AMD Fusion: How It Started, Where It’s Going, And What It Means

[Onus]

Very interesting article. Now we need to see some mainstream products coalesce out of this promising vapor-nebula. It can happen, but I'm sure I'm not the only one who has learned to not count chickens before they are hatched.

Show us.

 

[ashinms]

@army-ant7, all comes down to current gpu prices, the games you play and whether or not you're willing to overclock.[\quote]

...So how does that make AMD a master of none?

 

[billcat479]

Finally, a very good article.
It also gives enough information to understand why they designed their new cpu line the way they did if their main reason was to marry it to a GPU but put enough in it to work good enough as a stand alone product.
Good one, a very well done article. Thanks for a good read..

 

[Blandge]

What people need to understand is that AMD isn't even in the same ballpark as Intel in terms of size. They have a much more narrow scope of focus and competitor. Because of this, they are able to adapt and change more quickly than Intel. This is their primary advantage.

 

[beetlejuicegr]

AMD just bet it's entire company, the future of ATI (or what was the lovely discrete line at AMD), the future of their x86 platform, and their manufacturing business all on something that it wasn't sure it would even be around to see. They bet the farm on a dream.

Wrong! Well i will rephrase it as i see it from my point of view.

Amd lost the bet and now aims for a dream to come true earlier and save them all together.

One thing can save them, tablets! Seriously, Tablets are the best consumer item that could take their APU on a profit place, because, imagine a software developer keep writing his code and compiling on it...with an AMD APU. I am dead serious lol, sounds nuts BUT! think if they could make it,
an optimized OS for their APU on a tablet competing a PC on programming performance ...

(i wonder if that is even possible, but hey AMD dreams, i can dream too for them ^^ )

 

[rebel1280]

Why not just really push for the next gen game consoles and tablets? Low power, low heat, could embed a great graphics chip. I like Intel for my desktop but their IGP sucks, and it is most likely cheaper to manufacture a board with a single chip instead of graphics and CPU separately. And we all know XBOX could use better cooling!

 

[Blandge]

[citation][nom]rebel1280[/nom]Why not just really push for the next gen game consoles and tablets? Low power, low heat, could embed a great graphics chip. I like Intel for my desktop but their IGP sucks, and it is most likely cheaper to manufacture a board with a single chip instead of graphics and CPU separately. And we all know XBOX could use better cooling![/citation]

Consoles have razor thin margins. AMD won't make very much money, but it's still a worthwhile win for AMD. And Fusion APUs in their current form use far too much power to be successful in tablets. Maybe in a few years.

 

[eddieroolz]

I really appreciate this kind of article that highlights the amount of work done by engineers in the background. At first I didn't believe in AMD's purchase of ATi either, as I thought AMD was really starting to lose sight of itself.

In the end though, I must congratulate AMD for putting everything together and surviving that near downfall.

Meanwhile, I don't appreciate the comments pitting AMD cards against Nvidia. Take your bias and advertisements elsewhere.

 

[blazorthon]

[citation][nom]Blandge[/nom]Consoles have razor thin margins. AMD won't make very much money, but it's still a worthwhile win for AMD. And Fusion APUs in their current form use far too much power to be successful in tablets. Maybe in a few years.[/citation]

AMD is working on a low power/performance architecture (presumably to replace Brazos) based on Piledriver and it might be able to be used in tablets.

 

[megamanx00]

AMD's right, us programmers hate being told we need to change how we code something. I even hate it when I find out on my own that there's a better way to do something and I should be doing it . The faster a programmer can go from starting HSA to being productive with it will determine how often they use it. Once it requires minimal effort you'll see it used a lot more.

 

[tomfreak]

[citation][nom]blazorthon[/nom] Also, up to the 6670 still supports CF with Trinity. I totally agree with you in that they should change this, but the architecture doesn't need to match (obviously) for CF.They could have at least made the 7750 and 7770 compatible with it and I think that they should have, but oh well. Your point about letting it help very high end cards is a little poor in that the GPU simply isn't fast enough to help those cards more than a few percent. It probably won't be nearly a 10% improvement if it could and the CPU would probably hold everything back to a point where disabling the IGP and overclocking the CPU would probably have greater performance.[/citation]6670 is roughly the 25% of 7970 speed, IGP being half of that should be somewhere 12%, the inefficient crossfire scaling down to 5-10% gain is still worth something if it is done right, it may be worth more for 7800/7700 setups. But these definitely worth something for people to still run 5800/6800 series.

The crossfire support for the hybrid GPUs are poorly taken advantages compared to the traditional Twin GPUs. Almost every modern games are well supported on traditional crossfire. They could have at least make a similar discrete GPU to get better support on games, rather than crossfire across different architecture, 6670 is a HD5000 chip, it has poor tessellation performance, make note of that too. the problem here is the hybrid crossfire dont work half of the time.
it is not like they wont gain something else by introducing discrete VLIW4 GPU as they are generally more efficient, and all these power saving on VLIW4 are a big thing in mobile world, since most of the time GPU in mobile are a lot more inferior than their CPU.

A reliable crossfire with APU + GPU that work almost every game is a significant advantage over Intel's CPU + GPU combo. Trinity APU + VLIW4 GPU in crossfire is MUCH better gaming platform than Intel sandy bridge with VLIW4 GPU alone.


They got the asset, but the implementation is poorly executed.

 

[uglynerdman]

regardless of what it means, amd is in a bad position, this is not my opinion its just how it is. bad marketing and support for mobile or any good mobile parts, 2 companies are in a position to buy amd. all the engineers have been gutted and most of the r and d are gone. amd is not amd. bubye

 

[blazorthon]

[citation][nom]Tomfreak[/nom]6670 is roughly the 25% of 7970 speed, IGP being half of that should be somewhere 12%, the inefficient crossfire scaling down to 5-10% gain is still worth something if it is done right, it may be worth more for 7800/7700 setups. But these definitely worth something for people to still run 5800/6800 series. The crossfire support for the hybrid GPUs are poorly taken advantages compared to the traditional Twin GPUs. Almost every modern games are well supported on traditional crossfire. They could have at least make a similar discrete GPU to get better support on games, rather than crossfire across different architecture, 6670 is a HD5000 chip, it has poor tessellation performance, make note of that too. the problem here is the hybrid crossfire dont work half of the time. it is not like they wont gain something else by introducing discrete VLIW4 GPU as they are generally more efficient, and all these power saving on VLIW4 are a big thing in mobile world, since most of the time GPU in mobile are a lot more inferior than their CPU. A reliable crossfire with APU + GPU that work almost every game is a significant advantage over Intel's CPU + GPU combo. Trinity APU + VLIW4 GPU in crossfire is MUCH better gaming platform than Intel sandy bridge with VLIW4 GPU alone. They got the asset, but the implementation is poorly executed.[/citation]

Excellent points, I admit that, but again, any video cards such as Radeon 7800 and 7900 would probably be rather CPU-bottle-necked. Trinity is probably able to more or less keep up with Phenom II in gaming performance (maybe not the top Phenom II models, but at least with some of the upper models) and that simply isn't enough in some games. I also say that I would like to see AMD widen the CF support of their IGPs and low-end VLIW4 cards would make excellent replacements for Radeon 6600, 6500, and 6400. Just keep in mind that although Trinity's IGPs are pretty great, they are limited in CPU performance and would be more detrimental to gaming performance in at least some games using the IGP in conjunction with the high end discrete card than disabling the IGP for some serious overclocking.

 

[BoredErica]

[citation][nom]gondor[/nom]Why would it have anything to do with haters ? I use Intel CPU and AMD's graphics card. If I was to upgarde both components today, I'd go for Intel CPU and AMD graphics again but I'm watching AMD's APU efforts very closely and if AMD manages to come up with an APU that has the graphics performance of say HD5770 and CPU performance of a quad core Core2/Nehalem (Q9xxx series Yorkfield/Lynnfield) in a decent power envelope (65W would be awesome) it will be my next purchase instead. AMD appears to be right on the amrk for this kind of performance with Kaveri.I don't need more performance than that and I'd prefer to get rid of discrete GPU and its fan (which becomes noisy after a while). With single a item to cool (the APU itself) it's easier to get a decent (= quiet and efficient) cooling solution. The fact that HSA is going to offer great speedups with certain loads (data compression/decompression, media encoding etc.) is just an added bonus.[/citation]

Because anybody that dislikes AMD is getting shot down with like negative 50 downvotes on this site. That's why.

 

[blazorthon]

[citation][nom]dark_wizzie[/nom]Because anybody that dislikes AMD is getting shot down with like negative 50 downvotes on this site. That's why.[/citation]

The votes only go to positive or negative 20 and the only comments that are hidden from their low ratings are those that were wrong (and were addressed as to what was wrong with them) or are what might be a joke (see second comment of first page for example). Whenever a Nvidia article is published, people then complain about the same thing as you are except the companies are reversed regardless of the accuracy of the complaint. Both sides are full of fanboys.

 

[tomfreak]

[citation][nom]blazorthon[/nom]Excellent points, I admit that, but again, any video cards such as Radeon 7800 and 7900 would probably be rather CPU-bottle-necked. Trinity is probably able to more or less keep up with Phenom II in gaming performance (maybe not the top Phenom II models, but at least with some of the upper models) and that simply isn't enough in some games. I also say that I would like to see AMD widen the CF support of their IGPs and low-end VLIW4 cards would make excellent replacements for Radeon 6600, 6500, and 6400. Just keep in mind that although Trinity's IGPs are pretty great, they are limited in CPU performance and would be more detrimental to gaming performance in at least some games using the IGP in conjunction with the high end discrete card than disabling the IGP for some serious overclocking.[/citation]it is not bottleneck if people are running high resolution with full AAs, or may be multi-monitors gaming. What AMD should have done now is to build up a solid crossfire support first b4 the their newer CPU take advantage of the extra performance if 7970 is bottlenecked now. Think about this, by upgrading future AMD APU, I dont just upgrade my cpu speed, with good support of crossfire, I upgrade my Graphic performance as well. I can stay with 7970 and buy a newer AMD APU and get more fps performance than changing Intel CPU. With DDR4 coming IGP is going to get a lot of memory bandwidth using 1 channel per ram stick = 4 channels on most motherboard.

But the major point is still mobile GPUs, mainstream mobile GPU are bottlenecking the mobile platform on gaming. It is better to make a similar VLIW4 discrete GPU that work better with trinity APU using regular Crossfire

 

[blazorthon]

[citation][nom]Tomfreak[/nom]it is not bottleneck if people are running high resolution with full AAs, or may be multi-monitors gaming. What AMD should have done now is to build up a solid crossfire support first b4 the their newer CPU take advantage of the extra performance if 7970 is bottlenecked now. Think about this, by upgrading future AMD APU, I dont just upgrade my cpu speed, with good support of crossfire, I upgrade my Graphic performance as well. I can stay with 7970 and buy a newer AMD APU and get more fps performance than changing Intel CPU. With DDR4 coming IGP is going to get a lot of memory bandwidth using 1 channel per ram stick = 4 channels on most motherboard. But the major point is still mobile GPUs, mainstream mobile GPU are bottlenecking the mobile platform on gaming. It is better to make a similar VLIW4 discrete GPU that work better with trinity APU using regular Crossfire[/citation]

Increasing AA and such doesn't decrease the load on the CPU, it only increases the load on the graphics. Games such as BF3 MP with more than 32 players can still bring such a system to its knees simply because it needs more CPU performance.

 

[The_Trutherizer]

I think we should all appreciate the passion AMD has put into their goals.

 

[army_ant7]

Increasing AA and such doesn't decrease the load on the CPU, it only increases the load on the graphics. Games such as BF3 MP with more than 32 players can still bring such a system to its knees simply because it needs more CPU performance.

Yeah, I think that's what he was getting at blaz, that there some situations that could use the graphics boost and not really a CPU boost.

 

[Guest]

Why the hell would you want to Crossfire a Trinity APU with AMD's most expensive video card? The whole point of CF is to have the GPUs work in tandem, it makes no sense to shackle a $400 GPU to an $80 APU and expect them to link up and do great. It's not feasible anyway to take VLIW4 and GCN and force them to peacefully coexist on the same task. The engineering effort and money required would not come remotely close to paying itself off. Nobody with a brain is going to stick Trinity with a 79xx card - you'll use an FX-8350/i5/i7 for that.

 

[tntom]

Great article William Van Winkle.

I have for a while been wanting to understand how AMD felt about how Fusion has positioned them. and weather it carries unique IP that differentiates them from Intel and will allow them to compete with ARM. It only seem inevitable that Intel's fabrication process escalation is going to hit the wall and everyone will catch up to them. Intel's only advantage will be instructions per cycle on the CPU side only. But consumers will really be expecting a lot from smartphones by that time. Microsoft's overly simplistic Metro interface will only help Intel on the GPU side for one generation and then Intel will need a more advanced solution to compete with ARM and AMD's APU.

The question is, will Intel be ready with the the next advancement beyond silicon by then?

 

[blazorthon]

[citation][nom]RationalHatred[/nom]Why the hell would you want to Crossfire a Trinity APU with AMD's most expensive video card? The whole point of CF is to have the GPUs work in tandem, it makes no sense to shackle a $400 GPU to an $80 APU and expect them to link up and do great. It's not feasible anyway to take VLIW4 and GCN and force them to peacefully coexist on the same task. The engineering effort and money required would not come remotely close to paying itself off. Nobody with a brain is going to stick Trinity with a 79xx card - you'll use an FX-8350/i5/i7 for that.[/citation]

AMD managed to have their VLIW4 IGP in Trinity to Crossfire with VLIW5 with little fuss, so I fail to see your point in saying that AMD can't do it with GCN. Crossfire, unlike SLI, is software and has no trouble working with different architectures. Heck, Lucid could have AMD cards work in tandem with Nvidia cards and they're undoubtedly more different than two generations of AMD cards architecturally. Furthermore, if the A10s could provide a 10-15% boost (I don't think it would work out so well, but my opinion on that is irrelevant for this), then that's a 10-15% boost. I can't imagine anyone complaining about a 10-15% boost in graphics performance.

[citation][nom]tntom[/nom]Great article William Van Winkle.I have for a while been wanting to understand how AMD felt about how Fusion has positioned them. and weather it carries unique IP that differentiates them from Intel and will allow them to compete with ARM. It only seem inevitable that Intel's fabrication process escalation is going to hit the wall and everyone will catch up to them. Intel's only advantage will be instructions per cycle on the CPU side only. But consumers will really be expecting a lot from smartphones by that time. Microsoft's overly simplistic Metro interface will only help Intel on the GPU side for one generation and then Intel will need a more advanced solution to compete with ARM and AMD's APU.The question is, will Intel be ready with the the next advancement beyond silicon by then?[/citation]

If Intel continues improving their IGP performance by so much per generation, then they might surpass AMD in the next two or three generations if AMD doesn't get similarly great improvements (which is very possible). So far, Trinity is not a similarly great improvement over Llano, although it is a good improvement, especially considering that it kept the same process node.

 

[wh3resmycar]

biggest mistake from the acquisition was when they didnt absorbed the chip that would later be known as the adreno.

 

[fedelm]

Great Piece.

Congratulations!!!!

 

[hector2]

I'm pulling for AMD but, frankly, I think they're making a huge deal about Fusion because it's their only hope. They know they've lost the CPU-centric race and have the advantage in Graphics (at least for now). So a Graphics-centered approach gives them the best chance. They've come up with something that will keep them in the game.