Original Link: http://www.anandtech.com/show/1238
Intel 3.2E vs. 3.2EE vs. 3.2C: Comparing Baseline Performanceby Wesley Fink on February 12, 2004 2:46 PM EST
- Posted in
With the introduction last week of the Socket 478 Prescott, we now have no less than three 3.2 GHz Pentium 4 processors from Intel. As we move toward the expected April 2004 introduction of Socket 775 and a new revised Prescott, we will be including compatibility testing of the 3.2E and 3.2EE, in addition to the 3.2C in future Pentium 4 motherboard and system reviews. With 3 flavors of 3.2GHz P4, 3 different architectures, and potentially different performance, all benchmarks needed to be updated to allow a reasonable comparison of current and future benchmark results. The results of the updated tests were so interesting that we decided to publish the results. Some of the results are already well-known, but other results like Workstation Performance were a bit of a surprise.
This is not intended to be an in-depth review of the new Prescott processor; Anand published a definitive review of Prescott at launch. For an in-depth look at Prescott and a comparison to AMD processors, you should read Anand's Intel's Pentium 4 E: Prescott Arrives with Luggage. The scope of this article is much more basic - to update our standard suite of Motherboard Benchmarks with the 3.2C, 3.2EE, and 3.2E on the same Asus P4C800-E motherboard.
While the three 3.2GHz processors from Intel all run at the same clock speed, they are quite different under the hood.
|3.2GHz Pentium 4 Processors|
|L1 Cache||8k Data Cache +
12k instruction cache
|8k Data Cache +
12k instruction cache
|16k Data Cache +
12k instruction cache
|Memory Type||Dual-Channel Unbuffered DDR||Dual-Channel Unbuffered DDR||Dual-Channel Unbuffered DDR|
|Web Price 2/10/2004||Retail $289||Retail $957||?
We could not yet find the new 3.2E available for sale anywhere, but Intel has priced the 3.2E the same as the 3.2C. When it does arrive, the prices should be about the same as Northwood. To keep value in perspective, the Prescott and Northwood will sell for about the same price, while the P4EE is more than 3 times that price.
Performance Test: ConfigurationTo provide a baseline for future motherboard tests, we tested all three 3.2GHz processors using our latest Motherboard tests. The same Socket 478 motherboard, the widely available and excellent performing Asus P4C800-E, was used for all tests. This is our standard Socket 478 motherboard for memory testing, and it is the motherboard recommended by both Corsair and OCZ for running their fastest DDR550 memory.
|Performance Test Configuration - 3.2GHz Socket 478|
|Processor(s):||Intel Pentium 4 3.2E (Prescott, 1MB L2 cache)
Intel Pentium 4 3.2EE (512kb L2 Cache + 2MB L3 cache)
Intel Pentium 4 3.2C (Northwood, 512kb L2 cache)
|RAM:||2 x 512Mb OCZ 3500 Platinum Ltd
2 x 512Mb Mushkin PC3500 Level II
|Hard Drive(s):||Seagate 120GB 7200 RPM (8MB Buffer)|
|Video AGP & IDE Bus Master Drivers:||Intel Chipset Drivers|
|Video Card(s):||ATI Radeon 9800 PRO 128MB (AGP 8X)|
|Video Drivers:||ATI Catalyst 4.1|
|Operating System(s):||Windows XP Professional SP1|
|Motherboard:||Asus P4C800-E (Intel 875P - 478) Rev. 2.00|
It has been reported on other sites that the P4C800-E will not run the Prescott processor. We had no problem at all with the 3.2E Prescott on our P4C800-E. This particular board has been at AnandTech for about 2 months and is a late model Revision 2.00. Asus tells us that all versions of the P4C800-E will work fine with Prescott as long as BIOS is updated to version 1014 or later. You can find a CPU compatibility list for Asus motherboards at http://www.asus.com.tw/support/cpusupport/cpusupport.aspx.
All performance tests were run with the ATI 9800 PRO 128MB video card with AGP aperture set to 128MB with Fast Write enabled. Resolution in all benchmarks is 1024x768x32 unless otherwise noted. Results at 1280x1024 have also been provided where they are useful in comparing performance.
With the retesting required to compare Prescott, EE, and Northwood, benchmarks were updated with additional games and updates to other benchmarks.
Content Creation and General UsageWe have recently updated to the latest release of Winstone benchmarks. Veritest Multimedia Content Creation Winstone 2004 and Veritest Business Winstone 2004 are now the standard tests for system benchmarking.
Media EncodingDivx 5.1.1 has added support for additional extensions on both Intel and AMD processors. As recommended by both Intel and AMD, we have updated to Divx 5.1.1. We also updated to the latest XMpeg 5.03 as the front end for Media Encoding. Encoding is now benchmarked as a common dual-pass setup, so the results reported with XMpeg 5.03/Divx 5.1.1 are not comparable to earlier Media Encoding results.
GamesWe have added several new benchmarks to our standard Gaming tests. These include Halo, Microsoft's Direct X 9.0b game; Splinter Cell, a DX9 game; X2 Benchmark, a DX 8.1 game that includes Transform and Lighting effects; the DX9 Aquamark 3; and the DX 8.1 Comanche 4 benchmark. Since we have found that Comanche 4 can become video card limited at higher resolutions, we will only include benchmarks that run with 4X anti-aliasing enabled to differentiate system performance better using our standard ATI Radeon 9800 PRO video card.
We have dropped Yeti Studios DX9 Gun Metal 2 from our standard motherboard and system benchmarks, since there are many other DX9 choices now available that measure system performance variations better.
Heat and CoolingAll of our testing was done on air with Intel's stock CPU cooler supplied with the 3.2E. There is absolutely no doubt that Prescott gets hotter than either Northwood or Extreme Edition in our benchmarks. We did not measure temperatures, but the Heatsink felt much warmer during benchmarking Prescott. This was also the case with other HSF, like the Zalman 7000, OCZ Eliminator 2, and Thermalright SLK900. However, we never experienced a shutdown or throttle during several days of testing, and we did not have to use water-cooling to keep temperatures under control.
Our advice is to buy Retail with the Intel HSF or use only the best HSF with Prescott because it is definitely a hotter CPU than Northwood or EE. For those who will overclock, the Intel HSF is outpaced pretty quickly and you may need to consider water, phase-change, or other cooling methods. This should not be a concern for normal operation, however, since we had no real issues with the standard Intel heatsink.
Content Creation and General Usage Performance
As expected, the expensive P4EE with 512KB L2 cache plus 2MB L3 cache is the best performer in both Multimedia Content Creation and Business Performance. The Prescott and Northwood score essentially the same, despite the fact that the pipelines and cache are quite different.
Since the Winstone benchmarks are based on the applications used by almost everyone on a daily basis, computer users will not likely see any difference in Prescott and Northwood in most applications.
Gaming and Media Encoding Performance
Anyone who did not understand why Intel introduced a 3.4EE at the same time Prescott was launched only needs to look at gaming benchmarks. The 3.2EE dominates in virtually all the game benchmarks. However, considering all the concerns about the deep-pipe design of the Prescott, the Prescott performance is remarkably close to Northwood. Yes, Northwood is a bit faster, but the difference is generally small. In Comanche 4, however, Prescott is completely outperformed by Northwood; the results are not even close. The benchmarks were repeated several times in Comanche 4 to confirm the results.
Media Encoding is a big win for Prescott in an application of great importance to many. Prescott is a bit faster than either P4EE or the 3.2C in encoding. This may be the result of the Divx 5.1.1 support for the new SSE3 instruction LDDQU. Further evidence that SSE3 may be helping encoding performance is the fact that the 3.2E came in last of the 3.2's in our testing with Divx 5.05.
Workstation Performance was a bit of a surprise because processors with the largest on-chip cache are usually the best performers in this benchmark. However, in all the individual tests with SPECviewperf 7.1.1 except UGS, Prescott was the best performer. Apparently, there are aspects of the Prescott design that make it more efficient in SPECviewperf. This same pattern was also seen in Intel CPU comparisons on other chipsets and other motherboards. It is a big surprise that Prescott beats the EE in Workstation Performance benchmarks, even though the EE has 2.5MB of cache compared to the 1MB of the 3.2E.
Final WordsThe only goal of testing three different Intel 3.2 processors on the same reference motherboard was to establish benchmark results for future motherboard testing. However, it is hard not to compare the CPU performance differences in these tests. What perhaps stands out the most is that for all the negative press Prescott has received at launch, it is really a decent performer that is very close to Northwood, at least at the 3.2GHz speed. As Anand found in Intel's Pentium 4 E: Prescott Arrives with Luggage, Prescott scales faster than Northwood as speed increases. At 3.2GHz, performance is closer than we first thought it would be, and Prescott is certainly ahead of Northwood by the time you reach 3.7 to 3.8GHz.
Looking at individual performance categories from our standard benchmarks, it is easier to understand why Intel introduced another EE chip in the 3.4 speed. The one area where Prescott is poorest is gaming performance, an area already dominated by the Athlon 64. Prescott is slower compared to Northwood in this area, but the real difference is 0% to 5% in most cases. Northwood is not a great game chip either compared to A64, so the 3.2EE and 3.4EE fill that void. Unfortunately, the cost of bringing Pentium 4 gaming performance to Athlon 64 levels is very high, with the EE chips selling for premium prices.
For our standard "real application" benchmarks, Multimedia Content Creation Winstone and Office Winstone, Prescott and Northwood performed virtually the same in these baseline tests. You will not notice the difference between these similarly-priced chips. P4EE performs best in these benchmarks, as expected, but at 3 times the price, the P4EE is not really in the same league with Prescott and Northwood.
The areas that most surprised us were Media Encoding and Workstation Performance. We expected P4EE to lead in these benchmarks, but instead, Prescott was the top performer. Intel tells us that Divx 5.1.1 provides support for new SSE3 instructions, and at least in our configuration with an XMpeg 5 front end, Prescott leads in Media Encoding. Other Media Encoding benchmarks that use older codecs have been showing Prescott about the same as Northwood in encoding tests. Workstation Performance was generally dominated by Prescott, so those applications that depend on the types of operations tested in SPECviewperf will perform best with the Prescott 3.2E. The SPECviewperf 7.1.1 benchmarks were repeated on a VIA PT880 motherboard with all three 3.2 processors, and Prescott again dominated these benches, providing the top performance in all the 7.1.1 suite including UGS.
Prescott and EE compatibility will be a part of all future testing of Pentium 4 motherboards. With the testing that went into this comparison, you should have a better idea of how the 3 flavors of Pentium 4 compare at the same speed on the same reference motherboard. Prescott is not a very good gaming chip at 3.2 GHz, but it is closer to Northwood than many expected. If you can afford the price tag, P4EE is still the best choice for gaming among Intel processors. Northwood and Prescott are virtually interchangeable as measured in Content Creation and General Usage performance. For Media Encoding and Workstation Performance, Prescott at 3.2GHz appears to have an edge over both Northwood and the much more expensive 3.2EE.