The Most Controversial x86-Processor Of All Time
Two days ago Intel launched its new Pentium 4 processor, which was received by the masses with rather mixed feelings. Due to rather strange behavior in different benchmarks many people don’t really know what to think about the new flagship CPU from Intel. The majority of Pentium 4-reviews draws rather negative conclusions and even in my own article from Monday I wasn’t really able to see Pentium 4 as the great processor Intel wants it to be. However, Pentium 4 was able to score very well in 3D games and our MPEG4-encoding benchmark, it has a very compelling new design and will most likely show much better performance once software takes advantage of its new features and Intel brings up its clock speed. This is why I refrained from slamming this new CPU.
Record Reader Numbers Ask For Record Responsibility
On the very Monday of Pentium 4’s release the web servers of Tom’s Hardware Guide were under extreme load, making it difficult for many readers to download the pages of the Pentium 4 article. I would like to apologize for those inconveniences and also thank you for your faith in Tom’s Hardware, as we scored a new record of 1.413 million pages that day. I would like to express my full awareness of carrying a huge responsibility towards all those hundreds of thousand faithful readers who rely on the conclusions of my articles. Unfortunately new results out of the still ongoing Pentium 4 evaluation have urged me now to change my stance on how I see Pentium 4 and I want to get the word out without any hesitation.
Are Our Initial MPEG4 Results Not Meaningful?
Yesterday I received an alarming email from Toby Hudon, who expressed his doubts about the validity our MPEG4-benchmarking results:
I assume the source files to be encoded to MPEG-4 in most of your benchmarks are high bit rate MPEG-2 similar to what’s on most DVDs. One thing many people are unaware of is that the selection of what kind of iDCT is used to do the DECODE phase has a big impact on the final output.
FlaskMPEG has 3 settings for this, in the Export settings video tab:
The default is MMX. However, this tends to produce a lot of artifacts in the final MPEG-4 video because all the pixel values of the decoded frames are approximations. Thus when a second DCT transform is applied to convert it to MPEG-4 it tends to approximate again and produce really horrible artifacts in some cases. Using the IEEE decode eliminates most of these artifacts and produces an output that rivals most DVDs when set to about 20% of the original bit rate (1.5mbps for a 7.5mbps DVD like Matrix). This is on the large side for full movies but fine for TV episodes and music videos. The problem is, this is an FPU based operation, whereas the default decode is an integer or MMX-based operation. As we can see, P4 is much faster than Athlon at the default settings but what about in a pure FPU mode that most people doing high quality rips would use? I’d expect a reversal of scores more like the 3D Studio MAX results. Can you please run a small test comparing the speeds of the current top CPUs (Athlon 1.2 P3 1.0, P4 1.5) in each decoder mode? I’d like to see how they perform so I can pick the one right for my applications. |
I pointed out earlier, that I am very aware of the responsibility Tom’s Hardware is carrying as a premium technical publication. Every doubt that a reader expresses about our benchmarking scores is looked into immediately and with the outmost care. Toby’s concerns are making a whole lot of sense, because we obviously want the MPEG4-benchmark scores to be useful to those people who are actually doing DVD-rips. It is not hard to assume that the majority of DVD-rips are supposed to produce the best possible quality and so it obviously became questionable to me if our benchmarks, that were indeed carried out using the MMX-iDCT, are actually helpful to those who really do MPEG4-encoding. As a result I rushed to repeat the MPEG4-benchmarks runs with the IEEE-iDCT to find exactly the results, which Toby had been able to forecast by simply drawing logical conclusions.
The New MPEG4-Encoding Situation
Before I get into the new results, I would like to point out that our initially published MPEG4-encoding benchmark scores were not wrong, but I had used the default MMX-iDCT setting, assuming that it is the iDCT used by the majority of DVD-rippers. However, you may remember that those results published on Monday were showing Pentium 4 as the clear winner of the MPEG4-benchmark, which had been an important issue to me for drawing my final conclusions about Pentium 4.
Those are the initial results we published. Pentium 4 is obviously way ahead of the competition, seemingly due to its high memory bandwidth and surprisingly powerful MMX-engine.
The New MPEG4-Encoding Situation, Continued
Now let’s have a good look at the new results, scored by using the IEEE high quality iDCT with Flask MPEG.
It’s not hard to spot the difference. A highly overclocked Pentium 4 at 1.728 GHz and 108 MHz system bus is just about able to reach the scores of a normal Pentium 3 at 1 GHz. Athlon is almost double as fast!
Let’s also have a very good look at the speed comparison between MMX-iDCT and IEEE high quality iDCT.
While Athlon as well as Pentium 3 drop down to about 40% of the speed when using the high quality iDCT that’s necessary to produce MPEG4 videos of good quality, Pentium 4 plummets down to 20%!!! That means Pentium 4 needs FIVE TIMES as long to encode an MPEG4-movie to high quality! Looking at the absolute numbers shows that you’ve got to wait about 12 hours for an average movie to be encoded at high quality with Pentium 4 and only 7 hours with a fast Athlon processor. This is a difference of FIVE HOURS!!!!
New Results Force New Conclusions
It is obvious that Pentium 4’s mediocre x87 floating-point unit is responsible for those poor results. It is incredible, but Intel didn’t seem to care about Pentium 4’s FPU-performance whatsoever when this processor was designed. The Pentium 4-designers relied completely on the new SSE2-unit and the fact that SSE2-enabled software won’t use the x87-FPU anymore. I leave it up to you to decide if you consider this as a brave move of Intel or an ignorant one. I suppose that Intel doesn’t want to sell Pentium 4 processors to Quake 3 players only. However, for the time being Quake 3 is the only benchmark that shows above-average scores of Pentium 4.
I do certainly admit that MPEG4-encoding is only one of many tasks that could run significantly faster on Pentium 4 once new SSE2-code is used in the encoders. However, these are future tunes. Right now there is Quake 3 and then there is Quake 3, besides that there is only Quake 3. Have I mentioned that Pentium 4 is really good in Quake 3?
Let’s summarize the latest findings:
- Pentium 4 scores very badly in MPEG4 encoding benchmarks once the IEEE high quality iDCT is used.
- Pentium 4 runs current office applications slower than Pentium 3 and much slower than AMD’s Athlon, as proven in our Sysmark 2000 benchmark results from Monday.
- Pentium 4 scores worse than Athlon in Unreal Tournament.
- Pentium 4 is a very bad solution for compilations with gcc 2.95.2 under Linux.
- Pentium 4 scores very badly in 3D Studio Max , proving a sub-par FPU.
- Pentium 4 gets slightly beaten by Athlon in the 3D game MDK2.
- Pentium 4 is badly losing out against Pentium 3 as well as Athlon in clock-for-clock comparisons.
- Pentium 4 is currently the most expensive x86-system solution available.
- BUT Pentium 4 is really good at Quake 3 Arena! Honestly!
New Results Force New Conclusions, Continued
What bottom line is any sane person supposed to draw after those findings?
Right now there are hardly any important applications available that could show the strength of Pentium 4’s new SSE2 features. As long as this isn’t the case, Pentium 4 loses out against Athlon and sometimes even against Pentium 3, because Intel was obviously saving silicon by implementing only a rather weak FPU into Pentium 4. The other way out of this dilemma would be extremely high clock rates, but it will take a while until faster Pentium 4 processors will become available. Then there is the fact that it won’t even take a year until the successor of the current Pentium 4 will become available, which will ensure that the upgrade path of current Pentium 4 systems ends in a dead end road, because the ‘new’ Pentium 4 will use a new and incompatible Socket478.
If you now consider that Pentium 4 systems are the most expensive PCs money can buy right now, then I wonder how any customer could possibly justify the purchase of Pentium 4 within the next months.
I have to admit that I started off being a believer in Pentium 4 and I still respect Pentium 4’s future potential. However, right now I am genuinely disappointed. For the time being, I wouldn’t let any of my friends or family members buy a Pentium 4 system. It’s simply not justifiable.
Epilogue
I am continuing to benchmark Pentium 4 with even more benchmarks on more operating systems. I want to find out if Intel was just being bold to release a new processor with sub-par performance in today’s applications and call it “the fastest x86-processor in the world”, or if there is indeed a bit more to it than good scores only in Quake 3 Arena and beautiful future tunes. You can be assured that I will keep you posted.
I wish a lovely Thanksgiving!
Please read the Intel’s New Pentium 4 Processor article!