Benchmarks
- Intel Pentium IV 2.4GHz 'Northwood A'
- Intel Pentium IV 2.4GHz 'Northwood B'
- Intel I850E Motherboard
- 256MB Samsung PC800 RAMBUS RIMMS (2 x 128MB)
- Intel Stock heatsink
- Geforce 3 TI500 Card @ Stock speeds
- AMD Athlon XP2100 (1733 MHz)
- MSI KT3 Ultra-ARU, run in asynchronous memory mode for maximum performance
- 256MB Samsung DDR 2700, run at CAS2, strictest timings.
- Geforce 3 TI500 Card @ Stock speeds
Software
- Windows XP Professional Build 2600.xpclient.010817-1148
- NVIDIA Detonator XP 23.12 drivers
- Sisoft Sandra 2002 Professional
- Pifast v41
- Lame v3.91 MP3 encoding with RazorLame 1.15 front-end
- OcUK SETI Benchmark.
- PC Mark 2002
- 3DMark 2000
- 3DMark 2001SE
- Quake 3 v1.30
All benchmarks were conducted at 1024x768x32 100Hz with vertical sync' disabled. Benchmarks were run 3 times consecutively, an average score was taken. Both systems were configured for maximum performance. A fresh installation of Windows XP was used in both instances.
Benchmarks
As always we will start with Si Soft Sandra for our first round of testing - this doesn't give perfect world results but it does show nicely how a new CPU sits in with older CPUs.
As you can see the Northwood B CPU out performs the other 2 units which we are comparing with - this shows that the P4 can surpass the AMD XP2100 CPU easily - unlike its 100FSB sister. The explanation for these seemingly erroneous results lies in our discussion regarding pipeline length and work done per cycle.
The P4s large deep pipelines make it harder for the CPU instructions to pass through quickly and they have around 20 stages to navigate before they hit completion.
As Intel stated with the release of the Pentium 4 CPU the CPU is designed to be scaled - this is shown when you see the 2.4B which has a lot of room for the data to be pushed through. The P4 always wins on Drystone - this is impressive.
The Pentium 4 hoards bandwidth - the more present there is the faster the CPU can go as it gets fed the data quicker. The P4 coupled with the Rambus memory helps performance a lot.
If you have read the other Pentium 4 coverage on Hexus or other sites you will remember that the Pentium 4 has quad-pumped FSB - this means the P4 400Mhz of bandwidth - 3.2GB/s - but with the new 133 Bus P4s - it gives you 4.26GB/s bandwidth this is the bandwidth available between CPU and memory controller. This is similar to the AMD dual pumped FSB which will only give 266 MHz or 2.1GB/s of bandwidth - You can see the result of the quad pumped bus speed with the results.
The P4 takes a commanding lead with the default speeds. This lead looks impressive it isn't as high as it could be meaning if the AMD is only running at dual pumped FSB then you would expect the P4 to have a lot higher result. We only manage to extract 78% efficiency with RAMBUS compared to the Athlon's impressive 96% utilisation from DDR, albeit both in a buffered state.
The P4 totally beats the AMD in the synthetic benchmarks but will it do so well in the practical benchmarks? Well we can find out now.
We use Pifast as our practical benchmarks. Pifast calculates the constant Pi to X million decimal places using the fastest method possible. Pifast is very responsive to the change in CPU and memory subsystem.
As you can see that the P4B runs at a superior speed to this calculation this is impressive performance - the P4's coupled with the Rambus and the memory bandwidth gives a good performance but even though it can beat the Athlon XP - there is nearly 600 MHz in speed difference present here.
A Lot of you out there encode your CD collection so that you can have it on your system available to listen to in MP3 format. Here were encoding a 481MB WAV file into MP3 128 kb/s format. MP3 encoding has historically been a CPU-intensive activity it doesn't need high memory bandwidth. Let's see if this still holds true.
Since the LAME benchmark is CPU bound, clockspeed is what matters and the 2.4B has no advantage over the 2.4A. The extra bandwidth doesn't help things here.
We next ran the OCUK SETI benchmark, a rather tough work-unit with an angle ratio (AR) of 0.417. This one takes a while to complete, as it sifts through huge chunks of data in the hope of finding some signs of Extra Terrestrial existence. One advantage in this benchmark is its ability to display results to within 1/10000th of a second, we've rounded the results up to the nearest second.
The results between the XP and the P4A were very close - We thought that the P4 would out perform the AMD due to the high bandwidth. But the P4B wipes the table and beats both CPU due to the higher bandwidth with the 533 Bus. The extra power is shown by the bus speed and not by the clock speed. The SETI benchmark takes over 3 hours to run - this is why we are calling it close - VERY close.
All the tests so far show that the P4B is a sure improvement in speed over the AMD counterpart and even the Intel P4A unit. It is definitely a fast CPU.
Recently we have been using the latest Madonion benchmark PCMark 2002 from MadOnion.com. It is a similar benchmark to SiSoft Sandra in that it consists of a series of tests that represent common tasks in home and office programs, such as Adobe and other popular applications. They have 3 different results - CPU, Memory, and Hard Drive result. The aim for this benchmark is to do the same testing which 3DMark does for the Graphics age.
This benchmark seems to favour the Pentium 4 - we have often run this multiple times and we have come out with the same results. The only common factor within these results is that the hard drive results come out the same. (This is normal as the CPU doesn't affect the hard drive performance.)
The Athlon XP doesn't manage to keep up with this test - perhaps Intel worked very closely with Madonion with this benchmark software We're bewildered by this result - especially since the AMD platform does a bigger punch per MHz.
We now move on to another Madonion set of testing - 3DMark 2000 is a DX7 testing platform which even though we are now on a new version of DirectX this is still a key benchmark as a lot of games still use DX7. We also do testing with 3DMark 2001SE testing, this is a DX8 test, this is the latest benchmark from Madonion in the 3DMark series - which is one you all have probably run on your system. We will run all the benchmarks at 1024x768x16
As you can see this testing all the CPUs get very close performance there isn't much in it - this is very surprising. they both mirror each other - we are surprised that the results do not show off the P4 more - this could be down to the 23.12 drivers which we are running [these could favour the XP CPU more.]
We have already mentioned that when Doom 3 comes out this will be the most system hungry benchmark on the market. BUT until then this is Quake 3 - IDs massive first person shooter. It is a consistent benchmark and very few have lasted as long as iD Software's. We as always will run it in 512 fastest and 1024 quality Point release v1.30 was used in both instances.
So that's the fastest - how about the highest quality - on to 1024 quality.
If you have run any Quake 3 tests yourself you will know that nothing can beat the P4 in this test - as you can see from the graphs this is still the case. The P4 is the daddy of Quake 3 CPUs. No one can physically tell the difference between these FPS's but it is nice to show off about.
