Next Generation AMD Graphics (Volcanic Islands/HD9000) Finally Coming. (Official)

[SubTract]

I wonder if they'll release the frame pacing driver as a Beta before the official release of the 31st.

[TechSage]

7970 crossfire :P need to get that framerating driver though

[ajoy39]

I wonder if they'll release the frame pacing driver as a Beta before the official release of the 31st.

That drivers official release date is July 31st, so it will be out for a few months prior to these cards

[eshbop]

I wonder if they'll release the frame pacing driver as a Beta before the official release of the 31st.

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-AMD-Improves-CrossFire-Prototype-Driver/Battlefield-3

 

Check out that review, it's got prototype drivers for the 7990. It's from about 3 months ago but still's a pretty big improvement. Just to give you some idea.

[SubTract]

That drivers official release date is July 31st, so it will be out for a few months prior to these cards

I think you miss understood, I meant that I wonder if there will be a beta version of the driver out prior to the official release of the drivers on the 31st of July.

 

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-AMD-Improves-CrossFire-Prototype-Driver/Battlefield-3

 

Check out that review, it's got prototype drivers for the 7990. It's from about 3 months ago but still's a pretty big improvement. Just to give you some idea.

I've seen this, I have crossfire 6970's so I am kinda hoping they work well :D

[Muffinator]

Highlight from that article ^^^

 

"Specifications for the HD 9970 include 4096 stream processors, four geometry engines, 256 TMUs, 16 serial processing units, and 64 ROPs. Memory will be driven through a 512-bit memory bus, and GDDR5 memory."

[EmmaLong]

This news couldn't have come sooner.

[Bramm17]

Cant wait :D!! Me want so bad :D

[AdamThePole]

Well that is just lovely. I can't wait to go back to recommending 7950's to people again, haha. 

 

You should still be recommending single 7950's. Great card. 

[tabuburn]

The 9970 looks like is going to be an absolute beast, easily a titan killer.

http://www.tomshardware.com/news/AMD-Radeon-Volcanic-Islands-7790,22518.html#xtor=RSS-998

This article looks a little suspect IMO. 4096 Stream Processors is the same amount we get from the HD 7990 or 2 HD 7970's. The die has to be double the size of the existing HD 7970 to accomodate that many. Even with a process shrink, that's pretty hard to achieve without increasing the die size. The same thing would apply if the memory bus would be increased to 512-bit.

 

According to a more recent article done by xbitlabs, it more on the lines of

 

 

The Curacao XT graphics processor is expected to feature 2304 stream processors (36 compute units), 144 texture units, 48 render back ends and 384-bit memory controller.

http://www.xbitlabs.com/news/graphics/display/20130701195901_AMD_May_Unleash_Next_Generation_Radeon_HD_9000_Series_in_October.html

 

How this will translate to performance has yet to be seen but according to rumors I've read floating about, it should slot in between the GTX 780 and GTX Titan but at a lower price point than the GTX 780.

[Kuzma]

This article looks a little suspect IMO. 4096 Stream Processors is the same amount we get from the HD 7990 or 2 HD 7970's. The die has to be double the size of the existing HD 7970 to accomodate that many. Even with a process shrink, that's pretty hard to achieve without increasing the die size. The same thing would apply if the memory bus would be increased to 512-bit.

 

According to a more recent article done by xbitlabs, it more on the lines of

 

http://www.xbitlabs.com/news/graphics/display/20130701195901_AMD_May_Unleash_Next_Generation_Radeon_HD_9000_Series_in_October.html

 

How this will translate to performance has yet to be seen but according to rumors I've read floating about, it should slot in between the GTX 780 and GTX Titan but at a lower price point than the GTX 780.

If that's true nvidia need to either price slash or suffer.

[ERSCreeper]

I can't wait to sell me HD 7950 and get two 9950/9970's :D

[tabuburn]

If that's true nvidia need to either price slash or suffer.

I doubt they will. They've already got the head start and it has probably netted them a bit of cash. The most I can see is that since they came out first, they can afford to drop the price in time for the holidays at the end of the year.

[Kuzma]

I doubt they will. They've already got the head start and it has probably netted them a bit of cash. The most I can see is that since they came out first, they can afford to drop the price in time for the holidays at the end of the year.

Accord to that 9970 specification it will have bulldozer-esque modules (look at the serial processing modules) meaning that it will be like an APU in a gpu card ._. that would be a shit tonne of power and I feel like it'd wipe the floor with the titan.

[tabuburn]

Accord to that 9970 specification it will have bulldozer-esque modules (look at the serial processing modules) meaning that it will be like an APU in a gpu card ._. that would be a shit tonne of power and I feel like it'd wipe the floor with the titan.

If it really is Bulldozer-esque, it would be more on the lines of better GPGPU (computational) performance rather than just standard GPU (graphics) power. 

[Kuzma]

If it really is Bulldozer-esque, it would be more on the lines of better GPGPU (computational) performance rather than just standard GPU (graphics) power. 

Exactly, giving developers the option to use that extra computational performance could work out to far better than just standard gpu power

[tabuburn]

Exactly, giving developers the option to use that extra computational performance could work out to far better than just standard gpu power

I think it'd be something like NVIDIA's PhysX technology. Off-loading more of the computations to the GPU itself instead of the CPU. Not sure how this will impact performance though since that's adding an additional load to the GPU just like what PhysX does.

[Kuzma]

I think it'd be something like NVIDIA's PhysX technology. Off-loading more of the computations to the GPU itself instead of the CPU. Not sure how this will impact performance though since that's adding an additional load to the GPU just like what PhysX does.

It would be different to PhysX because they would be a lot more specialised for computing functions, it'd be almost like having a miniature computer o.o:

• Ram [x]
• Computing? [x]
• Graphics? [x]

It's only really missing storage and it could practically run by itself.

[tabuburn]

It would be different to PhysX because they would be a lot more specialised for computing functions, it'd be almost like having a miniature computer o.o:

• Ram [x]
•  
• Computing? [x]
•  
• Graphics? [x]
•  

It's only really missing storage and it could practically run by itself.

It'll still require a decent CPU to pace it. The computational part of it will be more on raw number crunching rather than actual processes that the system can use. Kind of like what Bitcoin and Folding requires.

 

AMD's strategy with their products is more on general usage scenarios (i.e. Jack of all Trades) instead of the highly segmented/specialized product stacks of Intel or NVIDIA. 

[Kuzma]

It'll still require a decent CPU to pace it. The computational part of it will be more on raw number crunching rather than actual processes that the system can use. Kind of like what Bitcoin and Folding requires.

 

AMD's strategy with their products is more on general usage scenarios (i.e. Jack of all Trades) instead of the highly segmented/specialized product stacks of Intel or NVIDIA. 

The decent CPU is obvious but I'm just saying that the computing could always be used for some extra graphical power, maybe physics, maybe something completely different. Also o.o I struggle to see how the 7990 is "Jack of all trades" .

[tabuburn]

The decent CPU is obvious but I'm just saying that the computing could always be used for some extra graphical power, maybe physics, maybe something completely different. Also o.o I struggle to see how the 7990 is "Jack of all trades" .

That's AMD's strategy with their highest tier cards since several generations back. Unlike NVIDIA, their solution to pack more power into a single card is putting 2 GPU's into them. The HD 7990 is kind of like a derivative of one of their workstation class cards (FirePro) that was also a dual GPU card. IMO it was a forced move to try and counter NVIDIA's GTX Titan in performance.

 

AMD cards are OpenCL based and are optimized with GPGPU usage and not just regular graphical performance. This is why a single HD 7970 is much more powerful than the Titan in OpenCL ang GPGPU applications such as Bitcoin mining, Folding, and possibly in the Adobe applications. NVIDIA does their cards differently, they have segmented them into different roles. GeForce series cards are optimized just for gaming (aside from the GTX Titan), Quattro series cards are optimized for professional applications such as 3D rendering, photo/video editing etc, and the Telsa series cards are optimized for computational use such as supercomputers. AMD only has 2 segments, the Radeon which is optimized for gaming but still has great performance in computational and rendering applications and FirePro which is strictly for professional applications.

[TechFan@ic]

AMD did double the stream processor count before, they did it on the 4870 to 5870 jump (55nm to 40nm), so it is doable.
Currently it's very hard to speculate because there are a lot of conflicting rumors but if the trend in the last decade is anything to go by, then expect a performance improvement of a full node jump, which is approximately 60%.

[tabuburn]

AMD did double the stream processor count before, they did it on the 4870 to 5870 jump (55nm to 40nm), so it is doable.

Currently it's very hard to speculate because there are a lot of conflicting rumors but if the trend in the last decade is anything to go by, then expect a performance improvement of a full node jump, which is approximately 60%.

Sounds about right. Looking at past trends, each generation change usually starts at around a 40% performance jump.

 

Current rumors suggests that AMD will still be using the 28nm process since their manufacturing partner won't being going down to 20nm until Q4 2013 or later so a doubling of stream processors will be unlikely.

[TechFan@ic]

Sounds about right. Looking at past trends, each generation change usually starts at around a 40% performance jump.

 

Current rumors suggests that AMD will still be using the 28nm process since their manufacturing partner won't being going down to 20nm until Q4 2013 or later so a doubling of stream processors will be unlikely.

Volcanic Islands is officially a 20nm architecture, & AMD already confirmed that the upcoming cards are Volcanic Islands based.

[Kuzma]

That's AMD's strategy with their highest tier cards since several generations back. Unlike NVIDIA, their solution to pack more power into a single card is putting 2 GPU's into them. The HD 7990 is kind of like a derivative of one of their workstation class cards (FirePro) that was also a dual GPU card. IMO it was a forced move to try and counter NVIDIA's GTX Titan in performance.

 

AMD cards are OpenCL based and are optimized with GPGPU usage and not just regular graphical performance. This is why a single HD 7970 is much more powerful than the Titan in OpenCL ang GPGPU applications such as Bitcoin mining, Folding, and possibly in the Adobe applications. NVIDIA does their cards differently, they have segmented them into different roles. GeForce series cards are optimized just for gaming (aside from the GTX Titan), Quattro series cards are optimized for professional applications such as 3D rendering, photo/video editing etc, and the Telsa series cards are optimized for computational use such as supercomputers. AMD only has 2 segments, the Radeon which is optimized for gaming but still has great performance in computational and rendering applications and FirePro which is strictly for professional applications.

Higher tier fair enough, thought you were talking about in general