Is This The XFX Radeon R9 390 Double Dissipation?

[GhostRyder]

Wow, if this is the R9 390 then we are getting near the launch of something or someone likes to mess with us. To me what I see more than anything is the fact this is a DD edition of the card (Referencing the name of XFX's cooler) and not a stock AIO/Blower because seeing this out well before launch could mean we will see custom versions out the gate this go round.

As far as XFX goes, they have had their ups and downs on the designs of their cooler recently. I believe the 6XXX series was pretty good while the 7XXX series had its share of issues with the VRM cooling and such. But I believe that all the R9 series coolers resolved that issue for the most part from what I have seen. I have a friend who has a pair of the 290 versions and they have been pretty good.

Looks like this is indeed R9 380, which is supposed to be rebrand of R9 290, which is the same size, have same cooler, have nothing to be developed, just bios flash for the name of the card - all fits.

Well judging from the picture this does not look to be an R9 380 mostly because the middle number in the 3XX part has a curve on the left side but not the right. Could be a problem with the picture but this looks more to me to be a 9 than an 8.

 

[AsRock]

I do the plate is not good enough, it's that simple, design is some what flawed to make it 2 slot. how ever if they moved some fins around so more air would hit directly on the VRM's and gave a little more space for a bigger \ beefier heatsink it be a none issue.

And yes they use the copper type.

Looks like they may of changed that heatsink, i would have to check as it's been a while since i looked at mine. Maybe in the next day i will has i need to swap out a motherboard anyways.

 

[Finners]

If you can afford new high-end GPUs all the time, surely you can afford a chassis which can fit standard PCI Express cards, no?

Its not about cost, I prefer the smallest chassis I can get away with, currently have a cooltek w2 with a radiator in the front and bottom.

Full tower cases just take up too much space for me.

 

[neliz]

Its not about cost, I prefer the smallest chassis I can get away with, currently have a cooltek w2 with a radiator in the front and bottom.

Full tower cases just take up too much space for me.

Some manufacturers like to place drivespaces or fans where they're not even supposed to be. These are PCI Express standards and manufacturers like Fractal are known to ignore them in order to provide a more appealing product.

 

[neliz]

Also, overheating of the VRM is largely related to the electrical design, not the components per sé. If the design is bad, it doesn't matter which component you place on it, too high temperatures will be an issue.

 

[Haytch]

I have never had any issues with reference cooling from AMD.
Sure, when I pump up the fan speed there is a lot of noise, but that doesn't bother me since I wear headphones with noise cancellation and maximum volume.
Reference cooling max's out under 40C under heavy load ever since my 2900XTX days up until my 290X.
Looking forward to the 390/x's, especially the reference cooling ones.

I always read about people complaining about the reference cooling, maybe I just always end up with a 'Super-fan', maybe people are not adjusting fan speeds, maybe people care too much about noise. I have always thought that with great power comes great noise

In-fact, the only reference coolers I ever had trouble with was the Nvidia range. I remember I had to modify my cooler on the 8800GTX, but that was a fun project

 

[BiggieShady]

Also, overheating of the VRM is largely related to the electrical design, not the components per sé. If the design is bad, it doesn't matter which component you place on it, too high temperatures will be an issue.

Components matter a lot because all components doing the same thing are rated differently (efficiency, maximum operating temperature etc.), and when integrated circuit replaces several components on the board then you have both better electrical design integrated inside a single component always with increased efficiency (less energy lost as heat).
For example one power phase designed with regular mosfets can output 70 amps while dissipating 60W as heat ... the same phase using integrated power stage can output 60 amps while dissipating only 11 watts as heat. That's why when you see VRM array with integrated circuits it's often with less phases for the same power capability and much less heat.
Also, it's not that rare for design to have many phases just to make each phase less hot, rather than to allow extra oc.

 

[the54thvoid]

I have always thought that with great power comes great noise

You should know better:

Best built/looking gfx card ever. (IMO)

 

[neliz]

Components matter a lot because all components doing the same thing are rated differently (efficiency, maximum operating temperature etc.), and when integrated circuit replaces several components on the board then you have both better electrical design integrated inside a single component always with increased efficiency (less energy lost as heat).

Did you read what I wrote? I said if the design is bad, it doesn't matter which component you put on there, it will generate too much heat. So if you have a reference design using D-Pak or PowerPak, and you create your own design with DirectMOS and 2 extra phases, it's not automagically going to give you better temperatures.

Also, it's not that rare for design to have many phases just to make each phase less hot, rather than to allow extra oc.

It will lower the load on each phase, but again, if they add 2 phases, on a badly designed PCB with, for instance 2 layers less and 2mm less board space, it will not matter how many phases you'll add, you'll get issues.

 

[BiggieShady]

Did you read what I wrote? I said if the design is bad, it doesn't matter which component you put on there, it will generate too much heat. So if you have a reference design using D-Pak or PowerPak, and you create your own design with DirectMOS and 2 extra phases, it's not automagically going to give you better temperatures.

Yes, I did read what you wrote. For some reason you take a bad design as a major cause of overheating VRM, which would be relevant if we were talking about motherboards. Graphics cards VRM designs don't vary that much, that's why I gave an example how different components on the same design can affect heat dissipation because, hey, 11 watts heat to dissipate is nothing and that's all the way at 60 amps at room temperature ... to power a 250 watt gpu you slap 5 phases without a heatsink and call it a day. The same overall VRM design, only the whole power stage is integrated on a chip.

It will lower the load on each phase, but again, if they add 2 phases, on a badly designed PCB with, for instance 2 layers less and 2mm less board space, it will not matter how many phases you'll add, you'll get issues.

Yes, pcb quality and layout is also important for good heat management ... and your response actually shows you didn't read or get what I wrote.

 

[neliz]

Yes, I did read what you wrote. For some reason you take a bad design as a major cause of overheating VRM

Yes, because I worked on graphics cards for 4 years at MSI, I kind of know what I'm talking about.

 

[xixou]

It looks like R8 380 to me.

R5/7/9 are used, there is space for R6/8/10 ^^

 

[BiggieShady]

Yes, because I worked on graphics cards for 4 years at MSI, I kind of know what I'm talking about.

Well that's great, then you can certainly tell me which component is cooler at operation this one https://www.fairchildsemi.com/datasheets/FD/FDMF6823C.pdf or this one http://www.sinopowersemi.com/RE_UPLOAD_FILE/20130808182117430.pdf
You should recognize the second one.

 

[GhostRyder]

It looks like site corrected this to be an alleged picture of instead the R9 380 sadly and not the R9 390. Though I guess the cooler and design will not change much, but at least it points towards something being closer to release.

 

[Xzibit]

If there is an air cooler for the 390 HBM it will more then likely be bigger heatsink plate due to the HBM.

This just looks like a XFX 290/X with a different shroud really.

 

[HumanSmoke]

Funny how you are extra careful with conditionals ...

Judging by the length ( scaling the PCI-E slot connector), it actually looks to have roughly the same dimensions as their existing 290....
..Interesting if this is a 390 - about time AMD allowed vendor designs on launch day.

And now you know why.

Note: Due to images being of really low quality, it was hard to figure whether the card was actually a R9 390 series or R9 380 series cards. While we were led to believe it as the R9 390 at first sight, turns out to be that the graphics card is actually the Radeon R9 380 series from XFX

BTW: I'm usually pretty assiduous in delineating supposition from fact. I've been following tech way too long to take virtually anything not supported by proof as factual.

 

[Serpent of Darkness]

If we assume that photos from Rumor Mills have some degree of inaccuracy and accuracy, we could conclude the following:

If the photos showing the R9-380 has a lower TDP draw in comparison to it's R9-290/290x cousin, it's possible and reasonable to believe in these images, the R9-380 is design this way because the card itself won't generate higher or equal temperatures on loads.

I believe that AMD has a predictable trend with each generation. AMD 7970 had Frame Time Variance Issues in CrossfireX. Frames were dropping and runt-frames manifested itself. This was more noticeable in the AMD 7990 Graphic Cards. R9-280 was released correcting these issues in the preceding generation. Frame Time Variance Curves displayed that R9-280 was less spikey, curves didn't hit zero, but they weren't ideal curves like NVidia Frame Time Variance Curves. Nevertheless, they were improvements.

R9-290x has GPU Core Frequency throttling issues. This is either due to an increase in temperatures on the GPU or VRam. So naturally, R9-380 (a revision of the R9-290x) is going to have temperature and power draw improvements. Power Draw is actually smaller in comparison to the R9-290x. So the nuclear reactor that we know as an R9-290x, is less radioactive in the R9-380. With this thought in mind, one can conclude that a massive air-cooled setup isn't necessary.

R9-380 doesn't have HBM which is one of many selling point for the R9-390 and R9-390x. Justification for the AMD Premium price tag that you'll see on the R9-390/x. R9-380 will most likely support D3D12.0. I don't believe R9-290x fully supports it, but I could be wrong.

Lets take into account that photos from Rumor Mills have shown R9-390x having a TDP Power Draw near R9-290x, but how accurate is that remains to be unclear. Those numbers won't be confirmed or verified until 3rd party benchmarks and test are made by TPU, etc... I would reasonably and generously suggest that R9-380 would probably have a 10% to 20% power draw less than R9-290x. Consider this an educational guess. In addition, it will probably produce 20% to 30% less generated thermal energy at the same amount of work outputs or loads with the air-cooled solution, on the R9-380.

The "actual" may produce better results than the things I have listed, exceeding the bar of expectations on the R9-380 higher than previously believed. If we think about it, AMD doesn't really need to produce another R9-290x variant and call it a R9-380, place more units on the market with no tweaks or improvements. Otherwise, the already flooded market of R9-290x will only increase with the volume of R9-380 products coming in. The only assumed difference between the R9-290x and the R9-380 would be the labeling in this case. My point is with the current situation, R9-380 "could" meet a higher level of expectations. In a sense, the R9-380 has to exceed the R9-290x, or AMD will just be increasing and investing stock to a flagship tier, one generation ago that consumers aren't buying anymore. Especially with the NVidia GTX 980 pulling a significant performance over it, and the previous generation, the GTX 780 Ti, was already pulling 10% to 22% average FPS over it in the same tier and generation as the R9-290x. I don't believe R9-380 is going to be sold for $100 to $200 dollars just to make up for losses... Especially if the average price of an R9-290x is roughly between $300 and $700 dollars.

 

[kn00tcn]

Its not about cost, I prefer the smallest chassis I can get away with, currently have a cooltek w2 with a radiator in the front and bottom.

Full tower cases just take up too much space for me.

full tower!? that's height, we're looking at depth (or if there are hard drive bays in the way)

the case can be vertically tiny or even a steambox (with pci-e riser to rotate the card) & fit extremely long cards just fine

 

[Casecutter]

It looks like R8 380 to me.

R5/7/9 are used, there is space for R6/8/10 ^^

When I first saw this I was like IDK... It just looked too much like a 290 layout in the heat pipes and other bits. So yea say a 380, that someone doctored.
Interesting thought as to R6/8/10 but that makes for an extremely crowded field

 

[RichF]

I hope someone will make a triple slot card to keep the noise down with one of the high-end 390s (single card, not dual GPU model).

Trying to squeeze all that cooling performance out of dual slot design makes sense if you're going to run Crossfire but not so much otherwise.

 

[btarunr]

I don't know if I should mention this for fear of anyone important actually catching on, but BTA has a tendency to 'correct' things in his news posts when he knows something, which in itself can be telling of a situation (often shortly before launches).

You holding out on us, man?

"So what's the deal, little fella?"

"Well that's classified."

"So I couldn't tell you, even if I wanted to."

"But it is pretty cool."

"So I might tell you."

"I'm gonna tell you."

 

[alwayssts]

"I'm gonna tell you."

...........

(For what it's worth, I see what you did there...love ya for it....you're a clever bastard dude.)

 

[NC37]

No such problems with HD7950 WindForce 3X and you know how old it is. And it's running almost 24/7. Though I do have custom made fan profile so it never spins like mad, making it nicer on the bearings I guess. I do keep it dust free and when cleaning with compressor, I block the blades from spinning.

Could be. Mine do hit 100% all the time. Just no way around it here. But they are pretty cheap fans. When they popped off I got to see that first hand.

I took some old PSU or CPU fans and spliced them on. Cut them out of their frames first, of course. Used gorilla glue to attach to original mounting bracket after I cut the middle spindle out. They spin slower but sit taller and its surprising how much better they cool. My 2nd card is a solid 10-20 deg C cooler. So these GPUs can run much cooler, its just the fans which suck.

 

[RejZoR]

100% ?! That can't be right. I even tried once running it with forced 30% fan speed. Some newer games caused some problems regarding temperature, but for the most part, games worked just fine at temperatures slightly below 90°C. For an absolutely silent experience it's not bad actually. Now I'm using my own custom fan curve that never goes beyond 55% (I think) fan speed. It's slightly audible, but if there is music or even quiet environmental effects, I don't really hear it anymore.

 

[Eric_Cartman]

The R9 390 series is going to overheat so bad with air coolers.

But that won't stop AMD from lying about the clock speeds and using the boost clock as the advertised clock speed.