Overclock.net banner

AMD 65 nm launches

4K views 47 replies 24 participants last post by  Mootsfox 
#1 ·
Good morning from the distant europe, here's what I found: AMD finaly launched Brisbane 65nm.

AMD Athlon 64 X2 Brisbane Model
Core Frequency L2 Cache TDP Pricing
5000+ 2.6 GHz 2x512KB 65W $301
4800+ 2.5 GHz 2x512KB 65W $271
4400+ 2.3 GHz 2x512KB 65W $214
4000+ 2.1 GHz 2x512KB 65W $169
 
#2 ·
Quote:
Sunnyvale (CA) - AMD fired up the third stage of a series of announcements this morning and announced a first batch of 65 nm desktop dual-core processors. Almost one year behind Intel, the new X2 family CPUs launch as a quartet that do not bump processor performance but lower power consumption and provide headroom for future clock speed increases.

65nm processors aren't quite news anymore. Intel has launched the industry's first 65 nm microprocessor, the Pentium D 955 shortly after Christmas last year. Since then, the company not only has shipped Pentium D 900-series processors with "Presler" core, but also millions of Core Duos ("Yonah"), single-core Pentium 4s 600s and Celeron Ds ("Cedar Mill) as well as the Core 2/Xeon 5100 family as dual-core and quad-core 65 nm chips. The crossover from 90 nm to 65 nm was reached several months ago and there are very few Intel processors that are built in 90 nm today (current Intel processor price list, PDF format).

AMD now makes the jump to 65 nm as well, perhaps not with quite as much buzz as Intel did last year. While the new processors use a different core code-named "Brisbane," there's no entirely new processor family and not even a new flagship model. Instead, there are four new models that partially bring back recently dropped model numbers and decrease the power consumption to a thermal design power (TDP) of 65 watts.

That does not sound too impressive, as the company already has been offering some 65 watt 90 nm X2 processors since the introduction of the socket AM2 in May of this year. Especially the 65 nm X2 4400+ (2.3 GHz) and the X2 4000+ (2.1 GHz), which had been available as a 65 watt (90 nm) processors before, may not spark excitement. However, progress can be seen with the two higher-end 65 nm processors: The X2 5000+ and the reborn 4800+, both previously rated at a TDP of 89 watts (110 watts for socket 939), now max out at 65 watts.

Samples of the new 65 nm processors, which AMD says are available in volume today, apparently have not been sent out to the media yet. So far, sites such as Tom's Hardware have been unable to confirm AMD's power claims, which go as far as stating that the new processors consume less power than Intel's Core 2 Duos in certain scenarios. For example, AMD told us that 65 nm X2 processors consumes only 7.5 watt when idle, while a Core 2 Duo chip runs at about 14.3 watts.

AMD said that it expects to reach a crossover from 90 nm to 65 nm - the time when it produces more 65 nm than 90 nm processors - in Q1 of next year. A full conversion is scheduled for mid-2007. The phrase "full," however, has to be taken with a grain of salt, as the company conceded that it will have to "support existing designs," which means that a range of 90 nm processors will remain available at least until sometime in 2008. While the company did not specify which of these 90 nm processors will survive until then, sources told TG Daily that these products will include all X2 processors that are available today - the 5200+, 5000+, 4600+, 4200+, 3800+ as well as the soon to be launched 5400+ (2.8 GHz), 5600+ (2.8 GHz), 5800+ (3.0 GHz) and 6000+ (3.0 GHz). The 5200+ and 5400+ will also be offered as 65 nm processors with a TDP of 76 watts.

A completely new 65 nm processor core is expected to arrive in early Q3 2007, sources said. The "Kuma" dual-core processors are related to the "Agena" quad-core processors and will debut with clock speeds between 2.0 and 2.9 GHz and a TDP of 89 watts. Other 65 watt upgrades will include the "Lima" core for single-core Athlon 64 processors (45 watt TDP) as well as the "Sparta" core for Sempron CPUs - which will drop the power consumption AMD's entry-level CPUs from currently 62 watts to 35 watts.

AMD's 65 nm Brisbane Athlon 64 X2 processors are produced in the firm's Fab 36 in Dresden, Germany. Combined with the use of 300 mm wafers, AMD says that it is capable of increasing the output of processors with the transition to the new production process. A company representative declined to comment on the number of its 300 mm wafer starts and how many more processors it is able to produce now. However, he mentioned that would be a "natural conclusion" that the production volume has grown through this transition.

65 nm processors, which - according to AMD - are shipping now, are available for the same price as their 90 nm sister models. The 4000+ models carries a tray price of $169, the 4400+ is available for $214, the 4800+ for $271 and the 5000+ for $301.
http://www.tgdaily.com/2006/12/05/amd_intros_65nm_cpus/
http://www.amdcompare.com/us-en/desk...=ADO4000IAA5DD
 
#5 ·
Quote:

Originally Posted by HrnyGoat View Post
They all have a 65W TDP? I would have expected better, especially if AMD has 35W 3800 X2s.
Those were cherry picked and were in -extremely- limited supply, I don't think I've ever actually seen one in stock or even heard of one in stock o_O

They could release lower TDP chips once they get enough or they might not even bother this time around, 65w is pretty low already
 
#8 ·
Quote:

Originally Posted by boredom View Post
Yup and it will stay like that even into K8L

K8 doesn't "need" massive amounts of cache, so it can get away with only 2x512kb
I had side by side two rigs, one with a 1.8GHz SC 512k L2 cache Athlon 64 and one at the same 1.8GHz sempron64 with 128k L2 cache. I'll have to admit that the performance difference is small, almost unnoticeable, but... I had this feeling that the sempron is a bit more sluggish, not as smooth as the athlon.
With the low amount of L2 cache is more like a bumpy ride, fast but not smooth.
 
#9 ·
128KB cache is massive. Not in responsiveness...but in SuperPI scores!


Example:
128kb cache 1.8ghz = ~ 54s
256kb cache 1.8ghz = ~49s
512kb cache 1.8ghz = ~46s
1024kb cache 1.8ghz = ~43s

See! You're losing nearly 15% efficiency!
 
#11 ·
Quote:

Originally Posted by CWell1337 View Post
That def. has me a little sour this morning. Come on, I love my 2MB L2 cache... are they going to release Opterons at 65nm anytime goon does anyone know?
Probably in Q1 '07... But there will also be native quad-core 65nm Optys.
 
#12 ·
:\\ Only 512mb of cache per core- Not cool. I would have liked at least 1MB per core.

Dunno if anyone is gonna buy these, seems like there isn't enough onboard cache.
 
#13 ·
Quote:


Originally Posted by Burn
View Post

:\\ Only 512mb of cache per core- Not cool. I would have liked at least 1MB per core.

Dunno if anyone is gonna buy these, seems like there isn't enough onboard cache.

People buy Manchesters and 512kb Windsors, don't they?
 
#14 ·
Quote:


Originally Posted by Burn
View Post

:\\ Only 512mb of cache per core- Not cool. I would have liked at least 1MB per core.

Dunno if anyone is gonna buy these, seems like there isn't enough onboard cache.

The cache isn't a very limiting factor in my opinion...
 
#15 ·
I dunno- I am of the opinion that more cache is better. Considering the C2D packs 4MB into its core, I think AMD really needs to begin stepping up to the plate and offering more high-cache solutions. L2 cache is really nice because it's super-fast, much faster than RAM will be, so more "stuff" is available for the processor to process, which leads to better performance.

See? And 512k really shouldn't be offered anymore- San Diago [/Ron Burgundy] offered 1MB, so why shouldn't Brisbane?
 
#16 ·
Quote:


Originally Posted by Burn
View Post

I dunno- I am of the opinion that more cache is better. Considering the C2D packs 4MB into its core, I think AMD really needs to begin stepping up to the plate and offering more high-cache solutions. L2 cache is really nice because it's super-fast, much faster than RAM will be, so more "stuff" is available for the processor to process, which leads to better performance.

See? And 512k really shouldn't be offered anymore- San Diago [/Ron Burgundy] offered 1MB, so why shouldn't Brisbane?

Darn it, I wanted to make an Anchorman reference!


I'm not saying that I'm glad AMD's gotten rid of higher-cache chips - just that people will still buy them.
 
#17 ·
I for one will more than likely wait until native quad-core comes out for AMD, then compare vs. the QX66600 and figure out which is the better buy. No 4x4 for me, Intel quad-core already kicks its butt.

I guess you like the Anchorman reference
 
#18 ·
Initial results for the native quad are promising, but then again, so were initial 4x4 results. =/

I'm in love with that movie. Massive amounts of comic genius stuffed into 2 hours.
 
#19 ·
Quote:


Originally Posted by Burn
View Post

I dunno- I am of the opinion that more cache is better. Considering the C2D packs 4MB into its core, I think AMD really needs to begin stepping up to the plate and offering more high-cache solutions. L2 cache is really nice because it's super-fast, much faster than RAM will be, so more "stuff" is available for the processor to process, which leads to better performance.

See? And 512k really shouldn't be offered anymore- San Diago [/Ron Burgundy] offered 1MB, so why shouldn't Brisbane?

I believe it was to lower manufactoring costs or something to that effect.

Dunno.
 
#20 ·
Quote:


Originally Posted by Dockery
View Post

I believe it was to lower manufactoring costs or something to that effect.

Dunno.

AMD nearly DOUBLED their income since the aquisition of ATI. I think they can afford an extra $5 per chip.
 
#21 ·
Quote:


Originally Posted by Namrac
View Post

AMD nearly DOUBLED their income since the aquisition of ATI. I think they can afford an extra $5 per chip.

Hey, that's what they said, not me!


To be honest, I hope they get this new ZRAM going with 2MB of cache for every core!
 
#23 ·
On the AM2 platform cache amounts to nearly no performance boost. Maybe 2% difference.

But heres something..

Quote:


For example, AMD told us that 65 nm X2 processors consumes only 7.5 watt when idle, while a Core 2 Duo chip runs at about 14.3 watts.

This is also AMD's first 65nm chip, and 65W is the base power consumption, instead of the 89W of 90nm. As 65nm ramps up in production more so im sure 35W variants and below will become available.

Edit:
Quote:


5400+ (2.8 GHz), 5600+ (2.8 GHz), 5800+ (3.0 GHz) and 6000+ (3.0 GHz). The 5200+ and 5400+ will also be offered as 65 nm processors with a TDP of 76 watts.

Not quite as low as the 65W Core 2 Duo or the slower AMD chips, but Thats a helluva lot lower than the current FX-7#'s. If they could implement them in 4x4 enough it might actually be worth it for the money... But that probably wont happen for months, if at all.
 
#25 ·
There is no reason for AMD to put more L2 cache when it doesn't really affect performance, more cache is better when talking about the C2D, for K8 it doesn't make as much of a difference, so I don't see why some people are so hung up on the 512kb part o_O

Yorkfield will have 12mb of shared L2, K8L will have 4x512kb... and you know why K8L will only 512kb? because it doesn't "need" more (surprise!)

 
#26 ·
Quote:


Originally Posted by Jori
View Post

On the AM2 platform cache amounts to nearly no performance boost. Maybe 2% difference.

Not quite as low as the 65W Core 2 Duo or the slower AMD chips, but Thats a helluva lot lower than the current FX-7#'s. If they could implement them in 4x4 enough it might actually be worth it for the money... But that probably wont happen for months, if at all.

Saying that AMD doesn't need more than 512kb of cache is ignorant. That's like saying that since I have a 5ghz Conroe, I dont need more than 512MB of RAM. The CPU physically cannot hold items over 512kb in the cache of an AMD CPU. AMD really needs to quit cheaping out and make 2x1MB L2 standard on all of their chips. Who wants to spend their money on a crippled chip?

So basically, the new AMD chips still put out more heat than a Conroe. I seem to remember people making fun of Intel 65nm tech saying that AMD 90nm still runs cooler. It looks like the tables have turned.

I still don't understand why AMD is wasting their time with 4x4. Why would anybody go out and spend $1000 or more on a pair of CPUs when they can buy a QX6700 for $1100 which runs cooler, performs better, uses less power, and is quad core? It just doesn't make any sense to me... People act like 4x4 is innovative. How? Intel already has something that is much better. 2x Clovertown anyone? 8 cores.

EDIT

Adding cache to the AMDs doesnt effect performance? So what happened to them saying that the added 512KB of L2 added about 200mhz of performance? If what you guys are saying is true, the Venice and the San Diego should perform the same. Think about it.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top