AMD Trinity- Reduced Radeon Cores

[LiquidHard]

http://images.tweaktown.com/news/2/..._series_cpu_specifications_confirmed_full.png

The specs leaks on APU Trinity platform indicate a reduction on the number of Radeon Cores.

Comparing the A8 class:
Trinity A8 - 256 radeon cores
Llano A8 - 400 radeon cores

Even tho Trinity IGP are HD7xxxD vs Llano IGP HD6xxxD with architecture differences.

Arent more radeon cores supposed to be better for graphics?

 

[Bobnova]

Generally, but one 7xxx core is significantly faster than one 6xxx core.
Not that much faster, but they do clock significantly higher as well. I expect the Trinity IGP to stomp the Llano.

 

[LiquidHard]

Generally, but one 7xxx core is significantly faster than one 6xxx core.
Not that much faster, but they do clock significantly higher as well. I expect the Trinity IGP to stomp the Llano.

1) Bob, say in real terms, how many fps difference before its considered a noticeable improvement e.g. 8fps increase for Trinity over Llano.

2) Would lesser Radeon Cores reduce heat from the igp at least a couple of degress celsius?

 

[Bobnova]

Noticeable really depends on the game/test, resolution, detail/aa/af/etc. settings, section, monitor and person.
Compare, if you'd like, the performance difference between a 5750 and a 7770. The 5750 has more shaders. It gets stomped.

 

[]-[itman]

Llano actually uses a series 5xxx type gpu and Trinity uses a series 6xxx type gpu. They are both of the VLIW type. The difference is that Llano has a 4 simple plus one complex SP configuration where as the 6xxx series has a 4 simple SP configuration. The reason for this was that the one complex unit was hardly ever used in gaming situations and so AMD cut that unit out so that you could fit more "useful" shaders in the same sized gpu. This is also why typically a 5xxx series gpu outperforms a similar 6xxx gpu in compute situations as the complex unit will be used for compute whereas the 6xxx gpu has to use more than one simple shader to get the same result. Add to that a few other tweaks and the 6xxx series was a much more efficient gpu for gaming.

Long story short, for gaming, Trinity is much more efficient than Llano. With that in mind, you also have to realize that AMD changed the sku hiearchy where the top Llano carries an A8 moniker but a top Trinity carries an A10 moniker. So if you want to compare, you should use the A10 trinity. The A8's from each line should be roughly comparable but you really can't make a direct comparison as they are filling different spots in the product line.

edit: BTW, series 7xxx Radeon processors use a whole new type of architecture which AMD calls GCN. Search anandtech.com for it as there's too much info to put in a forum post on how it differs from VLIW.

 

[LiquidHard]

Llano actually uses a series 5xxx type gpu and Trinity uses a series 6xxx type gpu. They are both of the VLIW type. The difference is that Llano has a 4 simple plus one complex SP configuration where as the 6xxx series has a 4 simple SP configuration. The reason for this was that the one complex unit was hardly ever used in gaming situations and so AMD cut that unit out so that you could fit more "useful" shaders in the same sized gpu. This is also why typically a 5xxx series gpu outperforms a similar 6xxx gpu in compute situations as the complex unit will be used for compute whereas the 6xxx gpu has to use more than one simple shader to get the same result. Add to that a few other tweaks and the 6xxx series was a much more efficient gpu for gaming.

Long story short, for gaming, Trinity is much more efficient than Llano. With that in mind, you also have to realize that AMD changed the sku hiearchy where the top Llano carries an A8 moniker but a top Trinity carries an A10 moniker. So if you want to compare, you should use the A10 trinity. The A8's from each line should be roughly comparable but you really can't make a direct comparison as they are filling different spots in the product line.

edit: BTW, series 7xxx Radeon processors use a whole new type of architecture which AMD calls GCN. Search anandtech.com for it as there's too much info to put in a forum post on how it differs from VLIW.

That is an informative post, thank you. Very clear explanation.

Any idea why 1 complex SP configuration was implemented in Llano in the first place when it was seldom used?

 

[LiquidHard]

[QUOTE=']
edit: BTW, series 7xxx Radeon processors use a whole new type of architecture which AMD calls GCN. Search anandtech.com for it as there's too much info to put in a forum post on how it differs from VLIW.[/QUOTE]

http://www.anandtech.com/show/4455/amds-graphics-core-next-preview-amd-architects-for-compute/3

Says in the above link Trinity will not be GCN. But AMD has labelled these IGPs 7xxx series haven they...

 

[]-[itman]

http://www.anandtech.com/show/4455/amds-graphics-core-next-preview-amd-architects-for-compute/3

Says in the above link Trinity will not be GCN. But AMD has labelled these IGPs 7xxx series haven they...

It can be confusing. When I posted before I posted in fairly general terms but I'll clarify a little bit.

First, AMD can name their processors whatever they want and they sometimes do so in a very confusing manner for end users. For instance, I mentioned before that 6xxx line of gpu's are VLIW4, but really, I believe only the 6950/6970 are VLIW4, the 6870/6850 are tweaked versions of the 58xx gpu's. Furthermore, everything below that is basically just a 5xxx gpu re-named to 6xxx (e.g. 6770=5770). I think they added a very minor feature or two when updating the naming but nothing worth mentioning. So when I say 7xxx/6xxx/etc. I am basically referring to the latest technology during that generational stretch.

With that in mind, AMD has taken that one step further with their APU's and basically given the gpu of the apu a moniker indicating a generation ahead of what the gpu is compared to the discrete counterpart. Why did they do this? Marketing, plain and simple. So when you see that Trinity has 7xxx series radeon and Llano has 6xxx series, they actually are based on the 6xxx and 5xxx series architectures respectively. Perhaps someday they will catch up, most likely this is just how it is going to remain.

As far as the VLIW5 architecture in Llano, it's simple, they had the VLIW5 architecture ready when designing Llano, the VLIW4 wasn't ready yet so that's what they used. There could have been other considerations such as wanting to use Llano for compute tasks as well but the market hadn't really made that transition yet so they decided to push on with VLIW4 in Trinity until we get GCN in the apu's.

P.s. Aren't acronyms fun?

 

[LiquidHard]

It can be confusing. When I posted before I posted in fairly general terms but I'll clarify a little bit.

First, AMD can name their processors whatever they want and they sometimes do so in a very confusing manner for end users. For instance, I mentioned before that 6xxx line of gpu's are VLIW4, but really, I believe only the 6950/6970 are VLIW4, the 6870/6850 are tweaked versions of the 58xx gpu's. Furthermore, everything below that is basically just a 5xxx gpu re-named to 6xxx (e.g. 6770=5770).

Yup you hit it right on the spot.

Given the renaming process, I found the LLano apu igps were based on the Redwood series i.e. 56xx. I have a HD5670 512mb ddr5 here. I know this is not listed in apu dual graphics supported discreted card but why not give it a try given its based on Redwood too..

Tried to dual graphics the HD5670 with the igpHD6530 but it didnt work.

Is this an architecture issue or just AMD locking the function away in the drivers they provide?