http://tweakers.net/nieuws/47131/AMD:-Fusion-processor-mag-uit-meerdere-chips-bestaan.html
I came across this.
can anyone translate it?
I came across this.
can anyone translate it?
Some new 'Fusion' details- but in german
- Thread starter Techno+
- Start date
Im from germany, but i cant translate this. Why do u think this is german? this is dutch.
Heres something new about Fusion in german: http://www.k-hardware.de/news.php?s=K3a9b96c5ccefcc32d753a1c3f88b2079&news_id=6518
Fusion will be a MCM design.
Heres something new about Fusion in german: http://www.k-hardware.de/news.php?s=K3a9b96c5ccefcc32d753a1c3f88b2079&news_id=6518
Fusion will be a MCM design.
sorry, it is dutch u r right (had to google around for the origins of the site), it says MCM but also has some more details, anyone from holland want to volunteer.
So, after all that marketing talk at the day they (AMD and ATI) merged, it turns out it'll be basically the same concept as Intel's "Nehalem" on-die IGP.
Presumably, this implies in my eyes that they are considering trying to put a MCM together earlier than H1 2009, so that Intel couldn't "beat them to market" with a similar solution for Nehalem. If they actually cancelled or delayed the single-chip Fusion aimed at early 2009, however, I wouldn't know what to think...
u r right Arun. This is only a temporary solution to beat intel to such a product, it might even come out in the beginning of 2008 ( maybe who knows). However, they would be soon going for a single chip version, since adding gfx extensions to x86 would only make sense for an on die GPU.
The Flemish Belgians not good enough for you?
Plus all the other people who (can) speak Dutch but don't live in The Netherlands (North and South Holland are just provinces of The Netherlands)...Anyway, the tweakers.net site sais the same thing as the German site. AMD's Guiseppe Amato suggests Fusion could be two dies in one package.
Last edited by a moderator:
Maybe this jives with recent statements which suggested that in hindsight AMD regretted doing "native" dual- and quad-core for their CPUs (from a marketing perspective). Intel was able to beat them to market with a two-die MCM with the same number of cores. Most punters don't know the difference (or even care). Maybe AMD learned their lesson.
Most people who buy processors don't even know what they are buying. I still have friends who compare processors by clockspeed.
The Flemish Belgians not good enough for you?
I meant anyone who can speak dutch
well i speak dutch (first time ive read dutch in about a year though), theres nothing really interesting in the piece (though cpus/gpus arent my field) apart from this made me giggleI meant anyone who can speak dutch
"amd now realize that a good product isnt automatically guaranteed sales"
aint that the truth, marketing is everything
oeLangOetan
Newcomer
Fusion's main targets are the smartphone, pda and umpc markets... ? Wasn't fusion supposed to bring high performance graphics integrated into the cpu.
No. That's just what some (well, tons of) uninformed dreamers were assuming. There's very little to be gained by integrating a CPU and GPU.
Sound_Card
Regular
According to AMD, this is the most logical step in the processor and they certainly see fit to do so.
Last edited by a moderator:
Killer-Kris
Regular
It's only time to do so because they need something to sell the consumer with now that the clockspeed race has died. Intel is trying to turn it into a multi-core race, while AMD is likely to try and turn it into a feature race.
So once again I would not expect high-end GPU integration but rather IGPs moving ondie. In the end this likely does mean the consumer gets a better/faster part (than non-integrated IGP) for less money, it's just not the sort of segment most people at B3D are interested in.
Killer-Kris
Regular
Oh and I think that perhaps their slide would have been a little more clear on the progression towards GPU integration if the second from last bubble was IMC rather than 64-bit extensions.
According to AMD, this is the most logical step in the processor and they certainly see fit to do so.
Sound_Card
Regular
I think it is a little silly to say that it is only time to do it is so they can try to sell the consumer something. Let's look at AMD beyond just being a busniess. I don't look at their 5.4 billion dollar investment as just a way to cut into the chipset and GPU markets. Their investing some big money into their GPGPU and Torrenza initiative (which Fusion is part of). I have a strong feeling that K10 is the last true CPU arc. from AMD and everything afterwards will be based off fusion, and in my opinion, I don't see a single thing wrong with that.
However my stance was taken regarding Silent_guy's quote "There's very little to be gained by integrating a CPU and GPU." Not sure if he is refering to high end graphics in the CPU, or their is just flat out nothing to be gained at all from intergration.
I was referring to high-end graphics.
The german site says it's a multichip module, so you can eventually keep the CPU and be able to switch the GFX-chip. That doesn't sound like one package to me.
Alright
Alright, I'll translate...
AMD: 'Fusion Processor is allowed to be multiple chips'
AMD has major plans for the integration of CPU and VPU. With the Fusion Project the chip maker envisioned a multi functional hybrid chip, but according to a manager of the company it isn't necessary that all functionality is housed in a single die.
According to AMD's Giuseppe Amato it's probably best to launch the Fusion processor in the form of a Multi-Chip Module. MCM is a single package of different slices of silicon so from a customer viewpoint he's buying a single chip while there are two or more cooperating under the hood.
Competitor Intel already used this technique to beat AMD with the release of dual- and quadcore processors which internally consisted of two single- and two dualcore CPU's.
AMD used to object to this approach, an integrated design can be quicker, cooler and more efficient.
That principle point of view has led to intel currently being the only producer of X86-Quadcore products, we will probably have to wait until September to see AMD's Barcelona four-valver (straight translation )
AMD has obviously realised that a good product doesn't automatically translate to good sales. The company made it known earlier it regrets not copying intel's move of packing two chips as one.
That this approach is looked at for the Fusion program doesn't seem like the most stupid move. If the seperate components already exist, the design and production of a MCM would be a lot cheaper than a newly designed chip.
With Fusion processors AMD targets the markets for smartphones, PDA's and UMPC's where efficiency is of the utmost importance.
This sector is currently supplied bij energy efficient ARM and MIPS chips, with which the x86 platform currently can't compete on a "Performance per Watt" level.
If AMD decides to supply MCM's instead of more advanced silicon, the company takes the risk of being outperformed on the performance level.
According to Amato it revolves around a reasonable solution: if Fusion MCM's perform well they will still end up as single Chips.
Now, I'm a long time reader of Tweakers.net (user number 237 actually) but I haven't read anything from the writer of the article (René Wichers.)
But in the first and second paragraph there is a clear contradiction which seems like something I would write when I was a student, as I copied stuff from different websites for my reports.
First paragraph talks about how efficient a single chip is, next talks about how important a single chip is.. and then the conclusion that MCM is not a single chip.. and that this actually might be bad ... but he already gave examples of how intel beat AMD to every party with it's MCM solutions.
shrug...
Alright, I'll translate...
AMD: 'Fusion Processor is allowed to be multiple chips'
AMD has major plans for the integration of CPU and VPU. With the Fusion Project the chip maker envisioned a multi functional hybrid chip, but according to a manager of the company it isn't necessary that all functionality is housed in a single die.
According to AMD's Giuseppe Amato it's probably best to launch the Fusion processor in the form of a Multi-Chip Module. MCM is a single package of different slices of silicon so from a customer viewpoint he's buying a single chip while there are two or more cooperating under the hood.
Competitor Intel already used this technique to beat AMD with the release of dual- and quadcore processors which internally consisted of two single- and two dualcore CPU's.
AMD used to object to this approach, an integrated design can be quicker, cooler and more efficient.
That principle point of view has led to intel currently being the only producer of X86-Quadcore products, we will probably have to wait until September to see AMD's Barcelona four-valver (straight translation
)AMD has obviously realised that a good product doesn't automatically translate to good sales. The company made it known earlier it regrets not copying intel's move of packing two chips as one.
That this approach is looked at for the Fusion program doesn't seem like the most stupid move. If the seperate components already exist, the design and production of a MCM would be a lot cheaper than a newly designed chip.
With Fusion processors AMD targets the markets for smartphones, PDA's and UMPC's where efficiency is of the utmost importance.
This sector is currently supplied bij energy efficient ARM and MIPS chips, with which the x86 platform currently can't compete on a "Performance per Watt" level.
If AMD decides to supply MCM's instead of more advanced silicon, the company takes the risk of being outperformed on the performance level.
According to Amato it revolves around a reasonable solution: if Fusion MCM's perform well they will still end up as single Chips.
Now, I'm a long time reader of Tweakers.net (user number 237 actually) but I haven't read anything from the writer of the article (René Wichers.)
But in the first and second paragraph there is a clear contradiction which seems like something I would write when I was a student, as I copied stuff from different websites for my reports.
First paragraph talks about how efficient a single chip is, next talks about how important a single chip is.. and then the conclusion that MCM is not a single chip.. and that this actually might be bad ... but he already gave examples of how intel beat AMD to every party with it's MCM solutions.
shrug...
Last edited by a moderator:
Similar threads
