Logo   Banner   TopRight
TopUnder
Transparent
Pentium 4 670 3.8GHz Performance Profile
Transparent
Date: May 26, 2005
Section:Processors
Author: Dave Altavilla
Transparent
Introduction and Specifications

 

Though clock speed is no longer a true measure of processor performance, it's hard to argue with a few hundred more MHz no matter which CPU architecture you'd consider.  Regardless, it's clear the industry as a whole, as well as the world's largest chip manufacturer, Intel, has now embarked on an effort to provide more computing power per clock cycle, focusing on IPC (Instructions Per Clock Cycles) metrics as better manifestations of overall performance.  As the PC and workstation marketplaces now look toward a future of dual-core processors, Intel continues a parallel effort of fleshing out its 6XX Sequence series of processors today with the introduction of the Pentium 4 670.

The Intel Pentium 4 670 processor is the result of a simple speed bump, one that takes the company's latest 2MB cache-infused Prescott core to 3.8GHz.  Historically we've seen plenty of headroom in the Prescott 2M core with much more manageable thermals, so it's no surprise that Intel has decided to add another high-end SKU to the mix.  In this showcase, we're focusing on strictly the performance characteristics that comprise the new 3.8GHz Pentium 4 670, so we'll keep the chatter to a minimum and cover some brief spec details before we move into our standard battery of tests.

Specifications: Intel Pentium 4 670 3.8GHz
Intel's High-End Mainstream Performance CPU

Intel Pentium 4 670 - 3.8GHz Processor

  

  • 3.8GHz Single Core Processor With Hyper-Threading
  • 800MHz "Quad Pumped" FSB
  • .09-Micron Manufacturing Process
  • 2MB On-Chip, Full-Speed L2 Cache
  • 16KB L1 Data Cache
  • Intel EM64T Extensions - 64-bit Computing
  • Enhanced Intel SpeedStep Power Savings technology "EIST"
  • Execute Disable Bit - Enhanced Security
  • Streaming SIMD Extensions - SSE2, SSE3
  • Supported Chipsets:
    i955X Express
    i945G Express
    i945P Express

    i925X
    i915
  • 1.25 - 1.4v Operating Range
  • LGA775 (Land Grid Array) Packaging
  • 115 Watts Thermal Design Power

With an 800MHz Quad Data Rate frontside bus and a 3.8GHz total clock cycle, the only thing that differentiates the Pentium 4 670, other than its performance, from the Pentium 4 660 is its 19X multiplier.  Other than that, the chip is identical in every way, from process technology to its core architecture and LGA775 packaging.  The nice thing about the Pentium 4 6XX Sequence processors are that they all come equipped with 2MB of full-speed L2 cache.  So although these are single-core CPUs that are only capable of multi-thread processing through Intel's "Hyper-Threading" technology, that core is beefed up with additional low latency on chip cache to assist in keeping its pipeline efficiently full.

Transparent
Test System and Vital Signs

 

Here we've detailed our test system configurations for the various platforms and processors that we have test results for in the following pages.

How we configured our test systems: When configuring the test systems for this showcase, we first entered system BIOSes and set each board to its "Optimized" or "High-Performance Defaults."  The hard drives were then formatted, and Windows XP Professional (SP2) was installed. When the Windows installation was complete, we installed the rest of the necessary drivers and removed Windows Messenger from the system.  Auto-Updating and System Restore were disabled, and we set up a 768MB permanent page file on the same partition as the Windows installation. Lastly, we set Windows XP's Visual Effects to "best performance," installed all of our benchmarking software, defragged the hard drives, and ran all of the tests.

Test System Specifications
Intel & AMD Death Match
SYSTEM 1:
AMD Athlon 4000+
(2.4GHz)
AMD Athlon 64 FX-55
(2.6GHz)
AMD Athlon 64 X2 4800+
(2.4GHz)
(Dual Core CPU)


ASUS A8N-SLI Deluxe
(NVIDIA nForce 4 SLI)

2x512MB Corsair PC3200 Pro
CL 2-2-2-5

GeForce 6800 GT
On-board Ethernet
On-board Audio

WD "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
nForce 4 Drivers v6.53
NVIDIA Forceware v71.89
DirectX 9.0c
SYSTEM 2:
Pentium 4 670 - 3.8GHz
Pentium 4 Extreme Edition
(3.73GHz)
Pentium Extreme Edition 840
(3.2GHz)
(Dual Core CPU)


ASUS P5ND2-SLI
(nForce 4 SLI Intel Edition)

2x512MB Corsair DDR2-667
CL 3-2-2-7

GeForce 6800 GT
On-board Ethernet
On-board Audio

WD "Raptor" HD
10,000 RPM SATA

Windows XP Pro SP2
nForce 4 Drivers v7.02
NVIDIA Forceware v71.89
DirectX 9.0c
Pentium 4 670 Vital Signs
CPU ID Readings and Thermals

A quick check with CPU-Z shows the Pentium 4 670 as expected.  We see a full 2MB of Level 2 cache accompanied  by 16KB L1 data cache.  Its core is set to operate with an 800MHz frontside bus (200MHz QDR) and a 19X multiplier.  In terms of instruction set support, the P4 670 supports MMX extensions up through SSE3 and of course x86-64 64-bit extensions for true 64-bit computation.  The testing scope of our article doesn't cover 64-bit performance, however.  If you are interested in 64-bit performance metrics for Pentium 4 6XX sequence processors, have a quick look at our launch article from February.  As it stands today, there isn't much use in covering 64-bit performance simply because there are so few mainstream applications written currently to take advantage of 64-bit architectures.

 

Next we took a look at the new Pentium 4 670's thermal profile, on an open air bench test setup utilizing a stock Intel heatsink with an ambient room temperature of about 68o F.

At stock speed and sitting idle, the Pentium 4 670 operates at a tepid 41o C baseline.  The thermal profile of the new P4 6XX sequence CPUs are a big improvement since the early days of the Prescott core.  

In our full load test, we've fired up both Prime 95 and a Folding at Home client instance in the background, which effectively pegs the processor core at 100% utilization.  After about 15 minutes of this stress test, we saw the new P4 670 stabilize at 60oC, which again is a solid improvement, since we've historically seen temps in the mid to high 60s even from lower clock speed Prescott cores.

 

 

Transparent
Overclocking and Sandra Quick Take

 

As we've come to expect from previous overclock outings with new Pentium 4 6XX CPUs, overclocking this new 3.8GHz-rated core proved to be both successful and exciting.  Our overclocking efforts were centered around adjusting the FSB of the chip higher, while keeping the core multiplier at its stock 19X setting.  We feel this is more indicative of an end-user experience because altering multipliers can get dicey, depending on the motherboard chosen.  Having said this, end users do have the option to break down the multiplier of the 670 to a lower multple and raise the FSB even higher than we did in the following experiment.

Intel Pentium 4 670 Overclocking Experiment
Simply faster at higher FSB speeds

    
Pentium 4 670 3.8GHz - Overclocked at 4.35GHz - Stock Air Cooled

Once again we were thoroughly impressed and dare we say "amazed" at the available headroom in this new Pentium 4 core.  We were able to realize a 917MHz FSB with a stable overall clock speed of 4.35GHz in our test motherboard.  Again this was performed at stock voltage, with a stock Intel cooler on an open air test bench.  This overclock appeared stable for about half an hour of heavy load under Prime95 testing, and temperatures at the core didn't exceed 63oC.  However, we feel it's important to underscore that processors, test systems, and conditions can vary greatly, so these may or may not be indicative of what you could achieve on your own.

Preliminary Benchmarks with SiSoft SANDRA 2005
Synthetic Testing

We began our testing with SiSoftware's SANDRA, the System ANalyzer, Diagnostic and Reporting Assistant. SANDRA consists of a set of information and diagnostic utilities that can provide a host of useful information about your hardware and operating system. We ran four of the built-in subsystem tests that partially comprise the SANDRA 2005 suite (CPU, Multimedia, Cache, and Memory) with the Pentium 4 670 installed into our test rig.  All of the scores reported below were taken with the processor and memory running at their respective default clock speeds of 3.8GHz and 667MHz CAS 3, 2, 2, 7 DDR2 Corsair Memory.



CPU Arithmetic Benchmark
Pentium 4 670 (3.8GHz)


Multimedia Benchmark
Pentium 4 670 (3.8GHz)


Memory Benchmark
Pentium 4 670
- DDR2-667

Sandra's CPU tests report the new 3.8GHz Pentium 4 670 high among its single-core peer group, with performance at the top of its class versus the reference numbers.  The same could be said for the results obtained in the Multimedia test, as well.  Finally, the SANDRA memory test shows solid performance, but frankly, we've gotten a bit spoiled with the type of bandwidth you can obtain from a system running a 1066MHz FSB.  Although we see the memory bandwidth scores here break 5100MB/sec, a 1066MHz FSB-based P4 would put up numbers well into the 6K range with the help of 667MHz DDR2 memory.

Transparent
PCMark 04 Tests

 

We'll continue on briefly with more synthetic benchmark testing, which on its own isn't a true measure of real-world performance. However, it does complement a well-rounded benchmark perspective of overall performance within a given system architecture.  PCMark 04 is next.

Futuremark PCMark04
More Synthetic CPU and Memory Benchmarks

"The CPU test suite is a collection of tests that are run to isolate the performance of the CPU. There are nine tests in all. Two pairs of tests are run multithreaded - each test in the pair is run in its own thread.  The remaining five tests are run single threaded. These tests include such functions as file encryption, decryption, compression and decompression, grammar check, audio conversion, WMV and DivX video compression."

With certain parts of PCMark 04's CPU test designed to take advantage of multithreaded computing resources, the two dual-core CPUs in this synthetic test show the potential that lies within for multi-core architectures.  The Athlon 64 X2 4800+ takes the lead by a sizeable margin, however, besting the fastest Intel Pentium 840 by over 10%. Additionally, the slightly higher clock speed of the new Pentium 4 670 edges it just ahead of the 3.73GHz Extreme Edition P4, allowing it to place third in this high-end processor horse race.


 

The following are details of the types of tests run by PCMark04's Memory Test Module.


"The Memory test suite is a collection of tests that isolate the performance of the memory subsystem. The memory subsystem consists of various devices on the PC. This includes the main memory, the CPU internal cache (known as the L1 cache) and the external cache (known as the L2 cache). As it is difficult to find applications that only stress the memory, we explicitly developed a set of tests geared for this purpose. The tests are written in C++ and assembly. They include: Reading data blocks from memory, Writing data blocks to memory performing copy operations on data blocks, random access to data items and latency testing."

Bandwidth is king in this test, and the P4 Extreme Edition 3.73GHz processor with its 1066MHz frontside bus handily takes the lead, followed by the new Pentium 4 670 with about a 10% shortfall.  From there it's nip and tuck between the Pentium 840 Dual Core CPU and the Athlon 64 contingent.  Oddly, we had difficulty getting the Athlon 64 X2 to complete this test without failing midstream.

 

Transparent
Content Creation Winstone and WB5 Office XP

To get this next batch of results, we used Veritest's Content Creation Winstone 2004 suite.  Before running these benchmarks, we patched the program to its latest version (v1.01), shut down any unnecessary background processes, and defragged the hard drive.

Content Creation Winstone 2004
Real-World Application Performance

The Veritest Content Creation Winstone 2004 test utilizes the following applications in its benchmark routine. For more information about this test, see this page:

  • Adobe Photoshop 7.0.1
  • Adobe Premiere 6.50
  • Macromedia Director MX 9.0
  • Macromedia Dreamweaver MX 6.1
  • Microsoft Windows Media Encoder 9 Version 9.00.00.2980
  • NewTek's LightWave 3D 7.5b
  • Steinberg WaveLab 4.0f

The Athlon 64 systems handily sweep this test, which is heavily dependant on raw CPU throughput.  While some of its tests, such as Windows Media Encoder, enjoy the benefits of multithreading and are heavily SSE optimized, other tests such as Adobe Photoshop, Adobe Premiere, and Dreamweaver, thrive on the high IPC and memory bandwidth of the Athlons. The new Pentium 4 670 comes in just ahead of the Dual Core 840 and slightly behind the 1066MHz FSB-based Pentium 4 Extreme Edition 3.73GHz processor.  While the EE and 670 have identical core architectures (although the 670 is clocked a few MHz faster than the 3.73GHz EE), its 800MHz FSB holds it back somewhat in this test versus its Extreme Ed. sibling, but the variance is negligible.

PC World's World Bench Office XP Module
More Real-World Application Performance

PC World Magazine's WorldBench 5.0 is a new breed of Business and Professional application benchmark, poised to replace the aging and no-longer supported Winstone tests in our benchmark arsenal. WorldBench 5.0 consists of a number of performance modules that each utilize one or a group of popular applications to gauge performance.  Below we have the results from WB 5's Office XP module, recorded in seconds.  Lower times indicate better performance.

In our WorldBench 5 Office XP test, clock speed plays a more significant role within the various architectures.  As such, the new Pentium 4 670 comes out on top, followed closely by the 3.7GHz Pentium 4 EE within a narrow 1-2% margin.  From there it's a close race between the two dual-core CPUs, with the new Athlon 64 X2 edging out the 3.2GHz Pentium Extreme Edition 840 CPU.  In this case, the multi-tasking strength of dual-core architectures certainly helps, but clearly Hyper-Threading is enough to push the P4s ahead in conjunction with their higher clock speeds.

Transparent
WB5 Photoshop 7 and WB5 Windows Media Encoder

 

As probably one of our more popular desktop professional application tests, WorldBench 5's Photoshop 7 test module is next.

PC World's World Bench 5.0: Photoshop 7 & Windows Media Encoder
More Real-World Application Performance

 

If our WorldBench 5 Office XP tests were all about clock speed, then it's fair to say that the Photoshop 7 WB5 test modules is anything but.  Here's another big sweep by the Athlon 64 delegation.  There's a 24-second delta (about 8%) between the fastest score put up by the new Pentium 4 670 and AMD's flagship single-core chip, the FX-55.  Even the A64 4000+ takes the 670 to task with about a 3-4% speed advantage.

 

We continued our testing with WB5's video encoding benchmark based on Windows Media Encoder 9.  In the Windows Media Encoder 9 portion of the WorldBench 5 suite, encoding times are again recorded in number of seconds to complete a conversion to Windows Streaming Media format.  Lower times indicate better performance.

WME9 is both SSE optimized and multi-threaded.  Here it's the battle of the Dual Core CPUs, with the new AMD Athlon 64 4800+ leaving all competitors completely in its dust.  The 4800+ clocks in a full minute faster than the Pentium EE 840 dual-core processor.  Probably more interesting is how heavily optimized for multi-threading this test is, as is evident in the scores between the Pentium 4 670 at 3.8GHz versus the dual core Pentium EE 840 with its two 3.2GHz cores.  The 840 is only faster than the 670 by about 3%, but when you consider its 600MHz clock speed deficit, the benefits of multi-threading and multi-core performance are obvious.

Transparent
Lame MP3 Encoding and Cinebench 3D Rendering

 

In our custom Lame MP3 encoding test, we convert a large digital audio file to the MP3 format, which is a very popular scenario that many end users work with on a day-to-day basis to provide portability and storage of their digital audio content.  In this test, we chose a large 223MB WAV file (a never-ending Grateful Dead jam) and converted it to the MP3 format.  We utilized the latest 3.97 version of the Lame executable and encoding DLL for this test. Processing times are recorded below. Shorter times equate to better performance.

LAME MP3 Encoding Test
Converting a Large WAV To MP3

Our Lame MP3 test seems to favor system memory bandwidth more than anything else, and the low latency of AMD's integrated memory controller allows the Athlon 64s to shine brightest with the new 3.97 version of Lame.  The Pentium 4 Extreme Edition at 3.73GHz with a 1066MHz FSB puts up a good fight, but from there the new P4 670 clocks in 11 seconds slower although due to its slower 800MHz FSB.

Cinebench 2003 Performance Tests
3D Modeling & Rendering Tests

We ran two sets of numbers with Cinebench, one in single-thread mode and another in the benchmark's multi-thread mode. Please note that single-core Athlon 64s are only capable of running the single-thread test, hence the missing blue bar graphs in the chart below for all but the new Athlon 64 X2 CPU. Hyper-Threading-enabled Pentium 4 and Dual Core Pentium EE 840 processors, as well as the new Athlon 64 X2 4800+, are fully capable of running this test in its multi-thread mode, however.

The Pentium 4 670 offers up our fastest single-threaded score for the Intel processors we tested, but the dual core Pentium EE 840 smokes the fastest single core with Hyper-Threading Intel CPU available.  Overall, AMD's single-threaded performance is stronger, but single-core Athlons can't compete with Intel Hyper-Threading.  Finally, on the contrary, the Athlon 64 X2 4800+ strikes the best balance in this test, with its dual core design besting the Pentium EE 840 and its single-threaded performance edging out the new Pentium 4 670, as well.

Transparent
Kribibench 3D Modeling

 

Next up, we ran the Kribibench rendering benchmark produced by the folks at Adept Development.  Kribibench is an SSE aware software renderer.  A 3D model is rendered and animated by the host CPU, and the average frame rate is reported.  We used two of the included models with this benchmark: an "Exploded Sponge" model consisting of over 19.2 million polygons and then its enormous "Ultra" model that comprises over 16 billion polygons.

Kribibench v1.1
Details: www.adeptdevelopment.com

The Kribi engine is based on 100% software rendering for a scene, which makes it a perfect CPU benchmark. It also makes heavy use of SSE instructions and is multi-threaded.  Here the dual cores take the lead by a hefty margin with the A64 X2 once again dishing out a beating.  In the single-core camp, the Pentium 4 670 does give the FX-55 a run for its money.

More of the same is noted here with Kribibench's massive "Ultra" model in play.  It's a photo finish this time for the Pentium EE 840 and the Athlon 64 X2. The same can be said for the P4 670 and P4 EE 3.73GHz chip.  However, the FX-55 seemed to run out of gas, unable to keep up with the fastest single-core but Hyper-Threading-capable P4s.

Transparent
3DMark05 CPU Test

 

It may not be an actual game, but 3DMark05's built-in CPU test is a DirectX-based game engine metric that's useful for comparing relative performance among similarly equipped systems.  This test consists of two different 3D scenes that are generated with a software renderer, which is very much dependant on the host CPU's performance.  This means that the processing loads normally reserved for your 3D accelerator are instead sent to the host processor.  The number of frames generated per second in each test are used to determine the final score.

Futuremark 3DMark05 - CPU Test
Simulated DirectX Gaming Performance

 

3DMark05 is another benchmark with parts that are multi-threaded and take advantage of either Hyper-Threading, Multi-CPU-based systems, or dual core CPUs.  We've always seen the Pentium 4 shine in this test.  However, with the introduction of the Athlon 64 X2 4800+, AMD currently holds the lead spot, followed closely by Intel's Pentium Extreme Edition 840 CPU.  From there things scale back to the Hyper-Threading-capable Pentium 4s, with the test leaning more in the favor of the extra memory bandwidth offered by the 1066MHz FSB-driven P4 Extreme Edition at 3.73GHz.

 

Transparent
Unreal Tournament 2004 - DX8 Gaming

Next we delved into the gaming side of things a bit and performed some low-resolution benchmarking with Epic's Unreal Tournament 2004.  We specifically used a "Low-Quality" game setting with UT2004, which allowed us to isolate CPU and memory performance with little burden on the graphics subsystem.

Unreal Tournament 2004
DirectX 8 Gaming Performance

Our first traditional gaming test reports a very different picture of performance versus many of the other types of benchmarks we've shown you thus far.  As you can see, dual-core and multi-threaded CPU resources afford no advantage in UT2004's aging game engine.  New forthcoming Epic titles will undoubtedly make use of multi-threading on some level, but for now, the Athlon 64s have their typical gaming edge over Intel's Pentium 4 EE 3.73GHz chip, and from there the clock speeds and relative performance results scale accordingly.

Transparent
Doom 3 - OpenGL Gaming

 

For our next game test, we benchmarked all of the test systems using a custom multi-player Doom 3 timedemo. We cranked the resolution down to 640 x 480 and configured the game to run at its "Low-Quality" graphics setting. Although Doom 3 typically taxes today's high-end GPUs, when it's configured at these minimal settings it's more CPU and Memory-bound than anything else.

Benchmarks with Doom 3
OpenGL Gaming Performance

 

With Doom 3 there is a bit more to tell, but because we're running these benchmarks in a "timedemo" script, the test results look a little bit like our UT2004 scores.  Doom 3 is actually multithreaded, with engine physics and AI being driven on a separate thread.  However, in a timedemo scenario, the engine is in playback mode, so AI is taken completely out of the picture and the effects of real-time physics calculations are much less prominent. 

Additionally this game heavily favors overall system memory bandwidth as its predecessor Quake has historically.  In this test the Pentium 4 670, P4 EE 3.73GHz, the new Athlon 64 X2 4800+, and the FX-55 are all tightly bunched together with barely a 2-3% difference between the slowest and fastest scores.  Looking at things further, the 4000+ isn't that far behind either.  The only CPU that is really out in the cold here is the Pentium Extreme Edition 840 dual-core processor, which at a 3.2GHz clock speed per core, just doesn't have enough muscle to keep up with the other high-end CPUs in this death-match.

Transparent
Performance Analysis and Conclusion

 

Performance Summary: The performance landscape is changing these days, slowly perhaps but literally right in front of our eyes.  Intel's new Pentium 4 670 held its ground as we expected it to, in general desktop Business app performance as well as Gaming scenarios.   For us the chip felt sort of like a "jack of all trades but master of none."  That is to say that where we saw strength amongst both single- and dual-core processors in non-multi-threaded applications, the P4 670 was often surpassed by one of the high-end Athlons or perhaps the 3.73GHz Pentium 4 EE with its higher-speed frontside bus, as we saw in our gaming tests.  On the multi-threaded end of things, the Pentium 4 670 offers solid performance indicative of single-core Intel processors that are Hyper-Threading enabled. On the contrary, depending on the application, the 670 doesn't keep pace with the higher-end dual-core offerings recently introduced from both Intel and AMD. 

In the end, depending upon if you're a "glass is half full or half empty" type, you could look at the P4 670 as striking a balance on the Intel side of the fence: between standard and multi-threaded or gaming and professional application performance, against the backdrop of price-point versus a more expensive dual-core Athlon or Pentium 4 EE 3.73GHz CPU.  However if you're a hard-nosed performance enthusiast, the metrics we've shown you today may have left you a bit thirsty.  We subscribe to the former mind-set actually, because there is no denying this processor is more than fast enough for just about anything you could conceivably throw at it currently.

According to our sources at Intel, the Pentium 4 670 processor is going to be priced at $851 in 1,000-unit quantities.  That puts this new P4 a healthy notch beneath the ridiculously expensive Pentium 4 Extreme Edition at 3.73GHz (currently $1000+ street price), as well as the recently announced Athlon 64 X2 4800+ (MSRP of $1001), which we have yet to see emerge in the retail channel.  Without question, the world is going the way of multi-core processors and multi-threaded applications.  The question is "when," and our response to that would be "now."  Obviously OS support for multiple threads in multi-tasking scenarios has been around for quite some time.  Furthermore. application-level support is growing literally exponentially, as new dual-core architectures are introduced by both Intel and AMD.  Not to mention Intel has been seeding the market for years now with its Hyper-Threading technology, now refined in the company's new P4 6XX sequence CPUs.

So where does this leave the new P4 670?  Again, we come back to the question of "balanced performance"... Ultimately it's up to you to make this call from what we've shown you within our benchmark matrix here.  We'd suggest that the P4 670 strikes a happy medium.  It's not the fastest desktop processor around, but it's certainly no slouch if you're willing to part with a limb or two to afford it.  At least you won't have to part with your first born, as is the case with the Pentium 4 EE 3.73GHz or Athlon 64 X2 4800+. 

 

Get into HotHardware's PC Hardware Forum
Tweaked, Overclocked and Ready To Rock!



Content Property of HotHardware.com