It comes as no surprise to anyone that
large companies often bend
the truth in order to peddle their wares to consumers. "
New and Improved" is a statement that can read on 25% of the
cereal boxes in a
grocery store at any given time. Razors are constantly updated with new lubricating strips and
extra blades and funky swivel tips. Every
product on the market is constantly undergoing revisions and refinements that give them 1% more functionality that is
marketed as some gigantic improvement.
But few companies pass that thin line that separates the over-inflated truth from the blatant lie. Those that do often have their bluffs called, and suffer the consequences of a consumer backlash. However, those companies who sell a majority of their products to consumers who wouldn't know better don't suffer at all, and thrive while shoving lies down consumer's throats wholesale.
One such company is Intel. However, Intel isn't satisfied merely lying straight to the consumer. They also lie to the sales people that populate stores such as Best Buy, CompUSA, Fry's Electronics, and other computer retailers. Thus, the people who help the consumer to make the decision unknowingly reinforce their lie.
Rather than go over Intel's consumer-oriented propaganda, I'd like to draw your attention to something you as a consumer may not see - The literature Intel gives to the retail employee. Now, before we start, you may already be asking, "Well, can't the person just not read this? And shouldn't he be informed enough to know it is bunk?" No, and not necessarily.
See, Intel offers 'incentives' to retail personnel in the form of massive discounts on Intel processors and motherboards. At the time of this write-up, Intel is offering a Pentium 4 2.53 GHz, a compatible Intel motherboard, and a copy of Windows XP Pro for $179. That's 33% the cost of the retail editions of those products through even the most competitive online retailer. In order to get this fabulous deal, though, the salesman must complete 6 Intel "training episodes," each containing a 5-question quiz. So sitting through the Intel BS-spiel and regurgitating the "correct" answers is something that thousands of sales reps across the United States and Canada are doing right now.
Ideally, these same salesmen would be well versed in the area of computers. They would know the architectural differences between the Pentium 4 and Athlon, what L2 cache is, and what components the front side bus connects. They'd know how to calculate theoretical bandwidth given bit-width of a channel and what it's clocked at, and return the answer in MB/s. Also, the salesman should know how to understand what this information means when they were selling computers, and to help the customer cut through the marketing mung that the consumer had been exposed to on TV or in magazines. These were all things I did when I used to work as a computer product specialist, and they're things I always made sure those I was working with knew.
However, this is not 1999. In the few years that have passed since then, I have watched computer stores slowly stop hiring those with solely computer experience, and focus on those solely with sales knowledge. That's not to say there aren't any well-informed types working the Home Office/Computer departments of retail outlets worldwide, but in my experience, those computer-savvy are few and far between.
A majority of the learning this new breed of sales reps gets is from websites like Cybersmart.com, and through the training sessions provided by the manufacturers. During manufacturer-run training sessions, the employees are fed fantastic food, given all sorts of free stuff ranging from pens to keychains, and then made to watch "informational" videos after their defenses have been lowered. So although they should be smart enough to see through the suspicious information, there's a high probability that they aren't.
Which brings us to Intel's outrageous "educational" material. I'm going to present three of the questions from a sampling of their quizzes, followed by their correct answer, and the statement Intel uses to justify the answer. I will then break down why they are utterly incorrect.
Episode 3, "Broadband", Question 1
Q: What element of the Intel Pentium 4 Processor keeps the broadband pipe large all the way to the processor?
A: The 533 MHz
system bus
With the 533 MHz system bus, content has a big, fat pipe all the way to the processor, so the broadband connection won't get backed up.
Overt Premises
- The 533 MHz system bus is big and fat
- The 533 MHz system bus is the only link leading from the Ethernet adapter to the processor
- The 533 MHz system bus will speed up broadband Internet browsing
Implied Premises
- Broadband connections contain so much data they have the potential to exceed the internal bandwidth of a computer that doesn't have the 533 MHz system bus
- Only the Pentium 4 has a 533 MHz system bus
Intended Conclusion
- If you're browsing the internet, you will be able to do so faster on the Pentium 4 than on any other processor out there.
Analysis
Overt premise 1 is correct. However, overt premise 2 is dead wrong.
Look first at the wording - "...content has a big, fat pipe all the way to the processor." The only way to interpret this is statement is that from the Ethernet cable to the processor, the data is on the front side bus, or, the data is traveling at the same speeds as the front side bus.
However, that's not the case. There are two possible ways for the data to travel in the computer, which is dependent on whether the Ethernet adapter is on the motherboard, or in a PCI slot. This little arrow diagram shows the two paths. Mousing over an arrow (---->) below will show you the bandwidth of that connection.
Ethernet adapter ----> Southbridge Chip --V-Link--> Northbridge Chip --FSB--> Pentium 4
Ethernet adapter --PCI Bus--> Southbridge Chip --V-Link--> Northbridge Chip --FSB--> Pentium 4
In the best case scenario (onboard Ethernet), the data still has to pass through Southbridge, across the V-link, through the Northbridge, and across the front side bus to get to the processor. The V-link only has a bandwidth of 266 Megabytes per second, while the FSB has a bandwidth of 4.2 Gigabytes per second. The V-link then becomes what is known as a bottleneck; regardless of how fast the FSB can get data there, it's still going to be slowed by the V-link. Therefore the second premise is false.
Only one premise of an argument needs to be false to be found an argument incorrect, however, I'd like to address implied premise number 1, because that is not only wrong, it's downright incorrect.
A 10/100 Ethernet adapter can transmit either 10 Megabits per second or 100 Megabits per second. Currently, cable modems only connect to Ethernet at the 10 Megabit per second speed. Completely ignoring the fact that most cable connections are rated for a maximum of 1.5 Megabytes per second, and assuming you're somehow able to download at 10 Megabits per second, the cable connection still only takes up .5% of the bandwidth of the V-link. There is absolutely no way a normal broadband Internet connection could saturate the bandwidth of the V-link, let alone that of the FSB. Even a T3 connection, at 45 Megabytes per second, only takes up 1% of the bandwidth of the FSB, and widely available T3-level broadband connections are more than a decade away.
The entire conception that a faster processor, or a wider FSB, can somehow speed up your Internet connection is a complete fallacy. The only benefit a processor could give is the ability to process the data faster once it gets there - And considering how little information makes up a web page compared to, say, a sort function in Excel or a filter in Photoshop, the difference between it and another processor would be unnoticeable.
It should be noted that the Pentium 4 is not the first processor touted by Intel to enhance the web experience. This practice started with the Pentium III, around the same time Intel began to lose their ability to bang the performance drum against AMD's offerings.
Conclusion
Intel's intended conclusion is invalid, and founded on at least two wholly inaccurate premises.
Episode 3, "Broadband", Question 2
Q: By keeping frequently accessed data close at hand, _______ brings a new level of performance to
online gaming?
A: 512 KB of
Level 2 cache
With 512 KB of Level 2 cache, data used to build the visuals of the game are accessed much quicker to virtually eliminate lag time, to keep the bad guys from eliminating you.
Overt Premises
- Level 2 cache keeps frequently accessed data close at hand
- The data used to build the visuals of the game are frequently accessed
- The data used to build the visuals of the game fit within the L2 cache
- The 512K of Level 2 cache on the Pentium 4 increases the performance for online gaming
- Lag time is almost totally eliminated by the L2 cache, to keep the bad guys from winning
Implied Premises
- Larger L2 cache will eliminate lag
- Lag effects your ability to play the game
Intended Conclusion
- The Pentium 4 will make you a better gamer
Analysis
Once again, a loaded promise - Buy the Pentium 4, and you will be able to better liquidate your opponents in online gaming. It's quite a statement.
Although tempted to quote figures on frame rates in order to test the veracity of this statement, the phrasing of their pitch makes frame rates moot. They specifically say, "(the) data used to build the visuals of the game are accessed much quicker to virtually eliminate lag time." They do not say that the L2 cache will increase frame rates and general game performance. Instead, they say "lag time."
There is only one term used in online gaming referring to lag, and that is the delay between the information from your computer reaching the game server and vice-versa. This is also known as ping, latency, or round-trip time. Due to the fact this type of lag is the type referred to in online gaming, and the fact that this question appears in the "Broadband" educational section, it can be implied that this is the type of lag Intel claims to "virtually eliminate" thanks to their L2 cache.
As we saw in the previous question, the benefits of the 533 MHz front side bus to internet connections are nonexistent. The L2 cache, something strictly inside the processor, is not going to have any effect on your lag either. Ergo, premise 4 is invalid, and the conclusion is therefore false.
Conclusion
Intel's intended conclusion is invalid, and founded on at least one wholly inaccurate premise.
Episode 5, "Speed", Question 2
Q: New digital video editing software has become simpler and...?
A: More processor-intensive
Digital Video editing is just one extremely processor-intensive applications that thrive on the faster Pentium 4 Processor.
Overt Premises
- Digital video editing software has become easier to use.
- Digital video editing software is extremely processor intensive
- Many processor-intensive applications thrive on the Pentium 4
- Digital video editing software runs better on the Pentium 4
Implied Premises
- Consumers care about the performance of processors in digital video editing software
- The Pentium 4 is faster than its competitors on all processor-intensive applications
Intended Conclusion
- The Pentium 4 is therefore faster than its competitors, period
Analysis
First, the overt premises. In the past several years, digital video editing has moved from being a niche of the professional market to something that home computer users are beginning to pick up. This is no doubt a result of the release of affordable consumer-grade digital video products. And it is very true that digital video editing, along with demanding high hard drive transfer rates and lots of memory bandwidth, requires a combination of good integer and floating-point CPU power.
However, the remaining overt premises tie in closely with the implied premises. To debunk them, we need to look at the processor's performance in at least two independent studies*, comparing the performance between the Pentium 4 and its closest competitor in several high-end applications.
The criteria of this analysis will be simple - price-performance ratio. Sure, you could have the highest performing component in the world, but what if it cost 10 times more than another product, which performs 9.99/10ths as well? There are obviously some applications where cost is not an issue, but in the real world, you odds of you not having some budget to meet are extremely unlikely. In each of these benchmarks, a lower score indicates the faster processor, since these are timed tests, and less time means faster speed.
This methodology will result in the processors ranked according to performance obtained per dollar spent. Intel does not claim that their processors meet this requirement; they simply claim better performance regardless of cost. However, I will provide both, allowing the reader to take in all of the information.
At the time of this write-up, the prices (all in USD, based on Pricewatch's lowest price) on the components being compared below are as follows. All Pentium 4 processors are the 533 MHz front side bus versions. All Athlon XPs are 133 MHz front side bus versions.
- Athlon XP 2800+ (2.25 GHz) - Currently unavailable. We will assume that the processor will debut at $400
- Athlon XP 2600+ (2.13 GHz) - $290
- Athlon XP 2400+ (2.0 GHz) - $180
- Athlon XP 2200+ (1.8 GHz) - $139
- Athlon XP 2000+ (1.67 GHz) - $82
- Pentium 4 (2.8 GHz) - $390
- Pentium 4 (2.53 GHz) - $218
- Pentium 4 (2.4 GHz) - $179
- Pentium 4 (2.26 GHz) - $174
Now, high-end applications would include video editing, of course. Adobe Premiere is one of the single most used video editing programs out there. Also, Photoshop ranks as high-end, since professional users often manipulate images exceeding 4000x4000 pixels, and performing even simple erases on segments of images that big is a massive tax on CPU and memory resources. Finally, there is Maya, the 3D rendering software that has taken the 3D world by storm and been used in countless movies such as Final Fantasy: The Spirits Within, Minority Report, and Spider-Man, from video games ranging from Grand Theft Auto III to Final Fantasy IX, covering basically every form of visual media that exists.
These 3 programs represent the most common of the high-end, covering video, 2D images, and 3D modeling and animation. So how does the Pentium 4 compare with the Athlon in these programs? The winner of each benchmark will be marked in bold, the second place will receive italics, and the rest will be listed in normal font.
Photoshop 7.0 - (smaller is better)
- Athlon XP 2800+ (2.25 GHz)
33 seconds * $400 = 13200 dollar seconds
- Pentium 4 2.8 GHz
35 seconds * $390 = 13650 dollar seconds
- Pentium 4 2.26 GHz
47 seconds * $174 = 8178 dollar seconds
Premiere 6.5 - (smaller is better)
- Athlon XP 2800+ (2.25 GHz)
263 seconds * $400 = 105200 dollar seconds
- Pentium 4 2.8 GHz
294 seconds * $390 = 114660 dollar seconds
- Pentium 4 2.26 GHz
360 seconds * $174 = 62640 dollar seconds
Maya - (smaller is better)
- AMD Athlon XP 2600+ (2.13GHz)
75 seconds * $290 = 21750 dollar seconds
- AMD Athlon XP 2400+ (2.00GHz)
79 seconds * $180 = 14220 dollar seconds
- AMD Athlon XP 2000+ (1.67GHz)
91 seconds * $82 = 7462 dollar seconds
- Pentium 4 2.53 GHz
91 seconds * $218 = 19838 dollar seconds
- Pentium 4 2.4 GHz
94 seconds * $179 = 16826 dollar seconds
- Pentium 4 2.26 GHz
99 seconds * $174 = 17226 dollar seconds
From this data we can see two things.
The first is that despite having a sheer clock speed advantage, the fastest Intel processor in each comparison nonetheless doesn't offer the performance of the highest (and in Maya, even the lower-end) AMD processor. Note that this does not prove the Intel chip is inferior, only that it is not as good at these specific programs. However, this can also be interpreted to say that the AMD CPU architecture is so well designed that it doesn't need a raw clock speed advantage to perform spryly. If you go strictly comparing Intel's best part to AMD's best part, the AMD part wins. Thus, the Intel part is not faster than its competitors.
The second is that there is no conclusive winner in the price/performance battle. Although AMD does beat Intel's high-end part in 2 out of the 3 tests, the price performance crown does go to the Pentium 4 in those same 2 tests. In the Maya test, AMD's 1.66 GHz part, the Athlon XP 2000+, manages to tie the Pentium 4 2.53 GHz in speed, while also offering the best price/performance ratio. The rest of the AMD processors get even faster, but lose the ratio battle once above 2 GHz. The results of this measure can only be deemed inconclusive.
Conclusion
So, if you use one definition of "better," AMD wins. If you use the other definition, there is no definitive winner. That is hardly faster than its competitors, period. Therefore, this argument is false.
Final Thoughts
As much as we may
dislike the truth, there are many companies willing to stretch the truth thin, or in some cases, turn it
inside out altogether. That's one aspect of
capitalism that, though
repugnant, is to be expected. However, we as consumers do have a
voice. We can become educated; spot the
inconsistencies and
mistruths that are fed to us, and
smack the hand that would deceive us.
Intel isn't a bad company. They helped revolutionize the world of processors, and we probably wouldn't be where we are today if it weren't for them. However, when a company has a deathgrip on the market, they often become slothful and rely on simple name recognition to help them sell more units than innovation, and that's bad for us. Innovation is what drives products to be designed stronger, cheaper, better, and faster.
Intel has become lazy, and that shows in their products. A relative newcomer to the consumer processor industry, AMD has only had a truly competitive product to match against Intel's since the K6-2 processor debuted in 1998, and they only began to trump Intel in the benchmarks when the Athlon exploded onto the market in 1999. The fact that a company with less resources managed to design a processor that can run 15% slower than their processors and match them in many applications is a testament to the fact that Intel lost the edge.
If Intel hit the drawing boards and really went to work, they would likely be able to stomp AMD into the ground. Instead, they've focused hundreds of millions of dollars trying to convince retail sales associates that their product is better, when in truth, it's equal at best. In turn, these same salespeople have been entrusted with finding the best use for consumers' money. That's just not right. There's nothing more ugly than when good products fail because an inferior product was marketed better.
So get out there. Get informed. Vote with your wallets. Visit the Intel website listed below, sign up, and go through their training episodes. There's nothing wrong with pretending to work at a store, and as long as you don't win anything, you won't hear anything from them. And then decide who's getting your money next time you buy a computer or components. Because how you decide will have an impact on the future of computing.
Intel RetailEdge Website:
http://www.intel.com/retailedge
(*) Performance benchmarks from:
Most of my knowledge about bus speeds, processor and system architecture, is gleaned from my past experience and years of reading, repeated straight off of the top of my head. Having gone over this thrice, I'm willing to vouch for the integrity of my statements, however, I may have slipped up somewhere. Don't be afraid to correct me if you catch something.