K6-2/300AFR — AMD — WikiChip
K6-2/300AFR was a 32-bit x86 K6-2-based microprocessor designed and manufactured in 1998 by AMD. Manufactured using a 0.25 µm process, this MPU operated at 300 MHz with a FSB of 100 MHz consumed 17.2 W. Note that K6-2/300AFR-66 is an identical model with a multiplier of 4.5 instead of 3 designed to support a 66 MHz bus instead.
Contents
- 1 Cache
- 2 Graphics
- 3 Features
- 4 Gallery
- 5 Documents
- 5.1 DataSheet
- Main article: K6-2 § Cache
L2$ can be 512 KiB to 2 MiB, depending on manufacturer and motherboard model. L2$ is off-chip.
Cache Info [Edit Values] | ||
L1I$ | 32 KiB
0.0313 MiB |
1×32 KiB 2-way set associative |
L1D$ | 32 KiB
0.0313 MiB |
1×32 KiB 2-way set associative |
Graphics[edit]
This SoC has no integrated graphics processing unit.
Features[edit]
[Edit/Modify Supported Features]
Supported x86 Extensions & Processor Features |
||||
|
- Auto-power down state
- Stop clock state
Gallery[edit]
-
Week 35, 1998
-
Week 42, 1998
-
Week 51, 1998
Documents[edit]
DataSheet[edit]
- AMD-K6-2 Processor Data Sheet; Publication #21850 Revision J/0, February 2000
Facts about «K6-2/300AFR — AMD»
RDF feed
base frequency | 299.![]() |
bus rate | 99.99 MT/s (0.1 GT/s, 99,990 kT/s) + |
bus speed | 99.99 MHz (0.1 GHz, 99,990 kHz) + |
bus type | FSB + |
clock multiplier | 3 + |
core count | 1 + |
core family | 5 + |
core model | 8 + |
core name | Chomper + |
core stepping | 0 + |
core voltage | 2.2 V (22 dV, 220 cV, 2,200 mV) + |
core voltage tolerance | 0.1 V + |
cpuid | 580 + and 58C + |
designer | AMD + |
die area | 81 mm² (0.126 in², 0.81 cm², 81,000,000 µm²) + |
family | K6-2 + |
first announced | May 28, 1998 + |
first launched | June 4, 1998 + |
full page name | amd/k6-2/k6-2-300afr + |
instance of | microprocessor + |
io voltage | 3.![]() |
io voltage tolerance | 7% + |
l1d$ description | 2-way set associative + |
l1d$ size | 0.0313 MiB (32 KiB, 32,768 B, 3.051758e-5 GiB) + |
l1i$ description | 2-way set associative + |
l1i$ size | 0.0313 MiB (32 KiB, 32,768 B, 3.051758e-5 GiB) + |
ldate | June 4, 1998 + |
main image | + |
main image caption | 300AFR, Week 33, 1998 + |
manufacturer | AMD + |
market segment | Desktop + |
max case temperature | 343.15 K (70 °C, 158 °F, 617.67 °R) + |
max cpu count | 1 + |
max memory | 4,096 MiB (4,194,304 KiB, 4,294,967,296 B, 4 GiB, 0.00391 TiB) + |
max storage temperature | 423.15 K (150 °C, 302 °F, 761.67 °R) + |
microarchitecture | K6-2 + |
min case temperature | 273.![]() |
min storage temperature | 208.15 K (-65 °C, -85 °F, 374.67 °R) + |
model number | K6-2/300AFR + |
name | K6-2/300AFR + |
part number | AMD-K6-2/300AFR + |
platform | Super 7 + |
power dissipation | 17.2 W (17,200 mW, 0.0231 hp, 0.0172 kW) + |
process | 250 nm (0.25 μm, 2.5e-4 mm) + |
series | K6-2 Desktop + |
smp max ways | 1 + |
technology | CMOS + |
thread count | 1 + |
transistor count | 9,300,000 + |
word size | 32 bit (4 octets, 8 nibbles) + |
The AMD K6-2 1998 Build – Riks Random Retro
So quite a while ago from work I managed to snag a few generic towers from the mid 90s. They had lived their lives as fax and voicemail servers in a fairly clean environment. I stored them and eventually used 1 of them to build the machine featured in my first video. That would be the P3 700 Mhz machine that I frequently use for gaming, streaming and all kinds of nostalgic goodies. However this page isn’t about that computer but rather what I did with this tower.
The beginnings of the AMD K6-2 Build
As it came from it’s previous humble life it was specced out with Socket 7 motherboard and a K6 266 MHz (66 FSB) processor. I didn’t think too much of it and stocked it away to return to at a later time. That later time recently came up as I realized I had the parts to basically make it into a 1998 gaming machine. Let me clarify what I mean with 1998 specifically.
The parts in the machine and what would later go in would target a general rule of having been released in 1998. There are some exceptions we will cover later on but that was my general idea. Let’s go through the parts.
The motherboard is a Tekram p5t30-b4e or in this guise also referred to as an Impression TX533.
The Tekram P5t3-b4e
It’s a Socket 7 motherboard featuring a wide array of options. Considering the era this was made for it’s clearly designed to be sort of an upgrade path or hedged bet on standards. We have both DIMM and SIMM memory slots, PCI and ISA slots as well as support for both ATX and AT power supplies. Clearly the intent being a motherboard that could fit many different scenarios. A bus speed of 66 MHz limits the upper ranges this can go to however but that is the tradeoff of having that Socket 7 processor support.
All in all, especially considering the application I found it in, it appears to primarily target the business users. The chipset in use is the Intel 430TX. By the time we got to the later processors this board could support Intel had moved on to their new Slot 1 design but AMD decided to stay behind there and squeeze out some more life out of the socket 7. Thus the K6-2 line was devised and kept growing.
The AMD K6-2 300AFR/66
That brings us to the processor for this build. The is the AMD K6-2 300AFR/66. A 300 MHz processor operating on a 66 MHz FSB. The multiplier for our motherboard is set to 4.5x to get those 66 MHz up to the internal speed of 300. Interestingly enough the documentation I found for this motherboard does not list the K6-2 as being one of the supported processors but since it runs at the same clock and voltage as the K6 (which is supported) it seems to have taken it just fine.
The K6-2 300 seems to definitely be more targeted to the budget user in mind especially with that 66 MHz bus speed. Comparing it to a Pentium 2 of the same clock it’s quite easily outmatched in most benchmarks. In games you may not notice a ton of difference but in raw performance it’s quite behind many tests.
Benchmarks from https://www.cpu-world.com/
But most of us aren’t just staring at benchmarks all day (right?) and it’s more real world use that matters. The point I wanted to make with all of this was just to stick with a representative processor from 1998 and the K6-2 300 is spot on for that. It’s something that you had a high chance of finding in a home machine so that’s what we are using here.
We need a couple of video cards to make this gaming rig fully sing. First off to handle the 2D workload we are using what actually came in the machine. A tested, tried and true 4 MB S3 Virge DX.
S3 Virge DX
This card actually does fully feature 3D acceleration but so few games took advantage of of the S3D (clever huh?) API. There certainly were more powerful cards from S3 later on but in this format the 3D is more of an afterthought. Sure you could play some games but the list is short and considering the excellent 2D performance and compatibility this card has it will be kept in that role.
Which of course brings us to what will give this computer most of the gaming chops it needs which is a 3D Blaster Voodoo 2 8 MB, model CT6670.
Creative Voodoo 2 8 MB
The Voodoo series from 3dfx is pretty legendary at this point. Ushering in the 3D accelerated future with gusto there was nothing quite like it. While other 3D standards attempted to take over and the popular DirectX and OpenGL mostly winning the race eventually at first there was no competition. More and more developers threw their weight behind 3dfx and their Glide API making them the king of the hill. At least for a while. The Voodoo 2 card continued what the Voodoo had started and may be considered the high point for the company as a whole. If you bought a Voodoo 2 on launch you were in for a treat with a wide range of games supported and for the time quite impressive visuals. So since I had one of these cards it seemed like a perfect fit for our 1998 gaming rig. The Voodoo 2 launched in, you guessed it, 1998. I love it when a plan comes together.
When playing games sound can be just as important so I wanted to use a good sound card for this machine but yet again sticking to that 1998 timeframe. To get the broadest compatibility and also give us excellent DOS audio I ended up with a Soundblaster AWE 64 Gold model CT4390.
Soundblaster AWE64 Gold
The AWE64 Gold may be one of the most compatible cards around when it comes to late DOS and early to mid 95/98 gaming. The ISA form factor allows it to easily work with DOS applications directly and having support for several of the Soundblaster standards plus OPL you won’t be wanting on that front. It may not be the best for every single scenario since it doesn’t use a true OPL3 chip as an example but rather an in house synthesized one but again that compatibility is there.
Windows audio is handled out of the box even with the retail version of Windows 98 installed on this machine and already on first boot audio was present and accounted for. The Gold part of the card stems from some upgraded components plus the RCA audio out jacks which are of course gold plated. This truly is a one and done card that supports just about everything. What it does lack are some of the more advanced 3d positional audio standards that showed up on cards later on. Such as EAX from Creative or A3D from Aureal. This computer does have room to fit another PCI sound card should we feel like running an even more complex setup to gain access to those as well.
As for storage the eternal challenge these days with retro computers are the spinning disks and their limited time left. One by one they succumb to the inevitable end result of being based on something mechanically spinning. They fail. The click, they won’t read and they just flat out stop working. Fortunately on the retro PC front we are now a bit spoiled for choice via the path of SD to IDE adapters. They were likely originally designed for industrial machinery which may still be reliant on IDE style drives or disks but serve us quite well. The idea is that your motherboard is none the wiser and just sees it as a disk like any other. Of course this allows us to use readily accessible SD or Micro SD cards as hard drives on our old computers.
Micro SD to IDE Adapter
There is the concern about read/write cycles and if these SD cards will hold up to the challenge we are putting them through. My take on that however is that a) the mechanical drives are failing any way and b) SD storage is dirt cheap so if one fails you are out a few bucks plus the inconvenience of restoring the system. In my case it’s a no brainer where spinning disks fail. I prefer the disks where I can just from the nostalgia and noise but from a modern solution aspect it’s an easy choice. Being able to take the SD card and plop it into your modern machine to transfer files or even try out different OS loads with ease makes it well worth it.
To actually mount the adapter in the computer this time I enlisted a friend with a 3D printer who was nice enough to make me a bracket. It’s not an exact fit for the adapter I have but a bit of double stick tape took care of that problem.
3d printed mounting bracket
A really interesting thing happened with the CD-Rom drive for this machine. First off the original drive that was in there seemed fine but then I shook it a bit and it basically sounded like a loose bag of marbles. Not something I wanted to have running in my machine if I could avoid it. So I switched to a spare drive I had with about the same specs. Well that one also ended up being a bust… so 3rd time was the charm and I ended up with something a little newer than intended but still does the trick. I tell you these retro computers always have something up their sleeve.
Which speaking about the next part being the power supply that probably was the biggest bummer with this build. The machine came with an AT style power supply but during the build I managed to put red against red on the power connector which likely proved fatal to the PSU. I’ll do some real testing on it to determine that for a fact but I consider myself lucky that I didn’t fry anything in the process. Live and learn and it pays to not be sloppy with these old machines! Our upgrade instead consists of a newer ATX style power supply which leads us to the next upgraded part.
The AT to ATX Power Switch
Normally an AT power supply is turned on directly meaning there is a cable that runs all the way from thee case to the PSU. That provides the mechanical breaker needed to close the circuit and turn the power supply and thus computer on. I’m sure I’m simplifying this greatly but that’s the way I understood it. An ATX power supply expects a power signal from the motherboard and the computer turns on with just that momentary button push. This particular cable bridges that gap and allows you to connect the mechanical switch of an AT style case directly to a motherboard spot to allow an ATX power supply to turn on. It works great and I found it on eBay for about 10 bucks.
End result with everything put together I think turned out great. It’s a nice 1998 machine with what you’d expect from that era.
The final product… missing the case cover
Final specs:
- AMD K6-2 300 AFR
- Tekram P5t3-b4e Motherboard
- 64 MB RAM
- 4 MB S3 Virge DX
- Creative Labs Voodoo 2 8 MB
- SoundBlaster AWE64 Gold
- 32 GB IDE to Micro SD Card Adapter
It was a fun machine to explore, piece together and finally use. Trying to stick to a year made it a bit more grounded compared to many other machines I have which usually are quite powerful for their task. Sure that gives you a very distinct edge as well but there’s something to be said for a more period correct machine as well.
If you’d like to see the 2 part build series as well as a livestreaming session featuring this very machine they are linked below.
I had a simple motherboard based on VIA chips, which according to the passport does not have K6-2 support. There were no problems with power supply, most boards with switching regulators have 2.2 Volts (undocumented). Here, there was no multiplication by 4.0x for 266 MHz on the board, but this is not a problem either: with a multiplication by 3.5x as 233, the processor started up. The problem turned out to be different: the BIOS does not really support K6-2 processors, the computer gets hung up on a program restart or using Ctrl-Alt-Del. Although the Reset button is overloaded well. nine0007
I tested one more board based on VIA chips (manufacturer ZIDA) and two boards based on TX chips, which do not support K6-2. It turned out that they do not really support these processors. I had to look for something inexpensive (the crisis hit everyone). I was offered a TX98 based on VIA chips (manufacturer ZIDA), however, the passport only indicates support up to AMD K-6-PR300, but it really supports K6-2 as well. I would like to especially note that this board from ZIDA was a success. I experienced almost no difficulties installing it and connecting other devices to it (twice I already experienced huge difficulties with individual ZIDA boards, since the company is clearly striving for miniaturization, alas, to the detriment of the arrangement of elements on them). In addition, there is a modification of TX98 with built-in Sound Blaster. It would be interesting to try too.
The processor does not get very hot, but it does not run. When the core power is increased to 2.3 Volts (according to some sources, it does not exceed the maximum permissible value), it runs up to 300 MHz, both at the bus frequency (75) and at the coefficient (4. 5x), and does not heat up more. The processor understands the coefficient and 5.0x (333 MHz), but hangs even when the core power is increased to 2.4 Volts, at 4.5×75 it also hangs.
According to InfoPro tests, the processor confidently overtakes both Pentium II — 233 and 300. With the coprocessor, things are somewhat worse, but not as deplorable as before. It was not possible to test 3D Now, since I only have a 2D accelerator and do not work in any 3D programs. Installing a driver with 3D Now support for Quake II without a 3D accelerator did nothing. nine0007
My friend, an ardent fan of Intel, having tested my AMD-K6-2-266, took AMD-K6-2-333 3D Now for himself, after adding the necessary core voltage, adding multipliers and flashing the BIOS on his motherboard ( if you have access to the Internet, time and skills — this is real). The only thing: the processor runs at 4.5×75, since 83 is not documented and the board does not want to work at this frequency with any processor.
Unfortunately or fortunately, the change of processors is happening so fast that at the time of writing this article AMD-K6-2-266 3D Now is leaving the market and longer overclocking tests of this processor will probably no longer be of interest to anyone. nine0007
Vladimir Tomkovich
Computer newspaper. The article was published in issue 46 for 1998 under the rubric Miscellaneous :: country of councils
AMD-K6-2 — serious competition to Pentium-II
Announced in Atlanta May 28
AMD’s new processor
be a good alternative
latest developments from Intel. AMD-K6-2
different from its competitors
only clock frequency, but also
another extension of the set
commands for processing 3D graphics and
sound, which was called 3DNow! it
positions it as the basis for
building relatively inexpensive
(compared to P-II)
multimedia systems. nine0007
The processor is manufactured according to
0.25 micron technology and contains
about 9.3 million transistors and
produced in Austin, Texas. He has
standard 321-pin ceramic
case and installed in
standard Socket 7 (although requires
Super7 specification support).
Data published on www.amd.com
testing new processors
brought me to the lung
confusion. Judge for yourself:
3D graphics performance on
Pentium-II-400 level, and for office
applications — slightly worse than P-II-300. nine0007
However, more detailed
configuration study,
which were produced
testing, something clarified.
The fact is that AMD-K6-2 supports
Super7 specification with bus frequency 100
MHz, and the P-II operates at a bus frequency of 66
MHz, at which it was carried out
testing. Agree that
such a difference in the exchange rate with
memory will give some gain
even without instruction set optimization
(namely 50%). Through a series
processor hardware tweaks
can do up to four
floating point operations
one beat. As a result, peak
new performance
processor (K6-2-333) reaches 1.33 Gflops
versus 0.33 for the Pentium-II-333 processor. On
other frequencies quadruple
the win is saved. Finally AMD
solved my old problem
with a weak coprocessor and overtake
its older competitor. nine0007
Additional instructions,
included in the new processor,
focused on improving
processor performance in
resource-intensive tasks such as
three-dimensional graphics. Basically they
designed to accelerate
pushing data between
processor and graphics
3D accelerator. As is known, this
the bottleneck of modern
computers, which is constantly
trying to expand: first
the AGP bus appeared, designed
to offload the PCI bus
from the boundless flow of graphic
information, then increased
local bus frequency from 66 to 100 MHz. nine0007
Currently known to me
only one type of motherboards
which can be installed
AMD K6-2 processors are «Apollo»
by Lucky Star on the VIA VP3 chipset.
The processor was created in close
cooperation with manufacturers
software (Microsoft) and
video adapters. Processor not
requires any additional
drivers to work with
existing operating
systems, however, for a full
support for 3DNow! update
you still need a driver. AT
in particular, for our beloved
new Wintel platform expected
version of DirectX 6.0, which should
appear in the next month. nine0007
According to AMD, using
new processor will allow
achieve fundamentally new
image realism level
and sound by increasing the frequency
frames when generating graphics with
high resolution. Of course it is
good news for the ardent
«Quakers», but computational
power can be directed in peaceful
purposes.
Financial difference
between AMD-K6-2 and Pentium-II looks
impressive enough.
AMD-K6-2-266 | ||
Processor | 255 | 148 |
System board | 105 (LS Intel LX430) | 68 (LS VIA VP3) |
Housing | 45 (ATX) | 25 (regular) |
TOTAL: | 405 | 241 |
Given that in our area
the money question has always been on
first place, systems on
processors K6-2 have a good
future. Vadim VINOKUR
Did you notice an error? Select it with the mouse and press Ctrl+Enter!
-
The editors of KV.