A somewhat interesting 486SX33! Not because of its performance (which was just fine when this CPU first came to market) but because of its production date. This particular CPU has been manufactured in the 25th week of 2000. That's years after the initial release.

The earliest 486SX CPU's I have (at 20 and 25MHz) are dated 1991. I guess most SX33's were sold around 1993.

These later produced models weren't used in consumer PC's but in industrial applications. > Read more

The UMC 486 CPUs are somewhat oddbal CPUs. Not all 486s are identical and some really perform different. For example the Cyrix Cx486DX2/66 is slower than the Intel 486DX2/66, however, the Cyrix Cx486DX4/100GP4 is faster than the Intel 486DX4/100. The difference lies in variations in the architecture and especially L1-cache memory (write-back vs write-through).

Both Cyrix and UMC designed their 486 in-house. UMC invested in streamlining the processor and optimize the execution of instruction. This resulted in a CPU that is clock-for-clock faster than Intel. Using Doom as benchmark the UMC Super33 is faster than a Cyrix DX2/5- and almost on par with an AMD Am486DX40.

The CPU you are looking at is a soldered QFP chip on a small motherboard. It's a very low cost solution which reminds of the later 286 and 386SX motherboards. Small, an optimised chipset that doesn't need much 'glue' (chips that connect the chipset to other parts of the computer) and no fancy features in general. This motherboard is dated 1995 and only has 6 ISA slots. Remember, in 1995 a decent PC would be a 100MHz 486 or Pentium 75. In March 1995 the Pentium 120 was the fastest x86 processor money could buy.

I have an identical CPU but in ceramic: UMC Green U5SD-Super33 which I have benchmarked on a fast Biostar 8433UUD-A motherboard using a PCI S3 Virge DX graphics card.

The UMC processors are not very common. They were late to the market and UMC faced lawsuits from Intel. > Read more

Identical to this Overdrive CPU. I've had this CPU for a long time and pulled it from a typical 486 system (Socket3, VLB, 8MB). I doubt the system had been upgraded with the Overdrive but instead was fitted with an Overdrive when the system was built new.

In essence the DX2ODPR is identical to a regular (write-through) DX2/66. However, the DX2ODP without R-suffix is different so you might double check your pin-out before using those. > Read more

The 486SX33 was a nice CPU back in the day. Certainly not the fastest one but it provided decent performance for its price point. Especially since most software didn't need a co-processor (FPU) which the 486SX didn't have.

This particular CPU is basically a one-of-million 486SX built in 1993. However, if you look closely the Intel logo is printed at an angle. I don't recall seeing other CPU's that have such a big deviation in its printed logo. From the looks of it, it seems like a genuine Intel part. Since the SX25 would usually run at 33MHz some SX33 CPUs are counterfeit, reprinted SX25 models. Some of these 'fake' CPUs are hard to distinguish from the real product. > Read more

A regular Intel 486DX2/50 that runs at 50MHz using a 25MHz FSB. It does it's job but if you could overclock this to 66MHz you could gain, in a VLB or PCI configuration, 25% more speed in Doom due to both the higher CPU clock and higher front-side-bus .

If memory serves me correctly this CPU was pulled from a Compaq system and was fitted with the smaller style heatsink (see right CPU on the second picture). However, the CPU on the left is also a 486 at 50MHz (but a regular DX) with a much taller heatsink. Both will run fine but this shows that back in the day there wasn't really a 'standard' for cooling CPU's. Keep in mind that heatsinks (and fans) started appearing with 486-class CPU's. Before that all CPU's were practically almost without heatsink. A few excepts occur in laptop situations or when dealing with a LGA-socketed R80286 in which the heatsink clamps the CPU down. > Read more

. > Read more

An uncommon upgrade chip for the 286! It upgrades a 286 to an IBM 486SLC2 that can run at 40, 50 or 66MHz with 16k L1-cache. In my case I think it'll run at 40MHz (it's clock doubled) since the FPU is rated at 20MHz. I have no documentation to verify, however.

Basically IBM upgraded their 386SLC 8KB L1 chip with 16KB L1-cache and the i486 instruction set to make a 486-class CPU on a 16-bit bus. Evergreen fitted that CPU on this little PCB along with a Cyrix FasMath FPU to create a beefy upgrade for your old system. > Read more

. > Read more

An Am486DX4/100 NV8T with an early 1995 date-code. I believe it's a fairly early Am486 100MHz part. Most of these CPU's are made after summer 1995. AMD did manufacture the Am486 for a long time: I also have a 16KB cache CPU from 2001.

This particular model is the, I believe most common, NV8T. The NV8T and SV8B feature 2x and 3x multipliers and the SV16B has 3x and 4x multipliers. The number (8 and 16) stands for the amount of L1-cache. The S-models have write-back cache which is faster compared to write-through.

Earlier models can have the V8T-marking (without N). These versions are the so called 'No ICE' CPU's. Intel allowed AMD to copy 286 and 386 CPU's to sell them as Am286 and Am386. The 486, however, was not allowed to copy so AMD had to remove Intel's in-circuit emulation (ICE) microcode. > Read more

> Read more

An E6-stepping of the AMD Am486DX2/66 processor. It has package number 25220 made by Kyocera. This is the 'N' version of the CPU meaning that it doesn't have Intel ICE. See the description of this Am486 for more information about the different numbers. > Read more

Identical to this Am486 except that this CPU is packaged in the Sumitomo-plant. Most Am486 CPU's are packaged by Kyocera. The symbol in the lower right corner indicates the plant. > Read more

A new-old-stock 'FOR OEM USE ONLY' processor from Texas Instruments. It fits in a 386 socket and is able to do clock-doubling like the 486DX2. This means it will run at 50MHz on a 25MHz motherboard.

Because this CPU has 8KB L1-cache it should perform quite well. Regular 386's don't have L1-cache at all and most budget 486 CPU's that fit in a 386 socket only have 1KB L1-cache.

Unfortunately this CPU doesn't have a date code so I can't really pin-point when it's made. The code FC2-63ANT9T is a batch code that is not unique to this particular CPU; all the CPU's from this batch have that same number. Texas Instruments released the TI486SXLC-33, TI486SXL-40, TI486SXLC2-50, TI486SXL2-50 for respectively, US$79, US$89, US$110 and US$149, in quantities of 1000 units. > Read more

A D-stepping of the original AMD Am486DX2/66 with package number 24361 made by Kyocera. The first models didn't have any additional numbers in it's model name. Later AMD made different versions which had write-back cache and a slightly modified chip due to a lawsuit with Intel.

I have different versions of the Am486DX2/66 in my collection.
Am486DX2/66 E6 stepping with package number 24361 made by Kyocera
Am486DX2/66N E6 stepping with No-ICE microcode and package number 25220 made by Sumitomo
Am486DX2/66N E6 stepping with No-ICE microcode and package number 25220 made by Kyocera
Am486DX2/66NV8T A stepping with No-ICE microcode and package number 25220 made by Kyocera
Am486DX2/66SV8B C stepping with No-ICE microcode and package number 25398 made by Kyocera

The V stands for 3,3V and the 8 for 8KB L1-cache. The T and B indicate if write-through or write-back is supported. Write-back cache is backwards compatible with write-through. Write-back offers better performance. All models with S are Enhanced Am486 CPU's that have several power-saving options (if the motherboard supports this). Also all models with S are equipped with the No-ICE (In-Circuit Emulation) microcode modification. All versions with N are also No-ICE. AMD was forced to remove Intel's ICE due a lawsuit.

Keep in mind that late models of the Enhanced Am486 don't include the S-marking despite being Enhanced. > Read more

I bought this CPU in a package deal along with a few other CPU's. Amazingly enough this CPU has the same sSpec number and production date as my other sSpec SX419 DX33. What are the odds . > Read more

. > Read more

. > Read more

. > Read more

. > Read more

A late DX2/66 (40th week of 1996). In late '96 486 at 100, 120 or 133MHz were considered low-end and Pentium's at 100 or 133MHz were commonplace with 166MHz being the top model.

An interesting twitch of this CPU is the indication of the model number. Every 486 has an indication of it's type and speed seperated by a dash, except for this particular CPU. On the CPU it says DX266 rather than DX2-66. > Read more

A regular 486DX33. Although not many people used the 'DX' part (which indicates it has an on-board co-processor) a lot of people thought that the 'DX' was faster. They would've been right if it was a 386 because the SX/DX suffix has a different meaning with 386's. With 386 class processors SX means a 16-bit external databus and DX a 32-bit.

Years ago I used to run a DX33 in which I later fitted a Cirrus Logic VLB video card. I think that the SX33/DX33 with VLB is the slowest combination that run's Doom fairly well. Of course the bigger levels in Doom II can bring down the FPS (frames per second) but the three episodes of the original Doom ran playable, especially back then. In hindsight I would recommend a DX2/66 because Doom runs a lot better on that CPU but if you don't know better the DX33 is fair enough . > Read more

Built in the 15th week of 1991, this is an early 486DX33. At that time, a 386 running at 25MHz or 33MHz was considered a modern PC. The 286 was already low-end and the 486 was high-end. Interestingly, the 486 logo lacks the DX suffix because the 486SX hadn’t been released yet. The first SX CPUs appeared in mid 1991 (I have one dated 21th week of 1991: i486SX-20. However, in the specification on the CPU it does already mention A80486DX-33. The earliest 486 CPUs (dated 1989) don't even mention DX at all.

In early 1991, this CPU offered ample processing power for most users, even though technology was rapidly advancing. Just a year later, in 1992, the 486DX2/66 was introduced which was twice as fast as the DX33. These were great systems back then, but unfortunately, VESA Local Bus (VLB) wasn’t available yet, so you’d be stuck with the slower ISA bus. For graphics-intensive applications and, especially games, the VLB bus truly made a difference on a 486.

The 486DX features an integrated floating-point unit (FPU) that significantly speeds up applications like CAD and spreadsheets. However, in most mainstream applications and games that run well on a 486, the FPU isn’t particularly noticeable. Running Doom or Windows 3.1 with or without the FPU doesn’t make a practical difference. 486s with the SX designation don’t have the FPU enabled.

In the 386 line, the DX and SX designations have different meanings. DX indicates a 32-bit external bus, while SX refers to a 16-bit bus. Despite this, both CPUs are internally 32-bit and neither features a FPU. In most cases, you can install a separate FPU in a dedicated socket. > Read more

. > Read more

In the picture you see the 40MHz Am486 fitted onto an Octek motherboard. A fine combination for most people in late 1993 because it was affordable. Performance-wise it was fine although one could get faster systems like the DX2/66 or even the first Pentium's .

The Octek motherboard in question doesn't have VLB-slots. If you wanted fast VGA you had to invest in VLB graphics. It greatly enhanced performance and made Doom playable on these systems. However choose your parts wisely because running VLB I/O controllers at 40MHz can sometimes be troublesome. > Read more

A KU80486SX-25 (in QFP package) fitted on an UMC 4SLUD motherboard from 1992. Typically these motherboards came with a soldered QFP CPU and an additional unpopulated socket. To upgrade these system an Overdrive processor could be fitted while keeping the SX25 soldered on-board. The Overdrive has an extra pin that disables the on-board CPU. I have DX2 ODP 50 Overdrive processor that could boost the system from SX25 to a DX2/50. > Read more

Made in week 28 of 1994 which makes it a somewhat late 486 DX2/50. In that time-frame the Pentium 90 was Intel's flagship and running a DX/2 66MHz, DX/2 80 or even DX/4 100MHz made more sense in 1994. However, these 50MHz parts usually run fine at 66MHz and this particular example even booted up at 80MHz without hassle .

Looking at time-frames it's interesting to see how scattered the early 90'ies were. In my database you can find both 286 and 486SX25 CPU's dated 1991. 1992 went on with 25MHz 286 and 386 CPU's but also provided the 486DX50. Performance between the slowest and fastest CPU is huge!

This CPU is currently fitted on a PCChips M601 motherboard. Quite typical for it's time. Especially at 66MHz and equipped with a VLB graphics card it's a fine combination . > Read more

> Read more

> Read more

The Evergreen 586 is based on the AMD Am5x86 -P75 which is basically a 133MHz 486. Because Evergreen fitted it's CPU with additional logic to control overdrive-mode, cache and multipliers it's adjustable to a number of scenario's you can find in older 486 sytems.

Usually you have to configure your motherboard to run 486DX CPU's (because the Evergreen has an FPU built-in). If you use the overdrive socket, which is often found on older motherboards that have an on-board soldered CPU, you have to switch the jumper over to 'OVRDRV' mode rather than 'NORMAL'. Depending on the final clock-frequency you can switch between 3x or 4x multiplier. Depending on your FSB you can calculate what your final CPU clock-frequency will be (4x25: 100MHz, 3x40: 120MHz, 4x33: 133). Usually it'll be 4x33 which gives 133MHz (original speed of the Am5x86-P75). At last but not least there is a WT/WB switch that controls Write-Through or Write-Back cache. Write-Back is faster but more complex. If the CPU proves to be troublesome then configure WT.

AMD named it's CPU 5x86-P75 because it's performance is, to a certain point, similar to a Pentium 75. Because the 5x86 is really a 486 (it has nothing to do with a 586, it's just marketing) it's floating point unit (FPU) is much weaker than that of a Pentium. It's ALU performance is quite on par with a Pentium 75 but games like Quake really show that the FPU of a 486 is much weaker.

Also keep in mind that it's still based on the 486 platform so the busspeed is still 33MHz. The Pentium 75 exceeds this with it's 50MHz bus. Despite that the PCI-bus of the Pentium 75 runs at just 25MHz (and 33MHz on the 486-platform) you'll see that games like Doom still run better on the Pentium 75.

I have benchmarked a stock Am5x85-P75: the results can be found here. The red-bar is the Am5x86-P75 and the grey bar is a Pentium 75.

Over the years I have also obtained a purple Evergreen 586 box with original diskette and removal tool for non-ZIF socketed CPUs. Two types of boxes exist: this purple one and a white one that looks like the cover of the manual. > Read more

Like my other Am5x86 but with different stepping. This one has the ADW stepping which is rated for 133MHz. My other chip is an ADZ-chip which should be able to reach 160MHz. > Read more

The normal/regular Am486 DX4/100 with write-through L2-cache. This version is slower than the SV8B marked Enhanced Am486.

I have another CPU like this that is 14 weeks older. > Read more

Just like this 486 but with write-back L1-cache rather than write-through. The SV8B marking on this CPU indicate it has the faster write-through cache. This CPU is also known as Enhanced Am486. > Read more

An 1995 model with NV8T marking. This is 8KB write-through L1-cache. Should have identical performance to the CPU without additional markings.

See this Am486 for more information about the processor markings. > Read more

> Read more

The Cyrix Cx486DLC plugs into a 132-pin 386 socket and will act like a 486 processor. However, because it has only 1KB L1-cache instead of 8KB it will not perform as fast as a 'real' 486.

You can use the Cx486DLC to upgrade a 386 system but keep in mind that older motherboards might not fully support it. The later model 386 motherboards usually work fine with this CPU.

I haven't benchmarked the 40MHz version yet but the Cx486DLC 33MHz version only gained 5% performance in Doom compared to a 386DX33. If you wanted a more serious performance gain (and stay on the 386 socket) the clock doubled Cyrix Cx486DRx2 33/66GP or the Texas Instruments Ti486SXL (that has 8KB L1-cache) will be in favour. > Read more

Much like my other Am486 DX2/80 except that this one doesn't have the HEATSINK REQ'D marking. > Read more

I like the way they present the overdrive CPU 'flying' out which rapid speed and taking the applications (running at rapid speed!) with it. The backside of the box shows a table with upgrade possibilities and which performance improvement you can expect.

Overdrives we're expensive back in the day and often you could install a regular processor as well. If you couldn't install a regular processor it was advisable to scratch your head twice before buying an overdrive; installing an expensive CPU on an old motherboard with old components like the VGA card doesn't always give the performance boost you wanted. > Read more

i486DX2/50 CPU in QFP package mounted on a Compaq CPU board for Compaq notebooks from 1994. > Read more

A 1992 486DX-50; a good performer back in '92 but with a high front side bus (FSB) which could be troublesome. The DX2/50 which runs at 25MHz FSB (clock-doubled) won't give you trouble but due the slower bus speed it will be slower with games or I/O. How much slower? See this link in which I benchmark another DX-50. The red bar shows DX-50 scores, the grey bar DX2-50 scores.

Benchmarks are run on a Biostar 8433UUD-A that runs the DX-50 at 33MHz PCI and DX2-50 at 25MHz PCI. > Read more

. > Read more

> Read more

... > Read more

... > Read more

> Read more

... > Read more

The Cx486DRx2 is a CPU upgrade for 386 systems. It uses the 486 instruction set and is pin-compatible with the 80386DX. It's performance per clock is not far off a normal 386DX but due the higher clock frequency and the 1KB L1-cache is runs noticeably faster than the AMD Am386DX/40.

This CPU uses clock-doubling technology which allows the CPU to run twice as fast as the FSB. If you use this CPU to upgrade a 33MHz 386 (with 33MHz FSB) it will run at 66MHz. A 386DX25 will be upgraded to 50MHz.

Performance-wise it's not comparable to a real 486DX2/66. Bear in mind that the Cx486DRx2 is a 386 class CPU with 486 instructions. In my benchmarks it positions itself between a 486SX/25 and 486SX/33. However, the 486 benchmarks are done on a high-end 486 PCI platform; the Cx486DRx2 did it just with an ISA Diamond SpeedSTAR24 (ET4000) which affects the Doom benchmark. > Read more

Almost identical to my other Cx486DLC 33GP except that this CPU is the 'AB' revision which doesn't have bugs with certain co-processors.

You can identify AB and BB chips by looking at the number in the right lower corner of the CPU. > Read more

A 0,35 micron 3,45v Am486DX/4 CPU with laser engraved markings. > Read more

Going along at a nice 100MHz. Nothing much to say about this CPU except that the Intel, AMD and Cyrix parts were more popular (and a bit faster) in the day. > Read more

Just like the DX/2 66MHz but a tad faster. > Read more

The DX4/120 is one of the faster true 486 processors available. It operates at 120MHz and its 40MHz front-side bus (FSB) is a nice improvement over the standard 33MHz FSB (depending on your motherboard / PCI divider). For example, in my case, the DX4/120 performs considerably faster than a 100MHz 486 because its PCI bus is also clocked at 40MHz. This almost surpasses the Am5x86 P75 (486 133MHz) and even outperforms the Am5x86 P75 in the graphics-intensive PCPlayer benchmark. However, on non-I/O-intensive calculations like Dhrystone (integer performance), the differences are smaller because they don't benefit from the higher bus and PCI clock.

This CPU was made in the 39th week of 1995. I don't have a price-list at hand but I guess it would be ever so slightly cheaper than Pentium 75. Both would provide decent computing power in 1995 for the average consumer. > Read more

Just like the other Am486 CPU's but with the 100MHz-barrier broken.

This model is also known as the Enhanced Am486DX4-S 100MHz because is has write-back L1-cache. CPU's with write-back cache are slightly faster than CPU's with write-through cache. You can recognize this via the SV8B marking.

The NV8T and SV8B feature 2x and 3x multipliers and the SV16B has 3x and 4x multipliers. The number (8 and 16) stands for the amount of L1-cache. The S-models have write-back cache which is faster compared to write-through. > Read more

AMD's 80MHz 486, nothing spectacular actually. However, Interesting to see is that Intel never made DX40 or DX2/80 486 processors. They always stuck to the more common front-side-bus (FSB) clock frequencies that were more reliable in the old days. At the time that the 486 was new it was difficult to get it running at a 40MHz FSB bus frequency but later motherboards could handle 40MHz easily so DX2/80 processors didn't cause trouble.

While configuring this CPU I also tried the settings for the DX4/120. It turned out that this DX2/80 can also run at 120MHz but unfortunately it was unstable.

Also note that this CPU is model NV8T which is a normal AMD 80486 CPU. The SV8B models are Enhanced Am486 CPU's that have write-back L1-cache. > Read more

Almost same as this Cx486DX2/80 but a few MHz slower. I still like the green heatsink . > Read more

Another 486 clone! This one is easily recognized because of the colored Windows logo on top. I think it looks quite nice but in reality it's just marketing. This processor is not optimized for Windows '95 but Windows '95 is programmed for the 80386 and this 486DX2/66 is compatible with the 80386 . > Read more

This is not the first 66MHz CPU in my database. In fact it's the third Intel 486DX2/66. The other two have a different sSpec number and one of them is faster than usual because it has write-back cache. > Read more

Same as the 66MHz part but somewhat slower. > Read more

This CPU has sSpec number SX411 and is produced in the 51th week of 1991. I have another 486SX25 SX411 that is made in the 16th week of 1994. The 486 architecture was used for quite some time with the oldest systems appearing in 1989. They were very expensive at the time and many people would still buy 8088 or 286 systems. In late 1991 the SX25 was a nice system to have in the average household. However, in 1994 it was already dated. > Read more

Quite much like the Kingston TurboChip but a different package and faster. I suspect this one has write-back L1 cache.

Very few AMD Am5x86-P90 (running at 160MHz) exist. AMD could make the P75 faster but didn't do so because it would hurt sales of the AMD K5. As far as I know, if an Am5x86-P75 (133MHz) has the ADZ marking it would have been able to run at 160MHz, with ADY at 150MHz and with ADW at 133MHz. > Read more

This Kingston TurboChip is based on the Am5x86-P75 which is an 486DX4/133 (yes, really quadrupled).

The Kingston TurboChip came in an upgrade package to boost a 486 to 'Pentium 75' performance. Hence AMD named the CPU Am5x86-P75.

The Kingston-solution runs fine but as seen in the benchmarks it is slower compared to the 'real' ceramic Am5x86-P75 or the Evergreen 586 133. The reason is that the Kingston TurboChip runs its L1-cache in write-through mode instead of the faster write-back mode. They probably configured it this way to be on the safe side with older chipsets. Early chipsets don't support write-back mode since this didn't exist on the first 486 CPUs. I have verified the L1-cache scheme via CheckCPU and by validating performance with the Evergreen 586 133. The Evergreen 586 133 has the advantage that L1-scheme options are switchable via a jumper so you can run it in either write-through or write-back.

CPU Identification utility v1.27.1 (c) 1997-2022 Jan Steunebrink
------------------------------------------------------------------------------
CPU Vendor and Model: AMD 5x86-133 P75 (X5)
Internal CPU speed : 134.2 MHz
Clock Multiplier : 4.0
Bus clock speed : 33.5 MHz
CPU-ID Vendor string: AuthenticAMD
CPU-ID Signature : 0004E4
|||||+- Stepping or sub-model no.
||||+- Model: Indicates CPU Model and 486 L1 cache mode
|||+- Family: 4=486, Am5x86, Cx5x86
||| 5=Pentium, Nx586, Cx6x86, K5/K6, C6, mP6
||| 6=PentiumPro/II/III, CxMII/III, Athlon, C3
||| F=Pentium4, Athlon64
||+- Type: 0=Standard, 1=Overdrive, 2=2nd Dual Pentium
|+- Ext. Model: Extends the Base Model to an 8-bit value
+- Ext. Family: Extends the Base Family value
Current CPU mode : Real
Internal (L1) cache : Enabled in Write-Through mode

The first 100MHz ticking part Intel brought onto the x86 market. The DX4/100 was actually a DX3/100 as it didn't clock quadruple but clock triple the front side bus frequency.

In many cases an old 486 couldn't be upgraded to the DX4/100. Not only because some motherboards didn't understand the clock doubling or tripling technique, also because the voltage requirements changed. Older 486 processors use 5V on the core instead of 3V for this DX4/100. Of course it did boot up at 5V but it can damage the chip. Not to mention the excessive heat output it generates at 5V!

This 486DX4/100 has no support for write-back cache. It can only do write-through. Write-back models have a &EW marking on the CPU. Write-through has &E. Write-back gives more performance. > Read more

Cyrix' 100MHz 486; just like their 66MHz and 80MHz version it's a reversed engineered part and generally didn't perform as well as Intel's DX4/100. Although basic mathematic calculations and Whetstone seem to go very well on the Cyrix 486 processors.

Also note this CPU requires less voltage. This is due a scaled down and improved manufacturing process that generally causes a CPU to scale better with less voltage and thus less heat output. > Read more

This is actually a Cx486DX2-80GP in disguise. It even boots up as a Cyrix CPU. > Read more

Just like IBM's 66MHz 486 but a bit faster.

As seen on the picture the processor has two gold colored strips on the side. I haven't seen any other CPU having those so I wonder why they are there . > Read more

First of all: nice heatsink! The green color gives this CPU some sort of extra touch.

Unlike AMD's 486, the Cyrix Cx486 was a reverse engineered Intel CPU. Cyrix did quite a good job on this but their products are generally a bit slower compared to Intel's and AMD's 486 processors. Although basic mathematic calculations and Whetstone seem to go very well on the Cyrix 486 processors. See the benchmarks down below!

The 486 of Cyrix was used in IBM and SGS Thomson (IT's ST) CPU's as well. No wonder they perform exactly the same. > Read more

An Overdrive processor with FSB of 25MHz. This one can upgrade a 16MHz 486 to 48MHz, 20MHz to 60MHz and 25MHz to 75MHz.

As seen on the top of the picture this CPU has voltage regulators on it to transform the 5V from the motherboard to 3,3V for the CPU. The CPU inside has been made on a narrower manufacturing process and thus it needs less power. The idea of Overdrive is to upgrade old systems that often only provide 5V. > Read more

An Intel Overdrive processor that can upgrade your existing 486SX33 or 486DX33 by removing the original CPU and replacing it with the Overdrive CPU in order to boost your system to 66MHz. A replace is possible because the DX2ODPR carries the R-suffix that stands for replacement. The Overdrives without the R-suffix (like this one) are fitted with an extra pin that allows an upgrade if your motherboard has a soldered 486 CPU with an empty 487SX socket. > Read more

IBM's version of the DX2/66 was actually made by Cyrix. On the bottom on the CPU a Cyrix logo can be found.

While AMD used Intels 386 architecture to build their 486, Cyrix used reverse engineering to develop it. The early Cyrix 486 chips are somewhat slower compared to Intel or AMD's 486 chips but sold for less money as well. IBM also made their own CPU's based upon the Intel 386SX architecture and sold it as 486DLC. Those CPU's are enhanced 386 CPU's with more cache and things like that. > Read more

AMD's version of the 486DX2/66. Runs practically the same as the Intel 486DX2/66. Usually the Intel part is a tiny bit faster but it's not noticeable in real life. > Read more

Unlike this 486DX2/66 it has no build on heatsink. This allowed people to stick their own heatsink/fan solution onto it as the 486DX2/66 didn't run happy without heatsink.

This version of the DX2/66 is the fastest Intel ever made. See the benchmarks where it is ahead on the other DX2/66. The performance difference can be declared because the DX2/66 came in two versions. A P24 and P24D. Difference is that the P24 only has write-through L1 cache and the P24D has write-back L1 cache, which is faster. > Read more

The DX2/66 was a popular processor in the 486 era. It has enough oomph to run Windows 3.1 and DOS games from that era. For instance, Doom runs a lot smoother on a system with a DX2/66 fitted instead of a DX33, especially when you're running a VESA Local Bus (VLB) graphics card.

This particular 486DX2/66 is less common because it has a nice blue heatsink attached. Most DX2/66 CPUs didn't have a heatsink mounted by Intel but a third-party heatsink or heatsink/fan solution. The heatsink is very similar to the Overdrive version except that it is ever so slightly taller. I never had cooling issues with both the blue and black (Overdrive) CPUs, however, that changed in summer of 2026 when I re-did my 486-benchmarks. This blue DX2 crashes after running a series of benchmarks. Adding a small fan solved the problem.

There are two major versions of the DX2. The P24 and P24D, of which the latter has a faster L1 cache mode called write-back that improves performance. The P24 version only offers write-through cache, which is slower. The 'SX731' (this model) is the P24 version. See this 486DX2/66, which is a P24D.

I have benchmarked this CPU in a fairly modern UMC PCI-based motherboard that allows me to incorporate benchmarks of other brands and faster types of CPUs as well. The Intel DX2/66 is generally compatible with every 486 motherboard out there, whereas the DX4 or Cyrix counterparts often need a more recent motherboard. Some of the benchmarks may appear wrong; a DX4/120 performing slower than a DX4/100 in Doom, but this is because the PCI bus runs slower on the 120MHz version due to the divider on the 40MHz PCI bus.

In the benchmarks, you see that with Doom, the DX2/66 is about 86% faster than the DX33, so the DX2 is not twice as fast, but it really does make a big difference.

Just like other 486-class CPUs, the DX version has the built-in floating-point unit (FPU). In 1993, the FPU wasn't really necessary for most people unless you used CAD or big spreadsheets. > Read more

The Intel Overdrive CPU's were available in various shapes and forms over the course of it's life. In the early days you could have, for instance, a 486SX at 25MHz that was soldered onto the motherboard. Because it was soldered you could not upgrade the chip. Instead, Intel decided to add an empty socket right next to the existing 486SX25 CPU that could initially house am Intel 487SX CPU.

In 486 terms the SX refers to the lack of a co-processor (FPU). A 486DX does have an integrated FPU. The 169-pin Intel 487SX was internally a 486DX (tradionally 168-pin) but with one pin extra that would disable the onboard 486SX25 so that the 487SX could take over and essentially work just like a 486DX.

The early Overdrive CPU with the name DX2ODP work the same way: they disable the on-board 486SX in order for the DX2 to take over. This way a 486SX25 could be upgraded to DX2/50. Because the original soldered 486SX could not be removed you could not sell this CPU for profit. An interesting way for Intel to increase sales by blocking you from entering the second-hand market with your old CPU!

Back in the day you could also buy a DX2ODPR CPU. The R-suffix means replacement and basically they are identical to a regular 486DX/2 but only with write-through cache since most older motherboards didn't support write-back cache.

Later Intel Overdrive CPU's had better/real functions available to upgrade an older system. For instance this 486 overdrive is equipped with a voltage regulator so it can be used on older 5V-only motherboards. Regular 486DX4's are 3.3V so they aren't suitable to run in older 5V motherboards. > Read more

Overdrive processors are build to 'Overdrive' a system with an existing surface mounted processor. In this case the Overdrive processor can be plugged in and will disable the on-board processor so it can boost an 486SX25 to an 486DX2/50.

If your motherboard has a normal zero insertion force (ZIF) socket you can probably install a normal 486DX2 as well. Today nobody would care anymore but back then you could save quite some money because Overdrive processors were more expensive. > Read more

The same as the 80486SX2/50 but with floating point unit (FPU) on-board. The FPU is used for floating point calculations and comes in to play with CAD and Excel. Back then games didn't really use it yet so Doom will run pretty much the same on a SX2. Quake does require a FPU but 486's are practically too slow to run Quake at a decent frame rate. > Read more

An unusual processor. Most people either bought the SX33, DX33 or went for the DX2/66MHz. The DX2/66MHz usually gave the extra oomph compared to the 50MHz part and was worth the extra cash.

The SX2 doesn't have an integrated floating point unit (FPU) like any other 486SX CPU. Besides that it is exactly the same as the DX2.

To get a higher clock frequency the SX2 and DX2 use a technique called 'clock doubling'. This involves a multiplier that sets the clock frequency at twice the frequency of the front side bus (FSB). In this case a SX2/50 uses a 25MHz FSB and the CPU itself runs at twice the speed. The DX2/66 uses a 33MHz bus and weirdly enough the DX4/100 uses a 33MHz bus as well. Marketing-wise the name DX4 sounded better than DX3 but technically the DX4 is actually clock tripled like the non-existing DX3 would be.

> Read more

Both the Am486DX40 and Cx486DX40 were tricky on early motherboards because of their high bus speed. Think again and look at this 50MHz part that requires a 50MHz front side bus.

I have benchmarked this CPU in a Biostar 8433UUD-A that runs the PCI-bus at 33MHz using a 2/3 divider.

A stout 50MHz front side bus sounds like it will fly but unfortunately it's not all glamour. It does perform better than a DX2/50 that runs with just a 25MHz FSB (and 25MHz PCI) but was usually not worth the trouble (at least back in the day). The safer bet would be the faster i486DX2/66 . > Read more

Same story as with the Am486DX40. It's not that this CPU ran unstable on my late UMC PCI motherboard but back then motherboards were often used at only 25MHz or 33MHz. 40MHz or even 50MHz required different (and slower) wait-states which causes an performance impact. Not to mention vesa local bus (VLB) slots that sometimes tend to be difficult at higher speeds.

The benchmarks are run with a 30MHz PCI bus in order to maintain stability. With 40MHz the system would lock up before even booting to the DOS prompt. > Read more

AMD's 40MHz 386 ran very well and was quite fast for a 386 because the front-side-bus (FSB) did its job at 40MHz as well. The Am486 at 40MHz was less successful. Ideally the 486 was plugged into a motherboard with Vesa Local Bus (VLB) slots. These slots were designed to run at 25MHz or 33MHz but could be troublesome when running at 40MHz or 50MHz. Because of this the 40MHz part was less common.

Also by the time the Am486DX40 was available most people wanted faster systems like a DX2/66. > Read more

DX, SX, what's in a name? The DX33 is almost like a SX33 except for the integrated co-processor. The co-processor (FPU, floating point unit) is used for floating point calculations. At the time when this CPU was new, regular applications didn't use the FPU so people had already enough power with the, much cheaper, SX 33MHz processor. After all Doom ran just as good on a SX as on a DX; Doom didn't use the FPU . > Read more

The SX33 was quite a good processor for a while. Both AMD and Cyrix made 33MHz parts as well but DX/2, DX/4 and even DX/5 parts from these companies can be found more easily. This is because AMD and Cyrix had to make their own 486 because Intel didn't gave the 'howto make a 486-manual' to AMD and Cyrix anymore. Both companies used the 386 as a foundation for their 486 processors and thus shipped their 486 on the market two years later. At that time the DX and DX/2 processors were becoming more popular.

Intel's SX33 processor was quite popular at the time. It is noticeably faster than the SX 25MHz. > Read more

A 25MHz 80486 SX processor isn’t particularly fast but it’s not the slowest 486 ever made (there are 16 and 20MHz versions as well). It runs just a couple MHz faster which is noticeable in applications. The SX25 could often be overclocked to 33MHz without any hassle. While 8MHz may not sound like much, it significantly improves system speed, especially since the front-side-bus (FSB) also reaches 33MHz. If the system has a VESA Local Bus (VLB) graphics card, the VLB bus also operates at 33MHz. This is an ideal situation because the processor can efficiently transfer data to the rest of the system as everything operates at the same speed.

The 486 offered several improvements over the 386DX. Both the SX and DX versions featured an internal cache, branch prediction and a five-stage pipeline. The internal cache (8KB) acts as a buffer for small data, allowing for faster access compared to regular RAM. With branch prediction the processor will 'think' in advance and get everything ready for the upcoming instructions. Pipelining is something that every modern CPU uses. See it as extinguishing a fire: without pipelining you have to fill a bucket of water, walk to the fire and dispose the water. With pipelining one fill the bucket, gives the bucket to the person standing next to him, he gives it to the person next to him and he gives it to the one that throws the water on the fire. This method of working will speed up the processor.

The 486SX did not have an internal co-processor (FPU) but in most cases it was not needed either. > Read more

20MHz; one of the slowest 486 chips available (there are 16MHz parts as well). Thankfully it didn't sell in big quantities because it is quite a lot slower than the SX25. It's only a difference of 5MHz but remember that the front-side-bus (FSB) is also 5MHz slower. This will cause the impact to be bigger. In the benchmarks below you can see the differences in the SX20, SX25 and SX33. Having a SX33 back in the day would make a person much happier than a SX20 .

The production date on this 486SX is 21th week of 1991 which makes it an early processor. I don't know the exact launch date of the 486SX but generally it's believed that it was mid-1991. > Read more

This is not a 486 like all the others. Technically it's a 386 with 486 instructions with 1KB L1 cache. It fits in a regular 386 motherboard (if the motherboard supports it) and runs a bit faster compared to a normal 386. Although it's not as fast as a 486DX33 because it only has 1KB L1 cache instead of 8KB.

I've read that the 'BB' version (as seen on the picture) could be buggy when using Cyrix FasMath EMC87 NPX, Cyrix FasMath Cx83D87 NPX or IIT-3C87 NPX co-processors. I never ran this CPU as a main system so I only tested it for a few hours and during that time it ran just fine. The 'AB' version does not have these problems. > Read more