[Computerbase] R9 390 (X) Roundup

[xthetenth]

Guess the Tri-X 390X isn't as good as the Fury Tri-X. This thing is dead silent in an R4.

It looks like the Tri-X is a great baseline, but the shorter PCB is a big help to take the Fury Tri-X all the way to the top.

 

[Gikaseixas]

Guess the Tri-X 390X isn't as good as the Fury Tri-X. This thing is dead silent in an R4.

Air flows freely at the end of the card, allowing much fresher circulation through the heatsink fins. Sapphire just perfected GPU cooling and others will follow

EDIT: started already and i bet Nvidia's partners will adopt this for Pascal

 

[96Firebird]

A lot of armchair engineers here...

Out of curiosity, how many of you have taken a class like thermodynamics or heat transfer?

The heat pipes not in contact with the GPU are fine, if indeed the pipes are in contact with each other. This is the big caveat, as air is a very poor heat transfer medium. However, a little space between the pipes can work. Thermal expansion of the pipes will occur, but the CTE of copper isn't much, especially below 100°C.

If the pipes are in contact, though, they will conduct almost as much heat away from the source as the heat pipes that are in direct contact with the GPU (well, in direct contact with the thermal paste).

Here is a quick FEA of a simple system like the card. This assumes contact between the pipes, however there is a fillet around the edges so it isn't full contact. I didn't take a screenshot, but the top view shows a slightly different gradient on the end pipes. The middle pipe has the most heat the farthest out towards the extensions, where the cooling takes place.

Any questions, let me know...

 

[Techhog]

A lot of armchair engineers here...

Out of curiosity, how many of you have taken a class like thermodynamics or heat transfer?

The heat pipes not in contact with the GPU are fine, if indeed the pipes are in contact with each other. This is the big caveat, as air is a very poor heat transfer medium. However, a little space between the pipes can work. Thermal expansion of the pipes will occur, but the CTE of copper isn't much, especially below 100°C.

If the pipes are in contact, though, they will conduct almost as much heat away from the source as the heat pipes that are in direct contact with the GPU (well, in direct contact with the thermal paste).

Here is a quick FEA of a simple system like the card. This assumes contact between the pipes, however there is a fillet around the edges so it isn't full contact. I didn't take a screenshot, but the top view shows a slightly different gradient on the end pipes. The middle pipe has the most heat the farthest out towards the extensions, where the cooling takes place.

Any questions, let me know...

I have one: Why are we still arguing about this when the cooler is badly designed for its purpose, no matter what the cause of that bad design is?

 

[3DVagabond]

Sapphire Nitro 390 = 38 dBA @ 70*C (Great result), total system power usage = 386W
PCS+ 390 = 36 dBA @ 76*C, total system power usage = 371W
Sapphire Tri-X 390X = 45.5 dBA @ 76*C = 430W
Asus R9 390X Strix = 51 dBA @ 89-91*C, total system power usage = 469-489W

So we will ignore the FACT that the Asus cooler performs poorly and argue it's fine? Look at the numbers. The cooler sucks.

Once again we have people arguing for the company with the best marketing.

 

[AtenRa]

So we will ignore the FACT that the Asus cooler performs poorly and argue it's fine? Look at the numbers. The cooler sucks.

Once again we have people arguing for the company with the best marketing.

I will speak about me, im not saying the ASUS cooler is the best design. I have only said that the pipes will still work even without direct contact with the GPU Chip.

 

[3DVagabond]

I will speak about me, im not saying the ASUS cooler is the best design. I have only said that the pipes will still work even without direct contact with the GPU Chip.

Still work yes, but they work poorly. Nobody says they do absolutely nothing, but people making it sound like they are just as effective as if they made direct contact, which is how that cooler is designed to operate (as the DC in the name stands for direct contact), are incorrect. The proof is in the numbers.

 

[AtenRa]

Still work yes, but they work poorly. Nobody says they do absolutely nothing, but people making it sound like they are just as effective as if they made direct contact, which is how that cooler is designed to operate (as the DC in the name stands for direct contact), are incorrect. The proof is in the numbers.

The heat-sink design IS Direct Contact, just because one or two pipes doesnt come to direct contact with the GPU chip doesnt invalidate the design and the naming.

 

[3DVagabond]

The heat-sink design IS Direct Contact, just because one or two pipes doesnt come to direct contact with the GPU chip doesnt invalidate the design and the naming.

I didn't say it did. Just that the heat pipes by design are supposed to be in direct contact with the heat source and 2 of the 5 aren't, thus reducing the efficiency.

 

[Flapdrol1337]

A lot of armchair engineers here...

Out of curiosity, how many of you have taken a class like thermodynamics or heat transfer?

The heat pipes not in contact with the GPU are fine, if indeed the pipes are in contact with each other. This is the big caveat, as air is a very poor heat transfer medium. However, a little space between the pipes can work. Thermal expansion of the pipes will occur, but the CTE of copper isn't much, especially below 100°C.

If the pipes are in contact, though, they will conduct almost as much heat away from the source as the heat pipes that are in direct contact with the GPU (well, in direct contact with the thermal paste).

Here is a quick FEA of a simple system like the card. This assumes contact between the pipes, however there is a fillet around the edges so it isn't full contact. I didn't take a screenshot, but the top view shows a slightly different gradient on the end pipes. The middle pipe has the most heat the farthest out towards the extensions, where the cooling takes place.

Any questions, let me know...

I think the most heat transfer from pipe to pipe is via the big mounting block directly on top of the pipes.

 

[Headfoot]

Case reopened.

Temperature differences could be caused by other factors (different radiant surface area, diferent spacing and position of the heatsink fins, different materials of the heatsink, different fans). Not necessarily because the heat pipes are not enclosed in a block. It could be this, but I did not see any demonstration of that (excluding the "it looks like it is worse" type of arguments).

Yep, it's everything but the fact that only 3 out of 5 heat pipes touch the thing that makes heat.

Bottom line, the ASUS one is the worst one by a mile. Yall ever heard of Occams Razor?

 

[KingFatty]

Do the heatpipes touch each other? In the picture from the earlier posting, it looks like there are lines/gaps between the pipes, so maybe that's what interfering with efficient heat transfer between the pipes and resulting in such bad numbers.

Also with heat pipes generally, what is the threshold temperature where the fluid inside the pipe will vaporize at the heat source? Maybe the gap between pipes prevents enough heat transfer to allow the warmest part of the heat pipe reach the threshold vaporization temperature (for those outer pipes that don't make direct contact), so the heatpipe is effectively neutered and instead of acting like an efficient heat pipe with vaporizing fluid, it's instead just acting like a boring metal heat conductor.

Heatpipe needs heat to evaporate the fluid:

 

[thesmokingman]

I don't understand the defensive Asus supporters. It's been common knowledge or should be for some time now that Asus on AMD is garbage.

 

[Gikaseixas]

Asus on AMD is garbage

I've been trying to say this all along, they're ok on Nvidia but don't work as they should on AMD.

The Strix cooler works better on NV high end I guess, I get around 73/74 degrees max on my card. The older ASUS DC2 design also had problems when bundled with 290/290x. I think these coolers are designed for Nvidia cards and adapted to AMD ones. That's only a theory

 

[LTC8K6]

The pipes are flattened, so there appears to be a gap, but there actually isn't.

How can the design work on NV but not on AMD though?

Is NV's die a lot bigger?

 

[LTC8K6]

The pipes are pretty clearly well in contact with each other and with the block.

http://www.vortez.net/articles_file/25546_asus-r9-290-direct-cu-oc-directcu.jpg

 

[Techhog]

The pipes are flattened, so there appears to be a gap, but there actually isn't.

How can the design work on NV but not on AMD though?

Is NV's die a lot bigger?

Yes.

Hawaii/Grenada: 438 mm²

Fiji: 596 mm² (and Asus's cooler actually does a good job with this one...)

GM200: 601 mm²

GK200: 561 mm²

So, Asus was lazy with the cooler design. It's that simple. They should have made a version which actually works for Hawaii instead of going the "one size fits all" route, since it clearly doesn't.

 

[Erenhardt]

The pipes are pretty clearly well in contact with each other and with the block.

http://www.vortez.net/articles_file/25546_asus-r9-290-direct-cu-oc-directcu.jpg

There will be a temp gradient. The further away from the heat source, the lower temps are.
Thermodynamics work in such a way, that the heat pipe cannot get hotter than a threshold depending on the phase-change fluid used in the heatpipe, as long as there is a fluid to be vaporized in the heat pipe.

So, in order to have a heatpipe warmed to the temperature required for effective vaporization (boiling) to occur, you need to have a source of heat that can deliver the energy.

Energy will flow from higher temp body, to lower temp body, with a gradient dependent on the thermal conductivity (resistance).

The heatpipes that do not have direct contact with the die will work with very limited capability as long as the heatpipes with direct contact have some phase-change fluid in them to be vaporized.

After reaching the point where the heatpipes with direct contact can't handle the energy they will overheat, then enough heat will transfer through the metal layers of heatpipes, cooler base, solder, TIM etc, to other heatpipes without direct contact with the core to rise their temperature to the boiling point of the fuild.

There is a reason why we don't use metal rods instead of vapor based heatpipes. Phase-change cooling is much more effective.

 

[Gikaseixas]

Fiji: 596 mm² (and Asus's cooler actually does a good job with this one...)

This
I'm mean, if the cooler is bad like some are saying, why is it so good with Fiji? Fiji/NV bigger die area touches more heatpipes while Grenada die being smaller makes contact with fewer pipes... might explain this phenomenon

 

[Techhog]

This
I'm mean, if the cooler is bad like some are saying, why is it so good with Fiji? Fiji/NV bigger die area touches more heatpipes while Grenada die being smaller makes contact with fewer pipes... might explain this phenomenon

The cooler is still bad either way, since they designed it to be universal and failed. All they needed to do was add a copper plate. Asus is 100% to blame here for their laziness.

 

[RussianSensation]

WOW, what an insane hot deal happening right now.

PowerColor PCS+ R9 390 for $225. Who is in on chipping in for Kleenex boxes to send out to GTX960 4GB owners? :'biggrin:
http://forums.anandtech.com/showthread.php?p=37754961#post37754961

 

[tential]

390 prices dropping like rocks. Looks good for holiday season

 

[RussianSensation]

390 prices dropping like rocks. Looks good for holiday season

Maybe the average PC gamer will wake up and stop buying GTX960 2/4GB. Nah. :hmm:

Warning issued for disregard of forum-specific posting rules.
-- stahlhart

 

[garagisti]

WOW, what an insane hot deal happening right now.

PowerColor PCS+ R9 390 for $225. Who is in on chipping in for Kleenex boxes to send out to GTX960 4GB owners? :'biggrin:
http://forums.anandtech.com/showthread.php?p=37754961#post37754961

Don't forget butthurt be gone.

That's a great price. Hopefully some offers may come on Fury X as well.

 

[Erenhardt]

Makes me think I should upgrade to freesync and r9 290/390.

I even prepered a lenghty PM to RussianSensation to help me make up my mind, but decided not to bother him.

I should stay strong and fight the itch...