Fuddo: AMD 45nm K10.5 scheduled to launch in 1H 2009

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar


Your takeing Newegg posts as legit?

Word, Playa.
 


LAME ARGUMENT. Quad-core is quad-core. The end result of MCM is far superior to what AMD can do monolithically. So to me, Intel made the right choice there.



OK.... All these hardware enthusiast sites must be doing it wrong!?
 


So you are saying that a Q6600 cannot at stock with not changing of anything cannot outmuscle a Phenom 9700? Although it has been shown that in an overall benchmarking it would take a Phenom 9900 to fully compete with the Q6600? I am sorry but you are being very blind right now.

Also Barcelona and Phenom are totally different chips. Memory bandwidth does nothing in gaming. In fact on the desktop the FSB never truly gets bottlenecked or saturated with the stuff normal people do like game, internet, encoding/decoding and music.

I can't wait until Intel releases its new Itanium just so we can get a glimps at their new QPI and IMC. It would finally get all of thr AMD fanbois to shut up about the "allmighty" IMC.
 


Trying to find the link to the revised processor roadmap. It lists Phenom FX91 and Phenom FX92 for socket 1207/1207+. Granted I'm hoping thats not a currently outdated roadmap. If they can get the speed up I say pick the platform back up just to spar with skull trail

 

The guy is a troll jimmysmitty. Just laugh at him. He is rediculous.
 


People have got to stop calling each other trolls, and they have to recognize that all of us can descend into fanboyism.

If the "allmighty" IMC were not worth it, then Intel would not be doing it. Behemoth corporations with huge market share only innovate when they face hurdles and/or competition. The Pentium M was forced on Intel because Netburst was not good enough for notebooks, but tolerable on the desktop.

The Athon X2 forced competition on Intel and led to Pentium M derived C2D's, because Netburst was fast becoming obsolete. Intel innovated conservatively, which is why they did not implement native quad core at 65nm or use IMC. Yet, native quad cores and IMC must be worth something when done at the right process, otherwise, you'd not see Intel go that route.

So yes, IMC is "allmighty" compared to a front side bus. Any failures on AMD's part are related to native quad core issues at 65nm with SOI. The Phenom core isn't that bad, it's just not clocked high enough natively and can't overclock well enough to please enthusiasts. I hope that changes at 45nm, at least after SOI. I really want to see two viable CPU companies, and not just to lower the prices of Intel's CPU's.


 


You just don't get what I mean do you? I mean that for servers, yes the IMC is better. But for us normal plebians the IMC doesn't have an advantage yet and probably still wont for a few more years or when games/software become more memory bandwidth intensive.

The fact is that continually using the IMC is better than FSB crap is a waste of good air. Right now the Q6600 sits on a 1066FSB and Phenoms IMC doesn't do much to help it vs that.

Yes Intel is moving towards it but not due to AMD. They started making an IMC with Timna remember? It just wasn't time to throw the FSB out yet.

I would say that now, with Nehalem, is the best time to make the switch. Looking at Windows and the software a IMC may help. But don't tout the greatness of the IMC as a reason why AMD is better. Tout that their CPU shows great performance vs the competition.

Using the IMC argument is kinda like saying I have a Viper with a V10 but that Corvette over there has a V8 and smokes me.
 


LOL too bad that 300MHz 'small bump' is about as far as most Phenom 9600BEs can be pushed. :lol:

And with a 'fairly large bump' on the Q6600 (lets say around 3.6GHz) it performs about 50% faster than the Phenom @ 2.7GHz... 😉
 


Um not according to what I could find on it, which is very little. According to what I can find it was to have both a IMC and a GPU integrated with the CPU. Yes it was to be used with RDRAM but if RDRAM has a memory controller then why, with the first Pentium 4's that used RDRAM, did they have a memory controller on the northbridge? Would seem redundant wouldn't it?
 


While this is true for the most part, there are still those nut jobs out there that will want to build "teh most uber l33t computer out there!!!@". A guy I knew a long time ago, use to use a Dual socket P3 system for his gaming, back when P3 was still the main Intel CPU(can't remember if they were Xeons or just regular p3's). He'd likely be the kind of person that would throw together a 4x4 or skulltrail system if money allowed him to do so.
 
As for the original posts at the beginning of the thread and estimated available dates, AMD has already shown working 45nm processors. I don't remember which site I read it on, probably anand's, but AMD does indeed have working 45nm cores. I don't really doubt working quad core parts this year, as that's where they're going to make more money, servers and higher end parts. The tri's and duo cores will come later, as people expect.
Whoever read the article on fudzilla didn't take into account that the quads will come first and just assumed the entire process and it's products won't be here until later.
AMD had a press release about it and I found it on Ars.
http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~124069,00.html
http://arstechnica.com/news.ars/post/20080304-amd-demos-45nm-server-and-desktop-processors.html
 


Sorry but its hard to believe AMD's press releases right now.

They could possibly have 45nm working but that doesn't mean it will be fully operational before the end of the year.

Hell Nehalem is already sampling which means we might start seeing benchmarks in about Mays time. AMD hasn't even gotten their B3 version Phenoms out for test sites to see if there is any change.
 


AMD did perform a demonstration of functional 45nm silicon at CeBit.

A bit more advanced than the "task manager" demo of Barcelona back in Sept of 06, and about of the same level as:

1) Penryn demos in Spring '07
2) Nehalem demos in Fall IDF '07.

Given that the first silicon of Shanghai was somewhere between end of January and end of February combined iwth the release date of Barcelona ("Oct '07"), Penryn (Nov '07) and Nehalem (Q4 -- pulled into Q3? ) -- where does that put Shanghai? Very very very very late (ie Dec 31) of 2008. At best.

There is no way AMD gets Shanghai out before Nehalem.
 
Yeah, it is looking increasingly likely that Nehalem will be launched before Shanghai, especially if the Nehalem launch is pulled forward to Q3.

This could have serious repercussions in regards to how the press/reviewers look at Shanghai, I'm sure AMD would much prefer to have it reviewed against Penryn rather than Nehalem.

Ed's take on the situation, he makes some good points:
http://overclockers.com/tips01304
 
I didn't mean to imply that AMD isn't screwed in general, I was just saying that they should still be able to launch quads based on 45nm before the end of the fourth quarter. Just thought people should remember that the tricores will be out _after_ the quads.
At this point, they are pretty much boned in the high end due to their complete and utter lack of execution. They never should have bought ATI for so much money because it meant they didn't have anything for R&D. And now they're the next thing to dead.
 


This might come as a shock to some, but I totally agree with you.

Nothing much was expected of AM2, it was basically a change to DDR2. We expected more of 65nm, in fact, all of AMD's best parts are still 90nm.

K10 was their counter-shot back at Intel, and they missed the mark.
 


http://www.digitimes.com/mobos/a20080317PD209.html
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3260



Yeah, not on the mark, but not too far off from it either. Sad but true on what Shanghai will be compared to if it's not out before Nehalem.

 


Nice links but still not what I would have expected so close to the "new" release date. I would have expected to see a few test done by a few sites to see if there was any change in performance at stock vs the Q6600. Not that I truly expect it but you never know.

And if Shanghai has to go up against what Nehalem is stated to be it might be interesting. It will be two chips on par with eachother and now one just has to have a better IPC than the other. No more IMC this or FSB that. Just which chip performs the best.
 

Personally I can't wait for teh fall-out of that!
Fanboys won't know what to do! :lol:
On a more serious note, it'll be interesting to see how Intel's implementation of IMC, HT link and monolithic multi-core's goes....
AMD will have a massive head-start in terms of experience with those, but will have much less experience of HK based CPU's (or, at least non-SOI), that Intel will have...
Although given the present situation, it'll be interesting to see how good Intel's R&D is... 😀
 


Yup. According to endeyn, AMD's HK/MG will still be a SOI wafer using HK/MG. Or thats what I got of it. Even though it will be IBM who creates the technology for AMD so it might be decent.

Look at Intels R&D this way. If Intel's software writing division were to break off into its own company it would be one of the largest software companies in the world. So their R&D division must get more funding than that. I am sure they have been working on that stuff for a while and have taken what they learned from Timna too although they never fully launched it.

I don't have a doubt that Intel will do fine with the IMC, HT and monolithic multi coress. Its just if the chip performs like stated.
 
I don't know that AMD has any advantage in terms of already having a monolithic core and to play with though ... because Intel is way ahead on the fundamental design principles wrt:

Transistor design - lower power and smaller cell size;

Cache and prefetch logic - cripes ... they wiped the floor with AMD considering the off-chip communication utilises FSB instead of HT Link system;

lithographic technology - their system is simply superior and smaller.

I think good old AMD will still be there ... perhaps picking up a bit of high end server business now they have a decent Barcy product ... but relegated to the budget zone again in terms of single socket ... Joe Public PC's.

Once Nehalem arrives they will likely lose the higher end server market ... this will see them staying in the low end unless someone pulls another DEC Alpha idea out of their ear.

Or I could be totally wrong.

You have got to ask why is Intel and Microsoft now working so hard together on multiple threading ??

Because more cores generates diminishing returns unless the work is truly being done by a number of parallel threads / cores.

All these touted cores otherwise become white elephants.

That is the next barrier ... unless the speeds can easily double (8 to 10 Ghz) in the next 18 months ... I don't see any breakthroughs in substrate / materials / design in that regard.

I also don't even see that Nehalem will be anything special ... just in terms of multiple socket performance.

A core2 tweaked with a memory controller is only going to be a bit faster on most tasks.

Nehalem also presents Intel with the perfect opportunity to close a hole in their microprocessor design that has been irking them for years ... OVERCLOCKING.

Mark my words ... they will lock these puppies up so you cannot overclock them.

It isn't hard to see the reasoning ... it isn't hard to see how.

Money ... Intel's entire line is made on such a good quality process they all overclock well ... much more so than any of the current AMD line.

Easy ... look at Phenom ... the memory controller speed is the issue with the dam thing ... tied to the L3 cache even makes it more difficult to overclock. Yeah the quality of the process might not generate as much headroom as Intel either ... but the point is the dam thing doesn't overclock easily (not counting the black).

Overclockers ... We are possibly now the enemy !!!

If you can't overclock the cpu then it is worth what is worth ... there is no hidden value.

Ask Why the FSB has been raised to 1333 and 1600 ?? There is little performance increase for a single socket ... but it is much harder to overclock.

LOL
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom