POPPUR爱换

 找回密码
 注册

QQ登录

只需一步,快速开始

手机号码,快捷登录

搜索
查看: 5898|回复: 72
打印 上一主题 下一主题

k8l同频性能提升20%?

[复制链接]
跳转到指定楼层
1#
发表于 2007-2-4 19:06 | 只看该作者 回帖奖励 |倒序浏览 |阅读模式
The representatives OF AMD today already attempted to form a public idea about the speed level of servers processors barcelona with four core. Their announcement will take place in the second quarter of the present year; however, more precise periods were not called, until now.

To correct situation decided associates, which along their own channels obtained information about the fact that AMD soon will start the preparation for starting the mass production of processors barcelona. Associates assume that already at the end of present month will start the following phase for the preparation of processors generation K8L to the mass production, and approximately during May these processors will find all signs of product series. If we add one month for the release level optimization of suitable chips, then June will become the month of the anoncement for corresponding opteron processors.

The related desktop processors generation K8L with four core will appear already in the third quarter 2007, then will have the cache in the third level dual core processors kuma. The dual cores processors without cache in the third level will be released also in the second half-year. Lastly will be released the single core processors spica, which will become base for family sempron. In the first quarter of 2008 the fraction of processors with architecture K8L in the assortment of AMD production will compose 60%. According to the preliminary estimations, the processors kuma will be up to 20% faster than present Athlon 64 X2 working at comparable frequencies.




原文
http://xtreview.com/addcomment-id-1381-view-amd-K8L-release-date.html



估计这个结论来自更早前的消息:

The representatives OF AMD already attempted to tell analysts and investors by statements about increase in the speed to 40-60% with the transfer to four core processor generation K8L; however, these forecasts related predominantly to the server processors with four core. Processors with two cores will be proposed to the majority of desktop systems in the following year, and supporters AMD would for sure be desirable to know, what increase in the speed will obtain the processors athlon 64 X2 after the passage to the architecture K8L.

For the beginning, let us check the assortment of dual core from this architecture. The processors athlon 64 X2 of new generation will exist in two varieties: with the divided cache in the third level (Kuma, previously Anatares) and without it (Rana, previously Arcturus). Both families processors will have socket AM2+ and allotted by 2 X 512 KB cache in the second level. Increase in the speed must ensure architectural changes, and also introduction of support DDR 2-1066 and expansions SSE4A. However, the influence of last factor will depend on the degree of the concrete optimization  under applications with this collection of commands.

The representatives OF AMD appeared in Japan before the local journalists. for Japanese associates it was possible to entice from AMD representatives  very important commentaries. In particular, it became known that  by the divided cache of the third level the processors athlon 64 X2 (Kuma) will work on the average on 20% faster than their predecessors generation K8. In the context of this statement it is complicated to understand that an  increase in the speed in 20% is ensured only by cache of the third level.
According to our estimations, the divided cache of the third level cannot exert a substantial influence on the speed of the processors generation K8L. In the case with the four core processors it will ensure increase in order 5% on the average. Dual core processors from its appearance will win less. the fact that the cache of the third level of shared into four core processor work as exchange of data between cores . The cache of the third level has sufficiently high delays. Increase performance from the cache of the third level will approach 20% only in those tasks, which actively work with memory using several cores simultaneously. However, the average value of increase will be modest, especially in the case with the dual core processors.

Thus, the named number “20%” has sense in the total increase of performance for dual core processors, ensured by all architectural innovations K8L in totality. will be these 20% sufficiently in order to fight with processors core 2 Duo from Intel? The answer will give the first tests, and also the data about the frequency range of processors K8L and their retail cost.


http://xtreview.com/addcomment-id-1134-view-dual-core-k8l-upto-20%25-faster.html

这个消息有待确认,当个笑话看吧(_(
但原文也说了,不能确定20%是k8l的提升幅度还是L3带来的提升幅度。
不过其也分析了靠L3平均提升20%不现实。


呵呵,20%的话还比较现实:unsure: ,不过这要是真的的话A可糗大了,前些日子还领先40%呢w00t)
2#
发表于 2007-2-4 19:09 | 只看该作者
沙发:) :)
回复 支持 反对

使用道具 举报

3#
发表于 2007-2-4 19:16 | 只看该作者
:lol: 看不懂英文的说……
回复 支持 反对

使用道具 举报

wangchunboss 该用户已被删除
4#
发表于 2007-2-4 19:22 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

5#
发表于 2007-2-4 19:31 | 只看该作者
对:mad: :(:mad:
回复 支持 反对

使用道具 举报

6#
发表于 2007-2-4 19:33 | 只看该作者
:loveliness:
想起ps3了

阴特儿还有有一个法宝没用了
那就是内置内存控制器
回复 支持 反对

使用道具 举报

G70 该用户已被删除
7#
发表于 2007-2-4 19:43 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

monsterrr 该用户已被删除
8#
发表于 2007-2-4 20:14 | 只看该作者
提示: 作者被禁止或删除 内容自动屏蔽
回复 支持 反对

使用道具 举报

9#
发表于 2007-2-4 20:18 | 只看该作者
:funk: :funk:
回复 支持 反对

使用道具 举报

10#
发表于 2007-2-4 20:55 | 只看该作者
20%比较现实,4核K8L貌似可以和4核肉一较高下了,AMD最大的问题还是工艺,如果能达到Intel的水平,K8L会比4核肉核心面积小很多,那样的结果就不用多说了.

可惜啊......
回复 支持 反对

使用道具 举报

11#
发表于 2007-2-4 20:58 | 只看该作者
原帖由 gzmagic 于 2007-2-4 19:33 发表
:loveliness:
想起ps3了

阴特儿还有有一个法宝没用了
那就是内置内存控制器



MC in die??呵呵,Intel上了那玩意频率也很难提高了.

AMD之所以上内置内存控制器是因为L2技术不如Intel,是成本,性能权衡的结果, 那玩意是目前影响K8频率提高的罪魁祸首.
回复 支持 反对

使用道具 举报

12#
发表于 2007-2-4 21:39 | 只看该作者
the processors kuma will be up to 20% faster than present Athlon 64 X2 working at comparable frequencies
up to:最大20%的性能提升~
要知道同频率Conroe的领先至少是平均20%的~
而且2.5GHz的频率更加决定了其命运

怪不得Intel既不急着出原生四核心,又不急着推广45nm高频~
原来是胜券在握(这个对于pro-I来说也不是什么好事)
真是很难想像AMD在07~08年还有什么花样,除了降价

09年的Fusion也就无非尔尔,之前我说的Intel的封装大法也被验证了,没准在08年就能出来~
真是为某社担心
回复 支持 反对

使用道具 举报

13#
发表于 2007-2-4 21:41 | 只看该作者
原帖由 HardCoded 于 2007-2-4 20:55 发表
20%比较现实,4核K8L貌似可以和4核肉一较高下了,AMD最大的问题还是工艺,如果能达到Intel的水平,K8L会比4核肉核心面积小很多,那样的结果就不用多说了.

可惜啊......


工艺水平?
如果AMD的工艺提高了,干爹还能像这样支持?
如果干爹不支持,AMD的工艺还能长进?
可见这就是平衡的结果,这也决定了AMD的工艺必然会比上不足,比下有余
回复 支持 反对

使用道具 举报

14#
发表于 2007-2-4 21:43 | 只看该作者
原帖由 HardCoded 于 2007-2-4 20:58 发表

MC in die??呵呵,Intel上了那玩意频率也很难提高了.

AMD之所以上内置内存控制器是因为L2技术不如Intel,是成本,性能权衡的结果, 那玩意是目前影响K8频率提高的罪魁祸首.


谁说集成内存控制器就不能提高频率?
power6照样是集成内存控制器的,频率比Conroe还高
难道您以为内存控制器和核心跑一样的频率啊?
回复 支持 反对

使用道具 举报

15#
 楼主| 发表于 2007-2-4 22:52 | 只看该作者
平均20%的提升,估计2MB L3是绝对提供不了的。


http://www.xbitlabs.com/articles ... hlon64-fx51_15.html


这个测的是2MB L3 的P4XE 3.2G 和普通北木3.2G,大概看了一下,除了一个项目提升到了16%外,其他几十个项目,P4XE相较P4的提升都在10%以下。

而k8l由于集成MC和A的缓存水平估计提升可能更小。:unsure:


提升20%很不错了,yonah=>Conroe也就提升30%:wacko:

不过,估计20%离个别人期盼的秒杀扣肉还很远,本来2.9G的话和c2q6700还有的一比,可是这两天又说降频成了2.6G,估计I当初“扣肉需要AMD追1年”的预言是实现不了了(_(


肥皂泡早点破了也好,免得发布时才知道打击太大,影响安定团结的大好局面:lol:

[ 本帖最后由 the_god_of_pig 于 2007-2-4 22:57 编辑 ]
回复 支持 反对

使用道具 举报

16#
发表于 2007-2-4 23:11 | 只看该作者
原帖由 the_god_of_pig 于 2007-2-4 22:52 发表
平均20%的提升,估计2MB L3是绝对提供不了的。


http://www.xbitlabs.com/articles ... hlon64-fx51_15.html


这个测的是2MB L3 的P4XE 3.2G 和普通北木3.2G,大概看了一下,除了一个项目提升 ...

增加多核下的共享cache和加大单核的cache效果是不一样的,而且K8L的改进又不是仅仅只有L3。
回复 支持 反对

使用道具 举报

17#
发表于 2007-2-4 23:23 | 只看该作者
原帖由 itany 于 2007-2-4 21:39 发表
the processors kuma will be up to 20% faster than present Athlon 64 X2 working at comparable frequencies
up to:最大20%的性能提升~
要知道同频率Conroe的领先至少是平均20%的~
而且2.5GHz的频率 ...

简单告诉你.双方的情况和你想象的正好相反.:devil:
回复 支持 反对

使用道具 举报

18#
发表于 2007-2-4 23:28 | 只看该作者
客观地说,整数做掉Conroe可能性不大,但是浮点性能做掉Conroe已经是意料中的事情了。
回复 支持 反对

使用道具 举报

19#
发表于 2007-2-4 23:33 | 只看该作者
原帖由 the_god_of_pig 于 2007-2-4 19:06 发表




原文
http://xtreview.com/addcomment-i ... L-release-date.html



估计这个结论来自更早前的消息:



http://xtreview.com/addcomment-id-1134-view-dual-core-k8l-upto-20 ...

单核平手的话,4核领先40%是很保守的估计:devil:
回复 支持 反对

使用道具 举报

20#
发表于 2007-2-4 23:41 | 只看该作者
原帖由 嘉蓝 于 2007-2-4 23:23 发表

简单告诉你.双方的情况和你想象的正好相反.:devil:

没错,AMD的工艺水平远胜intel,连IBM都不是AMD的对手。K8L的核心面积只有150mm^2,频率起码是4G。

[ 本帖最后由 acqwer 于 2007-2-4 23:43 编辑 ]
回复 支持 反对

使用道具 举报

您需要登录后才可以回帖 登录 | 注册

本版积分规则

广告投放或合作|网站地图|处罚通告|

GMT+8, 2025-1-28 00:46

Powered by Discuz! X3.4

© 2001-2017 POPPUR.

快速回复 返回顶部 返回列表