bye bye kentsfield

[confused1234]

http://www.pureoverclock.com/article37-4.html

hmmmmm octal-core power droooooool.................

lol im just a fellow AMD fanboy trying to raise his hopes>_<, but that would be kick ass if that could actually work!

 

[RichUK]

4x4 tech we already know about, and this is supposed to only work with the FX line which means a lot of money for this platform. Also most home user will not benefit from this technology, so as the article suggests the introduction of this technology will be just that, a niche.



However when K8L arrives that will be exciting, especially in 4x4 mode. Hopefully chip makers like nVidia and ATI might be able to license the rights to produce 4x4 chipsets, otherwise we will have to stick with AMD chipsets.

Its still a long wait for K8L though

I think I might miss Conroe and wait for the release of K8L, as I don?t fancy upgrading constantly, and quite fancy the idea of a single die quad core.

 

[DrMrLordX]

8 cores would provide a lot of theoretical computing power that would be difficult for end-users to fully utilize. Using 4 cores is difficult now, even for power users.

In the long run, the firm with the fastest individual core will win. Simply deploying more of them only helps to a point, and at least for now (and the near future), two cores seems to be "enough".

For this reason, I suspect that fast Conroe chips will the ones to get (as opposed to Kentsfield, 4x4 K8L, etc).

 

[Ika]

Originally posted by: DrMrLordX
8 cores would provide a lot of theoretical computing power that would be difficult for end-users to fully utilize. Using 4 cores is difficult now, even for power users.

In the long run, the firm with the fastest individual core will win. Simply deploying more of them only helps to a point, and at least for now (and the near future), two cores seems to be "enough".

For this reason, I suspect that fast Conroe chips will the ones to get (as opposed to Kentsfield, 4x4 K8L, etc).

you're looking at the future from our current perspective. Remember, though - a few years ago, dual-core wouldn't have helped at all. In several years, when multi-core is ubiquitous, developers will have written multithreaded apps for multi-core CPUs, meaning 4 cores CAN and WILL be utilized in the future.

 

[Keysplayr]

Whew!!! From the thread title, I thought Intel was dropping Kentsfield for a minute there.
No, no..... Kentsfield will be a sweet CPU. We have yet to see how this 4x4 thing plays out. By the looks of it, there is not much AMD can do ( they can do some things, but limited) until K8L. That should be very interesting.

 

[Regs]

It's software dependent. So that means software developers have to spend more money on development or education and training. Which means you won't see that any time soon.

 

[DrMrLordX]

Originally posted by: Aflac

you're looking at the future from our current perspective. Remember, though - a few years ago, dual-core wouldn't have helped at all. In several years, when multi-core is ubiquitous, developers will have written multithreaded apps for multi-core CPUs, meaning 4 cores CAN and WILL be utilized in the future.

That's not true. A few years ago, dual cores WOULD have helped the same people who were using dual-cpu systems. Multithreaded apps aren't a new development.

The problem is that the applications that can be multithreaded to the point that they can utilize 4 or more cores are few and far between, at least for power users and workstations users. Take a look at Duvie's Opteron 270 benchmark if you don't believe me.

http://forums.anandtech.com/messageview...atid=28&threadid=1830732&enterthread=y

The bottom line is that, outside of maybe some compiling tasks, even future multithreaded apps won't gain as much going from 2 cores to 4 cores as when going from one core to two cores. Going from 4 cores to 8 cores would be overkill.

Pegging two cores at 100% utilization off a single app is pretty easy nowadays with the right app, but 4 is hard to do, and that will continue to be the case unless AMD and Intel roll out some kind of speculative-threading/reverse HT tech like Mitosis.

 

[Fallengod]

Honestly. 2 cores is plenty imo. Come on now. I do pretty intensive stuff myself. Maya, Director, Photoshop, hard gaming, massive amount of mpeg2/dvd encoding with CCE etc... I do not think we have yet maxed out the ultilization of even 2 cores yet. The software out there still needs to progress some.

I can understand servers and such needing 4 cores but other than that... You think the average user needs 4 cores at this point in time? I dont.

 

[pcoffman]

Originally posted by: DrMrLordX
at least for now (and the near future), two cores seems to be "enough"

Maybe so. However, hardware as a rule precedes software. In order for the industry to start writing software that will take full advantage of quad-core and beyond, we need the hardware.

 

[pcoffman]

Originally posted by: DrMrLordX
even future multithreaded apps won't gain as much going from 2 cores to 4 cores as when going from one core to two cores. Going from 4 cores to 8 cores would be overkill

I hope you're wrong, and that software can be written that will take better advantage of multicore. The primary driver of performance in the future is going to be adding more cores and learning how to take advantage of those cores. It won't be clock speeds so much.

This is sort of what Bill Gates said at WinHEC:

"So what is the next big issue for our industry ... Well, it's multi-core ... Now, the software has always had the ability to run multiple threads in parallel, so we immediately get benefit from this capability. However, optimizing the software, particularly as you move up into large numbers of cores, as you move to 4, 8, 16, 32, that requires architectural changes ... the raw cost performance that we're going to see changing over the next four or five years won't be the rate of the past, it will only be a doubling, or at most a tripling it appears. Therefore, a lot of the performance gains will come from activating those additional cores".

 

[Dribble]

If the hardware exists the software will follow. tbh the big step was to go multithreaded at all (to make use of dual cores). Once you've got the hang of that then increasing the threading to make use of quad or octal cores for that matter isn't such a big deal.

/me runs up word 2007 beta, loads a document and see's it's using 20 threads

 

[TanisHalfElven]

4x4 sucks. i mean that. only including the expensive fx line in it means to use it you spend $2000+. there's really no point cuz opteron 2XX systems costing much less can be built.

now kentsfield is awesome. and if intel prices it right (hopefully no more than 2xsimilarly clocked conroe) then yay intel.
amd needs a attitude adjustment.

does anyone think the phrase "the empire strikes back" fite the current intel-amd situation

 

[Viditor]

Originally posted by: tanishalfelven
4x4 sucks. i mean that. only including the expensive fx line in it means to use it you spend $2000+. there's really no point cuz opteron 2XX systems costing much less can be built.

now kentsfield is awesome. and if intel prices it right (hopefully no more than 2xsimilarly clocked conroe) then yay intel.
amd needs a attitude adjustment.

does anyone think the phrase "the empire strikes back" fite the current intel-amd situation

To each their own...I'm thinking that a quadcore FX-xx processor and a PhysX coprocessor in the second CPU slot (with quad SLI as well) might be fun...

 

[DrMrLordX]

Originally posted by: pcoffman
I hope you're wrong, and that software can be written that will take better advantage of multicore. The primary driver of performance in the future is going to be adding more cores and learning how to take advantage of those cores. It won't be clock speeds so much.

I hope I'm wrong too, but I don't think that I am. Multi-processor (and therefore, multi-core) systems have been available for x86 workstation and server PCs for years, dating back at least to the PPro days. 3d rendering apps, Photoshop, CAD apps, compilers, and other apps have been aggressively multithreaded for years as well. Dual-core CPUs for the desktop are breaking ground that professional-grade computer platforms broke years ago, and believe me, there are serious problems involved with trying to get a single app intended for a single user to utilize many CPUs (or cores) simultaneously. You can get compilers to do it, and you can spawn multiple applications, but multithreading has certain limitations that can not be easily overcome.

It is possible that the proliferation of consumer-level multicore CPUs and the even-increasing need to squeeze performance out of new hardware (in order to sell said hardware) will finally spur additional progress in multithreading, but do keep in mind that coders and compiler monkies have been working for years to take advantage of dual and quad CPU platforms via multithreading. What we see now in highly-optimized professional apps should give us an idea of how hard it is to utilize 4 or more cores in a single app, and from what I've seen, pegging 4 cores at 100% CPU usage with a single multithreaded app is very difficult. I think you can do it with a compiler and maybe some 3d rendering and/or CAD apps.

Furthermore, everyone's going multicore, so the idea that the future lies in our ability to use multiple cores successfully is only partially true. He who has the best individual core wins. Both Intel and AMD will have quad core systems available for consumers within the next year or so. Whoever can populate these CPUs with superior cores and superior interconnects will win. It's still about clock speed and ipc.

 

[pcoffman]

Originally posted by: DrMrLordX

Originally posted by: pcoffman
I hope you're wrong, and that software can be written that will take better advantage of multicore. The primary driver of performance in the future is going to be adding more cores and learning how to take advantage of those cores. It won't be clock speeds so much.

I hope I'm wrong too, but I don't think that I am

I take the opposite view. I think the next 10 years are going to be very interesting indeed. And I think the cores shall keep proliferating.

there are serious problems involved with trying to get a single app intended for a single user to utilize many CPUs (or cores) simultaneously ... multithreading has certain limitations that can not be easily overcome ... What we see now in highly-optimized professional apps should give us an idea of how hard it is to utilize 4 or more cores in a single app

No one said it would be easy. There's no doubt that software and hardware engineers have big challenges in front of them. Not sure our points of view are that different.

 

[DrMrLordX]

Over the course of 10 years, sure, things will change. But we're looking at possible 8-core 4x4 systems by mid-2007. Unless AMD has their version of Mitosis up and running by then, uh . . .