AMD Radeon Fury X Series | HBM, Small Form Factor And Water Cooling | June 16th

But it also improves your desktop experience immensely for the rest of your life
I know where this road leads. Come 2016 new games with higher demands, you have to upgrade at a fiercer cadence to maintain that resolution.

I'll stay contented with my 1080p TV and couch for now.

Also 1080p 60fps ultra isn't guaranteed anyways with higher-end cards, much less 1440p and above.

Vrsqybl.jpg
 
AMD has stated that it runs at 50c under full load. They stated the board comes with 6 phase power for 400 amps. It runs at 275w but comes with two 8-pin power connections and the cooler is rater at dissipating 500w. So yeah, I do think it's going to be a killer overclocker.

50c is also probably a very conservative estimate. Gibbo at OCUK said he has not seen his Fury X go over 40c at load.
 
So the water cooling still has a fan that you need to attach somewhere, right?

Yeah, you need to mount it on the side of your case or somewhere.

I figured I waited this long to find out for sure if it has HDMI 2.0, I'll wait for someone to actually test one to see. No harm in doing that. 980Tis aren't going anywhere.
 
As long as the case has a 120mm air vent it should be fine. The 120 mm Nidec fan they use with the cooler is excellent, one of the best fans (not) available. Of course if it ramps up high it's not going to be silent, but according to AMD's slides it should be really quiet even at load.
 
I think some previews said the decibels under load for the water cooled fury is very low. I guess we'll see next week.

AMD has been pushing the OC benefits of HBM and fiji so I do think it will be a good OC architecture.

BTW are you 4K TV people using it on a desk or in the living room? I don't remember seeing anything below 40 inches which seems way to huge to use in a desktop setting.
 
I am using a 48" 4k curved display on my desktop. At first it seemed huge but for gaming I grew into it. For desktop apps I just don't use the whole screen.
 
Do you guys really think that Fury X will be able to overclock 15-30%?

I have seen some people say this, but it seems more like wishful thinking for several reasons. First, I think Fury X is already running at its maximum, and with its 4096 SP's, it runs really hot hence the watercooling. Likewise, it's the reason why we will only see watercooled Fury X and not a Fury X on air (there is of course Fury Pro which will run on air, but it's a cut down). It's also AMD's biggest chip, and it is very power hungry, and I think that increasing the voltage would make it throttle. Sure it runs cool but looking at Hawaii chips, they're not really known to be overclockers.

Hope I'm wrong, but I really don't think Fury X will be able to overclock as well as Nvidia's Maxwell's chips.
I think the watercooler is there not because it just runs so hot you'll absolutely need it, but because it provides an out-of-box $650 card that will outperform the TitanX/980Ti if not at stock, then definitely with an overclock. Basically, its their way to ensure that they've got the best performing card at the best price for the long wait to the next generation of cards.
 
Do you guys really think that Fury X will be able to overclock 15-30%?

I have seen some people say this, but it seems more like wishful thinking for several reasons. First, I think Fury X is already running at its maximum, and with its 4096 SP's, it runs really hot hence the watercooling. Likewise, it's the reason why we will only see watercooled Fury X and not a Fury X on air (there is of course Fury Pro which will run on air, but it's a cut down). It's also AMD's biggest chip, and it is very power hungry, and I think that increasing the voltage would make it throttle. Sure it runs cool but looking at Hawaii chips, they're not really known to be overclockers.

Hope I'm wrong, but I really don't think Fury X will be able to overclock as well as Nvidia's Maxwell's chips.

Obviously a better cooler helps to keep the temperature down for OCing, but it also depends on the chip and architecture. There's a point where upping the voltage more won't net any gain in clock speed because the thing will become unstable. From what I've read it seems GCN has a shorter pipeline compared to Kepler/Maxwell, which is a reason why it's not easy to reach something like 1.5GHz. It's kinda similar to CPUs with their IPC vs frequency trade-off.

This thread on B3D
discusses this topic a bit more if you're interested.
 
I think some previews said the decibels under load for the water cooled fury is very low. I guess we'll see next week.

AMD has been pushing the OC benefits of HBM and fiji so I do think it will be a good OC architecture.

BTW are you 4K TV people using it on a desk or in the living room? I don't remember seeing anything below 40 inches which seems way to huge to use in a desktop setting.

That's in direct contradiction to what the engineers have said, verbatim: "You'll be able to overclock this like there's no tomorrow" and "this card is an overclockers dream".

So I fully expect you'll be able to achieve at least 20% overclocks on this card, perhaps even in excess of 25%.
 
It has been said you cannot overclock the memory though. Granted 512gb/s is plenty, like edram plenty, it is still a bummer.
 
I don't kow if this has been posted before but the Fury Nano packs in some serious punch according to WCCFTech :

[...]2X the performance per watt and 2X the performance density of AMD’s previous flagship, the R9 290X. The result is a six inch long, 175W graphics card that we’re told is “significantly faster” than the 290X.

That's the key element for the calculation WCCFTech did further down in their article:

R9 290X’s peak FP32 = 5.6 TFLOPs, in other words 5600 GFLOPs, and its TDP is 250W.
Perf/W = 5600 GFLOPs/250W = 22.4 GFLOPs/W

We also know that the R9 Nano has 2X the perf/watt of the R9 290X.
Which means it’s 2X (5.6TFLOP/250W)
= 2X 22.4 GFLOPs/W
= 44.8 GFLOPs/W.
Thus the perf/watt rating of the R9 Nano is 44.8GFLOPs/W.

Incidentally we also have the TDP for the Nano, and that’s the last missing piece in the puzzle.

R9 Nano
Perf/watt = FP32 in GFLOPs (unknown) / TDP (175)
44.8 = FP32 (unknown) / 175
44.8 x 175 = FP32 (unknown)
44.8 x 175 = 7840 GFLOPs or 7.84 TFLOPs.

The article goes in stating that the real world performance should be around 7.3 Tflops instead of the calculated 7.84. The Titan's range was calculated as 6.14-6.6 Tflops. If true, it would be great news for ITX-enthusiasts like me who have been using a low profile GTX 750Ti so far. Pay attention though that they are just comparing FP32 Compute performances.

mDOhVZD.png


http://wccftech.com/fast-amd-radeon-r9-nano-find/
 
wouldn't that also mean nana > fury x?

because now it seems like the fury x is between the 980ti and the titan x.
The Fury X smokes both in FLOPS. Remains to see how gaming performance is for both. I do think AMD mean same performance as R9 290X for Nano, actual power consumption for the 290X is probably close to 300W.
 
The Fury X smokes both in FLOPS. Remains to see how gaming performance is for both. I do think AMD mean same performance as R9 290X for Nano, actual power consumption for the 290X is probably close to 300W.

but amd flops were always worth less than nvidia flops.
 
I don't kow if this has been posted before but the Fury Nano packs in some serious punch according to WCCFTech :



That's the key element for the calculation WCCFTech did further down in their article:



The article goes in stating that the real world performance should be around 7.3 Tflops instead of the calculated 7.84. The Titan's range was calculated as 6.14-6.6 Tflops. If true, it would be great news for ITX-enthusiasts like me who have been using a low profile GTX 750Ti so far. Pay attention though that they are just comparing FP32 Compute performances.

mDOhVZD.png


http://wccftech.com/fast-amd-radeon-r9-nano-find/

Wasn't the Nano the lowest end Fury card announced so far though? Seems like they could easily dethrone the 750Ti as the best bus powered card if they just went lower. After announcing so much why hide something like that? Did they overlook this? Does fury not scale so well at lower wattage? It's baffling.
 
I'm sticking to 1080p [32" HDTV], but I want awesome GPU because of VR [and long term future proofing of desktop gaming and eventual switch to 4K].

So far Nano sounds great, but price will determine much of its appeal. I will still most likely wait for 14nm+HBM2.
 
Wasn't the Nano the lowest end Fury card announced so far though? Seems like they could easily dethrone the 750Ti as the best bus powered card if they just went lower. After announcing so much why hide something like that? Did they overlook this? Does fury not scale so well at lower wattage? It's baffling.

You mean reducing Fiji's power consumption so much that it can compete with low end cards? It's a ~600mm² chip that uses a new memory standard, this thing is certainly not cheap to make. Doing something like that would mean financial suicide. Not to mention it'd start cannibalising AMD's other mid/low end cards.
 
I know where this road leads. Come 2016 new games with higher demands, you have to upgrade at a fiercer cadence to maintain that resolution.

I'll stay contented with my 1080p TV and couch for now.

Also 1080p 60fps ultra isn't guaranteed anyways with higher-end cards, much less 1440p and above.

Vrsqybl.jpg

That chart is off. I got a 290 and my average frame isn't 48. I'm between 55-60 on average with dips.
 
The Fury X smokes both in FLOPS. Remains to see how gaming performance is for both. I do think AMD mean same performance as R9 290X for Nano, actual power consumption for the 290X is probably close to 300W.

Yeah their flops estimations (and bear in mind they were correct with their Fury X TFLOPS prediction) is as follows:

Fury X: 8.6 TFLOPS
Nano: 7.84 TFLOPs

Titan X: 6.14-6.6 TFLOPS
290X: 5.6 TFLOPs
PS4: 1.84 TFLOPS
XB1: 1.23 TFLOPS


WCCFTECH said:
I should remind everyone again that FP32 compute doesn’t directly correlate with gaming performance. Especially when comparing different architectures such as AMD’s GCN and Nvidia’s Maxwell. FP32 reflects the capability of the chip’s compute engine and not necessarily the capability of the graphics pipeline.
So it should not be taken as a metric by which we can compare the gaming performance of different GPUs. Except when we’re talking about chips based on the same architecture. For example the R9 Nano, R9 Fury X and the console chips. In which case it can be a fairly accurate indicator for gaming performance.

Anyway, saw these on Overclock.net. Design of the card is the best they've made.

 
You mean reducing Fiji's power consumption so much that it can compete with low end cards? It's a ~600mm² chip that uses a new memory standard, this thing is certainly not cheap to make. Doing something like that would mean financial suicide. Not to mention it'd start cannibalising AMD's other mid/low end cards.

Sure, binning down that far would be insane, no argument here (you're right though, I did mean Fiji cards). But they've pretty much revealed their hand until what, Q3 2016? I guess I'm just disappointed that there isn't an answer to the 750Ti the year when I'm about due for an upgrade and I might end up getting a card I could've gotten a year ago. Also not likely to get an amazing deal out of the wait either since AMD isn't really putting much pressure on Nvidia in that segment with the rebadges. Also would suck for them if their response next year gets trumped by Nvidia's next refresh but that's too early to tell and not my problem anyway.

Oh well, I'll see what holiday sales and rebates will get me or see if I can tough it out another year.
 
50c is also probably a very conservative estimate. Gibbo at OCUK said he has not seen his Fury X go over 40c at load.

WTF. That is crazy, even when you stick watercooling on CPUs the temp rises above 40, as the temp at idle typically is above 30c.

That has to be some kind of mistake.
 
WTF. That is crazy, even when you stick watercooling on CPUs the temp rises above 40, as the temp at idle typically is above 30c.

That has to be some kind of mistake.

Well I told you fools about this, because the chip is so massive, it's child's play to cool as there is more contact surface area to dissipate heat.
I'm joking :) but that seems to be really low temp indeed
 
Sure, binning down that far would be insane, no argument here (you're right though, I did mean Fiji cards). But they've pretty much revealed their hand until what, Q3 2016? I guess I'm just disappointed that there isn't an answer to the 750Ti the year when I'm about due for an upgrade and I might end up getting a card I could've gotten a year ago. Also not likely to get an amazing deal out of the wait either since AMD isn't really putting much pressure on Nvidia in that segment with the rebadges. Also would suck for them if their response next year gets trumped by Nvidia's next refresh but that's too early to tell and not my problem anyway.

Oh well, I'll see what holiday sales and rebates will get me or see if I can tough it out another year.

Yeah, I see what you mean and I sort of agree. It would've been nice to see a full new stack down to the low end.

I can also see why AMD didn't go for it. 28nm is over 3.5 years old and in about a year or so we'll (hopefully) be moving onto 14/16nm. Designing an entire new line-up this late on a node that soon will be replaced doesn't make much financial sense. Especially considering AMD's R&D budget which they're also using for the development of Zen.

It seems like they just wanted to go for a halo product that gives them a bit of a marketing boost and a chance to blow some new life into Hawaii. It also gave them a chance to test the waters with HBM for the future. Power consumption aside, the 290s were already able to do relatively well against the 970/980 and with this "refresh" they're doing slightly better. For now they're betting a lot on Fiji and hoping to get by in 2015 so they can go all in next year. Assuming all things go well with the new node.
 
WTF. That is crazy, even when you stick watercooling on CPUs the temp rises above 40, as the temp at idle typically is above 30c.

That has to be some kind of mistake.

GPU dies are cooled much more efficiently. You're probably used to recent Intel quads that are crap at dissipating heat. Even with the new TIM they're miles behind GPUs partly because the die is so small.

Almost everybody with watercooling that covers CPU and GPU struggle with a much hotter CPU even though they're cooled by the same water (and the same water temp). If you have a watercooling loop over your CPU notice how "cold" the water is compared to the CPU. 60C should be hot enough to burn you.
 
WTF. That is crazy, even when you stick watercooling on CPUs the temp rises above 40, as the temp at idle typically is above 30c.

That has to be some kind of mistake.

Keep in mind that it's going to be very dependent on ambient temp. My idle temp varies a good 10c depending on what season it is and where I'm keeping the temperature in my house.
 
GPU dies are cooled much more efficiently. You're probably used to recent Intel quads that are crap at dissipating heat. Even with the new TIM they're miles behind GPUs partly because the die is so small.

Almost everybody with watercooling that covers CPU and GPU struggle with a much hotter CPU even though they're cooled by the same water (and the same water temp). If you have a watercooling loop over your CPU notice how "cold" the water is compared to the CPU. 60C should be hot enough to burn you.

Ya the new Intel's have just gotten worse and worse at overclocking. I'm sitting over here with my god-tier 2500k, that I was rocking 5.0GHz for 3+ years. I moved it down to 4.5 because I wanted to hopefully extend the life a little till we can get a decent CPU upgrade.

That being said, its been nearly 5 years with the CPU and going strong. At least the stagnation in GPU's isn't nearly the same.
 
GPU dies are cooled much more efficiently. You're probably used to recent Intel quads that are crap at dissipating heat. Even with the new TIM they're miles behind GPUs partly because the die is so small.

Almost everybody with watercooling that covers CPU and GPU struggle with a much hotter CPU even though they're cooled by the same water (and the same water temp). If you have a watercooling loop over your CPU notice how "cold" the water is compared to the CPU. 60C should be hot enough to burn you.

I'm watercooling a 290. It definitely is comparable to a cpu at idle. As soon as there is some load on the card the temps jump up to 50. At full load my card hits 75c.

Granted, I'm overclocking and I live in a warm region.
 
I'm watercooling a 290. It definitely is comparable to a cpu at idle. As soon as there is some load on the card the temps jump up to 50. At full load my card hits 75c.

Granted, I'm overclocking and I live in a warm region.

What is the watercooling setup like?
 
What is the watercooling setup like?

Essentially it is an AIO with a 120mm rad (arctic accelero hybrid II).

I also suspect I could fiddle around with the fan profiles to change that. Either way it is kind of hard to compare to the fury X given we don't know what kind of fan profile is used there.
 
Essentially it is an AIO with a 120mm rad (arctic accelero hybrid II).

I also suspect I could fiddle around with the fan profiles to change that. Either way it is kind of hard to compare to the fury X given we don't know what kind of fan profile is used there.
Interesting. They do advertise much lower temps for the 780, but the 290X can get really hot.
I expect reviews tomorrow

Embargo ends 1200UTC on the 24th.
 
This high end of a gpu is niche in itself. 144hz, gysnc, freesync, 1440p, and 4k monitors are all very niche in comparison in a 1080p monitor. If you only have a 1080p monitor you have no business buying this card.



can these cards run modern (2014) engines games at 1080p at over 144fps at ultra settings?
 
WOW, the fury is pretty small, I like that a lot. It's great too because the NANO is supposed to be even smaller and I've been hearing great things about it's performance.
 
So still no benchmarks yet?? If this card was soul crushing the 980ti and titan x and in performance in benchmarks and games, you would think that AMD would capitalize on it by releasing everything, thus potentially reclaiming market share or to at least make people more hesitant to by nvidia atm.

But since thats not the case it would appear that there hbm fury cards may perform the same as 980ti and titan X minus not having over 4gigs of vram.

There's my updated 0.02
 
can these cards run modern (2014) engines games at 1080p at over 144fps at ultra settings?

That's why I put 144hz in the post. 1080p high refresh rate is also pretty niche. I qualified post to mean 1080p 60hz. Most people with a 1080p 60hz monitor buying this card are planning on upgrading their display in short order.
 
So still no benchmarks yet?? If this card was soul crushing the 980ti and titan x and in performance in benchmarks and games, you would think that AMD would capitalize on it by releasing everything, thus potentially reclaiming market share or to at least make people more hesitant to by nvidia atm.

But since thats not the case it would appear that there hbm fury cards may perform the same as 980ti and titan X minus not having over 4gigs of vram.

There's my updated 0.02

Yeah this is also what i am thinking (but i really hope its a competing product so we can get better gpus faster and cheaper)
 
Top Bottom