backtop


Print 60 comment(s) - last by bart.. on Mar 2 at 9:03 AM


The latest AMD roadmap reveals some of the finer points about AM2 launch and schedule times
Athlon 64 5400+, FX-64, Energy Efficient Athlons and AM2 details revealed

AMD has been quite the busy bee lately.  The Athlon FX-60 launched a few weeks ago, the Opteron x85 series launches in a few weeks, and the whole platform will get an overhaul in early June.  AMD's latest roadmap reveals more about all of these projects and adds several more that have not yet been disclosed.

AMD's newest roadmaps claim the new 940 pin Socket AM2 will launch at Computex Taipei on June 6, 2006... that's 6-6-6 for those of you playing at home. The AM2 socket is not compatible with existing Socket 940 CPUs, but will allow AM2 processors to utilize up to DDR2-800. 

Athlon 64, Athlon 64 X2, Athlon 64 FX and Sempron will all use the AM2 socket.  The unified socket should alleviate many of the nightmares system builders experienced during the Socket 754/939 haydays -- and will also help AMD consolidate its manufacturing.
 
AMD Desktop Processor Roadmap 2006
Processor
Socket
Core
TDP
Launch Date
Athlon 64 FX-64
AM2
Dual
???
Q1'07
Athlon 64 FX-62
AM2
Dual
125W
6/6/06
Athlon 64 FX-60
939
Dual
110W
Now
Athlon 64 5400+
AM2
Dual
???
Q1'07
Athlon 64 5200+
AM2
Dual
???
Q3'06
Athlon 64 5000+
AM2
Dual
89W
6/6/06
Athlon 64 4800+
AM2
Dual
89W
6/6/06
Athlon 64 4600+
AM2
Dual
89W
6/6/06
Athlon 64 4400+
AM2
Dual
89W
6/6/06
Athlon 64 4200+
AM2
Dual
89W
6/6/06
Athlon 64 3800+
AM2
Dual
89W
6/6/06
Athlon 64 3800+
AM2
Single
62W
6/6/06
Athlon 64 3500+
AM2
Single
62W
6/6/06

You may notice the TDP on the Athlon 64 FX-62 has been increased to 125W. 

Like Intel, there was a large emphasis on TDP this time around.  The Athlon FX series is approaching thermal envelopes reminiscent of Prescott, and it seems clear to all parties involved that something more innovative needs to occur to counter that trend. 

AMD will also soon unveil a new set of desktop Sempron and Athlon 64s specifically geared for HTPC and other low noise / low power devices.  The goal is to build processors that do not need a lot of cooling -- specifically to reduce the noise of the overall system to the 27dBA range for idle and 30dBA under load.  One of the upcoming Windows MCE requirements for CPUs is that the processor fan not exceed 30dBA for MCE certification.

The Energy Efficient CPUs will have a slightly redesigned Sleep, Power and Suspend states.  These states are controlled by the operating system, but will require BIOS updates.   

AMD Energy Efficient Desktop Models
Processor
Socket
Core
Wattage
Launch Date
Athlon 64 4800+
AM2
Dual
65W
Q3'06
Athlon 64 4600+
AM2
Dual
65W
6/6/06
Athlon 64 4400+
939
Dual
65W
6/6/06
Athlon 64 4200+
AM2
Dual
65W
6/6/06
Athlon 64 4000+
AM2
Dual
65W
6/6/06
Athlon 64 3800+
AM2
Dual
35W
6/6/06
Athlon 64 3800+
AM2
Dual
65W
6/6/06
Athlon 64 3500+
AM2
Single
35W
6/6/06
Sempron 3500+
AM2
Single
35W
Q4'06
Sempron 3400+
AM2
Single
35W
6/6/06
Sempron 3200+
AM2
Single
35W
6/6/06
Sempron 3000+
AM2
Single
35W
6/6/06

The Energy Efficient Athlon 64 3800+ rated at 35W is not a typo -- AMD actually has a dual core 35W processor on the roadmap.  This bodes well for dual core Turions. 


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

I'm mad
By tjr508 on 2/17/2006 3:46:36 AM , Rating: 2
why is it that ever since intel released all those promotional pdf files about "performance per watt" people won't shut up about it. Sure it's great on laptops, expecially the really mobile ones, but who really cares if their desktop cpu is 90w or 110w. I don't see motherboard manufacturers dropping their price when they dont need the current requirements on their cpu power supplies so who really gives a damn. The high end desktop segment needs to be just that. high end. fastest. best work horse. gets the most done in the smallest time. i could care less if it consumes twice the power of a mobile chip.

Shame on intel for bringing up "PPW" expecially after the p4 and shame on AMD for now telliing us more about watts than performance on their desktops.




sorri i forgot to add
By tjr508 on 2/17/2006 3:53:35 AM , Rating: 2
As for the Athlon 64 fx 64, i can already see people buying an "athlon 64 fx" thinking they are going to get an "fx 64"


RE: sorri i forgot to add
By Visual on 2/17/2006 5:36:49 AM , Rating: 2
people who are that stupid usually dont get that rich... but if this actually happened, it'd be only just.


RE: I'm mad
By mircea on 2/17/2006 4:09:29 AM , Rating: 2
The more power it draws, the hotter it runs, the less you can overclock, the louder it gets because of cooling. Got it???


RE: I'm mad
By tjr508 on 2/17/2006 4:59:44 AM , Rating: 2
Are you suggesting x86 makers draw a line somewhere? make themselves a boundary? Sure low power has its benefits, the R&D involved eventually will lead to better performing chips, but for crying out loud, we havent seen a serious performance increase in two years now(FX51). Hell an old thunderbird 1.333 does great for most typical PC uses on XP and that was from late 2000 or early 01.

I am not about to call a dual core a performance increase just because they got the cores on the same chip. multiprocessor boards have been around forever. It is sad that we have come to a day where technology has stopped and now we must go back and change everything we thought we knew to gain a little performance by using more threads instead of faster ones.

It is nice to be able to stay on top of technology now a days with a two year old system, but i like to see bigger and better things come out of our industries and don't mind paying for them every couple of years or spending an extra $30 on top of the $300 chip for a quieter cooling solution.

Overclocking on the otherhand has a lot more to do with design than wattage. You can find a 130w p4 that will OC easily to 35% while a lot of 30w mobile chips won't be stable (or mobile) at 10% OC. On the other hand, makeshift mobile chips like the turion make great overclockers because they are more or less improvised mobiles. They are simply some of the best A64s with the lowest transients being worked well below their true capibilites to reduce power cunsumption so of course they will OC well.



RE: I'm mad
By dreddfunk on 2/17/2006 9:30:22 AM , Rating: 2
Look, I'm no expert, but performance-per-watt is a serious issue for a lot of markets: server, workstation, htpc, mobile, to name a few. It's become a big issue for big business, certainly: when you have hundreds of servers and thousands of workstations, the difference between a 130W CPU and a 65W CPU adds up to a lot of $$ over the course of a year, let alone the lifetime of the machine.

Of course it's not an issue at all for the enthusiast market, but, while prestigious, it's not a very large market compared to mainstream & business markets. It sucks that we haven't seen huge increases in CPU power recently but, given the amount of money Intel and AMD are throwing at the problem, it isn't for a lack of trying.


RE: I'm mad
By Viditor on 2/17/2006 11:31:21 PM , Rating: 2
For a server farm with 1000 servers, the difference comes to ~$57,000/year on just the power for the CPUs. Then you have to add the price of cooling, which will usually double that.


RE: I'm mad
By tjr508 on 2/18/2006 7:47:48 AM , Rating: 2
That is a good point though my initial rant was on "desktops" and not servers and laptops.

BTW modern heat pumps have a coef. of performance (COP) around 4 (This equates to a COP of 3 in reverse), so displacing (or moving) x units of heat energy will only require x/3 units of electrical energy. Therefore that 2x figure is a little off. Much more efficient pumps are available for buildings such as server farms. The figure of 4 is for standard home systems.

Maitnence and related costs are irrevalant here as they would be required with the smaller system.


RE: I'm mad
By NullSubroutine on 2/19/2006 6:49:14 PM , Rating: 2
AMD has a streamlined product lineup, this is why you see processors with lower tdp. also, unless i am mistaken, but companies buy DESKTOPs for their businesses. not every employee has a server or laptop to use, so your 'rant' on 'desktops' is null and void.

products that use less power require less power req (dont need huge power supplies), they require less captacitors and such on the main board as well.

products that use less power run cooler. products that run cooler last longer.

as for your rant on how they arent any faster than a thunderbird is completely baseless, what are you basing these benchmark conclusions on? how fast internet explorer shows you your porn?


RE: I'm mad
By Viditor on 2/20/2006 10:30:38 AM , Rating: 2
quote:
BTW modern heat pumps have a coef. of performance (COP) around 4 (This equates to a COP of 3 in reverse), so displacing (or moving) x units of heat energy will only require x/3 units of electrical energy. Therefore that 2x figure is a little off

For buildings you are correct...but remember that you are also increasing the displacement within a localised area (the case and the racks). Because the heat must be moved quickly, the COP is increased dramatically. I got the "double power cost" from a couple of friends that run server farms...


RE: I'm mad
By ddopson on 2/21/2006 10:06:35 PM , Rating: 2
Plus the fixed costs of that cooling equipment are not negligible.


RE: I'm mad
By del on 2/21/2006 4:56:54 PM , Rating: 2
Multiprocessor boards have been out for some time now, but multi-threaded software hasn't really been considered important until recently. Just look at Windows 98. Also, it's better to buy a dual-core processor and one HSF than two higher-clocked, and thus hotter, single-core processors and two HSFs. The only hope for increasing performance with conventional computers these days are dual-core (and later, quad- and multi-core) processors, 64-bit CPUs, and technologies like SLI. Silicon technology is reaching its limits. Maybe we will switch over to quantum computers when the conventional computing reaches the breaking point. Remember, 64-bit will also help, coupled with these serial RAM chips in another article here.


RE: I'm mad
By del on 2/21/2006 4:45:52 PM , Rating: 2
heh "shame on Intel"

...You don't know what you're talking about. Performance per watt is about watts electric, not watts thermal. What's the point of being the best work horse and all that when a lot of power is not used? Lower TDP is just a beneficial side effect of power-optimized CPUs. I'm surprised none of you noticed this. The more watts that are consumed by your power supply, the more money you have to pay each month for electricity. The only reason we still use 510 W power supplies is because video card manufacturers haven't followed suit with CPU manufacturers with technologies such as dual-core (and multi-core) and low-power designs.


RE: I'm mad
By ddopson on 2/21/2006 10:00:07 PM , Rating: 2
510W is just the maximum power rating for all the channels on your PSU combined. What really matters is its efficiency. Typical PSU's run from 75%-90% efficiency, which means that to generate 75W of DC power at 3.3, 5, and 12 Volts burns 100W of 120V AC power out of the wall.


Is there going to be a 35W dual-core?
By Furen on 2/16/2006 11:37:38 PM , Rating: 2
It looks to me like the 35W dual-core 3800+ is actually going to be a single-core CPU. If you look at the rest of the CPUs on this chart you'll notice that excepting this one all other 35W CPUs are single-core and all dual-core CPUs are 65W.




Moderated
By Knish on 2/17/06, Rating: -1
By KristopherKubicki (blog) on 2/17/2006 12:14:16 AM , Rating: 2
It's not a typo. The fact that it is the only 35W dual core processor on the roadmap is highlighted even. It should be a neat chip.

Kristopher


RE: Is there going to be a 35W dual-core?
By Furen on 2/17/2006 12:23:54 AM , Rating: 2
Interesting... then 35W dual-core Turions will be possible too. It's kind of nice to see something more concrete about that. Looks like AMD is tweaking the hell out of the K8 design in order to make it as power efficient as possible.

I wonder how these optimizations are affected by clock speed increases, since it looks like Merom is goint to gain around 30W of thermal dissipation when it goes to the desktop while only gaining a faster FSB and 333MHz-667MHz.


RE: Is there going to be a 35W dual-core?
By Viditor on 2/17/2006 12:46:06 AM , Rating: 2
quote:
Looks like AMD is tweaking the hell out of the K8 design in order to make it as power efficient as possible

I'll betcha these are the ones with the new strained silicon process. The rumour is that it requires fully depleted SOI, so there will probably be fewer models available until they sort out their supply issues on the wafers.


RE: Is there going to be a 35W dual-core?
By mlittl3 on 2/17/2006 2:04:39 PM , Rating: 2
Only reducing the clock by 200 MHz will not reduce the power dissipation by half. I don't think that is scientificly feasible.


RE: Is there going to be a 35W dual-core?
By mlittl3 on 2/17/2006 2:05:33 PM , Rating: 2
Oops, that post was meant for Viditor.


RE: Is there going to be a 35W dual-core?
By mlittl3 on 2/17/2006 2:06:27 PM , Rating: 2
Man, I'm having a bad day. I mean the post was meant for JackPack. Time for a nap.


By JackPack on 2/17/2006 6:22:16 PM , Rating: 2
True, but the power consumption vs. frequency curve isn't linear. I assume it starts to take-off somewhere along the 2.2 to 2.4 GHz range because that's where the TDP for Athlon 64 X2 changes from 89 to 110W.

35W dual-core 1.8 GHz should be doable -- should be similar to Turion X2. It may be a boutique part though.


By JackPack on 2/17/2006 1:04:51 AM , Rating: 2
It shouldn't be that difficult.

They can claim DDR2-800 to be a speedgrade. So AM2 X2 3800+ might be 1.8 GHz.


Heh...
By Doormat on 2/17/2006 12:37:05 AM , Rating: 2
The AM2 socket is released on 6/6/6. AMD is teh devil.




RE: Heh...
By Furen on 2/17/2006 12:38:51 AM , Rating: 1
It's going to be released at computex so if you wanna blame someone blame the organizers.


RE: Heh...
By blobguy on 2/17/2006 2:34:40 AM , Rating: 2
quote:
launch at Computex Taipei on June 6, 2006... that's 6-6-6 for those of you playing at home.


quote:
The AM2 socket is released on 6/6/6. AMD is teh devil.


How would you like that Date for your Birthday? (;


RE: Heh...
By gooser on 2/17/2006 9:29:28 AM , Rating: 2
HAHA!! I do! june 6, 1984.
A new AMD proc (AM2 style) for my birthday?? I think so.


RE: Heh...
By DigitalFreak on 2/17/2006 8:40:30 AM , Rating: 2
Or is that the latency from DDR2-800?


RE: Heh...
By butane317 on 2/17/2006 9:48:15 AM , Rating: 2
My thoughts exactly! That was hilarious.


RE: Heh...
By stromgald on 2/19/2006 11:42:34 AM , Rating: 2
Well, actually the number 6 is considered lucky in some asian cultures, so 6/6/06 could be considered a good day.


So what are the...
By Marlin1975 on 2/16/2006 11:25:04 PM , Rating: 2
So whats the mhz? Are they still the same 3800+ 939 = 3800+ AM2 ???




so X2 at 90um WILL reach 3GHz
By Hulk on 2/17/2006 12:07:41 PM , Rating: 2
I was wondering if AMD would be able to push this process to 3GHz. I guess they will. You can usually tell how far a process will go (usually) by how far people can overclock. I don't think many people are getting 3GHz out of the X2's on air cooling so perhaps AMD has been binning chips that can hit 3GHz.

?


RE: So what are the...
By bart on 2/17/2006 1:48:06 PM , Rating: 2
I wounder about this question to.

An AMD 3200+ socket 754 is clocked at 2,2 Ghz with 512kb of memory, and an AMD 3200+ socket 939 is clocked at 2 Ghz with 512 kb of memory. As I recall the answer to this was then that the + numbering is compared to a thunderbird b, and the same processor is faster on socket 939.

Following this, one could wonder if DDR2 at 800 mhz would make it reasonable to rename a socket 939 4800+ to a socket AM2 5000+.

So my guess is: Faster memory interface same processors

As a direct answer to the question, I think it’s unlightly the MHz of 3800+ 939 is equal to a 3800 AM2, but then again just im guessing.


RE: So what are the...
By mxzrider2 on 2/18/2006 2:05:22 AM , Rating: 2
reason= 939 is dual channel memory. that why


RE: So what are the...
By ddopson on 2/21/2006 10:02:55 PM , Rating: 2
FYI, DDR2 is slower, clock for clock. The standard allows for higher latency values than DDR (amung other things). But, on the other hand, this lets you clock the transfers faster. So 800Mhz DDR would spank 800Mhz DDR2, but there will never be 800Mhz DDR.


RE: So what are the...
By bart on 3/2/2006 9:03:45 AM , Rating: 2
People are speeking of DDR2 at 667 Mhz being about as fast as 400 mhz DDR. So in that case 800 Mhz DDR2 would be faster.

I think you have misunderstood my earlier post


Clocks?
By AggressorPrime on 2/17/2006 12:18:38 PM , Rating: 2
My sources indicate that the 5000 will be 2.6GHz and the FX-62 will be 2.8GHz. Therefore, I assume the 5200 will be 2.8GHz and the 5400 will be 3.0GHz along with the FX-64 being 3.0GHz? What will the FX have over the X@ besides overclockability then? If you know different clocks, please post!




RE: Clocks?
By AggressorPrime on 2/17/2006 12:19:16 PM , Rating: 2
Edit: X@=X2


RE: Clocks?
By MDme on 2/17/2006 7:57:52 PM , Rating: 2
IIRC the 5000+ is 2.4Ghz with the DDR2 Memory controller (the DDR2 i think was a speed grade (at least that's what AMD thinks) :)

so...

X2 5000+ will be 2.4Ghz with 2x1MB
X2 5200+ will be 2.6Ghz with 2x512?
X2 5400+ will be 2.6Ghz with 2x1MB

FX-60 2.6Ghz S939 with 2x1MB
FX-62 2.6Ghz AM2 with 2x1MB
FX-64 2.8Ghz AM2 with 2x1MB

my 0.02 cents


RE: Clocks?
By liquidaim on 2/18/2006 9:36:09 AM , Rating: 2
I disagree because the TDP for FX-62 is 125W. If it was clocked at the same speed, the TDP would remain the same.

I suspect that the 62 will be 2.8 and 64 will be 3.0 and it's TDP should be a whopping 140W if it's a 90nm chip. But since it's release is slated for Q1 07 it will most likely be 65nm.

baaahhh. its all speculation anyway. cant wait to get some benchies and finally make a decision on which chip to upgrade to from my 2600 barton.


RE: Clocks?
By KristopherKubicki (blog) on 2/20/2006 7:09:27 PM , Rating: 2
The 5000+ is 2.6GHz 2x1MB L2. You should be able to determine the other clocks/memory sizes from there.

Kristopher


RE: Clocks?
By stephenbrooks on 2/18/2006 2:53:46 PM , Rating: 2
There's some speculative clocks here:
http://www.c627627.com/AMD/Athlon64/
The way I understand this particular site, the clocks are speculative (particularly on AM2) but the model numbers and release dates are not (or not so much).


FX #'s
By kattanna on 2/17/2006 12:59:47 PM , Rating: 2
what annoys me most is that they have changed the FX ##'s from odd to even numbers.

i was REALLY wanting an FX-69 chip





RE: FX #'s
By AggressorPrime on 2/17/2006 6:16:23 PM , Rating: 2
But now you can have the Athlon 64 FX-64 64-bit CPU.
64-64-64-AMD!


RE: FX #'s
By Griswold on 2/18/2006 11:47:55 AM , Rating: 2
Triple Sixty-Four Bang Bang Baby!


RE: FX #'s
By KaerfSusej on 2/21/2006 8:03:23 PM , Rating: 2
You might have already figured it out, but for what it's worth, I believe that the odd numbered FX's are single core, while the even ones are dual-core. So it is possible you could get a FX-69 chip.


Athlon 64 3800+
By Questar on 2/17/2006 10:31:36 AM , Rating: 2
The exact same model number in both single and dual core?

Please tell me this is a typo. If not, AMD needs to smacked for being so stupid. If they want to be mainstream they can do crap like this to consumers.




RE: Athlon 64 3800+
By Furen on 2/17/2006 11:03:04 AM , Rating: 2
I'm pretty sure they'll use X2 for dual-core CPUs


RE: Athlon 64 3800+
By Plasmoid on 2/17/2006 2:08:37 PM , Rating: 2
I dunno... unofficial support for higher memory speeds, a better memory controller.

Hard to say really... im guessing it will be out well in advance though.


mhz dont matter
By mfed3 on 2/17/2006 10:42:25 AM , Rating: 2
why do people still get all bent out of shape over the mhz and speed ratings. they all have the same features, and plus you all know you're gonna read the benchmarks on these processors anyway when they come out. the smart person picks an architecture and then buys based on how much they want to spend, not the mhz rating.




RE: mhz dont matter
By stephenbrooks on 2/17/2006 4:34:04 PM , Rating: 2
MHz values at this stage would give us some idea of what those benchmark values are going to be. Additionally it would tell whether AMD has a lot of headroom in its current process or not.


RE: mhz dont matter
By Targon on 2/21/2006 11:49:02 AM , Rating: 2
The real issue is that it becomes more difficult to decide which processor is the best choice.

Which is faster overall, a socket 754 Athlon 64 3200+ with a single channel memory controller yet runs at 2.2GHz and has a 512K cache, or a socket 939 Athlon 64 3200+ with a dual channel memory controller, yet runs at 2.0GHz? Based on the speed rating, both should run at very close to the same speed, but the higher clock rate of the socket 754 version gives it an advantage in real world applications.

So, now we have DDR-2 versions. It will be more of a nuisance than a real problem if we need to check benchmark results for each new processor just to know what the real-world performance will be. Will DDR2-800 memory give enough of a performance boost to make it worth a bump in speed rating when compared to PC3200 memory? Then will we get into the situation where a socket 754 3700+ will be faster than a socket 940(AM2) 4000+(single core) due to the clock speeds?

Most of us don't mind not knowing the exact clock speeds, but we need time to learn how well the new chips perform compared to their speed rating. Clock rate is important, but it's not the only thing we look at. I've put off doing a system upgrade because I have been waiting to see what the performance of the DDR-2 versions of the Athlon 64 will be.


FX-62
By dsdsdk on 2/18/2006 12:29:53 PM , Rating: 2
FX-62 will be a 2666mhz cpu.. why are you saying 2800?




RE: FX-62
By stephenbrooks on 2/18/2006 2:52:00 PM , Rating: 1
Ah, but that was back when we thought AM2 was going to be limited at DDR2-667 and now it's DDR2-800. Hmm. Actually how the heck are they going to sort this out. What if you put 667 RAM with a chip that's clocked to a multiple of 200 and not 166? I think a RAM divider would appear, maybe not such a good thing.


RE: FX-62
By Falloutboy on 2/21/2006 12:33:29 PM , Rating: 2
Athlon 64s already use dividers with no performance penalty, the mem controller is still on die so it doesn't matter to the cpu what speed the ram runs at. I imagine the dualcore chips and highend single core will have official support for ddr2 800 while the lowend singlecore stuff will be ddr2 667 parts, but ethier will be able to run both speeds of ram, and prolly will overclock to higher speeds easily with a good board, just like you can run ddr600 on some current boards


35W Dual Core
By AggressorPrime on 2/17/06, Rating: 0
RE: 35W Dual Core
By stephenbrooks on 2/17/2006 4:35:16 PM , Rating: 2
That's true, but where does it say that in this article? There's no benchmarks :)


By RideFree on 2/21/2006 4:26:44 PM , Rating: 2
Kristopher, I didn't know you cared :-)




"I'd be pissed too, but you didn't have to go all Minority Report on his ass!" -- Jon Stewart on police raiding Gizmodo editor Jason Chen's home

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki