AMD (Advanced Micro Devices) or Intel? That is the question, eh? I've toyed
with the problem for a number of years and for the last couple of those years
I compared systems to determine whether the AMD based system was ready for the
enterprise yet (um, it isn't but that's food for another discussion). My most
recent experience with AMD is making me question whether AMD systems are even
ready for the home either.
I used to work with a very capable engineer named Bill Grable and he had a
working theory on electronic devices, which still seems reasonable in retrospect.
He postulated that every electronic device only had so much smoke built in at
the factory. The unit's lifetime, therefore, was regulated by how much smoke
you let out. When we'd bring him equipment from the field, he'd ask, "Did
you let all the smoke out of it?"
Bill, I woke up and all the smoke came out of my Athlon overnight!
Figure 1: The failed CPU. As you can see, the CPU paste only marginally
overran the mask. The failed area to the right of the chip is where the circuit
in the CPU die failed.
I know, I know, Intel is overpriced, P4 Pentiums under-perform and, besides,
they're another monopolistic behemoth. From the AMD side, most of the fondness
expressed for these processors comes from the crowd of people who have adopted
overclocking as their hobby. I'll admit, it's a fun thing to do…so long
as someone else is buying the hardware. There is a significant rush from realizing
you just pushed a 1.8 GHz processor up to 2.3 Gigs without a meltdown. Way cool!
Another bastion of AMD devotees points to the often significant difference in
price between a new Intel system and building the equivalent AMD solution.
But this particular article is about what happened to my monetary savings when
I tried to keep an AMD Athlon 2200 XP+ processor cool…at its rated clock
speed. Oh, and I should mention that the processor died in a most unpleasant
way on a Sunday morning. Perhaps it was the Sunday part that really raised my
ire since that meant a day of delay before I could begin to rebuild the system.
To start, this project was given a healthy boost by the wonderful sense of
timing of a good friend who happened to stumble over an excellent deal on the
processor. The money changed hands and I had that new processor, a nice, new
Gigabyte 7VAX-P motherboard (nice unit!) and a new ATI Radeon 9700 video card.
My geek gland was saturating my mind at the prospect.
From the start, this was designed to be an experiment while serving the purpose
of updating my 14 month old system. It's an 800 MHz PIII overclocked to a modest
896 Mhz. The plan was to take this older machine, set the clock back to its
design speed and replace our aging Windows 2000 server, a tireless old 500 Mhz
Intel system running with way too little memory, hard disk and too many services.
After a week of rebuilding and hardening the old desktop for Server duty and
archiving the file system, the Server promotion project was complete and I turned
my attention to the AMD project.
Almost immediately it was evident that there were going to be problems with
heat. The processor, obviously, ran much hotter than my previous system and
getting air through the chassis at a high rate was a real challenge. This case
was a 3 fan unit with an additional duct placed over the CPU socket in an attempt
to place cooler air right in the CPU environment. After replacing all the fans
in the chassis (including those internal to the power supply) I had still not
managed to improve the reliability of the system and heat caused lockups to
occur once every hour or so.
Figure 2: The circled area indicates where the jet of the failing CPU penetrated
and damaged the motherboard
Rather than abuse the processor, I shut the system down and ordered a new case
reputed to be amongst the best airflow systems on the market, the TermalTake
Xaser II (http://www.thermaltake.com/products/Xaser3Menu.htm).
With 5 case fans, rounded IDE and Floppy ribbon cables and the careful routing
of the power harness and SCSI cables, this unit proved almost up to the job
although processor temperatures remained well above 100 degrees Fahrenheit.
If these cases look familiar to you it may be because these cases are the visual
foundation for Alienware's line of high-performance desktop computers (http://www.alienware.com).
For 3 months, the system appeared to be working as expected but I had already
blown the extra dollar savings of going AMD in the attempts to get the environment
under control. My total investment at this point was $504.00 US (not counting
the video card). In fact, that cost is nearly $100 more than my cost would have
been for the equivalent Pentium 4 setup and I wouldn't have had to replace the
case to do the job.
Now, in these modern processors, the major players claim to have built in Thermal
Overload Protection and it is at this point only that I will throw away the
casual hobbiest's demeanor and solidly condemn the engineering principles in
use by AMD. You should know that it was nearly 2 years ago that I first watched
a demonstration video at Tom's Hardware (http://www.tomshardware.com/)
in which it was demonstrated that AMD's TOP system left much to be desired as
the result of an uncooled CPU often meant a ruined processor. The Intel units
seemed to work much more safely than this. I figured that surely, after 2 years,
AMD would have fixed this problem. After all, these little muffin fans we are
strapping to our CPUs DO fail!!!
I couldn't have been more wrong and the results of that assumption could barely
have become worse. The fact is that as the CPU overheated, it caused a number
of problems that resulted in a nice jet burn of the motherboard which then resulted
in a loss of load on the motherboard which then proceeded to drop power…including
power to the fan.
Figure 3: The bottom of the CPU die where strong evidence of heat and jetting
illustrates the final milliseconds of this processor's life
Subsequent tests showed the fan to be in great shape. In fact, this particular
unit is the Thermaltake Volcano 7 + fan/heat sink combo (http://www.thermaltake.com).
It's loud (having one under your desk is akin to having a squadron of jet fighters
idling just north of your knees), fairly expensive (@ $40 retail). Having the
load removed from the power supply by the cooking CPU just worsened the experience.
When I had finished examining the failed CPU and had started to wonder about
the condition of the motherboard I found I was becoming just about as hot as
the CPU had been. Since I wanted to survive the episode, unlike the CPU, I decided
to shutdown for a few minutes and head over to my favorite computer parts counter
and see their reaction. Here are a couple of my favorite comments from the discussion:
"Yep, AMD stuff isn't long lasting."
"Look around. See all those warnings taped to the wall? Go compare
how many are about Intel and how many are about AMD."
Interestingly, there was a single Watch out or else! statement about
the Intel processors. There were, however, a wide variety of warnings concerning
uneven stress of the die while attaching a heat sink, over/under greasing the
CPU and warnings that store policy dictates the sale of a cooling fan with every
CPU, all of them directed at the purchasers of AMD processors.
And my favorite comment?
"Do you mind if we keep that processor? That's the cleanest example
of a CPU burn we've seen."
In other words, nothing was obviously wrong with the way the CPU installation
had been handled. This CPU is now a quality artifact. The Burned CPU Museum
in that store has several damaged CPUs on display. Only one is an Intel and
it is of the Celeron vintage from a couple years ago.
I've learned the lesson and now I will strenuously argue the point that not
only are AMD processors deficient in stability for the enterprise, but the idea
that you save money by purchasing one is, by my experience, thoroughly debunked.
Figure 4: The GigaByte 8INXP motherboard
In the end, this experiment resulted in rebuilding the core of the system with
a Pentium 4, 2.4 GHz processor on a Gigabyte GA-8INXP1394 motherboard (http://www.giga-byte.com/MotherBoard/Products/Products_GA-8INXP.htm)
. The Retail Box processor comes with a quality cooling fan and heat sink. The
processor die has a much larger chip footprint affording a broader contact to
the heat sink AND the system for attaching the cooling unit is much, much better
than that used by AMD designs. It is a twin cam-lock lever system that you operate
easily after gently laying the heat sink within a couple guide standards (assembled
within the orange area of the picture in Figure 4).
Total cost to replace the failed components AND upgrade RAM? $430.00. Someone
with a more creative sense of mathematics will have to do some good work to
show me how the AMD solution was cheaper in the initial investment than a comparable
Intel system. I won't even demand that the equation account for the extra costs
incurred for high performance cooling units, cases and the tax on your personal
well-being from listening to the annoying whine of that huge cooling fan required
for the AMD system. Yes, the cooling system on the Intel is nearly impossible
to hear and I count that as an unexpected but important bonus. Across the room
is another AMD driven system which Dian uses for recording. She complains of
a background whine which is always present in her recordings. It's not the microphone
causing the whine. No, she, too, has one of those huge, high velocity fans and
the whine matches the frequency of that cooling unit.
 In the end, it's always your money and your opinion versus mine. But, if you
ask me, I'll tell you to save money, time and aggravation by simply turning
down any tempting offers on an AMD solution. In a couple more years, we'll re-examine
the issue and see if customers have forced them to learn anything about reliable
designs. Until then, help them with the lesson and simply don't buy one of their
half-baked solutions. Half-baked? Yep, you get to finish the cooking at home
when you buy from AMD.
|
