AMD R9 390X Coming With Cooler Master Liquid Cooler + Estimated Performance

[Faa]

This heat output will not magically go away just because the card is kept below its throttling temperature.

He never said it will go away >.<

[ivan134]

 

Clocking the GPU higher by around 10-20% without a bump to your voltage barely increases the power consumption, 690's/Titan-Z's aren't cut down because an extra 100-200MHz clock speed would bump the temperatures, can be still countered with a more agressive fan profile.

 

 

 

A 290 doesn't pull 275W.. A 290 is similar to a 780 power consumption wise, I never saw the 780 I had hitting 250W.

 

There's no such award >.< If there is, then it's AMD giving themselves an award.

 

Careful, your fanboyism is leaking through. Not only are you falsely disputing something with no evidence, but you're claiming if it's true, the "books must have been cooked". Sounds like Republican logic.

 

http://linustechtips.com/main/topic/254704-amds-hawaii-is-officially-the-most-efficient-gpgpu-in-the-world-to-date-tops-green500-list/

[Paragon_X]

For those joining now this Glorious thread. 


Segments of the thread. TL:DR: 

excitement about performance --> excitement about the AIO cooler --> Bashing AMD about the AIO cooler and that the cards its too hot that needs it -->defending the AIO cooler --> Argument about radiators and fins --> Arguement about TDP vs Power consumption --> Argument about what efficiency is --> Argument about how energy works --> Comparing the 390X to the sun --> Arguement on how Quantum mechanics can explain the 390X will wear itself off like the sun vs why it will not.

That was a fun ride.

[Khvarrioiren]

For those joining now this Glorious thread. 

Segments of the thread. TL:DR: 

excitement about performance --> excitement about the AIO cooler --> Bashing AMD about the AIO cooler and that the cards its too hot that needs it -->defending the AIO cooler --> Argument about radiators and fins --> Arguement about TDP vs Power consumption --> Argument about what efficiency is --> Argument about how energy works --> Comparing the 390X to the sun --> Arguement on how Quantum mechanics can explain the 390X will wear itself off like the sun vs why it will not.

That was a fun ride.

 

I'm actually baffled how much back-and-forth offtopic discussion there's been without the thread getting locked.

[s3ns3]

--> Comparing the 390X to the sun --> 

when a 390X dies it goes supernova

[othertomperson]

He never said it will go away >.<

 

Look at the post he quoted. When did I even mention the 290X overheating? I mentioned high TDP and my reasons for not going back to high TDP cards, relating my experience of GTX 580s that also didn't throttle but made gaming a very uncomfortable experience in summer. His response to this is "stick a better cooler on it." Well that won't fix any of the issues I have with a high TDP card.

[Humbug]

Look at the post he quoted. When did I even mention the 290X overheating? I mentioned high TDP and my reasons for not going back to high TDP cards, relating my experience of GTX 580s that also didn't throttle but made gaming a very uncomfortable experience in summer. His response to this is "stick a better cooler on it." Well that won't fix any of the issues I have with a high TDP card.

did you actually notice an effect on room temp? I can't tell with an overclocked R9 290. And i live in the tropics where it's hot anyway... I suppose it depends on the room and your proximity to the tower..

[othertomperson]

did you actually notice an effect on room temp? I can't tell with an overclocked R9 290. And i live in the tropics where it's hot anyway... I suppose it depends on the room and your proximity to the tower..

 

Of course I did. There was a *massive* decrease in my ambient room temperature between SLI 580s and SLI 970s.

[patrickjp93]

No... they're not. They have an extremely low detection limit and precision. They're very accurate instruments. Spectroscopy in general is very accurate and reliable. (Most detectors are accurate within 1-2%) They're actually used to quantify chemical gas leaks. (usually measured in g/hr) to within 2-3 g/hr depending on the compound. (So it can detect very very minor leaks.)

Mass spectrometers are based on a very different technology than this thermal camera as measure the various spectra of light which are emitted when a chemical is hit with specific wavelengths of light and electrical signals. They're not remotely comparable.

[patrickjp93]

Careful, your fanboyism is leaking through. Not only are you falsely disputing something with no evidence, but you're claiming if it's true, the "books must have been cooked". Sounds like Republican logic.

http://linustechtips.com/main/topic/254704-amds-hawaii-is-officially-the-most-efficient-gpgpu-in-the-world-to-date-tops-green500-list/

To be fair that supercomputer has made that list only due to having the most efficient cooling system in the world. It used to be that you needed 1.4W of energy spent on cooling for every watt spent on running the supercomputer itself (and this was a state of the art system). Now Microsoft builds data centers that only require 0.88W of cooling electricity for every watt spent on their computing. AMD is solely in the top spot because of a truly exotic and stunning thermal solution.

[sgloux3470]

Mass spectrometers are based on a very different technology than this thermal camera as measure the various spectra of light which are emitted when a chemical is hit with specific wavelengths of light and electrical signals. They're not remotely comparable.

 

Spectroscopy is a general term covering any interaction between mass and light, whether it be UV-vis or IR.

[Toddskins]

For those joining now this Glorious thread. 

Segments of the thread. TL:DR: 

excitement about performance --> excitement about the AIO cooler --> Bashing AMD about the AIO cooler and that the cards its too hot that needs it -->defending the AIO cooler --> Argument about radiators and fins --> Arguement about TDP vs Power consumption --> Argument about what efficiency is --> Argument about how energy works --> Comparing the 390X to the sun --> Arguement on how Quantum mechanics can explain the 390X will wear itself off like the sun vs why it will not.

That was a fun ride.

Very funny, and actually good summarizing.  

[Toddskins]

 I cannot recall who, but there was something said about both companies are still on a 28nm fabrication process (did I type that correctly?).

 

How do any of you know that AMD's new card cannot be on the 20nm? process?  Is there some objective data that dictates this piece of knowledge?

 

Thanks in advance.

[SGTWolfmann]

This makes me wish I was old enough to work. Damn government saying I'm too young to work. (Crosses arms and makes pouty face)

[Opcode]

 I cannot recall who, but there was something said about both companies are still on a 28nm fabrication process (did I type that correctly?).

 

How do any of you know that AMD's new card cannot be on the 20nm? process?  Is there some objective data that dictates this piece of knowledge?

 

Thanks in advance.

One thing worth noting is there's really not a 20nm process from TSMC or GloFo that's optimal for a 20nm GPU. It's better to stick with 28nm SHP or something along them lines. AMD will be using the 20nm node just not for their GPUs. I suspect Amur or something to use the 20nm low power nodes.

[s3ns3]

 I cannot recall who, but there was something said about both companies are still on a 28nm fabrication process (did I type that correctly?).

 

How do any of you know that AMD's new card cannot be on the 20nm? process?  Is there some objective data that dictates this piece of knowledge?

 

Thanks in advance.

TSMC has 20nm only for mobile chips. and judging by this cooler that the new card will require you can bet it won`t be on 20nm

[patrickjp93]

One thing worth noting is there's really not a 20nm process from TSMC or GloFo that's optimal for a 20nm GPU.

Well, they do, but they're calling it 14/16nm despite the fact it's really 20nm with FinFETs added lol. Meanwhile at Intel 22nm processes are reaching their end maturity and final figures are coming out as the flagship Haswell E7 Xeons hit the shelves. Soon Intel's 22nm fabs will be opened to use by others while true 14nm FF production is in full swing and heading towards the SHP node for Broadwell E5/E7 Xeons even as Broadwell desktop is heading out the door to retail warehouses in the next month or so.

[s3ns3]

Well, they do, but they're calling it 14/16nm despite the fact it's really 20nm with FinFETs added lol

 

emm no TSMC only promised 16nm. they said it would be avaible q1 2015. however they delayed againg till 2016. so amd went with global foundries but its still only 28nm. 

Edited February 20, 2015 by s3ns3

[Faa]

Careful, your fanboyism is leaking through. Not only are you falsely disputing something with no evidence, but you're claiming if it's true, the "books must have been cooked". Sounds like Republican logic.

 

http://linustechtips.com/main/topic/254704-amds-hawaii-is-officially-the-most-efficient-gpgpu-in-the-world-to-date-tops-green500-list/

Haha, the typical naivity of AMD fanboys. 

"In order to qualify for The Green500 List, a supercomputer must achieve performance at least as high as the 500th-ranked system in the current version of the Top500 List."

http://www.green500.org/docs/pubs/RunRules_Ver1.0.pdf

So even if your datacenter is better in terms of performance/watt you can't qualify, one datacenter cheating with having watercooling instead of air allowing for a higher peak, so tell me how exactly this is any proof of Hawaii being the most power efficient in compute?

980: 1700/170= 10 perf/watt

290x; 1473/286= 5.15 perf/watt

[ivan134]

Haha, the typical naivity of AMD fanboys. 

"In order to qualify for The Green500 List, a supercomputer must achieve performance at least as high as the 500th-ranked system in the current version of the Top500 List."

http://www.green500.org/docs/pubs/RunRules_Ver1.0.pdf

So even if your datacenter is better in terms of performance/watt you can't qualify, one datacenter cheating with having watercooling instead of air allowing for a higher peak, so tell me how exactly this is any proof of Hawaii being the most power efficient in compute?

980: 1700/170= 10 perf/watt

290x; 1473/286= 5.15 perf/watt

You're accusing them of cheating? Can you back that up?

[Faa]

You're accusing them of cheating? Can you back that up?

Not in that sense >.>

Also you're the guy who has to back his claims up, I'm still waiting for your proof Hawaii is the most power efficient GPU for compute. 

[Sonefiler]

No shit the Titan Z has higher heat output, it has two GPUs. The 290X throttling was because the TDP of the cooler wasn't high enough for the TDP of the card. This heat output will not magically go away just because the card is kept below its throttling temperature.

I'm saying that if Nvidia can make an air cooler for a 500 TDP, then AMD can make one for a 300W TDP. God you have a lot of chromosomes.

[Faa]

-snip-

Already did and I've provided the closest power draw of those cards. http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/20

Stop calling me a retard if you're thinking there's nothing more than GPU's on the floor in a datacenter.

[othertomperson]

I'm saying that if Nvidia can make an air cooler for a 500 TDP, then AMD can make one for a 300W TDP. God you have a lot of chromosomes.

 

Mods, I don't care how many times you edit this out, using Down Syndrome as an insult is not acceptable and I'm not going to keep a reference to that out of this post.

 

It doesn't matter how good the cooler is. I don't care if the card doesn't throttle. There are other reasons for not wanting a GPU with a high TDP which I've already gone into.

[zappian]

-snip-

 

Calm down guys ...

 

No need to get the thread locked.