Sign in with
Sign up | Sign in
Your question

AMD corrects Fusion delay rumor.

Last response: in CPUs
Share
January 19, 2007 10:47:03 PM

Over at Xbit an AMD spokesman was quoted as saying that Fusion is still set to sample at the end of 2008 and release in 2009.

Linkage!
January 19, 2007 11:10:39 PM

From the wording on that article it doesn't look like Fusion will dislodge the discrete GPU for a long time, at least for gamers.

Correct me if I'm wrong, but since Fusion is basically an ondie implementation of integrated graphics, does that mean it would use system memory? If that is the case, how is it ever going to come close to a discrete GPU which is generally one generation ahead (or even two! look at GDDR4) in memory technology?

Quote:
The company’s chief technologist believes that while standalone graphics cards – such as those developed by AMD’s graphics processor division ATI Technologies – have a lot of power, the vast majority of consumers hardly need them, but will certainly need 3D graphics going forward, particularly due to rather intensive usage of graphics capabilities by operating systems like Microsoft Windows Vista.


Oh please, any cheap $50 DX9 card can handle the 'rather intensive usage of graphics capabilities' in Vista Aero. :roll:

So just what exactly will Fusion excel in?
January 19, 2007 11:48:01 PM

Nice... this is probably AMD's answer to Intel's Integrate graphics line in business workstations.
Related resources
January 20, 2007 1:40:30 AM

Quote:
From the wording on that article it doesn't look like Fusion will dislodge the discrete GPU for a long time, at least for gamers.

Correct me if I'm wrong, but since Fusion is basically an ondie implementation of integrated graphics, does that mean it would use system memory? If that is the case, how is it ever going to come close to a discrete GPU which is generally one generation ahead (or even two! look at GDDR4) in memory technology?

The company’s chief technologist believes that while standalone graphics cards – such as those developed by AMD’s graphics processor division ATI Technologies – have a lot of power, the vast majority of consumers hardly need them, but will certainly need 3D graphics going forward, particularly due to rather intensive usage of graphics capabilities by operating systems like Microsoft Windows Vista.


Oh please, any cheap $50 DX9 card can handle the 'rather intensive usage of graphics capabilities' in Vista Aero. :roll:

So just what exactly will Fusion excel in?


That is the key to it. The first iterations will be IGP-based for power savings. They will use system RAM. The bigger issue for proliferation across markets will be modularity and a small enough process.

By the time this happens there will be DDR4 on the desktop so bandwidth shouldn't be an issue. These will still only make for mid-level desktops though since they would need to be dual (with one GPU) or quad (with two GPUs). It will more than likely be more successful for multisocket systems where the processing power can be divided amongst more modules.
January 20, 2007 11:11:49 PM

Quote:

So just what exactly will Fusion excel in?

HPC.
January 20, 2007 11:42:27 PM

Maybe they will come up with a special bus just for the GPU to access its own chunk of memory. Better yet, they should give it a dedicated bus AND some superfast special ram. Just for graphics.

Seriously, if this is what AMD bought ATi for - then they're bigger noobs than I thought.
January 21, 2007 12:28:55 AM

Quote:
Just wondering Baron, how come you are always into amd products when tests have proven conroes to be better, I have an amd myself and I am not saying that amd products are bad, and nor am I an intel fanboy, just kinda wondering


Because Intel are evil. :wink:
January 21, 2007 12:50:57 AM

It will dominate multimedia encoding and decoding and intense math applications and add a built-in FPU for when a discrete card is being used in gaming, but it wouldn't make a big difference in a lot of things. Of course this is on paper.
January 21, 2007 1:07:26 AM

Quote:
will if you think about it, its just like communism in the sense that it all works perfrectly in theory and on paper, but once you test it out, it's like 4x4, no actual performance gain, just better multitasking. I don't think that amd should really be introducing fusion until they have a real win against intel so they have some money to relax on. Torrenza seems like it will be a good seller for major businesses, but fusion seems kinda over the top to me


I must whole-heartedly disagree. It is nothing like 4x4, and if AMD can roll it out in '08 and do it well it would force Intel to counter. I think you're underestimating parallel processing power. Think about how much faster you can fold on a GPU than with a CPU.
January 21, 2007 1:50:03 AM

Quote:
oh no, I completely agree, its not going to be a major flop like 4x4, my point is that for the average user, it won't be worth the price difference, but for major gfx designs companies, fusion will offer un parraled (pun not intended) performance because of being able to connect the gpu and cpu better, but there are some things that I am kinda unsure about still. And one of those things is ram, I really don't like the layout of 4x4 so that it only has 2 dedicated ram slots per cpu, and if fusion is going to have a simular layout, you will have to buy extra memory just for the gpu, and that might be something that intel can get around to create a comback. But you're right again, as far as we know, intel has nothing in this field yet, and amd might get a chuck of the market share back


Fusion won't be feasible for a GPU replacement until DDR3. I don't see how Fusion will be more expensive than any other multi-core chip, though. The most important thing about Fusion is not even GPU replacement, I see that as nothing more as a bonus for low-power notebooks. I still contend that the performance boost that will accompany it is big, big, big for a number of areas. Think about servers with a built in parallel processing unit, supercomputers, etc.
January 21, 2007 1:51:45 AM

Quote:
Just wondering Baron, how come you are always into amd products when tests have proven conroes to be better, I have an amd myself and I am not saying that amd products are bad, and nor am I an intel fanboy, just kinda wondering



I post CPU news. There seems to be more AMD news accessible than Intel news.
January 21, 2007 2:02:49 AM

Quote:
will if you think about it, its just like communism in the sense that it all works perfrectly in theory and on paper, but once you test it out, it's like 4x4, no actual performance gain, just better multitasking. I don't think that amd should really be introducing fusion until they have a real win against intel so they have some money to relax on. Torrenza seems like it will be a good seller for major businesses, but fusion seems kinda over the top to me


Not to rain on your parade but it is known that E6700(2.66GHz) is faster than the QX6700 in many games (that aren't ready for quad core).

But back on topic.

Fusion's initial iteration is NOT designed for performance but power savings. AMDs GPU Fab partners are already at 65nm and at least 45nm and DDR4 is necessary for any "high-end" iterations(which will mostly be fp workstations and servers). They have already said that this WILL NOT ATTEMPT TO REPLACE PCIe GPUs.

It will allow AMD to provide mainstream performance in one package from CPU to GPU to chipset.
January 21, 2007 2:04:39 AM

Quote:
Just wondering Baron, how come you are always into amd products when tests have proven conroes to be better, I have an amd myself and I am not saying that amd products are bad, and nor am I an intel fanboy, just kinda wondering



I post CPU news. There seems to be more AMD news accessible than Intel news.

You first sentence is mostly false, your second sentence is mostly true.

So I make stupid posts about a "communication-error" you made? No, I know, I encourage other people to do it.

I knew I was screwed up.
:oops: 
January 21, 2007 2:13:04 AM

Quote:
Just wondering Baron, how come you are always into amd products when tests have proven conroes to be better, I have an amd myself and I am not saying that amd products are bad, and nor am I an intel fanboy, just kinda wondering



Maybe because you say AMD and then "Conroe" as if Intel doesn't exist without it so that could be it.

Can you ever stay on topic?
January 21, 2007 2:25:19 AM

I'm going to get some mojo going to get JumpingJack and Baron Matrix into the same elevator and then have the power go out for 12 hours. I think it'd make a good movie plot.
January 21, 2007 2:32:52 AM

Hollywood'd stretch it into a full 2 hours, maybe a series. Neither has lunch that day, there's some snakes, a car chase, some product placements - good for a mini-series possibly with a rewrite or two.
January 21, 2007 2:37:58 AM

I don´t see how a tv series almost exclusively consisting of scenes removed due to extreme violence and excessive swearing would please an audience?
January 21, 2007 2:41:53 AM

8O Protools. I'm jealous. I'm a Reason / Sonar / WaveLab guy. Someday I'll get Protools, dammit.
January 21, 2007 2:42:29 AM

LOL!
January 21, 2007 3:02:21 AM

Quote:
Fusion's initial iteration is NOT designed for performance but power savings. AMDs GPU Fab partners are already at 65nm and at least 45nm and DDR4 is necessary for any "high-end" iterations(which will mostly be fp workstations and servers). They have already said that this WILL NOT ATTEMPT TO REPLACE PCIe GPUs.

It will allow AMD to provide mainstream performance in one package from CPU to GPU to chipset.


So Fusion is not intended for the enthusiast then? How much power savings are we talking about over current integrated graphics? Most current IGPs use low end GPUs anyway and power consumption is already quite low.

In terms on integration between AMD/ATI Fusion is a good idea, but I don't see it catching on amongst enthusiasts.

Being a gamer and PC enthusiast in general, I never caught on to the Fusion hype, mainly because I couldn't foresee Fusion approaching the levels of a discrete GPU. By the time DDR4 is out, we'll be at GDDR5/6 or something... and games will be far more demanding in 2 - 3 years time than they are now.
January 21, 2007 3:16:52 AM

Fusion=MediaGX (without audio)
January 21, 2007 5:01:34 AM

performance and price, i really dont care ho winteresting something is unless it performs. if u wana hype it go ahead, ill wait for benchmarks thank you very much, i see how well 4x4 did compared to kentsfield, and hey even if the e6700, games arent even optimised for taking advantage of quad core cpus let alone dual socket, so wats ur point really
January 21, 2007 5:47:15 PM

Quote:
So Fusion is not intended for the enthusiast then? How much power savings are we talking about over current integrated graphics? Most current IGPs use low end GPUs anyway and power consumption is already quite low.

In terms on integration between AMD/ATI Fusion is a good idea, but I don't see it catching on amongst enthusiasts.

Being a gamer and PC enthusiast in general, I never caught on to the Fusion hype, mainly because I couldn't foresee Fusion approaching the levels of a discrete GPU. By the time DDR4 is out, we'll be at GDDR5/6 or something... and games will be far more demanding in 2 - 3 years time than they are now.


Cost-saving and lower power consumption is essential for HTPC / HPC / Labtops :wink:
!