questions about "hypertransport" and FSB in AMD 64

[nikko]

I have an AMD 64 3200+ running in an Asus K8V mobo. I ran a PCMark04 benchmark on it and got a score of 4168. Running the same benchmark on my work computer, a Pentium 3.0 Ghz I got a score of 4353. I'm not too concerned with scores, but I noticed something in the details that caught my eye.

In the report for the P4, PCMark says:

CPU Intel Pentium 4 2992 MHz

Physical CPUs 1

HyperThreading Available - 2 Logical Processors

FSB 800 MHz




For the AMD 64 3200+ it says:

CPU AMD Athlon(tm) 64 2203 MHz

Physical CPUs 1

HyperThreading Not Available

FSB 200 MHz



Now shouldn't the AMD chip be running at 800 MHZ on the FSB as well? And I know AMD doesn't have hyperthreading, but they have hypertransport. Isn't that the same thing? Doesn't PCMark recognize that? I'm just wondering if there is something I need to turn or enable to get this right, or if it *is* right but looks weird in PCMark. Any help would be welcome. Thanks.

 

[Schadenfroh]

differant archectectures, the Athlon64 has a on-die memory controller, which changes the whole compairson of the two up a bit

Text

And I know AMD doesn't have hyperthreading, but they have hypertransport. Isn't that the same thing?

no

 

[JackHawksmoor]

Now shouldn't the AMD chip be running at 800 MHZ on the FSB as well?

The front side bus is the connection between the CPU and the memory controller, which runs at 800mhz (really 200mhz x 4) on the newest Pentium 4s. Since the memory controller is PART of the Athlon 64, there isn?t really a FSB, or at least it?s part of the CPU, and running at the same clock speed as the CPU.

So whatever number that program is getting, it?s not really measuring the FSB on the Athlon 64. Maybe it?s measuring the real clock speed of the connection between the chipset and CPU?

And I know AMD doesn't have hyperthreading, but they have hypertransport. Isn't that the same thing?

They?re not related at all. Hyperthreading is a way for a single CPU to appear to be dual CPUs to the operating system, which then schedules two threads at once to run on what?s really a single CPU. That can help the Pentium 4 keep more parts of the CPU busy at once. Supposedly adding it took as much engineering as designing the rest of the CPU, and it?s been improved twice since it was first introduced. The Athlon 64 doesn?t have it at all, but it probably wouldn?t make nearly as big of a difference. The Pentium 4 has a super long pipeline-up to around 30 stages now, versus only 12 on the Athlon 64-hyperthreading probably just helps make up for some of the lost performance by having such a long pipeline.

Hypertransport is the name of a (I guess actual standard) connection. In the Athlon 64?s case, it connects the CPU to the chipset, and in dual CPU systems (Opertons), each CPU also has a separate hypertransport connection to the other CPU.

Doesn't PCMark recognize that? I'm just wondering if there is something I need to turn or enable to get this right, or if it *is* right but looks weird in PCMark. Any help would be welcome. Thanks.

It probably just doesn?t recognize it right. I wouldn?t worry about it, unless Office runs slow on your Athlon or something Whatever dumb score that benchmark comes up with, the Athlon 64 is actually supposed to run office stuff better than the Pentium 4, probably because of it?s integrated memory controller and short pipeline (it?s not as effected by branching code as the P4 is).

 

[NightCrawler]

200 mhz is the clock generator for the multipliers.

The memory controller is 400 mhz bi-directional or 3.2GB/s for single channel or 6.4GB/s dual channel.