AMD Radeon R9 295X2 8 GB

[birdie]

the real video you should watch instead of amd propaganda

Hey, you should have spoken a few words during filming this cause it's hard to compare to anything since there are no other audio sources we can relate to.

 

[darkangel0504]

 

[btarunr]

The background score in that video sounds like something out of Brazzers.

 

[W1zzard]

Hey, you should have spoken a few words during filming this cause it's hard to compare to anything since there are no other audio sources we can relate to.

I can't speak out of the video card (to have the proper noise levels you could calibrate on)

 

[TheBrainyOne]

With single GPU cards reaching 300W TDPs, I guess it's not unexpected that dual GPU cards are reaching 500W levels. Still, 2 pre-overclocked GTX 780 Ti cards would be a much better buy.

The card requires two 8-pin PCI-Express power connectors. Normally this configuration would be good for up to 375 W of power draw. AMD however has chosen to exceed the specifications, citing a minimum of 28A (=336 W) for each connector, which brings the total to around 750 W.

Is AMD crazy or what?! Was it that hard to put another 8 pin connector on there?!'

 

[Xzibit]

With single GPU cards reaching 300W TDPs, I guess it's not unexpected that dual GPU cards are reaching 500W levels. Still, 2 pre-overclocked GTX 780 Ti cards would be a much better buy.



Is AMD crazy or what?! Was it that hard to put another 8 pin connector on there?!'

They ran out of room

I believe its the same length of the 7990. They would have had to make it longer if they did

 

[TheBrainyOne]

I believe its the same length of the 7990. They would have had to make it longer if they did

Yeah. Because a guy purchasing a 500W 1500$ piece of hardware won't have a big enough case. Stupid AMD.

 

[Delta6326]

Great card! Sadly the whole card is ruined by coil wine!

 

[newtekie1]

the real video you should watch instead of amd propaganda

Look at how hot those 8-Pin connectors get, and I think we can even see the 24-pin in there and it is getting super hot too. I have a feeling this card is going to melt connectors.

 

[Mistral]

That thing is a beastly monster, and it serves to underline one thing: we needed freaking 20nm months ago...

 

[The Von Matrices]

You have to give AMD's engineers credit for this one. They crammed a 1024 bit memory bus and enough circuitry to provide 500W of power all on a normal height reasonably long PCB. One other site said the PCB is 14 layers, which is quite an achievement.

 

[15th Warlock]

I tip my hat to both AMD's and Asetek's engineering teams, they have accomplished what IMHO was impossible, taming two fully featured Hawaii XT cores using a single 120mms radiator, and designing a single board with 1024 lanes of DDR5 memory, they clearly had a no compromise design in mind, and they achieved that goal obviously.

Amazing performance, low noise levels and temperatures, what a big difference investing in a premium cooling solution has done for the beastly 290X core, still, IMHO, if you have the room for two cards it's a better deal than going for a dual GPU single card, but boy, what a fantastic piece of engineering this card is!

The writing is on the wall, this card is probably going to be as fast if not faster than Titan-Z, and at half the price again, see boys and girls? this is the reason why we need a healthy AMD to bring the heat to Nvidia and create a competitive environment, at $1500 I would personally not buy this card, but you can bet the bean counters at the green team are at full alert mode trying to figure how to compete with this card at this price point, 780X2? Who knows.

You have to give AMD's engineers credit for this one. They crammed a 1024 bit memory bus, and enough circuitry to provide 500W of power all on a normal height reasonably long PCB. One other site said the PCB is 14 layers, which is quite an achievement.

You posted this while I was writing my post, it's eerie how similar our conclusions on this card are, you know what they say about great minds....

 

[Xzibit]

The background score in that video sounds like something out of Brazzers.

TMI

Now we know where to find you when the news is delayed.

 

[the54thvoid]

In all honesty I can't be impressed. But that is down to personal preferences. If I wanted to go with a dual GPU solution I'd buy two standard 290x's and put blocks on them. I've not yet checked out the 290x crossfire comparisons but without water will these not be down clocking, making an apples to apples comparison difficult?

This is simply what the 6990 and 7990 was before, except this time they've had to slap a hybrid water solution on it. For this price a more genuine approach would be an EKWB dual full cover block with an integrated pump/rad combo. Given the current pricing on a single 290x, I don't think the price is 'unrealistic' (you want that, look to Titan Z) but I think the hybrid solution is a cheap one and the card is pretty ugly (not that it matters).

The power consumption is absolutely irrelevant in comparison as a 780ti sli solution is the same (accord to Guru3D). Performance wise, looks like AMD have clinched the single gpu crown back given that Titan Z results are yet to come out but looks likely NV will lose (on 780ti comparison).

If NV released a price comparable card (GTX 790, reduced compute, 6GB combined memory) then it would be a very good fight but as it is, in the dual GPU battle, $3000 Titan Z isn't even a competitor against a card half that price and better performing (arguably).

They're still both crap compared to a sensible single gpu/screen combo choice (he says stroking his Classified 780ti ).

And FTR the Guru3D benches at 4K (on the new 337 driver page) show the 4GB versus 3GB arguments look very tired.

 

[Casecutter]

not furmark, but something that mimics the peak load of the most demanding games

Yea, that's what I wanted to know nice to see not as a full-on stress test. Although, perhaps just as a "one-time" baseline it might be nice to see what it looks like in real "worst case" actually gaming in that same amount of time. Once we see that I/we can say "al-righty-then" that simulation appears to mimic what real world gaming is stressing.

And while such results are "glowing" being these are the first it's hard to say they have any true implication, we'll need to balance that against more cards/results.

As to this card... like all dual chip solutions they aren't my cup of tea... But for $1500 just knocking-off more or less an ARES II, it's not as not refined, evolutionary, or truly/adequately engineered. It almost feels like marketing drove this, and raised what we got… which always bad. Almost like someone in engineering said the PCB and layout was done and hanging around, but cooling needs "sophistication", which meant time and money. Marketing and bean counters said just get it checked off the list, quick and no more costs. Sadly I see it all too often… and it actually works against them.

 

[W1zzard]

Yea, that's what I wanted to know nice to see not as a full-on stress test. Although, perhaps just as a "one-time" baseline it might be nice to see what it looks like in real "worst case" actually gaming in that same amount of time

Drivers can detect Furmark and throttle the card, and people would cry "unrealistic" either way. I think the test I have is quite good and kinda represents worst case in realistic usage. Furmark would be just to show everything "omgz hot", which is not the point of this test. It would also affect the noise recordings, which are there to provide additional insight, because dBA numbers are not so easy to grasp

 

[ISI300]

Such a great review. Thanks, Wizz. Also thank god those cables didn't catch fire!

 

[mr2009]

why not run the middle tube water through the copper channel and back into the 2nd water pump?

 

[ISI300]

Good point, well made.

 

[TheBrainyOne]

why not run the middle tube water through the copper channel and back into the 2nd water pump?

Single 120*120*50 mm rad. There is only so much a single rad can do.

 

[W1zzard]

why not run the middle tube water through the copper channel and back into the 2nd water pump?

because they'd actually have to engineer something that you can't just buy from asetek. and I agree, that would be a much better solution

 

[Rahmat Sofyan]

Over at [H]ardocp - AMD Radeon R9 295X2 Video Card Review they also tested with 337 and noticed this.



Seams the new 337 drivers might be uping the thermal threshold of Nvidia cards to run faster.

so actually, with 337 drivers nvidia just overclocked their card silently?bs about the dx11 optimization?

 

[JBVertexx]

Do you know if there was a difference in temperatures between the two GPUs? On the thermal image, one looks hotter than the other. Given they are connected to the CLC loop in series, you would expect that. I'm just wondering which GPU the card temperature reading is measuring.

 

[W1zzard]

Do you know if there was a difference in temperatures between the two GPUs? On the thermal image, one looks hotter than the other. Given they are connected to the CLC loop in series, you would expect that. I'm just wondering which GPU the card temperature reading is measuring.

The temperatures are slightly different indeed, refer to page 28. The sensor readings in the video are from the primary (hotter) GPU

 

[Hilux SSRG]

Great review W1zzard. Overall I think Amd did a good job for an x2 card. Shame about the coil noise and lack of 3rd 8xpin. Amd botched it there.