• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

ISA Rockwell v.34 modem

With the release of the 1.7 GHz Pentium 4, discussions comparing various Intel and AMD CPU's have been the hottest topic on the forum. At the same time, several new forum members have made bald attempts to discredit accepted facts about the current offerings on the CPU market. Because the vast majority of their posts deal with this topic, and are filled with errors and half truths, few experienced members take them seriously. Still, there is the potential for newer ones to be mislead. It never hurts to set the record straight, even if most people have heard it before. So:

AMD or Intel for a new system? Or more specifically, P4, P3, Celeron, Athlon, or Duron?

To answer this, we must repeat the three criteria for judging a system: Price, Performance, and Reliability. Extremely accurate price and performance data are readily available from any number of verifiable, objective sources on the web, so this is what we will address first.

(For these comparisions, we will exclude the P3, Celeron, and Duron processors. Their price/performance ratios are hardly debatable by now: the Duron 800 at $48 is faster on any platform than the $86 Celeron 800. And the P3/800EB, at $146, costs three times as much as the Duron while delivering, at most, 15% more performance. Intel is reportedly developing a .13u "Tutalin" P3 with minor enhancements, but until then, the P3 and Celeron are moot in a serious CPU discussion. And the Duron, while proving higher value than any CPU in the history of the universe, is not available at high enough speeds to seriously challenge the P4. So we'll stick to the Athlon and P4.)

For performance data, we can defer to the oldest and most respected hardware reviewers on the web: Anand and Tom, in their P4/1.7 reviews. We will limit ourselves to real world benchmarks, which means excluding theoretical memory bandwidth tests and MadOnion stuff; these benchmarks have their place in exposing underlying factors, but are not appropriate in a product comparison. In the case of SYSMark 2001, we will just take the scores for Internet Content Creation and Office Productivity (the "average response time" results always mirror these scores anyway.)

Anand's P4/1.7 Review

P4 1.7 GHz leads in:

SYSMark 2001 Internet Content Creation
Quake3 Arena
Mercedes Benz Truck Racing

Athlon 1.33 GHz leads in:

SYSMark 2001 Office Productivity
CSA Research OfficeBench Constant Computing
Unreal Tournament
Serious Sam
WAV --> MP3 encoding

Tom's P4/1.7 GHz Review

P4 1.7 GHz leads in:

SYSMark 2001 Internet Content Creation
Quake3 Arena
Dronez
Mercedes Benz Truck Racing (exact tie!)
WebMark 2001
FlaskMPEG Encoding

Athlon 1.33 GHz leads in:

SYSMark 2000
SYSMark 2001 Office Productivity
CSA Research OfficeBench Constant Computing
Unreal Tournament
Evolva
MDK2
Mercedes Benz Truck Racing (exact tie!)

As we can see, the Athlon 1.33 always edges out the P4 1.7 by one or two benchmarks. On Anand's tests, the Athlon won five of eight benchmarks, while Tom had the Athlon winning seven to the P4's six. Not a huge performance difference, to be sure, but significant because even though Anand and Tom selected quite different benchmark suites, they both managed to show the Athlon as slightly faster.

(Some might argue that the P4's SSE2 optimizations are not fully supported in today's applications. That is certainly true, however, the same could be said for the Athlon's 3DNow! extensions. Besides, a smart computer hardware purchase is one that achieves maximum value today. The industry advances so fast that it is foolish to try to predict what will still be useful a year from now. Ultimately, longevity is simply a function of today's performance degraded over time. The industry pays little or no respect for hardware that was supposed to last a long time. Newer innovations always bury the old ones -- disposable computing.

In the case of SSE, 3DNow, and similar instruction sets, can anyone honestly say that having 3DNow! on their K6-2 gave it a longer useful life? Or what about that Katmai P3/450 -- did SSE make it stack up better to today's $46 Duron 800's? Obviously not, illustrating a key point: buy whatever gives you the best bang for your buck today. Let tomorrow bury today.)

With these respected sources, we can safely say that the Athlon 1.33 GHz is at least slightly faster than the Pentium4 1.7 GHz on the balance of modern real world tests. Now, let's move on to price. (The lowest PriceWatch figure will be our guide here):

There are two ways to compare prices between these platforms. One is the chip prices alone, and the other is to include the cost of an average motherboard and supply of RAM, and case/power supply if it is needed. For this comparison, we will assume that a specialized Intel-approved P4 case and power supply (with the new electrical standards) will cost about as much as a standard ATX case/PS that is AMD approved. Actually, the P4 case will likely cost more, but it's a minor detail. We will also assume an AMD-aproved Socket A HSF will cost as much as a P4 fan.

Pentium4 1.7 GHz $372
Intel D850GB Mainboard $159
2 x 128M RDRAM @ $85 each $170
------------------------------
$701 - Total

Atlon 1.33 GHz $194
Epox 8K7A AMD 760 Mainboard $134
256M PC2100 DDR SDRAM $98
-------------------------
$426 - Total

As we can see, a P4 1.7 GHz setup will cost roughly 65% more than a (slightly) higher performing Athlon 1.33 GHz setup. If we simply compare the cost of the chips, the Intel price gouging rises to 92%. Obviously, the Athlon beats the P4 on a pure price/performance basis. Few argue that. The only factor left is reliability.

The CPU industry is unique in that long term reliability is rarely needed or designed for. Even the best of today's engineering results in a chip with a rated life of about ten years. This situation is acceptable because the value of computer hardware depreciates so fast that in even three years, most products are nearly worthless. New disposable hardware replaces the old.

Short-term reliability, however, is important in reducing down time and increasing productivity. CPU's themselves are notoriously reliable, with extremely low RMA rates. As I and many other computer consultants and resellers can attest, CPU's are sent back deffective even less than SDRAM chips, which carry a lifetime warranty. These RMA rates are easy to measure and, for CPU's, usually end up being statistically insignificant. It appears both Intel and AMD have excellent quality control methods that weed out bad batches of chips.

System stability is an aspect of short term reliability that is not so easy to measure. It tends to be a hot topic among computer hardware enthusiasts, and it is hard to quantify like performance. Are systems based on Intel processors and their chipsets inherently more stable than those based on AMD processors and their chipsets?

Logically, the first question to be asked is, what is system stability? In this age of complex, bloated operating systems, rushed software applications, myriad hardware possibilities, and user ignorance and frustration, computer crashes are common. Everything from a simple lockup to an "Illegal Operation" error to the dreaded Blue Screen of Death is accepted as par for the course in the computer industry. But the difference between a stable system and an unstable system is that the stable system will exhibit these symptoms far less frequently, and will tend to do so in response to specific, identifiable problems, as opposed to random, intermittent conditions.

What is the main cause of system instability? By far, software: operating systems, applications and drivers are so large and complicated these days, and share so many of the same resources, that it becomes difficult for the OS to manage them and keep small errors contained. Still, defective, poorly designed, or overclocked hardware can contribute to the mess (e.g. Intel's Pentium Classic FDIV bug, their i820-MTH, or their P3/1.13 GHz, respectively).

So are Intel-based systems with Intel chipsets any more stable than AMD-based systems with AMD or VIA chipsets? In short, no.

Why?

1) Not a single research study by an objective outsider, at any time or place in the universe, has established that either platform provides a modicum of added stability over the other. There are three reasons for this. First, no knowledgeable computer hardware expert seriously believes that there is enough of a stability difference between the platforms to warrant such a comparison. Second, because computer crashes are so difficult and time consuming to track, diagnose, and record, and because of the inherent challenge in defining the test parameters (i.e. what is normal system usage and how do you reproduce it?), no controlled scientific study has ever been undertaken. Third, the financial resources necessary for such broad research basically exclude any smaller groups from attempting it, leaving the job to industry heavyweights such as Dell, Intel, and Microsoft, who have a vested interest in maintaining the status quo of ignorant consumer brand recognition.

So, seeing as there is no hard evidence in favor of either platform, we can only pronounce them equal. Certainly we cannot go with a "gut" instinct or base our decision on the financials or market status of either company. Like a court of law, both are innocent until proven guilty. Therefore, the discussion should end here. Just in case it doesn't:

2) The so-called "issues" with AMD/VIA systems are usually not such.

There is currently a grand total of one (1) compatibility issue with VIA chipsets -- the SBLive! resource hogging problem. This problem can also potentially affect Intel i815-based motherboards. Without the slightest support from Creative Labs, VIA has gone ahead and developed a fix with the help of the hardware community. It will be included in their latest 4-in-1 service pack, which they say will be released this week. Considering the strange conditions that trigger the problem, and how Creative Labs has been completely uncooperative in resolving something they are at least partly reponsible for, VIA's response time here has been excellent. The issue only surfaced, I believe, less than two months ago, and a patch was found within weeks. No high-profile campaign by famous hardware sites. No public recall. No defective hardware revisions. Just a simple software fix. Actually, there hasn't been a single, repeatable bug in VIA's recent chipsets that has required a new hardware revision or complete recall. Intel, take note.

There is also the widespread myth that VIA/AMD chipsets are incapable of reliable AGP 4x operation. This is absolutely false. Every single AGP 4x card in the universe works perfectly in that mode on the KTxxx chipsets. The only "issue" is with nVidia's recent Detonator3 drivers, which contain a subtle bug that in some cases renders the feature unusable on VIA chipsets. But since every other AGP 4x card words perfectly, and since nVidia cards also did with the Detonator2 drivers, the bug is obviously on nVidia's side. Besides, AGP 4x has been shown to provide absolutely no real world performance benefit over AGP 2x. There is no stability or performance loss versus Intel systems.

Regarding "thermal issues", it is true that Athlon systems dissapate more heat. Heat, however, is only an issue once it affects stability, or forces you to spend more on cooling. In the case of the Athlon, it does NOT. The chip is perfectly stable under normal operating temperatures, and compatible SocketA HSF units are available for under $5.

There were some reports of "core cracking" on early SocketA chips due to improper HSF installation. Having built over 200 AMD systems, I can tell you that it would take a MAJOR screwup to crack a core. Unless the person installing the HSF is extremely rushed and negligent, there is no reason it has to be a problem. AMD's step-by-step picture guide has gone a long way toward helping inexperienced do-it-yourself builders calm down and take it easy during tricky bits like this. Regardless, the same delicate procedure should be followed on both Intel and AMD systems, to prevent chipping and scratching.

To best set up a stable AMD/VIA system, you follow the same procedure as with any Intel system:

- purchase and install an approved power supply and CPU fan
- install the HSF according to AMD's step-by-step picture guide, which is the best method for Intel systems as well (apply some kind of thermal transfer compound in a thin layer before starting)
- locate the latest non-BETA flash BIOS and apply it
- set the appropriate BIOS settings according to the motherboard manual and/or manufacturer's website FAQ
- install windows
- download and install the very latest non-BETA chipset drivers (VIA 4-in-1 service pack and USB filter for AMD systems; Intel INF Updater and Ultra ATA Storage Driver for Intel systems)
- go to WindowsUpdate.com and get every single Microsoft security fix and product update
- download the very latest non-BETA drivers for your sound, video, modem, network, and other peripherals
- install these

The above procedure should be followed by anyone wanting a stable system -- Intel or AMD regardless. As you can see, an AMD/VIA system demands no significant effort over an Intel system, provided you are interested in maximum performance and stability out of both, and are not content to simply install Windows and start hammering away.

VIA's 4-in-1 service pack is an elegant solution to a terrible lack of support on Microsoft's part. It demands all of thirty seconds to install, and is no more time consuming that Intel's own INF Updater. 4-in-1 issues only arrise when people feel compelled to upgrade to the latest VIA release at the drop of a hat.

"But why do I see so many problems reported with AMD systems these days?"

The reason you see so many forum posts asking for Athlon help is simple -- many people are buying Athlons!

Smartly, few people who visit these forums are building P3 systems. I'd say that, at most, 10% of the AnandTech systems built this month were Intel based. It stands to reason, then, that only 10% of problems reported will involve Intel systems, while the vast 90% bulk will involve AMD systems. This has nothing to do with the reliability of either platform. It's simple statistics.

By the same token, the Celeron 300a looked like a terrible chip because about 30 months ago, all you saw on this forum was people struggling to get their Intel BX board to work with some stick of RAM or a weird SBLive! IRQ problem. The flavor of the month always looks bad when you head to the support forums, and the Athlon is no exception. But statistically, competent system builders have as many problems with Intel systems as with Athlons. It just happens that Athlons are popular these days.

There is a real danger here of certain people taking their limited personal experiences to be the absolute authority on a given topic. In consumer research, this is referred to as "small sample syndrome" -- if Bill tries Acme Auto Repairs once and receives poor service, he will tend to believe Acme Auto Repairs always provides poor service, regardless of their proven track record. There's a strong tendency for people to give more weight to negative hearsay than positive hearsay. It only takes a few bad recommendations from ignorant people to sway a large number of persons into going with the "safe, quality brand".

Those whose weak minded brand loyalty lead them to launch desperate attacks against alternative platforms from AMD and VIA play perfectly on the pre-conceived (and usually wrong) notion that "you get what you pay for". In reality, the computer industry has consistently shown this to be false -- you can just as easily get much more (i.e.. AMD Duron, ATi Radeon LE) than you pay for or considerably less (RDRAM, P4 1.3 GHz).

Take it from me, and the handful of other experienced resellers and consultants in this forum: AMD and VIA systems are no less reliable than Intel systems.

3) AMD/VIA systems have been shown by top hardware reviewers to be incredibly stable, easily matching Intel platforms.

AnandTech, bar none the most respected, unbiased computer hardware reviewer in the universe, has examined Athlon, P3, and P4 chips since they were first released. Has AnandTech, or any other respected hardware reviewer, ever in the past two years criticized the stability or reliability of a shipping AMD product or platform? Reading through the archives, we can find nothing but glowing praise beginning with the Athlon and the first production-level AMD 750 boards. On the other hand, Intel's supposed world class reputation has been tarnished several times recently, particularly by two high profile hardware recall fiascoes, one of which required public intervention from the hardware community.

There is actually no more effective endorsement of the stability and reliability of AMD/VIA platforms than the fact that AnandTech uses them as the sole platform for the web serving of its main site, entrusting them with a loaded job that is tied to their own success as an enterprise. What have been the results of this arrangement to use both Intel and AMD based systems behind the scenes at AnandTech?

"Eight months ago we asked the question of whether or not Athlons could be used as servers, and the answer was an astounding yes. . . They did such a good job that we devoted our remaining dual Xeon based webserver (formerly www5.anandtech.com) to hosting the AnandTech Forums which have gone through an extreme growth-spurt of their own."

"We'll keep on adding more boxes to the server farm as the needs grow, but for now we're definitely happy being powered by both AMD and Intel based servers; how's that for the best of both worlds?"

Now obviously, there are other methods for testing system stability. But we're hard pressed to find anything more fitting than devoting both platforms to the duty of running one the largest non-adult, non-commercial sites on the Internet. And if AnandTech says that AMD's Athlon can be taken seriously as a stable server processor, what does that mean for the vast majority of computer users who simply want to run one on their desktop?

Another important piece of evidence is AnandTech's November KT133 Motherboard Roundup, where the majority of boards displayed BX-level stability, and a handful from Microstar, ASUS, and ABit displayed incredible reliability in a 24 hour torture test, with the cheap MSI K7T-Pro-2a crashing a grand total of zero (0) times, making it (at that time) the most solid modern motherboard on either side of the Intel/VIA fence.

Also see Anand's FIC AD11 AMD 760 review. This board, one of the first AMD 760 products produced by a second-tier brand and still one of the cheapest AMD 760 boards, ran an amazing 34hr before crashing under the torture test.

And from Anand's ASUS A7V133 review: "We used to run the stability tests for 24 hours, but last time we had to run for 48 hours before we finally gave up on the MSI K7T Turbo crashing. This time around we also ran for 48 hours to prove that the A7V133 is just as stable as the K7T Turbo. There is no doubt the [VIA KT133A] A7V133 is an incredibly stable motherboard."

Tom's Hardware, historically the most visited hardware site, said in their recent KT133A Motherboard Roundup, "The most important finding was the enjoyable fact that each of the tested boards ran 100% stable even at the fastest possible memory timing settings. VIA's upcoming DDR chipsets may not look too impressive right now, but the Apollo KT133A is a matured, fast and solid product that offers good performance."

Tom's Hardware also recently published AMD Processors Vs. Intel Processors - Facts and Lies. Among other things, they made the following points:

"AMD Processors are significantly less expensive than Intel processors although they are at least on par in terms of performance. - FACT"

"AMD processors are incompatible. - LIE

Not that the average guy who just heard that phrase would know what the heck 'incompatible' is, but it sounds really bad, doesn't it? Well, even the people who do know that 'incompatible' means that a product wouldn't work reliably with other components (which of course is bad) are wrong if they accuse AMD's Athlon or Duron processors of it. In our labs we are testing all kinds of Athlon platforms with all kinds of different components and I can definitely say that I cannot see any difference between the compatibility of AMD products and platforms compared to the same from Intel."

"Chipsets for AMD processors are inferior to Intel chipsets. - LIE

Yeah, sure, the earth is flat and politicians are honest ... I am still amused when I see people posting the above message in news groups or as their response to articles. How many more times does Intel need to screw up their chipsets (i820, MTH, ...) until you guys get the message? . . . Incompatibilities are more a problem of the motherboard BIOS than of the chipset right now. Thus both chipset makers, Intel as well as VIA, are actually in the same situation."

From the above sources, we have addressed the three deciding factors in CPU evalutation: performance, price, and reliability. And even though most of us already knew the conclusion, it bears repeating: AMD systems are currently the better choice.

Modus
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top