Intel and Conroe are no match for what AMD has planned!

[thestain]

http://www.dailytech.com/article.aspx?newsid=2642

When you combine this tech news with the AM2 platform and double sockets, what AMD is doing is making a lot more sense! Who knows, maybe video cards and their heat and noise will be a thing of the past as will Ageia PhysX cards and.. it makes Intel's Conroe and Kentfield look like really old tech.

Added Edit:

Finally a glimpse at the technology behind AMD not being all that concerned about Conroe!

Yes, short-term concerned, but AMD had a long term answer in place and now has a way of bridging performance for the enthusiast to that longer term answer and keep Dell and Alienware happy while they are doing it. While only promising dual support for now for the very high end FX-62 and above, i would be very surprised if AMD did not have plans to bring this technology to the mainstream desk top market in the next year or so.

Wow!

Some may yawn, others Scoff, but this solution that AMD has come up with along with IBM, CRAY and others could really change the cpu/gpu/ppu landscape a lot more within the next year or two than the great performing stand alone Conroe. Imagine not buying Video Card boards or PhysX Cards anymore, but just dropping the processors into the sockets provided. No, I do not know how this will all work.. it is at a "rocket science" level after all, but.. it is neat to see what AMD has been up to and that AMD did not do some short term trick that many thought they might do, but simply said, that is nice Intel, you have a nice cpu there, but our new platform over-all will be much better than what Intel can offer, even with its adored little Conroe CPU.

My question is what does that mean to us, the DIY assembling enthusiast?

Will we be able to pick out the various cpus, gpus, ppus, accelerators and do with them as we please or will they be sold in packages that need to be loaded together?

Either way, AMD has not responded to Conroe with some sort of short term trick, but rather by setting the stage like the Rock used to, Conroe and Intel, "Can you smell what AMD has cooking?"

 

[kknd1967]

We know that already
AMD is building their new fab and R&D centers on Mars and importing alien technology to their new products. Intel does not stand a chance.

 

[classy]

If the dual socket was for all am2 dual cpus and just not FX ones I would say great move. But in its present form its just bs to be honest. They should have just said we don't have a damn thing to say until next year. Lets see two FX chips $2300, one dual socket board $300-500, being bent over with no lubrication believeing the high end gaming story, priceless.

 

[jlbenedict]

*yawn*

 

[Absolute0]

Ya the title is a little exaggerated huh, all i got from this article is eventually AMD will do some multiprocessor work with good communication and low latencies. No crap, that's what AMD 64 was built for from the ground up, superior multi CPU scalability & communication.

 

[thestain]

edited initial post to clarify why AMD's answer to Intel and Conroe and Core 2 Duo, merom, kentfield, etc.. was a good one in the long run and it really better explains the moves AMD has made recently.

In the Long Run, the impact 'Terrenza" has on AMD's cpus and platform, as well as on motherboards, gpus, ppus, etc.. will have a larger impact on the future of pc's, workstations, and servers, than a fast, great performing, dead end solution provided by Intel.

 

[Henny]

yea, and this is no match for Itanium.

LOL

 

[F1shF4t]

Interesting, would be quite nice for "cheap" (well compared to high end multiples socket) encoding pcs, like drop in a dually in one, and a co processor in the other which is speciolised for encoding and callculations. Will be usefull in the more proffecional market, ie encoding, rendering etc, i dont see how that will effect games unless u can drop a physics co-processor in there.

 

[jedisponge]

Originally posted by: thestain
edited initial post to clarify why AMD's answer to Intel and Conroe and Core 2 Duo, merom, kentfield, etc.. was a good one in the long run and it really better explains the moves AMD has made recently.

In the Long Run, the impact 'Terrenza" has on AMD's cpus and platform, as well as on motherboards, gpus, ppus, etc.. will have a larger impact on the future of pc's, workstations, and servers, than a fast, great performing, dead end solution provided by Intel.

You mean the idea of adding on a second, identical piece of hardware hasn't been done before? Amazing!

However if you really think about it, there's a point where price outweighs the performance gained, and in this case (from all the information released so far) the fact that this is currently only for fx chips make it way too expensive to justify purchasing. Nowadays, people are still thinking twice about buying sli and crossfire due to doubling the price (ie 100% increase) paid for video cards with a 30% or so gain in performance. Is the average Joe really going to be willing to spend an extra 1k for this? Keeping diminished returns in mind, will the added performance really be worth it?Not only that, extra support costs have to be paid for it as well (some might need to upgrade your power supply, and in certain situations a rise in your electricity bill), though I think the demography this technology is aimed at won't really mind those extra support costs. Again, will the average person really be able to do this?

I mean, a reason why Core Duo 2 is causing such excitement is due to its increase in performance yet (again, with the current information that we have) pricing within limits. Again, I'm not saying this technology is bad. In fact I think it's great. But lets think about it. The only people who are really going to be interested are people who are in the market for fx chips. This is a small demography in comparison to the rest of the market. Now, a smaller proportion of that demography are the only ones who are really going to be able to obtain these systems. So one has to ask, is it really worth it?

 

[NoStateofMind]

oh this has got Intel in a fury!!

















NOT! LOL

 

[SunnyD]

Best way to describe this would be... Opty in Socket 1, Insert GPU of choice in Socket 2... no more bus bottleneck, so to speak. All your 3D geometry processing is doing ... er, right on the CPU ... er ... I mean GPU.

Basically the article says it, but not in plain english. This will essentially be a very high speed version of PCIe or other bus technologies, except made from hypertransport. There will be a drawback though - you can only use this for the "processor" portion of whatever addin you want. IE: For a GPU, you'll have only the GPU, though it will have direct access to system memory at full speed and low latencies not being "bridged" from bus to bus. What you WON'T have is (unless it's on-die) a dedicated framebuffer, RAMDAC and/or TDMS, voltage regulators, and of course, the physical connection to the monitor (VGA/DVI).

About the only place other than a GPU that this would shine would be a PPU or some other fixed-function FPGA that has nothing to output to the outside world. Again, it would be perfect for a GPU aside from the major limitations above. It would work well for high definition audio processing, aside from the above.

Perhaps it would work well for media-stream decoding... hardware based MPEG1-4/h.264/Whatever else is coming out, along with Dolby/Surround processing... which this doesn't actually output to the real world but to the video card post-processing. For both encoding and decoding.

It has some interesting potential, but it's not the holy grail of computing. Like I said, it's basically really fast PCIe... for SOME devices.

 

[Henny]

LOL, if AMD's future plans are in response to Conroe then they'll be killed by Intel. We could say that "Conroe is no match for what Intel has planned".

 

[TanisHalfElven]

haha. amd is losing it thinking ppl will buy 2 fx chips when a 2.66 GHz mainstream conroe beats the living crap out if the fx-62

 

[irwincur]

This is actually more important in the server world where it has been speculated that AMD will release co-processors for any number of tasks. It is interesting however to see that they are looking at also incorporating this in the consumer level of equipment with GPU and PPU integration.

It is ideal for the HyperTransport design and well, it is much more forward thinking than Intel's current design. AMD's ability to be agile in development may provide them with massive rewards in the long run.

 

[DarkKnight69]

Originally posted by: thestain
edited initial post to clarify why AMD's answer to Intel and Conroe and Core 2 Duo, merom, kentfield, etc.. was a good one in the long run and it really better explains the moves AMD has made recently.

In the Long Run, the impact 'Terrenza" has on AMD's cpus and platform, as well as on motherboards, gpus, ppus, etc.. will have a larger impact on the future of pc's, workstations, and servers, than a fast, great performing, dead end solution provided by Intel.

You are nieve. Dead end solutions. Whether or not you care to say it, Intel developed a better chip (it appears) and I dont think will all the money they have they will be sitting on their asses for the time AMD is developing.

I am excited for quad bit processing...

 

[lifeblood]

AMD's vision and plans are all very well and nice, and in the long term are much more important to the company success, but dammit, what about K8L? How does AMD plan on making its core architect more productive than Intel?s core architecture? They keep throwing out tidbits like decreasing the L1 cache but adding an L3 cache, but I want more info. I know patience is a virtue, I guess I'm just not virtuous.

AMD did a serious end run around Intel with dual core and IA-64. With Torrenza AMD is again doing the same thing. In a way it is going back to the way things used to be (remember x87 math coprocessor?), but it opens new possibilities. Hopefully it will take off. The PPU is a perfect use.

 

[n19htmare]

How will this help when software makers still have not implemented the instructions need to take advantage of this so called supercomputer. Heck, only a few hand full of applications support dual core and now its quad core... what a waste of power and useless cores.

Not to mention that you'd be dropping on knees for TWO processors, premium Mobo, double the ram... and so on.

On top of that, people like us are really concerned about overclock.

2 FX 62 chips $2000 with possibly no choice of overclocking

1 X6800 $999 with more headroom than you need...

It was nothing but a deperate move on AMDs part.

 

[Viditor]

Originally posted by: n19htmare
How will this help when software makers still have not implemented the instructions need to take advantage of this so called supercomputer. Heck, only a few hand full of applications support dual core and now its quad core... what a waste of power and useless cores.

Not to mention that you'd be dropping on knees for TWO processors, premium Mobo, double the ram... and so on.

On top of that, people like us are really concerned about overclock.

2 FX 62 chips $2000 with possibly no choice of overclocking

1 X6800 $999 with more headroom than you need...

It was nothing but a deperate move on AMDs part.

Actually, there should be plenty of room for overclocking...the 4x4 is actually designed for it.
And the same benefits hold true for dual dual as they did for dual when there were NO apps that could utilize it...namely multitasking.

 

[lifeblood]

The 4x4 is a rather niche product. A very expensive niche product.

We should be grateful AMD is forcing the multicore switch. It was chicken and egg; Intel was unwilling to go dual core without supporting software, companies were unwilling to write dual core capable software without a customer demand. Same thing with 64 bit software. As a result we are starting to see payoff in the form of software that takes advantage of dual core. As time goes on we will see more and more payoff. Sometimes it takes someone willing to take a risk to move things forward.

 

[CreepieDeCrapper]

i think you have to visualize the next 3-5 years out to get a handle on what's happening... imagine 4-6 sockets for a variety of processors

1 or 2 dedicated to regular system tasks (like our current cpu's)
1-4 for handling graphics and gaming tasks... for instance, what about splitting up the graphics rendering tasks by having a separate processor for the various stages of frame rendering... maybe 2 for current frame rendering and 2 for the next frame, etc.
1 physics rendering processor
1-2 for encoding
the list goes on...

you could customize a system very nicely depending on what you were using it for by putting in the right combo of processors... kind of like a "mix and match" mentality to computing

very nice imo

however, overall pricing for multi-socket mobo's and processors would have to drastically decrease to bring this into the mainstream... hence thinking 3-5 years ahead

 

[mamisano]

Well, I think at this point with Quad-SLI type graphics systems the CPU and interconnect bandwidth are becoming bottlenecks. With the 4x4 you can have 4 GPUs, 4 CPU cores and increased bandwidth between both subsystems.

 

[robertk2012]

You can already buy a mobo with built in video and sound coprocessors. Pci-e still has twice the bandwith needed for one high end card...........so what is the big deal.....AMD is trying to reinvent the wheel.....not a good idea.

 

[Stoneburner]

AMD is either going off on some open source model for hardware development, or is pretending to until they come up with a new architecture.

 

[JEDIYoda]

Originally posted by: thestain
http://www.dailytech.com/article.aspx?newsid=2642

When you combine this tech news with the AM2 platform and double sockets, what AMD is doing is making a lot more sense! Who knows, maybe video cards and their heat and noise will be a thing of the past as will Ageia PhysX cards and.. it makes Intel's Conroe and Kentfield look like really old tech.

Added Edit:

Finally a glimpse at the technology behind AMD not being all that concerned about Conroe!

Yes, short-term concerned, but AMD had a long term answer in place and now has a way of bridging performance for the enthusiast to that longer term answer and keep Dell and Alienware happy while they are doing it. While only promising dual support for now for the very high end FX-62 and above, i would be very surprised if AMD did not have plans to bring this technology to the mainstream desk top market in the next year or so.

Wow!

Some may yawn, others Scoff, but this solution that AMD has come up with along with IBM, CRAY and others could really change the cpu/gpu/ppu landscape a lot more within the next year or two than the great performing stand alone Conroe. Imagine not buying Video Card boards or PhysX Cards anymore, but just dropping the processors into the sockets provided. No, I do not know how this will all work.. it is at a "rocket science" level after all, but.. it is neat to see what AMD has been up to and that AMD did not do some short term trick that many thought they might do, but simply said, that is nice Intel, you have a nice cpu there, but our new platform over-all will be much better than what Intel can offer, even with its adored little Conroe CPU.

My question is what does that mean to us, the DIY assembling enthusiast?

Will we be able to pick out the various cpus, gpus, ppus, accelerators and do with them as we please or will they be sold in packages that need to be loaded together?

Either way, AMD has not responded to Conroe with some sort of short term trick, but rather by setting the stage like the Rock used to, Conroe and Intel, "Can you smell what AMD has cooking?"

Gotta love that AMD all the way mentality!!

Way too funny!!

 

[F1shF4t]

For now it might not make sense to do this, but same as pci-e in the future it will find its use. How many people were whining about change to pci-e from agp.