News view more...
again it's been a really long time since my last update. I really appologize. My appologies especially go to those true fans who still give me new exhibits, for free. I'd like to mention one distinct donor, Mr. Fritz Weld, who again -after years- sent me a huge pile of truly remarkable chips to give me a trigger for continuing spreading the spirit of historical computing. So here is this year's first update - again hoping this won't be the last one:
The 486DLC family was a truly remarkable family which gained a very ambiguous reputation. I can remember the sad face of my classmate who proudly showed me his brand-new machine in 1994, showing a 486 label on its front. After opening the case, the computer looked pretty much like a plain 386DX computer. The only notable things inside were a then-rarely found VLB on a 386 board and this Cyrix CPU being labeled as "486DLC". In fact, this concept was a very clever one from today's perspective: Porting a some-kind-of 486 core to a mature and inexpensive platform like the 386 was back then, adding some cache to make it speedy, and still being able to add a cheap 387 CoProcessor if one needed a floating point unit.
This is the Cyrix incarnation of its 486 generation CPU. Unlike AMD, Cyrix introduced its own implementation of a 486 chip, including all the trouble caused by being "different" from Intel chips on software that relied on specific hardware characterisitcs which was common back then. Although being largely compatible with the i486 instruction set, you still would call this chip some kind of hybrid: Yes, it was able to execute 486 instructions and had a 32bit bus as well; nevertheless this chip suffered from its typically 386-ISA-based low-cost plattform it usually was plugged into. Interestingly, this variant was released without any speed marks on it and without its labels written in gold letters.
Texas Instruments TX486DLC
Although the Texas Instruments chip was basically identical to the abovementioned Cyrix 486DLC, it still featured a much larger cache which heavily pushed the chip's performance. 8kb instead of just 1kb made this chip keep up with a "genuine" Intel 486 chip (in a 25 MHz SX Variant though). Still more than with the Cyrix chip, the 386DX plattform became a major bottleneck.
Anyway, it didn't last a single year from then until a true 486 became available from Cyrix, ST and Ti. Being a fab-less company, Cyrix had to depend on other chip manufacturers to make their chips. Instead of paying them, Cyrix always chose a different aproach by allowing the manufacturing fabs to sell the chips under their own brand name.
I hope you enjoyed this update. Promising any updates freuqence would probaly severely harm my credibiliy, so I just say: World is turning, and with every turn there are more CPUs to be discovered :)
it's been a really long time since my last update. I really appologize. Lots to do @work, got married and alike. No more promises concerning frequent updates, but, however, cpu-museum.de was kept alive and is still running. In fact I was quite suprised that spammers finally seem to refrain from filling up my froum, so out of pure surprise and really bad weather outside these days I decided to spend some time on my almost forgotten hobby. Actually there are some facts to mention regarding the past alomst 3 years since my last update: Yes, Hard disk sizes have increased. The time I wrote my last update, I was using something around 100GB in my fileserver, together with a -can't remember- 850MB or 4GB SCSI hard disk in my desktop for noise reasons. Don't worry, there already were >100GB drives out there back then. Today, there's a ~4TB file server remembering all TV trash from the past days up in the attic. No more CRT monitor, but a cheap flat 19" display. However, I noticed hardly any change in CPU frequencies, nor architecture, but only in power consumption and number of cores. I finally got myself a portable computing thingy, a really neat Intel-Atom powered NC10 NetBook which seems to me to be a future trend (just like became true with my Pentium-M desktop back in 2004 :P ). Well, enough said, here's the long-waiting update (hopefully not the only for this year):
In my last update I was looking forward to some late K6 chips. As a lucky coincident these chips now turn 10 years old, so here they are:
AMD K6-2/233 AMZ
This is one of the eraliest K6-2 chips available. Compared to the K6, the K6-2 got a revolutionary new implementation of a floating point variant of MMX, called 3Dnow! which provided vector calculation support in hardware for floating point numbers. Unfortunately there was only little support from software vendors the days when the K6-2 was new, so in reality the difference between a K6 and a K6-2 wars almost null, and so became the price difference in later days. However, 3Dnow! became a victim of competitor’s marketing and especially market shares since its original aim was revolutionary: Providing a floating-point vector unit in a CPU enabled efficient calculation of modern 3D and multi-media content which has become ubiquitous in today’s software.
AMD K6-2/366 AFR
The 366 MHz variant is a really interesting incarnation of the K6-2, because AMD was still using the Socket-7 bus to transmit all computational data. With only 64kB L1 cache and a bus speed of only 66MHz (which was introduced back in 1993 with the Pentium-66) this CPU is heavily I/O bound and probably hardly worth the extra money unless you really knew why you need the 366MHz type. And so did the technically skilled people and chose this chip for its high multiplier which –together with the new lower voltage core and upcoming flexible mainboards- allowed nice overclocking.
AMD K6-2/550 AGR
The last model of this family. Interestingly this chip remained available in stores for quite some time, mainly in lowest-cost desktops. With the 550MHz incarnation, the old Socket-7 became a severe bottleneck which could only be overcome by adding the L2 cache to the CPU and move away traffic from the bus. In fact, there actually were K6-2 chips with L2 on-die cache (labelled K6-3 and K6-2+), but these were targeted at the notebook market in order to make the mass market move on to the Socket-A platform with the Athlon and Duron chips.
So that’s it for this time, hope to provide more new chips in more regular intervals from now on :)
recent months have showed a true change in CPU manufacturer's policy and roadmaps, introducing multi-cores instead of single higher-speed cores, and with apple's switch away from PowerPC towards Intel's Pentium-M based product line (now called "Core") which has its roots reaching back to the Pentium-III/ Pentium-II/P-Pro, the old i686 architecture received another important incarnation and appreciation. Along this long way from P-II up to Pentium-M there has been a loyal companion named "Celeron" for more than 8 years now (since 1998), even if Celeron's way has not always been identical to i686. Today, both the Pentium-M as well as the Pentium-IV have a Celeron counterpart, namely Celeron-M and Celeron-D. Not only for this reason we will have a short trip back to the earlier days when the Celeron matured:
Intel Celeron 533MHz PGA (Mendocino)
The 533MHz variant was the last to be based on the second core of the Celeron, after the infamous Covington (which was a Pentium-II lacking the whole L2-cache). Interestingly this chip managed to get this chip design to ramp up to 533MHz while the Pentium-II which was based on the same core quit at only 450MHz. During the short period of Katmai-based P-III chips, the last of the Celeron-Mendocino chips probably had the fastest caches of all Intel chips around, and if you were computing on small working sets the Celeron was not too far behind its greater brothers. With the Celeron-533 fading away from the computer market, the PPGA package (introduced with the Pentium-MMX and Pentium-166) which had been around for years disappeared and was replaced with the much less robust FC-PGA package which allowed for better cooling measures of the ever-shrinking core sizes.
Intel Celeron 566MHz FC-PGA
Back in 2000 when the new Coppermine-based Celeron pieces appeared, nobody seemed to make friends with them. The public focus was on the Athlon/Pentium-III battle, the K6-2/3 was still serving the low-end market, and the brand new Duron seemed to grab the whole rest of the mid-range home user market. On a close view, accounting for the pure technical data of this chip, it was just not sensible to have a celeron at all. However, 6 years later, the Celeron-Coppermine, and especially the Celeron-533 made perfect sense: For obvious reason the Mendocino era was over, and the Celeron delivered the MS-certified business computer market for computationally low-powered desktops. While the Athlon was a monster regarding power consumption (and the Duron to some extent somehow as well), and the mainboards for AMD based systems weren't the most relieable in past years, this chip just delivered acceptable computing power together with low power consumption on a quite modern mainboard platform; just try to attach >130GB hard disks to a K6-2 mainboard or try to find DDR- or RD-RAM based K6 boards. I still have a Celeron-533 in use here, and it perfectly fits my other system, a Pentium-M-Desktop, from a noise level :)
Intel Celeron 700MHz FC-PGA
The 700 MHz incarnation of the Celeron showed the Celeron's most pressing problem of that days: The acceptably fast core on a fast plattform was connected via an outdated bus speed, and the on-chip cache was no longer able to compensate for this. The 700MHz part was one of the last Celeron chips that were using a 66MHz bus, and in fact any other increase in core speed hardly made sense as the core would just wait a little bit longer for data from the rest of the system. However, it was sold, and however in very rare cases it still made sense to have a Celeron-700. Fortunatelly with the 800MHz version Intel switched to a 100MHz bus speed to finally make this chip competitive again. The rest was acceptable power consumption paired with SSE -- which nobody used back in the days it was sold.
late 90ies and early 2000s were a key era for AMD to gain a respectable market share, laying the foundation for AMD's current success in the 64bit x86 market. For that reason the next update will give an introduction to late K6 and early AMD Athlon developments. So make sure to check back from time to time for some more nice ancient computing devices :)
The Intenet has been a nice peaceful place for many years, and it was a great pleasure to publish content, give people the possibility to build a collector's community and give everyone a place for exchanging information or just talking.
Seemingly these days are past, and while my personal inbox is drowning in >3000 spam mails daily I harldy find time to separate nice mails from useless garbage, and even filters reach their limit.
In recent days there have been more posts than ever in this forum, though unfortunetly 90% of the posts were spam again. And while I was cleaning my inbox and deleting spam posts, I had to delete more than 500 registered bot-"users" whose only intent was to be listed in the memberlist with a signature pointing to garbage again. Garbage everywhere, cleaning eats up so much time so that every time I want to care for the museum I end up being frustrated, and giving up before I created one single line of content or one single picture, not to mention the countless emails and posts I didn't answer, which I heavily regret.
To give me back some time and make this forum a little more spam-resistant I decided to put some restrictions on this forum:
- If you live in russia and have a russian email address, please get another adress that does not end with ".ru". Any registrants with their email address ending with ".ru" will be discarded without notice. sorry.
- If you come from Kiev, don't enter anything in your profile about your location. People with "Kiev" in their profile will get deleted as well because seemingly most spammers use Kiev as their home location in their profiles.
- If you sign up and have a link in your signature, you will be deleted if you don't post a single forum contribution within the first 2 weeks after registration.
- Only registered users will be able to create topics and reply to posts in "Techtalk" and "Helpdesk".
- If you feel like contacting me (-> Grampa ), use PM in this forum, not email. If you still chose email, please just be patient, I'll try to get that pile sorted once per month and you'll hopefully be answered within some weeks.
... I wish I didn't have to put these rules on this forum, but I just can't handle all the garbage destroying the forum without these anti-spam measures. I hope you -and especially my visitors from russia- don't mind these restrictions. I hope this forum will become a peaceful community place again without any spamming interruptions. Enjoy, and hope to see you again in this finally peaceful forum :)
Christmas is over, new year has begun, and finally some time to add some new pieces to my museum. This time it's some mid-90ies 486 class processors by AMD:
When AMD released this nice part back in late 1994, it was a truly reasonable solution giving nice performance at low prices compared to the far more expensive 100MHz Intel 486DX/4, not to mention the brand new (but flawed) Pentium. With the outbreak of Pentium-FDIV-PR-disaster in late 1994 the Am486DX2-80 was a nice alternative for all those confused customers that preferred not to buy an Intel chip. Unlike all later chips this special 486DX2-80 variant still required 5 Volts like early Intel 486 chips did, so this chip ran even on older mainboards -provided that they supported a 40MHz bus- which made them nice upgrade parts as well. 40MHz bus speed traditionally lead to problems in earlier DX40 days, mainly related to inappropriate waitstates for memory and VLB cards, but as of late 1994 mainboards had matured so that the 40MHz bus became only rarely an issue. This nice part was produced in early november 1994.
AMD Am486DX2-80 V8T 3Volt
With the arrival of higher clock frequencies the need for smaller chip structures with lower core voltages became increasingly apparent in order to increase reliability and reduce power consumption. Starting from the 80MHz model, AMD 486 chips were mainly sold as 3Volt versions which required modern mainboards. Apart from its voltage, there was only little difference to any other plain DX2 chip. This particular chip was produced in mid-november 1994
AMD Am486DX4-100 NV8T 3Volt
AMD reached the 100MHz mark in september 1994, more than half a year after Intel's DX4-100. It introduced a higher multiplier of 3, which makes it run at 3x33MHz = 100MHz on a 33MHz bus. The DX4 nomenclature was eventually borrowed from intel; originally the "DX3" was intended to provide a multiplier of 2.5, so the number after "DX" does not show the actuall multiplier like in DX2 days. This chip was made in May 1995.
AMD Am486DX4-100 SV16B 3Volt
With the 486DX4-100 AMD 486 chips became available in two very different versions. Unlike all earlier 486 chips, this one (like the Intel counterpart) was equipped with a double cache size, i.e. 16kB of unified cache which provided a neat performance bonus. Both the later 120MHz model and the Am486DX5/Am5x86-P75 were based on this core. This nice piece was produced in August 1995.
With the arrival of a new year I hope to finally find more time to care for the museum. I recently acquired quite a pile of interesting chips from forum user "minuteman" who discontinued his own museum. So make sure to check back from time to time for some more nice ancient computing devices :)
long time no see, though never forgetting my loyal visitors :) This time I am seizeing the sad opportunity of a recent end-of-life announcement for a small update:
Via Technologies Cyrix-III 667MHz (Samuel-1)
Via Technology recently announced end of life for their socketed C3 processors. There has been quite a troublesome story behind this chip until Via introduced their "brand new" Cyrix-III. After buying the remains of Cyrix Corp. along with Cyrix' exceptional 6x86 out-of-order CISC design from National Semiconductors, Via's first Cyrix-based design "Joshua" turned out non-competitive. Taking Cyrix's name, Via decided to give its newly aquired Centaur design a chance which essentially was a Socket-7 C6-2 adaption to Socket-370. The first incarnation -Samuel-1- can definetly be considered intermediate, lacking a level-2 cache and its FPU running at half speed. This CPU particular CPU was introduced in Summer 2000 and runs at 667 MHz.
Via Technologies C3 733MHz (Samuel-1)
Despite the fact that Via's C3 chips never reached the mass market, it was still able to hold a constant market share, notably in the low-noise low-energy sector, like silent living room entertainment PCs or gateways. Via recognized what customers liked best about the c3 -the exceptionally low power consumption- and while Intel was moving away from Socket-370, Via introduced Epia, a silent x86 platform with reduced form factor for small PCs and soldered-on C3 chips which still seems to go strong. This particular C3 is one of the last implementations of Via's early Samuel-1 design. Lastest C3 variants go well up to 1.4 GHz with new dual-cpu and dual-core solutions being announced. After more than 5 years of availability, Via put an end to Socket-370 for two reasons: Via's license agreement with Intel for socket-370 ends in early 2006, and there are hardly any new socket-370 boards available except Via Epia boards. Latest incarnations of the C3 design are all introduced as soldered-on chips on Via boards, and obviously there are still another some years of innovation based on the C3 design to be expected.
Strongly looking forward to the winter/christmas season, there will finally be more time to spend on my museum again, so there will be many more updates during the next weeks. Hope to see you again here in the news forum :)
Continuing the series of computational prominent anniversaries, there has been (or will be) another birthday:
- ATI has celebrated its 20th anniversary, even though its actual birthday is in August. One of their first products was a card called VGA Wonder, an 8-bit ISA card for IBM PCs which is shown in the picture. ATI's Rage64 which appeared about 10years ago and received several extensions like basic 3D support became something like ATI's breakthrough. With its Radeon Series ATI has become one of the three market leaders for graphics solution for PCs and Apple Mac today.
When ATI was founded back in 1985, computers were radically different from today from a graphics capabilities perspective. While today's Graphics cards provide computing power almost as high as supercomputers from early 80ies. PC graphics chips from mid-80ies until 1990 were very basic, mostly capable of text mode only or showing a small pallete of colors on low screen resolutions, clearly inferiour to home computers like Commodore's Amiga. With the appearance of graphical user interfaces on the PC platform like OS/2 or Windows, chips appeared that could handle basic drawing tasks in hardware, like drawing lines or filling rectangles. But still, most graphical calculations had to be done by the central processing unit.
Even though Intel's 80386 was already introduced in 1985, it didn't play an important role for PC business during its first years. 8088- and 8086-based PCs were sold well until 1990, and the 80286 was a viable option for enthusiasts. IBM's original PC was equiped with an 8088, a stripped-down version of the 8086 with an 8-bit external bus to speed up the design of mainboards and keep costs down. The 8088 was offered in several packaging variants and speed grades: grey or dark ceramic (don't know of any white ceramic models), ranging from 5MHz up to 10MHz. The 8088 show was made in 1986 and runs at 8MHz max.
Intel's 8086 in plastic package with full 16bit external bus provides a better connection to the system bus and peripherals, thus leading to slightly increased performance. From an architectural point you can clearly see its historic descent -- a reduced non-orthogonal CISC approach with many instructions allowing only specific addressing modes, and some instructions always storing their result into %eax which is a typical accumulator-based approach that is usually found in simple microcontrollers, much like a Motorola mc68hc11. In addition memory accesses are done using segments and offsets, quite the opposite of easy linear memory access. However, Intel's x86 family became an unequalled success story, and still today any Pentium-4 or Athlon-64 relies on the same instruction set that the 8086 introduced. The 8086 on display was made in 1985 and runs at 5MHz max.
If you ever come across a working 8088 or 8086 system, give it a chance and try to run some basic graphics programs like a GUI -- with an unaccelerated graphics card you can actually see how lines are drawn, pixel by pixel, almost like slow motion :)
... there are some more anniversaries to celebrate, so make sure to come back ;-)
There have been some company's and product's birthdays during the last weeks which made me think about showing you the computational environment those products and companies arose from:
- Microsoft has become 30yrs old these days. Interestingly this important anniversary has passed almost completely without notice. Back in 1975 when Micro-Soft was "founded" by Bill Gates and Paul Allen mainly to take the chance to provide an interpreter for the programming language basic for the not-yet available Altair 8800 which lateron became a milestone in computer history as the first home computer. Bill and Paul created a Altair-8800-Simulator for a PDP-10 just by taking the specs of the Altair8800 and developed the basic interpreter which was released in spring 1975, though their BASIC still lacked a boot loader and the name Microsoft wasn't yet officially registered. The boot loader was written by Bill Bates lateron and published in "Computer Notes" magazine as --hardly believable when looking at current MS software-- the smallest an fastest boot loader for the Altair. Another interesting thing about this release was that it was _sold_ -- in earlier days, software was bundled with the computer, or you wrote programs on your own or made copies from other sources. This new model of selling software soon lead to the term "software piracy" which was first addressed in Gates' "An Open Letter to Hobbyists" in 1976.
Intel D8080A '74 '79
Computing infrastructure was scarce, slow, huge and expensive back when Bill and Paul wrote their basic. Home computers were de-facto non-existant, but some companies just laid the foundation for the coming age of home computers, like Motorola just released their 6800, the grand-father of many successful chips found in Commodores, Ataris or even early Sun Workstations and Apples. Intel was still young and started designing CPUs only some years before. The 8080 which was built into the Altair 8800 was something like a milestone in PC-history, and even current Pentium-4 or Athlon-64 machines owe much to this early chip. The original Altair 8800 was powered with a 2MHz intel-8080 chip and 256Bytes(!) of RAM. The chip shown is certainly not the exact model that was initially used for the Altair, but in later models iD8080 and second-sourced C8080 and P8080 from AMD were used as well. There has been a huge variety of 8080 chips over its lifetime, and many other companies made 8080s second-source. The chip shown is a late dark-ceramic 8080 from 1988 which runs at 2MHz max.
The 8080 wasn't only available in dark or white ceramic packaging, but also in Plastic. You can easily tell which material the chip housing is made of by the first letter: "C" stands for white ceramic, "D" stands for dark ceramic, and "P" stands for plastics. The "A" suffix denotes a chip revision; non-"A" 8080 chips are exceptionally rare.
Siemens SAB 8080 A-C
This Siemens-8080A-C is a typical second-source 8080 that was made by Intel licensing the core to other companies, like it is done today with IP cores like ARM. Unfortunatelly this chip suffers from corrosion, so maybe some stainless material would have been a better solution for a then-horribly-expensive thing like a CPU. However, I bet noone at Siemens ever though this chip would still live after such a long time. In the mean time Siemens' semiconductor division turned into Infineon. Note the old "Siemens-Halske" logo -- ancient :)
Unlike the Siemens chip which was never used for Altairs, some late Altair 8800 models were using a second-sourced 8080 by AMD. Back in these days, there wasn't as much rivalry between AMD and Intel like today. I bet if Intel knew AMD would grow that strong, they would surely have refused AMD as licencee. This AMD 8080 is in plastic packaging, runs at 2MHz max. and was made in 1986.
... to be continued these days with more birthdays -> click here.
this time I want to show you a little bit of CPU history of the last seconds of the past century:.
Intel P-III 450/512/100 SEC (Katmai)
Basically this CPU was not much different from the good old Pentium-II. While back in Pentium-Classic days MMX was packed together with a cache doubling to make every one feel it was faster, the Pentium-III initially offered almost no benefit. just being quite a bit more expensive than an equally-clocked P-II. While MMX was co-using FPU registers and could be used transparrently without changes to the Operating system, SSE introduced a number of additional registers that operating systems should know about when doing context switches, to avoid strange system behaviour. So SSE was a marketing gag back then, ending up making the Internet faster and more colorful. Today it shows that SSE was very beneficial for the x86 architecture as the number of applications processing vector data has massively increased, namely video and audio processing as well as heavily-parallel array operations. SSE was lateron adapted by AMD and received several extensions for additional data types and operations named SSE2 and SSE3.
Intel P-III 500/512/100 SEC (Katmai / Boxed)
Pentium-III's initial implementation was based on the earlier Pentium-II cores with external Cache chips soldered onto the module's PCB. It was available in 450-600MHz speed grades called "Katmai". Like earlier chips from Intel, P-III cartridges were optionally available as "boxed" variants with a cooler + fan attached like this one. Katmai was Intel's last slot-only based x86 CPU; With its successor Coppermine, the cartridge became no more than an optional adaptor, which disappeared once Socket370 became popular. The P-6 architecture lived on in Coppermine and Tualatin and eventually survived P-IV heavily modified as today's Pentium-M.
There will be some prominent IT Company and product anniversaries these days, so next update will be related to the "computing environment" of those days. Make sure to come back again!
This time I'll give you a short overview of mid-late 90ies AMD chips:
AMD K5-PR75 ABR /w heat-spreader
On the first view, this one looks extremely unspectacular: Another Socket-5/7 chip that just came too late to be competitive, and performance slightly behind Intel counterparts on a PR-Basis. Going deeper into its history, there has been an interesting story behind it reaching far back to the 80ies and AMD's 29k architecture. The 29k series was a load/store design inspired by Berkeley RISC design, featuring an enormous register set of no less than 128 registers with a flexible register window architecture. Somehow during mid-90ies AMD dropped its 29k line (probably low acceptance or alike) in favour of x86 architecture. Right before that in 1993 Intel won a lawsuit against AMD cloning Intel chips. Together with Intel's x86-to-Risc translation and Out-of-Order Execution attempts introduced with the Pentium-Pro AMD adapted the 29k RISC core to x86, which resulted in the K5 line of Socket-5/7 chips which was introduced in 1996.
AMD K6-266 AFR
Right before dropping AMD's K6 line of processors which were "borrowed" from NexGen's Nx686, AMD switched to a new manufacturing technology to reduce power consumption and reach higher clock rates. That way, late K6 chips required only 2.2Volts and ran quite cool; it turned out that with this technology its successor K6-2 was able to go beyond 500MHz. This nice piece was made in autumn 1998.
AMD K6-2/300 AFR 66
A truly unique K6-2 variant: Running at 300MHz, this one was unable to be operated at the new higher system bus speeds of 100MHz. As a result, these were better avoided unless you really needed a 66MHz bus variant (which was unlikely as most 2.2Volt mainboards supported higher speeds) or the tiny difference in price was of any relevance. However, if you were lucky, you could still operate it at 100MHz bus. This processor was made in autumn 1998.
AMD K6-2/533 AFX
Almost last of a series of successfull budget chips, before AMD entered back into the high-performance market again with the Athlon. With its 533MHz this chip ran pretty close to its limits, and with its tiny 64k cache on-chip there was hardly any difference to a 500MHz as it was heavily slowed down by its 95MHz system bus. This one was made in december 2000
Next week there'll be an Intel update, so make sure to come back again.