The Pentium III was a great success for Intel. It was a good, solid chip with excellent software support. But Intel was fighting a fierce battle with AMD on both the low and high end, so Intel released their Pentium IV. More of a psychological victory over AMD than a technical one, the Pentium IV makes for an interesting story.
Clock speeds are an interesting measure. In earlier times, clock speed was the main factor in determining the processing power of a PC. However, with CPU architectures advancing and branching off in different directions, an accurate comparison can no longer be made across CPU families simply using clock speed. The Pentium IV has a pipeline that with a different design than that of the Athlon. Intel calls this technology "Hyper Pipelined Technology." What it does is increase the length of the pipeline through which instructions flow. By lengthening a pipeline, a processor can process fewer instructions per clock cycle, but can run at a greatly increased clock, which, in the end, will make up for the tradeoff in pipeline length, and eventually allow the Pentium IV to have much greater performance than it could have had with a short pipeline and lower clock speeds.
Longer pipelines are not the only change the Pentium IV boasts. It has 256KB of Intel's "Advanced Transfer Cache." ATC is L2 cache that uses a 256-bit data path between the CPU and the L2 cache, and can pull data from the cache in one clock cycle. Along with this important technology is the Pentium IV's Quad 100MHz FSB. It allows for a 100MHz bus, as is used on the Pentium IV, to perform effectively as a 400MHz bus. It works much in the same way the DDR memory for Athlons runs at 133MHz, but with a doubled data-rate, allowing for "266MHz effective" performance. For the most part, this technology relies on RAMBUS memory, because even though PC133 memory is now supported by the new I845 Pentium IV chipset, it can not offer the sort of performance RAMBUS memory can.
A feature many people do not take into account is how a heatsink is attached. The way a heatsink is attached has gone basically unchanged since the first ZIF (Zero Insertion Force) sockets were used. Clips on the socket were used to attach the CPU. This was fine until CPUs required bigger, more tightly connected heatsinks and the CPU cores were moved from the bottom to the top of the CPU for better heat transfer. With AMD CPUs, there have been many horror stories of people cracking their CPU core while attaching a heatsink. Also, in removing these tightly connected heatsinks, I have heard stories of people damaging their motherboards while prying certain types of CPU fans off.
Intel has fixed this with the Pentium IV using the latest version of the ATX standard. CPU fans are now attached with mounting brackets, so all you need do is sit the CPU fan on top of the chip, and then fasten clips with little or no effort required. This is by far the most important innovation of the Pentium IV, and one AMD would do good to implement with its CPUs.
Pentium IV 1.9GHz (Socket 423)
256 Megabytes Kingston RDRAM
Western Digital 7200 RPM 10.2 Gig Hard Drive
3Com 3C905TX-C 10/100 NIC (PCI)
400 Watt ATX 2.03 Power Supply
Gigabyte GF3000 GeForce 3 64MB AGP
Slackware 8.0 with Kernel 2.4.14 and XFree 4.1.0
Athlon XP Configuration:
Athlon XP 1800+ (1.53GHz)
Soyo Dragon Plus! Motherboard
256 Megabytes CAS2.5 Crucial PC2100 DDR SDRAM
Pentium IV Socket 478 Configuration:
Pentium IV 1.7GHz
Gigabyte 8TX Motherboard
In order to test both the board's stability and speed, I ran three sets of Linux kernel compiles on this board. One is a normal, "uniprocessor" make, or make -j1, which is the default. This uses one process, and does not always maximize system usage. I then did make -j2, which spawns a second process. The last test I run is with make -j3, spawning two extra processes. I do this for several reasons -- to find the "sweet spot" for the board/CPU, as well as to stress the system
as much as possible when trying to rate its stability. Also, the kernel is extremely useful as a measure of integer performance. In order to compile the kernel, I untarred kernel 2.4.6, ran "make config" and used the default values.
|Kernel 2.4.6 Compile Times (Minutes:Seconds)
Lower numbers are better
|Athlon XP 1800+ 1.53GHz||2:59||2:51||2:49|
|Pentium IV 1.7GHz (Socket 478)||3:47||3:40||3:38|
|Pentium IV 1.7GHz||3:47||3:40||3:49|
|Pentium IV 1.9GHz||3:30||3:24||3:23|
We see here the Pentium IV is easily outpaced by the Athlon XP 1800+, even when the Athlon is clocked much lower. We see AMD's claim of being equivilent to an 1800+MHz Pentium IV is verified here, which is unfortunate for the Pentium IV. Also we verify that the Pentium IV in its previous form, Socket 478, and the smaller Socket 423 version are the same CPU in different packaging.
POVRay is a multi-platform raytracing program. It is a very floating point intensive task and serves well to help measure the floating point performance of a CPU. For more information on this benchmark, head to the official POVBENCH homepage. The command to run for this benchmark, once you obtain POVRay, you run povray -i skyvase.pov +v1 +ft -x +mb25 +a0.300 +j1.000 +r3 -q9 -w640 -H480 -S1 -E480 -k0.000 -mv2.0 +b1000 from the command prompt. Results are in seconds.
Lower numbers are better
|Athlon XP 1800+ 1.53GHz||14 seconds|
|Pentium IV 1.7GHz (Socket 478)||23 seconds|
|Pentium IV 1.7GHz||23 seconds|
|Pentium IV 1.9GHz||19 seconds|
Here again, despite the higher clock speed and the rating of the Pentium IV, the Athlon XP beats the Pentium IV by a large margin. The Pentium IV's FPU is clearly weaker, relying heavily on specific optimizations for it to perform well.
Distributed.net client benchmark
Distributed.net is a distributed computing network that works on various distributed computing contests. The contests use primarily integer numbers while performing their tasks, and therefore serve as an excellent benchmark for overall integer performance of properly optimized software.
|Distributed.net Client Benchmarks
Higher numbers are better
|CPU||RC5 Core 6||OGR Core 0|
|Athlon XP 1800+ 1.53GHz||5,419,596 keys/sec||11,514,790 nodes/sec|
|CPU||RC5 Core 8||OGR Core 0|
|Pentium IV 1.7GHz (Socket 478)||2,439,327 keys/sec||7,342,364 nodes/sec|
|Pentium IV 1.7GHz||2,448,429 keys/sec||7,433,902 nodes/sec|
|Pentium IV 1.9GHz||2,757,055 keys/sec||8,321,706 nodes/sec|
In the distributed.net tests, we see the same pattern -- the Pentium IV, at a higher clock speed, is beaten by the Athlon XP, illustrating again how invalid clock speed is as a measure of speed between different types of CPUs.
Quake III Timedemos
Quake 3 Timedemos are perhaps the best way to measure 3D Gaming performance under Linux. Timedemos used the four.dm_66 demo included with the latest version of Quake 3 Arena. To run a timedemo, hit the "~" key, type timedemo 1, followed by demo four.dm_66 - once this completes, hit "~" again to see your results. High-quality results were done by turning texture and color depth to 32-bit, filtering to trilinear and texture detail to its highest setting. The tests are run at screen resolutions of 640 by 480, 800 by 600, 1024 by 768, 1280 by 1024 and 1600 by 1200.
|Quake 3 Arena Timedemos (Frames Per Second)
Higher numbers are better
|Athlon XP 1800+ 1.53GHz||218.0||214.2||184.3||125.3||90.2|
|Pentium IV 1.7GHz (Socket 478)||206.0||205.9||181.4||124.4||89.6|
|Pentium IV 1.7GHz||206.5||205.8||182.8||125.3||90.2|
|Pentium IV 1.9GHz||221.8||218.6||186.2||125.3||90.2|
|Athlon XP 1800+ 1.53GHz||217.5||197.2||146.2||97.2||71.2|
|Pentium IV 1.7GHz (Socket 478)||203.6||191.5||144.6||96.7||70.9|
|Pentium IV 1.7GHz||205.2||194.3||146.1||97.3||71.2|
|Pentium IV 1.9GHz||219.5||199.7||145.4||97.3||71.2|
In Quake 3 testing, we find the Pentium IV and the Athlon XP on more even ground. At lower resolutions where performance is more dependent on the memory and CPU subsystem -- because the video card can more than handle the data at that resolution -- they score nearly identical results. As we get to higher resolutions where the video card is the bottleneck, the scores are almost exactly the same. These higher resolutions are where gamers buying this new hardware run their games, and we see that the video board is most important here.
The price gap is not the enormous factor it once was between Intel's Pentium IV and the Athlon XP line of CPUs from AMD. Right now a Pentium IV 1.9GHz can be had for around $270, while an Athlon XP 1800+ is available for around $200. Seventy dollars isn't exactly pocket change, but it isn't the kind of enormous gap we are used to. Even still, without significantly lowering the price of its CPUs while at the same time increasing clock speed, Intel just can't compete with the value of AMD processors at this point.
It isn't that the Pentium IV is a bad CPU -- I wouldn't call any of the scores in this batch of tests slow by any means -- it is that, for its performance, its price is unwarranted, especially when you add in the cost of a motherboard and the more expensive RDRAM required to get the most performance from a Pentium IV. If you are a staunch supporter of Intel, then you won't be entirely disappointed with the Pentium IV, but you might not get as much bang for the buck as you would with an Athlon-based chip. The Pentium IV 1.9GHz is available through vendors on Pricewatch for around $270.