AMD Up To Today

Picking up where we left off, this episode about AMD looks at its more recent history and how it has played a role in tech.

Learn more about your ad-choices at https://www.iheartpodcastnetwork.com

See omnystudio.com/listener for privacy information.

Episode Transcript

Speaker 1 (00:04):
Welcome to tech Stuff, a production of I Heart Radios,
How Stuff Works. Hey there, and welcome to tech Stuff.
I'm your host, Jonathan Strickland. I'm an executive producer with
How Stuff Works and I heart Radio and I love
all things tech. And in our last episode, which was
requested by listener Stephen, I left off with a m

(00:26):
d's history in ninety six, when the company found itself
reeling when Intel decided to cut ties. Up to that point,
a m D had been in an agreement to act
as a second source for Intel designed chips. Intel would
get a licensing fee and a m D would be
able to manufacture chips based on Intel's designs. But when

(00:49):
a m D chips started to do better in performance
tests than the Intel originals, things changed. Until ended a
ten year agreement several years early, and the two companies
would enter into a lengthy court battle that would ultimately
go all the way to the U. S. Supreme Court.
But we've got a few years to get through before

(01:10):
we get to that. So Intel had introduced the eighty
three eighty six microprocessor in nineteen eighty five, a year
before it severed the agreement with a m D. The
three eight six, as it was known, was a thirty
two bit micro processor. It could run most older code
designed for its predecessors, and was capable of faster clock rates,

(01:33):
meaning it could run more operations per second, and it
had a greater data bandwidth, which means it could run
operations on larger amounts of information than the earlier chips could.
Intel an a m D had set the stage for
creating the standard in computer processing, and now Intel was
determined to stand alone. A m D, for its part,

(01:53):
have been designing its own version of the three eight six,
called the get Ready for It a M three eight six,
but Intel's decision to end the agreement through things into disarray.
Intel argued that their agreement with a m D only
covered the eight eighty six through the e D two
eighty six family of micro processors, and that the three

(02:15):
eight six and later iterations were excluded. A and D
obviously disagreed with this interpretation of their agreement, claiming that
all x eighty six derivatives were covered under this ten
year plan that they had struck with Intel back in
nineteen eighty two. At this same time, things were shifting
in the PC market. You might remember from my episode

(02:38):
about early computer systems that there used to be a
ton of different types of PCs on the market in
the late seventies and early eighties, each with its own
hardware and operating system. The ones we think about today
are Windows based machines and Mac computers as far as
personal machines are concerned. But up through the early nineteen eighties,

(02:59):
the field was much more crowded. You had companies like Tandy, Commodore,
and Amiga and others competing with Apple and IBM. However,
by the mid nineteen eighties the field had thinned out significantly.
IBM had secured valuable deals with corporations, becoming known as
the computer of choice for office workstations. Apple maintained a

(03:21):
more niche market of users interested in the creative power
of the Macintosh. Everyone else sort of began to fade away,
and this left the IBM PC and it's compatible clones
with the lion's share of the market. In fact, by
the time Intel was trying to block a m D,
the IBM PC market share had grown to about eighty

(03:43):
four percent of all personal computers, so this was a
really big deal. A m D had helped cement the
X eight six chip as the go to microprocessor for computers,
and now it looks like Intel was going to run
away with the whole thing. Things were starting to smell
a little anti competitive. Am D did still have an

(04:06):
agreement to the underlying instruction set for the X eight
six family of processors, so in some respects a m
D was still in the game, but Intel wasn't going
to share the actual physical design of the three six
microprocessor with a m D. So the engineers at a
m D set out to reverse engineer the three six

(04:27):
and build their own version of it while the legal
battle continued in the courts. Reverse engineering alone is a
pretty fascinating subject. The basic concept is fairly intuitive. You
take a technology and you examine it closely and you
figure out what makes it tick, how does the tech
actually do whatever it does. Then you go back and

(04:50):
you build your own version of that technology based upon
your understanding of how the starting example worked. So you're
not starting off with some sort of blueprint or set
of plans or instructions. You're sussing it out on your
own based on existing instances of the technology. So it's
a bit like detective work. Between reverse engineering and the

(05:11):
legal battles. It would be years before a m D
could bring its own three eight six chips to market.
The company began releasing its version starting in nineteen, and
once again a m D S version of Intel's chips
were clocking in at a faster clock rate than the competition.
Intel's three eight six chips maxed out at thirty three

(05:32):
mega hurts, whereas a m D S could hit forty
mega hurts. The legal battles continued, and a m D
began to invest in designing its own microcode for chips.
Intel's next microprocessor was predictably the eight four eight six,
and a m D created its own version the A
M four eight six. Some A M for eight six

(05:56):
chips had Intel microcode from the x A D six
agreement and others had a m D microcode, making it
a little confusing, and all of them were outpacing Intel's
version of the same chip. Even the top of the
line microprocessor in Intel's forty six line was left behind.

(06:16):
The fastest four D six from Intel had the top
clock speed of one hundred mega hurts. A m D
S version was able to reach speeds of one D
twenty mega hurts. Now this was in When that legal
battle I talked about finally concluded, the courts found in
favor of A m D, granting the company some royalty

(06:37):
free use of some of Intel's patents and awarding A
m D millions of dollars in the process. But the
whole endeavor had taught the engineers at A m D
a valuable lesson. While they had won this battle, there
was no guarantee that things would remain stable between Intel
and a m D, so the company did release another
x E D six derived chip. This one was called

(06:59):
the A m D five x eight six, or five
by eighty six if you wanted to think of it
that way. And you may be thinking, ha, I never
heard of a five eight six computer. I'm pretty sure
that Intel switched over from four A D six to pentium,
and you would be right. The A m D five
x eight six chip was based off the same architecture

(07:20):
as the A M four eight six microprocessor, but it
did manage an even faster clock speed out of the box.
It was a hundred thirty Mega hurts, but original equipment
manufacturers or o e m s in the biz could
get an even faster version than maxed out at a
hundred fifty Mega hurts now according to Tom's Hardware, a website,

(07:40):
which is, by the way, a great resource if you
ever want to learn everything there is to know about
just about any computer component you can think of. The
A M four, A D six, and the five X
eighty six processors also moved the floating point unit or
FPU over to the central processing unit or CPU itself.
Up until then, it had been customary to have separate

(08:02):
CPUs and FPUs that would connect to each other through
the motherboard. So I guess it's time to give a
quick explanation about what these things actually mean. The CPU,
I'm sure you've all heard of, right, It's sort of
the head manager of your computer. It executes basic instructions.
In the event that the instructions require the use of
a specialized chip like a graphics processing unit also known

(08:26):
as a GPU, the CPU can delegate those tasks to
the appropriate hardware. It's a high functioning component of a computer.
We often refer to it as the brains of the computer,
but really it's just calling the shots at the highest level.
The floating point unit carries out instructions on what are
called floating point numbers. A floating point number is a

(08:47):
workaround for a particular problem, which is how a computer
represents real numbers. The range of real numbers is infinite,
but a computer can't handle that. A computer has a
limited capacity, so programmers use floating point numbers, so called
because the decimal point has no fixed number of digits
that have to appear before or after it. This allows

(09:09):
programmers to represent numbers separated by many orders of magnitude.
You can have incredibly large numbers paired with incredibly small
numbers using this approach. UH typically you would use a
variant of scientific notation for those really big or really
small numbers. However, this does mean that much of the
work computers do happens as approximations rather than as precise calculations,

(09:34):
and this introduces the possibility of error. The more you
are approximating something, the less accurate or precise it's going
to be, particularly as you perform more calculations based on
previous approximations. As these approximations start to add up, you
can potentially get further and further away from a correct

(09:55):
or true answer. But never mind that that's a discussion
for a different episode. Now, after the four a D six,
Intel came out with the first Pentium processor. So why
did Intel change things up? Why did Intel go from
four D six to pentium? Because the Pentium still followed
the x A D six architecture and instruction set and
spoiler alert, so do today's computers. So why would Intel

(10:19):
choose pentium instead of sticking with the naming convention it
had created. Why wasn't it the five eight six? Well,
the main reason was, as I'm sure many of you
have guessed that companies like A m D were the
cause of this. Intel decided because of a m D,
Intel couldn't trademark a number. Intel couldn't have five eighty

(10:40):
six trademarked. You can't just trademarket basic number like that.
So if it had stuck with the numbering system, a
m D could then come out with its A m
D five a D six and with its reputation for
outpacing Intel's comparable chips, that could hurt Intel's sales. But Pentium,
that was different because Pentium was a name. You can

(11:02):
trademark a name, and that's what Intel did. It trademarked
the term pentium, which prevented a m D and other
competitors from using that name on their own chips. So
now it added a marketing concern for these competitors. How
would they be able to market their own chips and

(11:22):
compare them against Intel's chips without using a trademarked name
that they did not have the rights to. It was
kind of throwing a monkey wrench into things. Now. The
way companies got around this was to include a number
that they referred to as PR, which essentially stood for
pentium rating. The number next to the PR designation would

(11:45):
indicate the comparable pentium clock speed that the chip in
question would be most like. So if you came out
with a microchip and you gave it a PR rating
of one hundred, what that tells the end consumer is
that the chip you have put out is equivalent to
an Intel pentium processor that has a clock speed of

(12:06):
one mega hurts. So it's kind of a way of
getting around the fact that they could not call their
own chips their variance of the Pendium processor. Now it
was clear that Intel was going to put up a
fight and resist as much as it could. It would
make little sense for a m D to depend solely
upon being a second source for Intel. Chips, particularly when

(12:28):
Intel wasn't really interested in cooperating fully, and so A
m D began work on designing its own x eight
six based microprocessor, which would be released in n and
it became known as the A m D K five.
Well why was it called the K five? Well, by
A m D s reckoning, it represented the fifth generation

(12:51):
microprocessor family that A m D had built, the other
four being second source Intel chips, but the K five
was a totally new architecture that was based on the
x A D six instruction set. So why the K Well,
because K is also the letter that starts the word kryptonite,

(13:12):
the substance that could bring down Superman. And I think
we can all guess who was Superman in this particular scenario.
A m D designed the K five entirely in house,
and it was the first x A D six processor
from A m D to have architecture designed by the
A m D team itself, as opposed to either following

(13:33):
Intel's detailed instructions to make a clone of their chips
or through reverse engineering and existing Intel microchip. The K
five copied some elements from the earlier A M twenty
nine thousand microprocessor that was a risk or R I
s C microchip the company made a few years earlier.

(13:53):
I talked about that in the previous episode, and I
think that was a pretty good choice. It gave them
a starting point to work from, and they were able
to really build on that and make a success out
of it. The K five's design was a little bit complicated,
and that placed limits on how much clock speed a

(14:13):
m D could get out of it. But at the
same time, the A m D engineers had made the
operations really efficient, so while it might have a technically
lower clock speed than a competing microprocessor, this increased efficiency
helped balance things out so that at the end result
it seemed like the K five was actually faster than

(14:34):
its counterparts that technically had higher clock speeds. Yes, the
other microprocessors could run more operations per second, but K
five's efficiency was such that was able to make up
for that lost ground. Now, when we come back, i'll
talk more about a m d s experiences in the
nineteen nineties and beyond, but first let's take a quick break.

(15:02):
A m D would follow up the K five with
a microprocessor called the now wait for it, the Case six,
But the K six wasn't designed by a m D engineers,
nor did it follow the K five architecture. Instead, A
m D acquired another microchip manufacturing company called next Gen
an e x G E N. Next Gen was getting

(15:24):
ready to release a CPU it called the n X
six eight six, but then A m D swooped in,
bought up next Gen, and then repurpose the as yet
unreleased n X six eight six to become the K six.
A m D marketed it as an alternative for Intel's
Pentium two processor, claiming that for less money you could

(15:45):
get the same level of performance, and that was mostly true,
though the Pentium two had some advantages over the Case six,
namely a better math coprocessor or FPU. At this point,
the K six and the variance I'll talk about in
a second we're still compatible with Intel designed motherboards. The

(16:05):
Case six was also cheaper than the pent Um two chips,
and so the Case six became a popular choice for
both O E M s and people building their own machines.
A m D would follow up the Case six with
the Case six two and the Case six three in nine.
The Case six three paired two hundred fifty six L

(16:26):
two cash memory on the CPU die in an effort
to speed up processing and increasing the amount of data
the CPU could access at any given time. The Case
six too, was phenomenally successful, so much so that some
analysts estimated that seventy percent of the under one thousand
dollar PC market in nine had a m D K

(16:49):
six two chips powering them. So, if you were building
a computer on a budget and you wanted to get
the most umph for your dollars, chances are you were
going with a m D. The company would also release
the Case six two plus in the Case six three
plus in two thousands. These were microprocessors meant specifically for
the mobile market, and they'd be the final entries in

(17:12):
the Case six line of CPUs. Meanwhile, Jerry Sanders, whom
you might remember from the first episode, he was the
first president of a m D. He was a co
founder and at this point was the CEO of the company,
was riding high. He predicted astronomical share prices for the
company in the near future. He continued the company's practice

(17:33):
of building fabrication plants at a breakneck pace. He was
building plants to manufacture microprocessors and semiconductor chips all over
the world. A m D had been incredibly aggressive in
building and staffing these fabrication facilities in order to meet
the demand for microprocessors and actually to anticipate the next

(17:54):
demand for them, and Sanders had adopted a reportedly lavish lifestyle,
maintaining an off us in Beverly Hills, which is pretty
darn far from Silicon Valley and the headquarters of a
m D. I guess he never really gave up his
dream of going into the recording industry, but a lavish
lifestyle might be fine if things continued to go well

(18:15):
for the company, and sadly that would not be the case.
Sanders spending also seemed to trickle its way into the
corporate culture of a m D overall, with executives and
high ranking salesforce professionals spending greater amounts of money to
curate an image of luxury and sophistication. Spending was getting
out of hand, and a lot of that spending had

(18:36):
to do with those fabrication facilities. According to Autika Raza,
who had led next Gen before a m D had
acquired that company and then later became the president and
chief operating officer or CEO of a m D. Sanders
was in a bad habit of building fabrication facilities too
far in advance, at least according to Raza's analysis. His perspective,

(19:01):
that is, Roza's perspective, was that the company should hold
off building new facilities until the need was there. Sanders
was building them ahead of the game, but that would
mean that a m D was constantly raising money to
build out the next facility in advance of any revenue
it was generating, and if the industry were to ever dip,

(19:22):
then it would leave a m D over extended. So
Raza wanted to take a different route. He wanted to
use revenues from current successes to fuel expansion on an
as needed basis. In other words, you don't need to
go out and build a new fabrication plant until the
demand requires you to do it. Raza, who at one
time had been viewed as a potential successor to Sanders,

(19:44):
found himself in direct disagreement with the founder, and he
would actually leave a m D in n reportedly after
a massive falling out with Sanders, with whom he would
never speak again. Now his successor was a guy named
Hector Ruez, who had up to that time been heading
up a division over at Motorola. Ruez was first wary

(20:06):
of taking this job. It was more technically oriented industry
than he had been used to, and he knew about
Sanders and his reputation of alienating senior level staff. And
he saw that there had been a string of chief
operating officers, several of whom were rumored to be groomed
as the heir apparent to a m D who had

(20:28):
subsequently left the company. But he figured that Sanders might
have issues relinquishing control to others, that this could cause issues,
but Sanders was still a very impressive person. A m
D was an impressive company, so Hector decided to take
the job. Then Jerry Sanders would retire in the early

(20:48):
two thousand's and Ruez would take over the company, and
he began to clean house. He got rid of several
top executives who had been around for quite some time
in the Sanders era, and he started bring in new people,
new talent. So Ruiz also saw that the market was changing,
and while a m D was being innovative in CPUs
and other microchips meant for personal computers, it was really

(21:12):
making most of its profits from selling flash memory not CPUs,
and so he started to refocus the company to that endeavor,
but he also found that a m D was holding
an odd place in the market. Ruez would write a
book about his experiences, stating that Sanders had created this
sort of weird paradox in a m D because Sanders

(21:34):
had a real can do attitude, a a never say
die approach to business. But at the same time, no
one in the company ever seemed convinced that a m
D could really go toe to toe with Intel, that
a m D would always be a tiny company compared
to Intel, that could never really take over as the
leading microchip manufacturer in the industry. That that was just

(21:58):
sort of this underlineing philosophy at a m D. And
as possibly because a m D had built its business
largely on being a second source chip company. So Rule
has tried to change things, directing a m d s
efforts at not just flash memory, but also developing premium
processors for stuff like Internet servers, which were just starting

(22:18):
to become a serious thing at the time. Now, around
that same time, a m D and Intel faced off
again in courtrooms. This time it was in the European Union.
A m D complained to the European Commission that Intel
was engaging in anti competitive behavior, violating the law, primarily
through what am D described as abusive marketing campaigns. A

(22:40):
m D even tried to use legal means to secure
documents from a separate case against Intel. This one was
brought against Intel by a company called Intergraph. But then
the Intergraph case eventually settled out of court and things
were obviously still very choppy between a m D and Intel,
despite the fact that they still had this cross licensing agreement.

(23:03):
The next chip from a m D, the Case seven,
better known as the Athlon processor, changed things up again.
Now the details get pretty technical, but an easy thing
to understand is that the company was able to push
clock rates up to one giga hurts. A m D
also began to manufacture its own motherboards, anticipating that the

(23:23):
day might come when compatibility with Intel's motherboards would come
to an end. So, hey, what the heck is a motherboard?
I mentioned it a couple of times in this episode.
A motherboard is just a printed circuit board. It's sort
of the highway system for information inside a computer. The
motherboard typically has connectors into which you can plug other

(23:44):
circuits like a CPU as a circuit, so you can
plug a CPU into a motherboard or a GPU a
graphics processing unit. The motherboard provides the physical connections between
all these different components so that these circuits can send
proper command ends to the right places. Now, not all
CPUs or GPUs for that matter, are compatible with all motherboards.

(24:07):
Motherboards can accept certain types of CPUs and not others,
and that's one of the reasons it's really important to
research first before you set out to build your first computer.
It's entirely possible to pick up sweet components that look
great on paper but ultimately won't work together because they're incompatible.
So a m D set out to build its own

(24:29):
computer platform. But in this case, Intel was able to
outperform a m D. While a m d S processors
were blazing, the motherboard chip set as a whole wasn't
quite able to match Intel's four four zero b X component. Still,
it showed that a m D was going to push
hard to compete with Intel. A m D also introduced

(24:51):
a new line of chips designed for entry level machines.
These were running on a similar architecture as the Athalon processors,
but that will lower clock speed. They called the new
line of processors DURAN, and they competed against Intel's Cellern
line of processors meant for the same market. A m
D upgraded the Athlon family steadily year over year with

(25:12):
names like Thunderbird, Palomino, Thoroughbred, and Barton. With each chip,
a m D built upon what it had learned from
the previous generation. The components sizes got smaller, Thoroughbred and
Barton were built using a one thirty nanometer process and
the clock speeds were climbing past two giga hurts. A
m D was optimizing the architecture for memory access. Things

(25:33):
were going pretty smoothly, and then a m D dropped
a bombshell. The company that had built a business out
of being a second source chip manufacturer actually beat Intel
to the punch by releasing the first consumer oriented sixty
four bit x eight six processor, the Athalon sixty four. Now,

(25:54):
I've been explaining a lot of basic computer concepts here,
so why not include sixty four bit? Ver is thirty
two bit? So the consumer focused processors up to that
point where thirty two bit processors. That means the processors
were able to work with data units that were thirty
two bits wide. Now, remember a bit is a single
unit of information. It can be either a zero or

(26:17):
a one. Eight bits is a byte or an octet,
and thirty two bits would be four octets wide. A
thirty two bit system can handle a range of two
to the thirty second power number of values. So if
we want to describe all the values that a thirty
two bit number can describe, and we start with the
number zero, we would go all the way up to

(26:40):
four billion, two hundred ninety four million, nine hundred sixty
seven thousand, two hundred nine five. That's the range of
values a thirty two bit system can handle. Now, as
the name implies, a sixty four bit system can handle
a data width of sixty four bits, And you might
be tempted to think that that means it can handle

(27:01):
twice as much data as a thirty two bit system,
but that's not how binary works. A sixty four bit
system can handle a value range of two to the
sixty four power of values, which is more than eighteen
quintillion values. That is a very big number, much much
bigger than the eight and a half billion or so

(27:24):
that would be twice the thirty two bit range values,
so you're not talking about doubling, you're talking much much
larger than that. So a sixty four bit system can
perform many more calculations per second. It can also support
more RAM. A thirty two bit system maxes out at
four gigabytes of RAM, or two to the thirty second
power bytes of memory. A sixty four bit system would

(27:47):
max out at least in theory at eighteen XO bytes
of RAM, which I can't describe as anything other than
a crap ton of random access memory. But six four
bit CPUs can't quite reach that theoretical limit, and they
max out in the terrabyte scale, not the exo byte scale. Still,

(28:08):
that's a lot more memory than thirty two bit systems
can handle. Now, sixty four bit systems had been around
since the nineteen sixties, but had only seen use in
academic settings and internally in various companies. No one had
yet made a sixty four bit processor for the general
public before A. M D and Microsoft released a sixty

(28:28):
four bit version of Windows that such processors could leverage.
And just to be clear, a thirty two bit system
can't run sixty four bit software, but most sixty four
bit systems can run either a thirty two bit or
a sixty four bit version of operating systems. Now, I've
got some more to say about what a m D
has been up to, but first let's take another quick break.

(28:56):
Am D, for the first time, had been the first
to market with a microchip innovation. This led to Intel
licensing the sixty four bit instruction set from a m D.
Ah how the tables have turned Now, Intel, so used
to being the entity to define standards was instead having

(29:16):
to follow the lead of the upstart company. No never
mind that both Intel and am D had been around
since the late nineteen sixties, and a m D was
really just a year younger than Intel was. I can
only imagine things were tense in some of those meetings
over at Intel headquarters, and a m D wasn't done
knocking the socks off computer nerds like me. In two
thousand five, the company released the athlont X two micro processor,

(29:42):
which was the first x a D six dual core processor. Now,
these days, multi core processors are the norm for many
computer systems and even handheld devices, but this was brand
new for the consumer market back in two thousand five,
So what the heck is a dual core or multi
core proces sessor. Now, I always like to use the
analogy of a math class that has one superstar pupil

(30:07):
and then a bunch of smart math students who don't
quite measure up to superstar status. The superstar pupil represents
a single core CPU that is significantly powerful. The smart
math students represent a multi core processor. Each individual core
of this multi core processor is less powerful than the

(30:28):
super strong single CPU, but collectively those students can tackle
some problems and solve them faster than the superstar. And
we refer to those types of problems as being parallel problems,
and that the cores are all executing operations in parallel
with each other rather than in sequence. So here's the example.

(30:49):
A math teacher hands out a pop quiz. The superstar
has to answer eight questions on the quiz, all eight.
The smart math students, of which there are eight, must
each answer just one of those questions. So students one
gets questioned one, student two gets questioned two, and so on.
So who finishes first? Now, while the superstar might get

(31:10):
through a couple of problems before any of the classmates
have finished his or her individual problem, Ultimately, the class
is going to finish First, they solved the test in parallel,
each taking one part of the problem. So even though
the superstar is technically better at math than they are,
they can't answer those questions in sequence as quickly as

(31:34):
the group can in parallel. Now, it's important to note
that not all computational problems are parallel in nature, So
for those problems, a really powerful single core processor is
going to do better than the multi core approach, and
a m D s early dual core processor couldn't work
on the same thread at all, but one core could

(31:54):
work on a thread of operations while the other core
worked on unrelated computational problems, and that sped things up. Overall,
both the sixty four bit consumer processor and the dual
core innovation were phenomenal achievements in the world of consumer computers.
A m D will never quite catching up to Intel's

(32:15):
marketing with the whole Intel inside thing, was proving itself
to be a capable and competitive player in the space,
at least on a technological level. Business Wise, things were
a bit less peppy. A m D was producing more
chips than it could sell, and that was probably part
of that whole crazy fabrication plant strategy Sanders had pursued

(32:36):
in the nineties. They were literally making more chips than
they had orders for. In two thousand one, a m
D posted a net loss of sixty one million dollars,
but the following year it was incredible. It was a
loss of one point three billion dollars. In two thousand
three it was another two seventy four million dollar loss.

(32:57):
This is not a trend you want to see, Tenue. Now,
while the company was introducing innovations, it was still battling
its nemesis, Intel in the courtrooms. A m D brought
another anti competitive suit against Intel in two thousand four
two thousand five, this time in the United States. The
complaint was forty eight pages long and accused Intel of
using a monopolistic approach to strong armed companies to work

(33:21):
with Intel rather than with a m D. At this point,
a m D had several lawsuits against Intel pending in
various courts, and in two thousand nine, Intel bargained a
settlement agreement with a m D. Intel executives promised that
their company would abide by a list of rules to
avoid anti competitive practices. Now, according to c NET, The

(33:43):
settlement included a payout to a m D to the
tune of one point to five billion dollars wolf That
certainly can help in an era where the company is
losing money through sales. Intel also would introduce its famous
tick talk Strata G, in which the company would first
design a new microchip architecture, typically by reducing the size

(34:06):
of the individual components from the previous generation's architecture and
then cramming more components onto a single chips So, in
other words, you say, let's take the design from the
last generation of microchips, make everything smaller, add more to it,
and release that. Then they would follow this up with
the talk part of the cycle. They would dedicate research

(34:26):
and development to find out how to best optimize the
new smaller components to create a new architecture that makes
the best use out of that. So the tick is
the new architecture or the new the smaller components, and
the talk was the new optimization of that. Each generation
of chips represented either a take or a talk. This

(34:48):
helped reduce risk and expenses on Intel's research and development
and helped the company mount a counter attack against a
m D. A m D got aggressive in the wake
of their innovation. In two thousand six, the company acquired
a graphics card company called A t I Technologies Incorporated
for more than five billion dollars. A t I had

(35:10):
launched in the mid eighties in Canada and had become
known for their graphics processing units, and for a while
a m D would market graphics processing cards under the
A t I brand name. In fact, in many ways,
a TI continued to perform as if it were a
subsidiary company and not a true part of a m D,
something that in hindsight, critics have suggested was a problem.

(35:33):
According to an Ours Technica article that was titled up
the Rise and Fall of a m D. Highly recommend
you read that, by the way, it's a two part article,
and it's fantastic. People within the company tended to gravitate
toward either the CPU side of the business or the
GPU side of the business, and both sides were competing
over the same set of resources. Now, competition within a

(35:54):
single company isn't always a great thing, and it led
to tension within a m D. As well's delays in
product development. A m d S CPU quality was starting
to slide as well. The Optron processor called Barcelona didn't
ship on time, and when it did finally come out,
it had a bug in the design that, when fixed,

(36:15):
slowed the chips performance speed by about ten percent. A
few years later, the Bulldozer processor had similar issues. In retrospect,
some engineers fault the acquisition for dividing the focus of
the company and the lack of an overall roadmap for
being the reason that the company was reeling a little bit. Meanwhile,

(36:36):
PC sales in general were slowing down as the world
began to shift more toward mobile computing. A m D
found itself in choppy waters again. Ru has managed to
take care of one big problem, the fabrication facilities that
were making far too many chips for a m D
to sell. He arranged a deal with a group of
investors from Abu Dhabi to sell off a m D

(36:58):
s fabrication plants. The idea was that m D would
negotiate production contracts with this new company, and that new
company could also accept fabrication contracts from other manufacturers, since
the production capacity for all the fabrication plants exceeded what
am D needed. Not long after that, Ruez would step
down as CEO. Dirk Meyer, who had worked on the

(37:21):
design of a m D s K seven chip, became
the new CEO. Alan Remember when I said A m
D and Intel settled that lawsuit in two thousand nine.
One reason a m D might have agreed to come
to the table with a settlement was that Intel lawyers
were claiming the agreement between a m D and Intel
to cross license the X eight six instruction set was

(37:43):
only valid if a m D was both the designer
and the fabricator of the chips. But now am D
was outsourcing fabrication and that, according to Intel's lawyers, was
in violation of the agreement. So it's possible a m
D came to the table to negotiate a settlement in
order to avoid a judgment on that point. Meyer would

(38:04):
serve as CEO from two thousand eight to two thousand eleven.
He was effectively removed from the position by the A
m D board analysts at the time, We're a bit surprised.
Meyer had been focusing on the traditional CPU market and
making a MD competitive there, with plans to address the
mobile market a little bit later further down the road.
He wanted to get the CPU thing right first and

(38:25):
then switch over to mobile. Now, it's possible that the
board objected to that strategy and wanted someone who would
lead the company to compete in the mobile space more aggressively,
as that was the perceived area for growth. This was
an era where it became clear that mobile was going
to be the future of computers. After a CEO search,
Rory Reid was selected to lead the company. Read diversified

(38:48):
a m d S approach beyond the PC market, and
Read was able to guide the company into entering new
markets while lowering operating costs. In two thousand fourteen, he
would step down as CEO. He said that that was
the plan the whole time, that he was there just
as sort of an interim CEO to make some business
level changes to a m D and get the company

(39:09):
on the right track. But he didn't have a deep
background in engineering. A m d's next and current CEO did,
and that is Lisa Sue. Since the nineteen nineties, Lisa
Sue has worked in the semiconductor industry. She started over
at Texas Instruments on the technical staff. She's also worked
at IBM and Free Scale Semiconductor. Before she joined a

(39:32):
m D. She served as the chief operating officer before
being named the new president and CEO of the company,
and under her leadership, a m D has done rather well.
In two thousand seventeen, the company had a revenue of
five point three three billion dollars. That was a growth
over the previous year. Also, and remarkably, seventeen would be

(39:54):
the first year that a m D would post a
full year of profitability, meaning there were no quarters where
they post in a loss. While the company would come
out profitable in previous years, it always had quarters that
had a loss in those years, so things really had changed.
Not just a few years ago, lots of people were
ready to write off a m D. The company was

(40:14):
posting massive losses and it cut back jobs. It looked
like it over extended itself. It looked like it was
just not going to measure up against the competition, and
the product quality appeared to be slipping. But more recently
things have seemed to turn around, and perhaps we've yet
to see the greatest achievements from a company that was
able to shock the world by beating Intel to the punch.

(40:36):
Who knows what they might do next. When that wraps
up these episodes about the history of a m D.
Thanks again, Stephen for sending in that request. I greatly
appreciate it. I hope you guys enjoyed learning more about
this semiconductor and microprocessor company. They are fascinating. They continue

(40:57):
to be fascinating. So uh, that's that for that story.
If you guys have suggestions for future episodes of tech Stuff,
whether it's a company, a technology, maybe a personality in tech,
whatever it may be, why not send me an email
about it. The addresses tech Stuff at how stuff works
dot com. You can pop on over to our website
that's tech stuff podcast dot com. You're gonna find an

(41:20):
archive of all of our previous episodes, links to our
social media presence, as well as a link to our
online store, where every purchase you make goes to help
the show. And we greatly appreciate it, and I'll talk
to you again really soon. Tech Stuff is a production

(41:41):
of I Heart Radio's How Stuff Works. For more podcasts
from my heart Radio, visit the i heart Radio app,
Apple Podcasts, or wherever you listen to your favorite shows.