Upcoming AMD processors (within the next 2-3 months)

[JBS]

I am thinking of upgrading my AMD CPU, but I'm not sure if I want to get something that is currently available, or if I want to wait for an upcoming CPU.

Is there a way to find out what AMD is going to release soon (within the next 2-3 months), so I can make a more well-informed decision.

 

[Markfw]

What do you have now ? and what is making you think that you need to upgrade ? And why is not Intel a choice ?

 

[JBS]

I am using an AM2+ motherboard, so I would need a new board if I went with Intel.

Also, I am just much more familiar with AMD processors and not at all familiar with Intel.

 

[BlueAcolyte]

nothing, really. 45nm's coming out late this year.

 

[myocardia]

AMD just released their newest CPU's, the Phenom 9x50's. It should be quite some time, possibly the end of this year, before they release anything else, unless the rumors of a dual-core Phenom turn out to be true (which they don't seem to be).

 

[Extelleron]

Phenom 9950 (2.6GHz) will be coming out soon.

That's about it till 45nm late this year.

 

[bryanW1995]

there's no need to wait for phenom 9950 when the 9850 has an unlocked multi.
@ OP, what are you current system specs? what mobo/psu/etc?

 

[Viditor]

Upcoming stuff:

1. All current 9x5x chips will be 45nm before year's end (meaning that 125w will be 95w)

2. Adding the 8450e (Tollman TriCore), possibly very soon (by summer)

3. 9150e, 9350e, and the 9950...also in the next 6 months or so (but we don't know for sure when)

 

[ther00kie16]

wiki goodness

Athlon 6000 series "Kuma" is also coming out with 1600 HT.
Looks like the 9950's voltage is lower, but that's about it.
Interesting to see if the budget quad core (9150e) will overclock anywhere close to 2.4ghz for a 33% oc. If it does, it'll be one heck of a performance/cost chip.

 

[danielackerman]

the bulldozer core cpus are going to be extremely interesting to see in action. ive got a 6400 right now and couldnt be happier, paired with an 8800gtx there is nothing that my system cannot handle well. id say get the 6400, save your dough and get a bulldozer core when it hits the streets.

 

[Viditor]

Originally posted by: danielackerman
the bulldozer core cpus are going to be extremely interesting to see in action. ive got a 6400 right now and couldnt be happier, paired with an 8800gtx there is nothing that my system cannot handle well. id say get the 6400, save your dough and get a bulldozer core when it hits the streets.

To play Devil's Advocate here...
We don't really know how the Bulldozer will do. It is to be a "from-the-ground-up" new architecture for AMD, and not a modiification of the current uA.
That said, I agree that at the very least it will be "extremely interesting to see in action"...

 

[Idontcare]

Originally posted by: Viditor

Originally posted by: danielackerman
the bulldozer core cpus are going to be extremely interesting to see in action. ive got a 6400 right now and couldnt be happier, paired with an 8800gtx there is nothing that my system cannot handle well. id say get the 6400, save your dough and get a bulldozer core when it hits the streets.

To play Devil's Advocate here...
We don't really know how the Bulldozer will do. It is to be a "from-the-ground-up" new architecture for AMD, and not a modiification of the current uA.
That said, I agree that at the very least it will be "extremely interesting to see in action"...

But I thought that Russian dude told us all about it in the other thread?

What, you didn't believe it either?

 

[danielackerman]

heres hopin... if amd doesnt blow us all away with the bulldozer core im not so sure their future will be very bright. bulldozer seems to be amd's last chance to grab some major market share and make money at the same time.

 

[Extelleron]

Originally posted by: danielackerman
heres hopin... if amd doesnt blow us all away with the bulldozer core im not so sure their future will be very bright. bulldozer seems to be amd's last chance to grab some major market share and make money at the same time.

Thing with Bulldozer is it is not coming anytime soon. From what we are hearing at this point, Bulldozer looks like 2010 for actual availability. Note that AMD has stated they will be sampling Bulldozer on 45nm in 2009, but sampling is quite different from releasing to the public.

I'm pretty sure that Bulldozer will be competitive with Nehalem, but the problem is it will face 32nm Westmere, which should improve performance slightly and improve performance-per-watt and allow for higher clocks.

Now if AMD can get to 32nm in early 2010 and release Bulldozer @ 32nm, then they may be able to compete performance wise. That's what I think might happen, even though it violates AMD's Pipe strategy; AMD will release "K10.5" 45nm products in 2008 & 2009, with up to 8-core & 1MB L2 cache per core, sample Bulldozer on 45nm, then actually produce Bulldozer @ 32nm in 2010. That's assuming AMD can hit 32nm in early-mid 2010, which given AMD's track record is quite an assumption of course.

 

[Idontcare]

Originally posted by: Extelleron
Thing with Bulldozer is it is not coming anytime soon. From what we are hearing at this point, Bulldozer looks like 2010 for actual availability. Note that AMD has stated they will be sampling Bulldozer on 45nm in 2009, but sampling is quite different from releasing to the public.

I'm pretty sure that Bulldozer will be competitive with Nehalem, but the problem is it will face 32nm Westmere, which should improve performance slightly and improve performance-per-watt and allow for higher clocks.

Now if AMD can get to 32nm in early 2010 and release Bulldozer @ 32nm, then they may be able to compete performance wise. That's what I think might happen, even though it violates AMD's Pipe strategy; AMD will release "K10.5" 45nm products in 2008 & 2009, with up to 8-core & 1MB L2 cache per core, sample Bulldozer on 45nm, then actually produce Bulldozer @ 32nm in 2010. That's assuming AMD can hit 32nm in early-mid 2010, which given AMD's track record is quite an assumption of course.

Somebody's been reading TheINQ

http://www.theinquirer.net/gb/...d-roadmap-restores-bit

I find your synopsis a far better approach to rationalizing the possible future though, Charlie is still in rant mode. Somebody sh!t in his wheaties again.

Bulldozer being pushed out to 32nm is bad news for AMD in the short-term, but if it is planned this far in advance then it doesn't have to mean bad news for AMD in the long run.

The 10% layoffs is part of that, getting lean to survive the next 2 years of lean times. If AMD pares itself down so it can avoid bankruptcy while dealing with the financial realities of sub-10% marketshare then they can still do alright and make a "comeback"...but that kinda requires Intel to do another Netburst on themselves in the meantime too.

Its bleak, I can't imagine what the morale is like in the fabs.

 

[nyker96]

by the time that is introduced, intel probablly already got some higher binned 45nm quads out that's running close to 4ghz.

 

[Extelleron]

Originally posted by: Idontcare

Originally posted by: Extelleron
Thing with Bulldozer is it is not coming anytime soon. From what we are hearing at this point, Bulldozer looks like 2010 for actual availability. Note that AMD has stated they will be sampling Bulldozer on 45nm in 2009, but sampling is quite different from releasing to the public.

I'm pretty sure that Bulldozer will be competitive with Nehalem, but the problem is it will face 32nm Westmere, which should improve performance slightly and improve performance-per-watt and allow for higher clocks.

Now if AMD can get to 32nm in early 2010 and release Bulldozer @ 32nm, then they may be able to compete performance wise. That's what I think might happen, even though it violates AMD's Pipe strategy; AMD will release "K10.5" 45nm products in 2008 & 2009, with up to 8-core & 1MB L2 cache per core, sample Bulldozer on 45nm, then actually produce Bulldozer @ 32nm in 2010. That's assuming AMD can hit 32nm in early-mid 2010, which given AMD's track record is quite an assumption of course.

Somebody's been reading TheINQ

http://www.theinquirer.net/gb/...d-roadmap-restores-bit

I find your synopsis a far better approach to rationalizing the possible future though, Charlie is still in rant mode. Somebody sh!t in his wheaties again.

Bulldozer being pushed out to 32nm is bad news for AMD in the short-term, but if it is planned this far in advance then it doesn't have to mean bad news for AMD in the long run.

The 10% layoffs is part of that, getting lean to survive the next 2 years of lean times. If AMD pares itself down so it can avoid bankruptcy while dealing with the financial realities of sub-10% marketshare then they can still do alright and make a "comeback"...but that kinda requires Intel to do another Netburst on themselves in the meantime too.

Its bleak, I can't imagine what the morale is like in the fabs.

Interesting read but I'm not sure I agree with the title - AMD writing off the next 2 years or so and asking everyone to kindly wait until 2010 is not inspiring confidence in me and I'm sure shareholders are not happy either. AMD cried wolf once in 2006 when they asked us to wait until Q3 2007 for Barcelona, which would bring them back to competition. Obviously Barcelona was not such a chip and now in 2008, they are asking us to wait until 2010 for a competitive core.

The 6/12 core CPU should keep them fairly competitive in the server area, although I don't think it will give them the performance crown (depending on how good Nehalem ends up being, of course). Native 8-core Nehalem on 32nm (which should be out by the time a 12-core AMD CPU comes out) w/ SMT should defeat a 12-core K10. And it would definitely be cheaper to produce. AMD also comes into problems with power consumption; given that Quad-core 45nm will be 95W TDP in most variants, clocks on 12-core parts will need to be very low to keep TDP at a respectable level.

On the desktop arena, cores beyond 4 don't matter much at all, so the niche products with >4 cores won't come into the picture here. K10.5 should bring performance parity w/ Penryn but Nehalem will likely be a good deal ahead. Depending on how effective Nehalem's SMT is, it is not inconceivable that a dual-core Nehalem could defeat a quad-core Deneb CPU in multi-threaded benchmarks. Considering Nehalem & Deneb are roughly the same size in quad-core form, dual-core Nehalem will be significantly cheaper to produce, even when compared to Propus w/o L3 cache.

Bulldozer is AMD's last shot and as the article mentions, they will have to do what it takes to stay alive the next two years - they have their work cut out for them. Deneb and Propus will sell at the right price, but what kind of price can AMD afford to sell them for? As we are seeing with Phenom right now, AMD is getting less competitive with pricing; at the point they are now, they cannot price Phenom much lower and still have it be worth selling. They would be doing fine if their parts were slower but cheaper to produce, but the problem is their parts for the forseeable future will need to be price competitive with Intel parts that are much cheaper to produce. That's just not a good way to do business.

 

[Idontcare]

Originally posted by: Extelleron
Depending on how effective Nehalem's SMT is, it is not inconceivable that a dual-core Nehalem could defeat a quad-core Deneb CPU in multi-threaded benchmarks. Considering Nehalem & Deneb are roughly the same size in quad-core form, dual-core Nehalem will be significantly cheaper to produce, even when compared to Propus w/o L3 cache.

This has weighed heavily on my mind for months now, as no doubt it has weighed heavily on the minds of Hector and friends too.

Effective SMT seriously upsets the tenuous balance in this "threads per socket war" (core war will no longer be an appropriate title for the era once SMT enters).

 

[bryanW1995]

I'm excited at the prospect of smt, but I still am withholding judgement on it until I see some actual performance numbers. If it scales at 10-20% then it's a failure imho, but if it goes to 70-80% then it likely will bust out the whoopin' stick on shanghai.

 

[Phynaz]

According to the slides AMD presented today, Bulldozer is not a 2010 part.

Bulldozer - if it still even exists - is now 2011 at the earliest.

 

[Extelleron]

Originally posted by: Phynaz
According to the slides AMD presented today, Bulldozer is not a 2010 part.

Bulldozer - if it still even exists - is now 2011 at the earliest.

Well the new roadmap stops at H1 2010, so that is not representative of all that will be released in 2010.

But it's definitely a far cry from the sunny roadmaps presented in July last year when Barcelona was going to rock C2D in Q3 2007 and Bulldozer was coming out in 2009.

 

[Viditor]

Originally posted by: Extelleron

Originally posted by: danielackerman
heres hopin... if amd doesnt blow us all away with the bulldozer core im not so sure their future will be very bright. bulldozer seems to be amd's last chance to grab some major market share and make money at the same time.

Thing with Bulldozer is it is not coming anytime soon. From what we are hearing at this point, Bulldozer looks like 2010 for actual availability. Note that AMD has stated they will be sampling Bulldozer on 45nm in 2009, but sampling is quite different from releasing to the public.

Yes...but if they are sampling in H1 09, then we should see retail availability in H2 09.

I'm pretty sure that Bulldozer will be competitive with Nehalem, but the problem is it will face 32nm Westmere, which should improve performance slightly and improve performance-per-watt and allow for higher clocks.

Now if AMD can get to 32nm in early 2010 and release Bulldozer @ 32nm, then they may be able to compete performance wise. That's what I think might happen, even though it violates AMD's Pipe strategy; AMD will release "K10.5" 45nm products in 2008 & 2009, with up to 8-core & 1MB L2 cache per core, sample Bulldozer on 45nm, then actually produce Bulldozer @ 32nm in 2010. That's assuming AMD can hit 32nm in early-mid 2010, which given AMD's track record is quite an assumption of course.

 

[jones377]

Originally posted by: Viditor

Yes...but if they are sampling in H1 09, then we should see retail availability in H2 09.

for someone who claims to follow AMD so closely....