lol... AMD: Intel\'s quad-core parts could consume 20 to 50 percent more power

Edison

AMD, Intel square off in quad core


Rick Merritt   
EE Times
(09/25/2006 4:39 H EDT)  


SAN JOSE, Calif. — Advanced Micro Devices and Intel Corp. will square off over quad-core processors this week as each side tries to position itself as leading the race for multicore x86 processors.

Intel is expected to demostrate at the Intel Developer Forum Tuesday (Sept. 26) its Clovertown and Kentsfield server and desktop processors which essentially pack two existing dual-core server or desktop chips inside a multichip module. Intel is expected to demo working versions of the chips and share some information on the performance, power and availability of the CPU.

Tom's Hardware Web site described Kentsfield as a 2.6-GHz chip with a 1.3-GHz front side bus. It has one 4-Mbyte shared L2 cache per die and four 32-Kbyte L1 caches per core. The chip was found to be in a B revision and fifth stepping.

By contrast, AMD will ship quad-core Opteron server CPUs in mid-2007 that link four cores on a single die. That reduces latency Intel's parts will exhibit when a core on one die has to go through an off-chip memory controller to talk to a core in the other die, said John Fruehe, AMD's business development manager for Opteron.

AMD uses the cache coherent HyperTransport bus integrated on its CPU cores to link each core on a single die.

Unlike Intel, AMD will not be ready to demostrate its quad-core CPU or deliver performance data on it until the end of the year. "We taped out the processor, I believe, in August," Fruehe said.

AMD will have quad-core products that fit into the same power dissipation specs as its existing dual-core CPUs—120W, 95W and 68W. Intel's quad-core parts could consume 20 to 50 percent more power, Fruehe said.

An Intel spokesman said the company's quad-core CPUs will not consumer more power than its Pentium D Extreme Edition, the last of the line of high-end, single-core desktop CPUs. Like AMD, Intel offers its processors at several power consumption levels.

BOSS

没什么意义,因为工艺上差一代,实际上能耗差不多

Asuka

2个125W的火炉放在一起，也好意思说别人

ly7707

原帖由 BOSS 于 2006-9-26 10:46 发表
没什么意义,因为工艺上差一代,实际上能耗差不多

AMD的.65也不咋低

BOSS

amd 65nm和intel 65nm能耗基本同一级别.

FENG950

考虑晶体管数量的话，也叫同一级？

BOSS

谁会去数晶体管数量?是不是合算些啊:lol:

FENG950

使用时没人会数，不过考虑能耗控制的技术含量这是不可忽略的东西。否则老家伙们个个不都是节能先锋了？:wacko:386用老工艺制造也比Conroe省电，Conroe和它都差了N个级别了。

[ 本帖最后由 FENG950 于 2006-9-26 11:54 编辑 ]

Prescott

不知道4x4功耗比Kentfield高多少。

kknd1967

够过冬了

原帖由 Prescott 于 2006-9-26 12:47 发表
不知道4x4功耗比Kentfield高多少。

itany

现在AMD的四核心连个影子都没有，还好意思说功耗。干脆功耗也使用PR值（Energy Efficiency Referance，EER？）算了。
  现在Intel的四核心比双核芯只增大了20-50%，已经很好了。况且四核心照样可以做到85W以下，Q6600就是。

popwangyuII

A记爱好者会说：

65nm的残次品先出，不然先把好的拿出来卖旧的90nm库存怎么卖？

popwangyuII

原帖由 itany 于 2006-9-26 17:52 发表
现在AMD的四核心连个影子都没有，还好意思说功耗。干脆功耗也使用PR值（Energy Efficiency Referance，EER？）算了。
  现在Intel的四核心比双核芯只增大了20-50%，已经很好了。况且四核心照样可以做到85W以下， ...

:lol: 谁告诉你没有影子啦，图纸多呢。

Edison

原帖由 popwangyuII 于 2006-9-27 16:03 发表
A记爱好者会说：
65nm的残次品先出，不然先把好的拿出来卖旧的90nm库存怎么卖？

就好像Intel的Yonah。

fayerlxy

clovertown的TDP不是出来了么，2.66G 100W, 2.66G以下80W
1.6G低压版才50W。

我倒要看看K8L会如何.

itany

先前说四核心控制在85W以下很容易，事实证明了吧！