Time certainly does fly when you
are clipping along at 3GHz, doesn't it? It has been a
whole 5 months since
Intel
released their last processor speed bump. Intel
actually let this one ride a little bit it seems. In
early February, AMD let loose with their new "Barton" core
based
Athlon XP 3000+ (2.166GHz), in an effort to keep pace
with Intel's brute force approach to processing. Don't
misunderstand us, there certainly is nothing wrong with
"brute force", when it comes to computing. However,
efficiency is also the name of the game and while it is
impressive to see a 3GHz CPU like the Pentium 4, in mass
production, it's equally as impressive to see a processor
like the AMD Barton core, keep step with the P4, while
clocked almost 1GHz lower. The Pentium 4's super deep
"hyper pipelined" architecture, that allows it to hit such
impressive clock speeds, is also a source of some
inefficiency within the chip. It seems you can't
really have your cake and eat it too, because along with
this architecture comes the inherent latency associated with
a deeper pipeline. Regardless, like all things in
life, there is a balance; a balance between raw clock speed
and processing efficiency. Intel certainly seems to
have found that balance and like AMD, they are on a
never-ending quest to improve upon the existing
architecture.
Another way to enhance P4
performance, is by providing the CPU with more bandwidth via
associated bus interfaces. Simply cranking up CPU
clock cycles, will certainly bring you incremental
performance. However, giving the processor enough
bandwidth to get out to other system resources and
peripherals, is also critical for overall throughput.
Otherwise you'll have the equivalent of a jet fueled Funny
Car, sitting in neutral, red-lining the engine and going
nowhere fast. As such, we've witnessed both Intel and
AMD taking System Bus (or Front Side Bus) speeds up a notch
or two, along the way of their respective roadmaps. At
2.26 and 2.4GHz, Intel took the Pentium 4 to a 533MHz System
Bus, with a Quad Pumped 133MHz reference clock. This
time around, at 3GHz, Intel gives us an 800MHz System Bus,
with a Quad Pumped 200MHz reference clock. However, we
are reminded once again, of a lesson taught to us by our
Asian friends, that of "yin and yang", or balance. To
compliment this extra bus bandwidth Intel also brought forth
Dual Channel DDR400 memory support.
If this is beginning to sound
like a complete system controller overhaul, you would be
correct. Along with this new 800MHz System Bus capable
P4, comes Intel's i875P chipset, more commonly referred to
as "Canterwood". And while they were at it, Intel
brought many of the peripheral interfaces of their new
chipset, up to snuff as well. The following is a
HotHardware.Com overview and performance analysis, of
Intel's new Processor and Desktop Platform, based on their
new Pentium 4 3GHz with an 800MHz System Bus and the D875PBZ
- i875P driven motherboard.
 |
|
Specifications of the Pentium 4
3GHz Processor |
|
Hyperthreading and a
neat clean 800MHz System Bus |
|
- Clock Speed 3GHz
- 800MHz "Quad Pumped" Front Side Bus
- Hyperthreading Technology for increased performance
in Multi-tasking and Multi-threaded applications
- .13 micron manufacturing process
- 512K on chip, Full Speed L2 Cache
- Rapid Execution Engine - ALU clocked at 2X frequency
of core
- 128bit Floating Point/Multimedia unit
- "Hyper Pipelined" Technology for extremely high
clock speeds
- Intel "NetBurst" micro-architecture
- Supported by the Intel® i875P and i865G chipsets,
with Hyperthreading support
- Internet Streaming SIMD Extensions 2
- Intel® MMX? media enhancement technology
- Memory cacheability up to 4 GB of addressable memory
space and system memory scalability up to 64 GB of
physical memory
- Support for uni-processor designs
- 1.525V operating voltage range
Obviously this quick specification list is a
repeat for the most part, with one exception. Intel's
new 3GHz Pentium (note the even 3GHz number, rather than
3.06GHz), is now characterized to run on an 800MHz System
Bus with a Quad Pumped 200MHz reference clock and DDR400
Memory Bus. Who would have thought we would see an FSB
of 200MHz as a "stock" clock speed on any motherboard?
Ahhh, but technology does roll on, doesn't it? This
one enhancement alone should boost performance nicely,
especially in system bandwidth demanding applications.
|
|
|
Specifications of The i875P Canterwood
Chipset |
|
|
800MHz System Bus,
Dual Channel DDR400, "Native" SATA - You've got
to love it |
With the launch of Intel's new "Canterwood"
chipset, we're given a new Northbridge and a new Southbridge
to work with, along with all of the latest creature
comforts. What, no passenger side airbag?
PAT -
Performance Acceleration Technology:
Intel has optimized the performance of the
Canterwood chipset's memory controller and improved on
latency characteristics in two key areas, DRAM Chip Select
and Memory Controller Access.
There is a resultant 2 clock improvement for
Canterwood, which should help somewhat. Intel is
calling this "PAT", Performance Acceleration Technology.
Frankly this feels like a bit of Marketing spin. We
should also note that for some reason, although we could set
2, 2, 2, 5 timings, for the PC3500 DDR Memory we used in
testing at 400MHz in other boards, we could not get beyond
2, 3, 3, 5 timings with this new Intel D875PBZ board.
It seems as though Intel's new chipset is a little less
forgiving with respect to aggressive memory timings.
However, we're sure Memory OEMs will be targeting full
performance with this chipset, so it probably won't be long
before better memory is available, that can support
aggressive timings like we've seen with the i845PE.
CSA - Communications Streaming
Architecture:
Both Canterwood and the upcoming Springdale
chipset share a new Networking technology from Intel called
"CSA". CSA is essentially a direct link to the
Northbridge System Controller, versus utilizing the PCI bus.
This allows Intel's Pro/1000CT MAC/Phy controller direct
access to system memory, with higher overall bandwidth at
266MB/sec, twice that of PCI, as well as reduced CPU
utilization. The benefits are obvious.
-
Direct Access to System Memory
-
Prioritized Access to GbE traffic
-
1.5V Interface
-
Low Latency Design
-
266 MB/s dedicated to Networking I/O
-
Reduced CPU Utilization
-
Better management of simultaneous data
streams
-
Lower GbE component power
-
Bursts of network traffic handled with
ease
We had a relatively short window of time to
test this new Intel platform and as a result, we're not able
to provide you back-up factual analysis to support these
claims. However, we can report that during testing and
routine use on a 100Mb Ethernet Network, the Pro/1000CT
Controller performed admirably with excellent throughput and
reliability.
A Closer Look, System Setup and Sandra
|
