Tuesday, April 26th 2016

AMD Radeon Pro Duo Performance Numbers Leaked

Ahead of its launch, performance numbers of AMD Radeon Pro Duo were leaked to the web by Expreview. Pitted against the $620 GeForce GTX 980 Ti, the $1,499 dual-GPU monstrosity is about 32 percent faster at 1080p (which sees CPU saturation), and on average 59.4 percent faster at 4K Ultra HD. The card is noted to conjure up more than playable frame-rates for all the games Expreview tested at 4K. Frame-rates were as low as 46 fps, and as high as 61 fps, indicating that the Pro Duo is the go-to single-card solution for Ultra HD. Find the review in the link below.
Sources: Expreview, KitGuru
Add your own comment

69 Comments on AMD Radeon Pro Duo Performance Numbers Leaked

#26
xvi
Silas WoodruffWhat's stoping people from buying a second 980TI, come out cheaper and most likely with more performance than AMD's dual gpu offering.
AsRockThe true killer of this card is the over zealous price tag, it's some $400 over the marker. As why would you buy it when you could just get 2x980Ti's plus more goodies :p.
Simply because it is among the most performance you can get from a single, enthusiast card. The Pentium 4 Extreme Edition was absurdly expensive, but it didn't stop enthusiasts from buying it.
Posted on Reply
#27
AsRock
TPU addict
xviSimply because it is among the most performance you can get from a single, enthusiast card. The Pentium 4 Extreme Edition was absurdly expensive, but it didn't stop enthusiasts from buying it.
And here's me thinking that AMD wanted to actually make money.
Posted on Reply
#28
ensabrenoir
.............everything in this article is 100% totally believable.......


Posted on Reply
#29
Darksword
Performance is pretty lacking. It's only 35% faster than a STOCK 980 Ti. We all know how well the 980 Ti's can overlcock. Once you factor that in, the difference would only be about 10% faster - and only on games that support CrossFire.

Posted on Reply
#30
MrGenius
cdawallIf nvidia had a dual gpu card they would have compared that I'm sure they however don't.
What do you mean "if"? I'm pretty sure they do. Isn't it called a GeForce GTX TITAN Z?

Or do you mean they just don't have one worth comparing it to? If so how is that true? Just curious.
The Radeon Pro Duo currently has no competitive landscape. As we mentioned earlier, NVIDIA did not launch a dual-GPU graphics card based on its GM200 silicon, and as such, the price-cut $1,500 GeForce GTX TITAN Z is the closest competitor, though based on the older "Kepler" architecture.
www.techpowerup.com/reviews/AMD/Radeon_Pro_Duo_Preview/

AMD Radeon Pro Duo up to 51 Percent Faster than GeForce GTX TITAN Z
Posted on Reply
#31
Kanan
Tech Enthusiast & Gamer
This card also has 3x 8 Pin connectors, so it can pretty easily be overclocked to 2x 1000 or 1050 or even more MHz, so consider that the card runs with its breaks on, it can be a lot faster. As someone mentioned 2x Fury X has a higher score, expect this card to be as fast as 2x Fury X when overclocked or better said, run in unlimited power mode, because exactly this is whats keeping this card from going all out on 1000 MHz.

The 295X2 is different than this card, because it's not throttled in any way, thus the big difference in power consumption. I expect this card to be 80-90% faster than stock 980 Ti / Titan X, if run without throttling. SLI isn't scaling nearly as good as Crossfire, so it doesn't change anything on that matter. Maybe the reason why Nvidia isn't doing SLI/dual cards anymore.
Posted on Reply
#33
MrGenius
I'm so tempted to make a car analogy. But we know how well those go over. Let me put it another way. WTF does old have to do with it?

And on another note. If people don't quit misusing the word "throttle" I'm gonna to start slappin' some bitches!
Posted on Reply
#34
the54thvoid
Super Intoxicated Moderator
MrGeniusI'm so tempted to make a car analogy. But we know how well those go over. Let me put it another way. WTF does old have to do with it?

And on another note. If people don't quit misusing the word "throttle" I'm gonna to start slappin' some bitches!
After throttling his wife the man jumped in his car, opened the throttle and escaped the scene of the crime. This of course was only a stage in a PC game the man was playing but stopped when his fps dropped due to the card throttling at high temps.

While I'm at it, it's in stock at Scan:

www.scan.co.uk/products/8gb-sapphire-radeon-pro-duo-pcie-30-1000mhz-hbm-1000mhz-gpu-8192-streams-aio-liquid-cooled-vr-ready-
Posted on Reply
#35
Basard
And another thing... Why not run the fan at max speed while Furmarking? Then you could run the cores at full speed..... OH! I know, the Power consumption would near double. Smoke and mirrors.
Posted on Reply
#36
Slizzo
MrGeniusWhat do you mean "if"? I'm pretty sure they do. Isn't it called a GeForce GTX TITAN Z?

Or do you mean they just don't have one worth comparing it to? If so how is that true? Just curious.

www.techpowerup.com/reviews/AMD/Radeon_Pro_Duo_Preview/

AMD Radeon Pro Duo up to 51 Percent Faster than GeForce GTX TITAN Z
Well Mr. Genius, that's because as pointed out to you earlier, the Titan Z is old. It contains Kepler cores, not the Maxwell cores that the Titan X and GTX 980Ti contain.
Posted on Reply
#37
N3M3515
DarkswordPerformance is pretty lacking. It's only 35% faster than a STOCK 980 Ti.
At 1080p, that is...
...and on average 59.4 percent faster at 4K Ultra HD
I'm no expert but i guess whoever buys a $1500 graphic card is not going to play at 1080p...
Posted on Reply
#38
cdawall
where the hell are my stars
N3M3515At 1080p, that is...
...and on average 59.4 percent faster at 4K Ultra HD
I'm no expert but i guess whoever buys a $1500 graphic card is not going to play at 1080p...
Don't have to be an expert to notice people ignoring the obvious.
Posted on Reply
#39
MrGenius
I should not compare the performance of my 2016 supercar to another supercar made before 2016. Because the one made earlier is too old. And it should be a given that, because mine is newer, it should perform better. Even though they are both supercars and have far more commonalities than dissimilarities. This makes sense how?
However unlike the 290 series, which allowed a GPU temperature of up to 95C, AMD has clamped down on the R9 295X2’s GPUs, only allowing them to reach 75C before throttling.
www.anandtech.com/show/7930/the-amd-radeon-r9-295x2-review/3
Under FurMark we find our load temperatures unchanged. While FurMark undoubtedly puts a greater load on our 295X2, the card’s power throttle has kept the card from generating too much more in the way of heat, keeping temperatures in check and within the limits of the CLLC.
www.anandtech.com/show/7930/the-amd-radeon-r9-295x2-review/17

"The 295X2 is different than this card, because it's not throttled in any way".

Is that a fact? I was not aware of that.
Posted on Reply
#40
GhostRyder
MrGenius
I should not compare the performance of my 2016 supercar to another supercar made before 2016. Because the one made earlier is too old. And it should be a given that, because mine is newer, it should perform better. Even though they are both supercars and have far more commonalities than dissimilarities. This makes sense how?
www.anandtech.com/show/7930/the-amd-radeon-r9-295x2-review/3

www.anandtech.com/show/7930/the-amd-radeon-r9-295x2-review/17

"The 295X2 is different than this card, because it's not throttled in any way".

Is that a fact? I was not aware of that.
Because Furmark is an unrealistic GPU stress test tool and cards are designed to hold back from that. Many cards are even programed when detecting Furmark input to throttle so they don't damage the GPU because Furmark pushes them beyond the normal scope. They have limits, every card does otherwise many people would kill cards on a regular basis. Its just not as strict as cards like the R9 Nano have stock which is what I believe they are talking about above.

The other thing is its not fair to compare it to the Titan Z. Its an old architecture and unlike cars the support for the card (Kepler) has been reduced which can cause unfair scenarios against a card fresh off the block (Unlike a car). There is no dual GPU GTX 980ti or Titan X so it can only be compared to the 980ti (Or realistically, a pair of them).
N3M3515At 1080p, that is...
...and on average 59.4 percent faster at 4K Ultra HD
I'm no expert but i guess whoever buys a $1500 graphic card is not going to play at 1080p...
You would be surprised... I have run into people at events (LAN parties) as recent as 2 weeks ago who have 2 Titan X's and run on a 1080p monitor (And no, that's not their portable monitor!).
Posted on Reply
#41
Kanan
Tech Enthusiast & Gamer
MrGenius"The 295X2 is different than this card, because it's not throttled in any way".
Is that a fact? I was not aware of that.
What I meant with that was that the 295X2 runs always on 1018 MHz unless in unrealistic (heat) scenarios (which is even more than 290X's 1000 MHz, as to why it's called "295") and the Radeon Pro Duo does not always run on it's maximum speed (1000 MHz). It works more like 2x Nano's, but a tad better because it's water cooled and the lower temperatures help the card clock down less to keep the 375W TDP target.
GhostRyderBecause Furmark is an unrealistic GPU stress test tool and cards are designed to hold back from that. Many cards are even programed when detecting Furmark input to throttle so they don't damage the GPU because Furmark pushes them beyond the normal scope. They have limits, every card does otherwise many people would kill cards on a regular basis. Its just not as strict as cards like the R9 Nano have stock which is what I believe they are talking about above.

The other thing is its not fair to compare it to the Titan Z. Its an old architecture and unlike cars the support for the card (Kepler) has been reduced which can cause unfair scenarios against a card fresh off the block (Unlike a car). There is no dual GPU GTX 980ti or Titan X so it can only be compared to the 980ti (Or realistically, a pair of them).


You would be surprised... I have run into people at events (LAN parties) as recent as 2 weeks ago who have 2 Titan X's and run on a 1080p monitor (And no, that's not their portable monitor!).
Well they didn't compare the cards at AMD, but I think they are comparable, because the Titan Z's Kepler (GK110) architecture and the GCN 1.2 (or Gen 3) of Radeon Pro Duo aren't so much different, both architectures are relatively old and the Titan Z is still available, so... but needless to say, that the Titan Z stands no chance, because it runs with really low clocks. The 295X2 already demolished that card back when both cards were new.
Posted on Reply
#42
medi01
DarkswordIt's only 35% faster
There was a reason to mention CPU bound absolutely out of context 1080p resolution in TPU article.
I knew it, I KNEW IT!
AsRockThe true killer of this card is the over zealous price tag, it's some $400 over the marker. As why would you buy it when you could just get 2x980Ti's plus more goodies :p.
You make it sound as if dual Ti was faster at 4k than Furys. Nope, not really:
iyd.kr/753

Not to mention formidable power consumption of Pro Duo.

It is, what it is, "I have too much money, what's the best I can buy" thing.
Posted on Reply
#43
las
DarkswordPerformance is pretty lacking. It's only 35% faster than a STOCK 980 Ti. We all know how well the 980 Ti's can overlcock. Once you factor that in, the difference would only be about 10% faster - and only on games that support CrossFire.

Exactly. Every custom 980 Ti pretty much hit ~1500/2000. At this speed, it's about 30-35% faster than a reference card.

A custom 980 Ti, out of the box, is ~20% faster than reference.

Tbh I don't see the point in buying a dual 28nm chip card now. Very soon we will see single GPU's beat it. I would def wait for vega or big pascal instead. Dual gpu cards have never sold good in the first place, I don't like them, often they overclock very bad too because of power delivery.

Also, would like to see 1440p results. 1440p at high Hz is fairly popular atm. And GM200 seems to do better in 1440p than Fiji. At 2160p they are more even. But I think 2160p/60fps is at least one generation away still, in demanding games (without lowering details). I know Fury X CF and 980 Ti SLI will go to it's knees in many games at this res, with maxed graphics..
Posted on Reply
#44
cdawall
where the hell are my stars
lasExactly. Every custom 980 Ti pretty much hit ~1500/2000. At this speed, it's about 30-35% faster than a reference card.

A custom 980 Ti, out of the box, is ~20% faster than reference.

Tbh I don't see the point in buying a dual 28nm chip card now. Very soon we will see single GPU's beat it. I would def wait for vega or big pascal instead. Dual gpu cards have never sold good in the first place, I don't like them, often they overclock very bad too because of power delivery.

Also, would like to see 1440p results. 1440p at high Hz is fairly popular atm. And GM200 seems to do better in 1440p than Fiji. At 2160p they are more even. But I think 2160p/60fps is at least one generation away still, in demanding games (without lowering details). I know Fury X CF and 980 Ti SLI will go to it's knees in many games at this res, with maxed graphics..
2 of almost any top end card will run 4k at good frame rates. Dual 390's do excellent as long as the age old CF requirements are met.
Posted on Reply
#45
las
cdawall2 of almost any top end card will run 4k at good frame rates. Dual 390's do excellent as long as the age old CF requirements are met.
Nah, it won't with decent IQ in my experience. I've tried both Sapphire Fury Tri-X CF and Gigabyte G1 980 Ti SLI at 2160p, and alot of demanding AAA games required settings to be lowered "alot" unless you accept fps dips into the 20-30s, and I won't. I don't see the fun in 2160p if lowered settings are required for smooth gameplay. Then I'd rather stay at 1440p with maxed details and much higher fps.

Even the 3-4 year old Crysis 3 will make these setups go to it's knees with decent IQ at 2160p.

Witcher 3 at 2160p with Zero AA and GameWorks:
media.gamersnexus.net/images/media/2015/amd/fury-x-w3-4k.png

Avg. fps seems "fine" for many people (on 980 Ti SLI / Fury CF), but the minimums are what matters for smooth gameplay, and they are bad across the board.

390 CF is not viable for 2160p unless you play less-demanding games only.
Posted on Reply
#46
cdawall
where the hell are my stars
lasNah, it won't with decent IQ in my experience. I've tried both Sapphire Fury Tri-X CF and Gigabyte G1 980 Ti SLI at 2160p, and alot of demanding AAA games required settings to be lowered "alot" unless you accept fps dips into the 20-30s, and I won't. I don't see the fun in 2160p if lowered settings are required for smooth gameplay. Then I'd rather stay at 1440p with maxed details and much higher fps.

Even the 3-4 year old Crysis 3 will make these setups go to it's knees with decent IQ at 2160p.

Witcher 3 at 2160p with Zero AA and GameWorks:
media.gamersnexus.net/images/media/2015/amd/fury-x-w3-4k.png

Avg. fps seems "fine" for many people (on 980 Ti SLI / Fury CF), but the minimums are what matters for smooth gameplay, and they are bad across the board.

390 CF is not viable for 2160p unless you play less-demanding games only.
I can run all high settings for most games at 4k with a pair of 290's. Leaving aa/af off and run the rest turned up. I have yet to see drops in the 20's.

That being said I don't play tw3 not my style of game so could care less what it scores.
Posted on Reply
#47
las
cdawallI can run all high settings for most games at 4k with a pair of 290's. Leaving aa/af off and run the rest turned up. I have yet to see drops in the 20's.

That being said I don't play tw3 not my style of game so could care less what it scores.
It was an example. Alot of games will not run good at 2160p with 290 CF. I guess you play older and/or less demanding games then, or sacrifice IQ. Like I've already mentioned, I've tried 980 Ti SLI and Fury CF at 2160p, both are much faster than 290 CF, and they had trouble at this res.

290 CF is on par with a single reference 980 Ti at 2160p.
295X2 is faster than 290 CF and this card is 10% faster than 980 Ti ref and 10% slower than custom 980 Ti at 2160p.
www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/23.html
Posted on Reply
#48
cdawall
where the hell are my stars
lasIt was an example. Alot of games will not run good at 2160p with 290 CF. I guess you play older and/or less demanding games then, or sacrifice IQ. Like I've already mentioned, I've tried 980 Ti SLI and Fury CF at 2160p, both are much faster than 290 CF, and they had trouble at this res.

290 CF is on par with a single reference 980 Ti at 2160p.
295X2 is faster than 290 CF and this card is 10% faster than 980 Ti ref and 10% slower than custom 980 Ti at 2160p.
www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_XtremeGaming/23.html
The 295x2 has not been tested with a new driver if you read the actual review you will notice how the only multi card setup in the whole thing is a pair of 970's. AMD has improved a good bit with drivers and games that support xfire are offering near 100% scaling at 4k. I play games such as fallout 4, crysis 3, carmageddon, the borderlands franchise, the metro franchise etc. All of them have been playable at 4k on a pair of 290s (well to be fair a 390 and 290 combo)
Posted on Reply
#49
las
cdawallThe 295x2 has not been tested with a new driver if you read the actual review you will notice how the only multi card setup in the whole thing is a pair of 970's. AMD has improved a good bit with drivers and games that support xfire are offering near 100% scaling at 4k. I play games such as fallout 4, crysis 3, carmageddon, the borderlands franchise, the metro franchise etc. All of them have been playable at 4k on a pair of 290s (well to be fair a 390 and 290 combo)
The 295X2 uses 16.1 WHQL Crimson, performance is pretty much the same as current. Many of the games you play are what i consider less demanding. But several of them will make your setup go to it's knees at 2160p if you don't lower IQ. Crysis maxed will for sure, and that game is over 3 years old now. Alot of games won't run on 390 CF at 2160p unless IQ is lowered.

By playable you mean 20-30 fps like consoles or?

Then I understand.
Posted on Reply
#50
Slizzo
lasThe 295X2 uses 16.1 WHQL Crimson, performance is pretty much the same as current. Many of the games you play are what i consider less demanding. But several of them will make your setup go to it's knees at 2160p if you don't lower IQ. Crysis maxed will for sure, and that game is over 3 years old now. Alot of games won't run on 390 CF at 2160p unless IQ is lowered.

By playable you mean 20-30 fps like consoles or?

Then I understand.
Metro is incredibly demanding, Fallout 4 is not light, but lighter than others. Carmageddon can be kind of crazy (it is on my now modest rig). Borderlands is for sure lighter.
Posted on Reply
Add your own comment
Nov 22nd, 2024 20:41 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts