalong]

Source? You really know what you're talking or you're trolling?



----------

Let's wait now more info for the upcoming 295x/390x then. We know barely nothing about the card at this time.


----------

A bet of mine: in order to battle the earlier forms of Maxwell that is still at 28nm, they needed to make a bigger GPU in order to remain competitives at high end, and they know that sooner or later the Big Maxwell will come to market. And, due to 16/20nm not coming soon as they expected, they needed to built an architecture to work as a bridge for the Volcanic Islands to the Next-Gen at current node, and this bridge is PI GPUs.

One thing we know is that, opposing to Nvidia's decision, AMD is building bigger GPUs as they can using smaller die sizes as they can. It results on hotter and more power hungry chips(due to higher transistor density), as we saw on 290x.
Build a GPU bigger than Hawaii with so few architectural improvements and virtually no reduction on power consumption per SP could result on a Video Card with a inappropriate total power consumption. IMO that's why AMD claimed for Asetek help in order to cool this coming up Video Card.

Would be good to see the impact of cooling solution in the thermal and power performance of the card. If the water cooling cuts out 50W for the power consumption of this big card and lower 40 degrees of temperature, it will be a good decision. Summed with HBM, it can bring a outstanding reduction on power used by this GPU.

About the GPU itself: No more than 512bit is needed(and 512 bit is already an exaggeration that only serves for 4k purposes), and i wish they build a not so complex PCB in order to make the 390x/295x easier to be made. If the GPU is exactly twiced Tonga with no more architectural improvements, then there is no much besides more performance to wait from it. I personally want to see what new things Fiji brings to the table. I want to see a 4,5GB Vram 384bit Video Card, if HBM clocking permits the GPU be fed with the proper bandwidth.

http://wccftech.com has the rumored specs. Estimate based on that is about 40% more. Not anything concrete for sure.

I also agree with your 28nm point of view. 20nm is around the corner for AMD: Faraway Islands. Personally, I think Nvidia might still have better power consumption this round though.

 

[3DVagabond]

http://wccftech.com has the rumored specs. Estimate based on that is about 40% more. Not anything concrete for sure.

I also agree with your 28nm point of view. 20nm is around the corner for AMD: Faraway Islands. Personally, I think Nvidia might still have better power consumption this round though.

I don't subscribe, so I can't see behind the pay wall, but I'm pretty sure "Faraway Isl." was Charlie trolling those who plagiarize his work. Faraway Isl. is from the Pokeman universe.

 

[tidy]

.

 

[JDG1980]

No more than 512bit is needed(and 512 bit is already an exaggeration that only serves for 4k purposes)

Why would anyone buy these monstrous GPUs if not for 4K or other ultra-high resolutions? At 1440p, a R9 280X is usually going to be enough.

 

[tollingalong]

I don't subscribe, so I can't see behind the pay wall, but I'm pretty sure "Faraway Isl." was Charlie trolling those who plagiarize his work. Faraway Isl. is from the Pokeman universe.

I don't pay either it's one of their with the pic on upcoming cards. Either way I'm taking the rumors with a grain of salt on AMD cards since they're some time away.

 

[fleshconsumed]

+1

Backplates are awesome.

What the hell do you need backplates for? Unless you have a windowed case and you enjoy viewing the innards of your PC on a regular basis the backplate is completely useless, it just traps the heat from the top of the card.

 

[DiogoDX]

What the hell do you need backplates for? Unless you have a windowed case and you enjoy viewing the innards of your PC on a regular basis the backplate is completely useless, it just traps the heat from the top of the card.

Backplate actually helps in cooling. My 5970 had termal pads in the backplate for memory chips. Look at the accelero xtreme 4 that have a huge backplate that cools vrm for behind.

And will look nice in my 760T.:biggrin:

 

[Lepton87]

Why would anyone buy these monstrous GPUs if not for 4K or other ultra-high resolutions? At 1440p, a R9 280X is usually going to be enough.

I don't agree, I had a single Titan driving a 2560x1440 monitor and while it was usable it could still choke at times so I added a second one. Bear in mind that a Titan is significantly faster than a 280X about 40% faster. At first I wanted to upgrade to a new card but a single card that has the performance of 2xTitan/285X2 is nowhere in sight. I guess it can't be done at 28nm, we need to wait for a 20nm flagship and if the situation with 28nm repeats itself then the first newcomer will be a mid-size die with the performance of a Titan x1.5(kind of like 680 was to 580 but it wasn't even 50% faster closer to 40% if that) and about a year later a big die comes with hopefully over twice the performance of a Titan. Such a long wait is unacceptable to me especially because there would be no performance per dollar improvement. So for about 18 months of waiting I would gain maybe 10% of performance and no hassle inherent to multi GPU solutions and what's irrelevant to me lower power consumption. I decided that I'd rather take that performance now and enjoy it then wait for so long. All because of late and expensive 20nm process. Or do you guys think that twice the performance of Titan is possible on 28nm process?

 

[DiogoDX]

http://wccftech.com has the rumored specs. Estimate based on that is about 40% more. Not anything concrete for sure.

I also agree with your 28nm point of view. 20nm is around the corner for AMD: Faraway Islands. Personally, I think Nvidia might still have better power consumption this round though.

Tonga is 8% better than Tahiti with same clocks and bandwidth by hardware.fr review. If the rumored 4224 sps (66 CUs = 50% more tham 290X) is confirmed we will see 50% or better performance.

 

[garagisti]

So, the rumors of AMD building a bigger chip seems to be right...

I think the guy who was first with information on 295x2, was the one mentioning this tidbit, big chip.

 

[Bateluer]

I don't agree, I had a single Titan driving a 2560x1440 monitor and while it was usable it could still choke at times so I added a second one. Bear in mind that a Titan is significantly faster than a 280X about 40% faster.

I call bull on this. I ran a 7950B from September 2012 to November 2013 driving a pair of 1440p monitors, and not a single game I ever loaded on them was unplayable using stock Ultra quality settings. If I'm not mistaken a 7950B is slower than a 280X. Your Titan should have had no problems at all at 1440p.

 

[Lepton87]

I call bull on this. I ran a 7950B from September 2012 to November 2013 driving a pair of 1440p monitors, and not a single game I ever loaded on them was unplayable using stock Ultra quality settings. If I'm not mistaken a 7950B is slower than a 280X. Your Titan should have had no problems at all at 1440p.

Play for example metro LL with everything maxed out or C3
See this:

http://www.hardocp.com/article/2014...rix_platinum_video_card_review/7#.VBjZpPmSyPs

about mid-40fps average without even maxing out AA, I need to have about 60fps average at the minimum, their criteria for play-ability are too lax for me. Everyone is different.

Watch dogs is also well below 60fps on today's single fastest cards

http://www.hardocp.com/article/2014...rix_platinum_video_card_review/4#.VBja1vmSyPs

780TI OC gets 49.3fps almost playable by my standards but why buy such an expensive card if 2x290 would get the job done well better for way less money. I know that I payed even more for my Titan, a moment of stupidity an impulse buy when I found out that my now ex- wife was cheating on me. I agree that if you are willing to forgo expensive AA methods in lieu of post processing AA like FXAA/SMAA and turn down some resource hogs(UBER sampling in the Witcher 2 takes the cake) that don't improve image quality max you can comfortably play on one fast card like 290(X)/780(TI)/Titan as for 280X you would need to further turn down setting as such cards are better suited to 1080p but certainly usable at 2560 but not maxed out in the most demanding games.

 

[Bateluer]

Play for example metro LL with everything maxed out or C3
See this:

http://www.hardocp.com/article/2014...rix_platinum_video_card_review/7#.VBjZpPmSyPs

about mid-40fps average without even maxing out AA, I need to have about 60fps average at the minimum, their criteria for play-ability are too lax for me. Everyone is different.

Watch dogs is also well below 60fps on today's single fastest cards

http://www.hardocp.com/article/2014...rix_platinum_video_card_review/4#.VBja1vmSyPs

780TI OC gets 49.3fps almost playable by my standards but why buy such an expensive card if 2x290 would get the job done well better for way less money. I know that I payed even more for my Titan, a moment of stupidity an impulse buy when I found out that my now ex- wife was cheating on me. I agree that if you are willing to forgo expensive AA methods in lieu of post processing AA like FXAA/SMAA and turn down some resource hogs(UBER sampling in the Witcher 2 takes the cake) that don't improve image quality max you can comfortably play on one fast card like 290(X)/780(TI)/Titan as for 280X you would need to further turn down setting as such cards are better suited to 1080p but certainly usable at 2560 but not maxed out in the most demanding games.

Watch Dogs is shoddily optimized; its an Ubi-Soft game. Won't comment any further on it.

Its important to remember that you also don't need much AA at 1440P and high. So even though 40fps is playable on a single player FPS, you could simply drop the AA to 2x or even off entirely.

With my 1440p display, the only way for me to get games to drop below 60fps is to crank AA up to 8x.

 

[Subyman]

If AMD doesn't do a proper update of their architecture then we'll start seeing liquid coolers strapped to their single GPUs.

 

[f1sherman]

If AMD doesn't do a proper update of their architecture then we'll start seeing liquid coolers strapped to their single GPUs.

They are at the point where Intel had been at the dawn of NetBurst era.

With Tonga I was expecting Maxwell like perf/W advancement, but all they gave was Kepler.

 

[3DVagabond]

If AMD doesn't do a proper update of their architecture then we'll start seeing liquid coolers strapped to their single GPUs.

Have you not even looked at the OP?