leilei wrote:I thought the Cyrix5x86-133 was the most powerful CPU available for Socket3?
Yes, I was going to mention the Cyrix 5x86 133 and Am5x86-P75+ as the other contenders for the title, but I decided to keep it short. I had used the qualifier "might be", so I figured it was clear that it wasn't the undisputed title holder.
AMD sold Am486s that used FSBs of up to 40 MHz, which is why they had a DX 40, DX2 80, and DX4 120. They also sold an "Enhanced" Am486 series that came in DX2 66 and 80 MHz, and DX4 75, 100 and 120 MHz. They differed from regular Am486s by using write-back cache instead of write-through, and later doubled the cache from 8 KB to 16 KB. They can be identified by an "8B" or "16B" (for write-back), instead of "8T" (write-through) after the DX2/4 label. For instance, my A80486DX2-80NV8T has 8 KB of write-through cache and runs at 3.3 volts (indicated by the "N").
The Am5x86-P75 was an Enhanced Am486 running at 133 MHz (33 x 4), and the Am5x86-P75+ ran at 150 MHz (50 x 3). If your motherboard could handle a 50 MHz FSB, the extra 67 MHz of internal clock speed and extra 17 MHz of FSB, combined with a 16 KB write-back cache, gave it performance somewhere between a P75 and P90, so it was probably close to the performance of an 83 MHz Pentium OverDrive. The Cyrix 5x86 133 also performed better than a P75 in most applications. All three are pretty bad ass Socket 3 CPUs.
Today entirely the maniac there is no excuse with the article.
I just took a closer look at the Engineering Samples I recently acquired, and I noticed something odd on four of the Pentiums. They are listed as "A8050260-90", "A8050260-120" and "A8050266-133".
Early Intel CPUs often had a prefix in front of the model number, like "D8086" or "P8085". A prefix of P referred to a plastic package, and C to a ceramic package (not to be confused with a C in the middle of the model number, where it refered to "CMOS" to differentiate it from the HMOS models). Starting with the 286, most desktop CPUs had a prefix of "A", so it's normal to see "A80286", "A80386" and "A80486". Because you can't trademark a number, Intel's competitors released their own, compatible 386s and 486s, using model numbers such as "Am386" or "Cx486". To stop that practice, Intel called their 80586 processor "Pentium" and trademarked the name to prevent copycats. Even the model numbers on the CPUs weren't "A80586", but "A80501" for the original 5 volt, Socket 4 Pentiums, "A80502" for the 3.3 volt, Socket 5 CPUs, and "A80503" for the 2.9 volt, Socket 7, Pentium MMX CPUs. This continued with the Pentium Pro (A80521), Pentium II (A80522 for Klamath, A80523 for Deschutes/Tonga, A80524 for Dixon), and some Pentium IIIs (A80525 for Katmai, RB80526 for Coppermine). These codes were phased out during the Pentium III era, so most Socket 370 Pentium IIIs, and every generation since, just have straightforward statistics (such as "1000/256/133/1.75" on my Tualatin Pentium III). The A80xxx numbers are still used internally at Intel, but it looks like they're only printed on engineering samples now.
The normal versions of the processors I mentioned above read "A8050290", "A80502120" and "A80502133". Most of my CPUs don't have a hyphen, but there is sometimes a hyphen before the MHz rating, such as "A80502-90". So... what does "A8050260-90" mean?
I have a 387 that specifies a MHz range for the 386 it can be paired with ("A80387DX 16-33"), but it's not very likely that these CPUs are specifying a MHz range of 60-90, 60-120 and 66-133. Since Pentiums used a Front Side Bus of either 60 or 66 MHz, it looks like the first number is the external frequency and the second number is the internal frequency. This might have been the for the benefit of the people testing the engineering samples, but it's also possible that Intel intended to label their shipping CPUs this way. In those days, you had to set the FSB and multiplier with jumpers on the motherboard, so it made sense to put the FSB in the model number. Intel may have decided to drop the FSB from the model number before shipping the Pentium since the math to figure out whether CPU speed was a multiple of 60 or 66 wasn't very hard, and it wouldn't be necessary for disambiguation unless the Pentium reached 300 MHz (60 x 5 or 66 x 4.5), which it never did.
So yeah, I'm actually much more of a hardware nerd than a software nerd.
Today entirely the maniac there is no excuse with the article.
My trading buddy who used to work at Intel gave me a bunch of new CPUs and other old computer stuff today. Among the CPUs were two new engineering samples: another Pentium 133 with sSpec Q0773 instead of the Q0772s that he gave me previously, and another Slot 1 marked Intel Confidential (Q818ES).
The Slot 1 CPU was the most interesting because it has a Pentium II cartridge casing, but the top says 80525PY500512, so it's a 500 MHz CPU with 512 KB of L2 cache. The Pentium II stopped at 450 MHz, so there was the tantalizing possibility that it was an unreleased Pentium II 500, but 80525 is the code for Katmai, which is a Pentium III core. The difference between Pentium II and Pentium III, for the curious, is that the Pentium III introduced SSE (which was known as Katmai New Instructions, or KNI, at the time), whereas the Pentium II only supported MMX. For P2s, Klamath (350 nm) had a die shrink to Deschutes (250 nm), whereas with P3s, Katmai (250 nm) is a new core with new instructions which was later shrunk to Coppermine (180 nm) and Tualatin (130 nm).
So, despite the Pentium II casing, the model number seems to indicate that this a 500 MHz P3 Katmai, not a 500 MHz P2 Deschutes. It's possible that Intel hadn't decided to brand Katmai-based CPUs as Pentium IIIs when the cartridge for this engineering sample was printed, or they may not have produced any Pentium III cartridge casings yet. I think it's really cool that this CPU says Pentium II on the front and has Pentium III details printed on the top.
Today entirely the maniac there is no excuse with the article.
